| History log of /src/sys/ufs/lfs |
| Revision | Date | Author | Comments |
| 1.7 | 07-Jan-2025 |
andvar | s/remaing/remaining/ s/containg/containing/, mainly in comments.
|
| 1.6 | 09-Feb-2024 |
andvar | branches: 1.6.2;
fix spelling mistakes, mainly in comments and log messages.
|
| 1.5 | 11-Dec-2005 |
christos | merge ktrace-lwp.
|
| 1.4 | 24-Feb-2004 |
wiz | parameter with two es. From Peter Postma.
|
| 1.3 | 05-Jul-2001 |
toshii | branches: 1.3.22;
Fix typo. s/extention/extension/
|
| 1.2 | 10-Apr-1999 |
perseant | branches: 1.2.14;
Change the reference to "newlfs" in the CHANGES file to the correct "newfs_lfs"
|
| 1.1 | 15-Mar-1999 |
perseant | branches: 1.1.4;
New CHANGES files that describes briefly all nontrivial changes made to
the LFS since the 4.4lite2 code was merged into NetBSD.
TODO updated to remove everything marked DONE in 4.4, and add in a list
of more current things to do.
Get rid of comments about the cleaner syscall code and missing fragment
support from README.
|
| 1.1.4.1 | 21-Jun-1999 |
thorpej | Sync w/ -current.
|
| 1.2.14.1 | 24-Aug-2001 |
nathanw | Catch up with -current.
|
| 1.3.22.3 | 21-Sep-2004 |
skrll | Fix the sync with head I botched.
|
| 1.3.22.2 | 18-Sep-2004 |
skrll | Sync with HEAD.
|
| 1.3.22.1 | 03-Aug-2004 |
skrll | Sync with HEAD
|
| 1.6.2.1 | 02-Aug-2025 |
perseant | Sync with HEAD
|
| 1.3 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.2 | 08-Jun-2013 |
dholland | branches: 1.2.10;
Split the definitions suitable for userland out of ulfs_inode.h into
lfs_inode.h. Since fsck_lfs, newfs_lfs, and lfs_cleanerd want to reuse
the inode structure for their own internal use, and some of them share
parts of the kernel code as well, the best way forward is to provide a
relatively sanitized header that doesn't bring in stray material.
Shuffle a few other definitions around so that lfs_inode.h depends
only on lfs.h.
Install lfs_inode.h into /usr/include.
|
| 1.1 | 12-Jun-1998 |
cgd | branches: 1.1.188; 1.1.198;
Rework the way kernel include files are installed. In the new method,
as with user-land programs, include files are installed by each directory
in the tree that has includes to install. (This allows more flexibility
as to what gets installed, makes 'partial installs' easier, and gives us
more options as to which machines' includes get installed at any given
time.) The old SYS_INCLUDES={symlinks,copies} behaviours are _both_
still supported, though at least one bug in the 'symlinks' case is
fixed by this change. Include files can't be build before installation,
so directories that have includes as targets (e.g. dev/pci) have to move
those targets into a different Makefile.
|
| 1.1.198.2 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.1.198.1 | 23-Jun-2013 |
tls | resync from head
|
| 1.1.188.1 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.2.10.1 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.3 | 15-Mar-1999 |
perseant | New CHANGES files that describes briefly all nontrivial changes made to
the LFS since the 4.4lite2 code was merged into NetBSD.
TODO updated to remove everything marked DONE in 4.4, and add in a list
of more current things to do.
Get rid of comments about the cleaner syscall code and missing fragment
support from README.
|
| 1.2 | 29-Jun-1994 |
cgd | New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
|
| 1.1 | 08-Jun-1994 |
mycroft | branches: 1.1.1;
Update to 4.4-Lite fs code, with local changes.
|
| 1.1.1.1 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite for reference
|
| 1.2 | 06-Jun-2013 |
dholland | branches: 1.2.2; 1.2.10;
Update the line-count standings.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.2.10.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.2.10.1 | 06-Jun-2013 |
yamt | file README.wc was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.2.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.2.2.1 | 06-Jun-2013 |
tls | file README.wc was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.10 | 11-Dec-2005 |
christos | merge ktrace-lwp.
|
| 1.9 | 01-Apr-2005 |
perseant | branches: 1.9.2;
Protect various per-fs structures with fs->lfs_interlock simple_lock, to
improve behavior in the multiprocessor case. Add debugging segment-lock
assertion statements.
|
| 1.8 | 26-Feb-2005 |
perseant | branches: 1.8.2;
Various minor LFS improvements:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statvfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
|
| 1.7 | 23-Feb-2003 |
perseant | branches: 1.7.2; 1.7.8; 1.7.10; 1.7.12;
Fix a buffer overflow bug in the LFS_UBC case that manifested itself
either as a mysterious UVM error or as "panic: dirty bufs". Verify
maximum size in lfs_malloc.
Teach lfs_updatemeta and lfs_shellsort about oversized cluster blocks from
lfs_gop_write.
When unwiring pages in lfs_gop_write, deactivate them, under the theory
that the pagedaemon wanted to free them last we knew.
|
| 1.6 | 17-Feb-2003 |
perseant | Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now
(there are still some details to work out) but expect that to go
away soon. To support these basic changes (creation of lfs_putpages,
lfs_gop_write, mods to lfs_balloc) several other changes were made, to
wit:
* Create a writer daemon kernel thread whose purpose is to handle page
writes for the pagedaemon, but which also takes over some of the
functions of lfs_check(). This thread is started the first time an
LFS is mounted.
* Add a "flags" parameter to GOP_SIZE. Current values are
GOP_SIZE_READ, meaning that the call should return the size of the
in-core version of the file, and GOP_SIZE_WRITE, meaning that it
should return the on-disk size. One of GOP_SIZE_READ or
GOP_SIZE_WRITE must be specified.
* Instead of using malloc(...M_WAITOK) for everything, reserve enough
resources to get by and use malloc(...M_NOWAIT), using the reserves if
necessary. Use the pool subsystem for structures small enough that
this is feasible. This also obsoletes LFS_THROTTLE.
And a few that are not strictly necessary:
* Moves the LFS inode extensions off onto a separately allocated
structure; getting closer to LFS as an LKM. "Welcome to 1.6O."
* Unified GOP_ALLOC between FFS and LFS.
* Update LFS copyright headers to correct values.
* Actually cast to unsigned in lfs_shellsort, like the comment says.
* Keep track of which segments were empty before the previous
checkpoint; any segments that pass two checkpoints both dirty and
empty can be summarily cleaned. Do this. Right now lfs_segclean
still works, but this should be turned into an effectless
compatibility syscall.
|
| 1.5 | 13-Jul-2001 |
perseant | Merge the short-lived perseant-lfsv2 branch into the trunk.
Kernels and tools understand both v1 and v2 filesystems; newfs_lfs
generates v2 by default. Changes for the v2 layout include:
- Segments of non-PO2 size and arbitrary block offset, so these can be
matched to convenient physical characteristics of the partition (e.g.,
stripe or track size and offset).
- Address by fragment instead of by disk sector, paving the way for
non-512-byte-sector devices. In theory fragments can be as large
as you like, though in reality they must be smaller than MAXBSIZE in size.
- Use serial number and filesystem identifier to ensure that roll-forward
doesn't get old data and think it's new. Roll-forward is enabled for
v2 filesystems, though not for v1 filesystems by default.
- The inode free list is now a tailq, paving the way for undelete (undelete
is not yet implemented, but can be without further non-backwards-compatible
changes to disk structures).
- Inode atime information is kept in the Ifile, instead of on the inode;
that is, the inode is never written *just* because atime was changed.
Because of this the inodes remain near the file data on the disk, rather
than wandering all over as the disk is read repeatedly. This speeds up
repeated reads by a small but noticeable amount.
Other changes of note include:
- The ifile written by newfs_lfs can now be of arbitrary length, it is no
longer restricted to a single indirect block.
- Fixed an old bug where ctime was changed every time a vnode was created.
I need to look more closely to make sure that the times are only updated
during write(2) and friends, not after-the-fact during a segment write,
and certainly not by the cleaner.
|
| 1.4 | 17-Nov-2000 |
perseant | branches: 1.4.2; 1.4.4; 1.4.6;
Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes.
(PR #11468). In the case of fragment allocation, check to see if enough
space is available before extending a fragment already scheduled for writing.
The locked_queue_* variables indicate the number of buffer headers and bytes,
respectively, that are unavailable to getnewbuf() because they are locked up
waiting for LFS to flush them; make sure that that is actually what we're
counting, i.e., never count malloced buffers, and always use b_bufsize instead
of b_bcount.
If DEBUG is defined, the periodic calls to lfs_countlocked will now complain
if either counter is incorrect. (In the future lfs_countlocked will not need
to be called at all if DEBUG is not defined.)
|
| 1.3 | 15-Mar-1999 |
perseant | branches: 1.3.8;
New CHANGES files that describes briefly all nontrivial changes made to
the LFS since the 4.4lite2 code was merged into NetBSD.
TODO updated to remove everything marked DONE in 4.4, and add in a list
of more current things to do.
Get rid of comments about the cleaner syscall code and missing fragment
support from README.
|
| 1.2 | 29-Jun-1994 |
cgd | New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
|
| 1.1 | 08-Jun-1994 |
mycroft | branches: 1.1.1;
Update to 4.4-Lite fs code, with local changes.
|
| 1.1.1.1 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite for reference
|
| 1.3.8.1 | 22-Nov-2000 |
bouyer | Sync with HEAD.
|
| 1.4.6.1 | 03-Aug-2001 |
lukem | update to -current
|
| 1.4.4.1 | 27-Jun-2001 |
perseant | Import of what I've been calling "LFSv2", that is, LFS with some features
added that require changes to the on-disk data structures. These include:
- 64-bit time in everything but inodes
- User-specified segment offset, and segment size no longer
restricted to PO2.
- Serial number on segment summaries in addition to timestamp, and
a new volume identifier, to make roll-forward feasible without
fear of finding old data and thinking it was new.
Although I think this version works at least as well as what's on the trunk,
we're not done yet; hence this commit is going in on a branch and not on
the trunk. Enhancements that are not here yet include fragment addressing,
like FFS does, instead of block addressing.
|
| 1.4.2.1 | 24-Aug-2001 |
nathanw | Catch up with -current.
|
| 1.7.12.1 | 19-Mar-2005 |
yamt | sync with head. xen and whitespace. xen part is not finished.
|
| 1.7.10.1 | 29-Apr-2005 |
kent | sync with -current
|
| 1.7.8.1 | 10-May-2005 |
riz | Pull up the following revisions (requested by perseant in ticket #1281):
1.8 sys/ufs/lfs/TODO
1.75 sys/ufs/lfs/lfs.h (via patch)
1.74 sys/ufs/lfs/lfs_alloc.c (via patch)
1.49, 1.51 sys/ufs/lfs/lfs_balloc.c (1.51 via patch)
1.78 sys/ufs/lfs/lfs_bio.c
1.62 sys/ufs/lfs/lfs_extern.h (via patch)
1.156 sys/ufs/lfs/lfs_segment.c (via patch)
1.48 sys/ufs/lfs/lfs_subr.c
1.101 sys/ufs/lfs/lfs_syscalls.c
1.163 sys/ufs/lfs/lfs_vfsops.c (via patch)
1.134 sys/ufs/lfs/lfs_vnops.c (via patch)
1.61 sys/ufs/ufs/ufs_readwrite.c (via patch)
1.20 libexec/lfs_cleanerd/clean.h (via patch)
1.52 libexec/lfs_cleanerd/cleanerd.c (via patch)
1.41 libexec/lfs_cleanerd/library.c (via patch)
1.4 regress/sys/fs/lfs/newfs_fsck/Makefile
1.2 regress/sys/fs/lfs/newfs_fsck/mkfs_mount
1.2 regress/sys/fs/lfs/newfs_fsck/smallfiles
1.3 sbin/fsck_lfs/bufcache.c
1.3 sbin/fsck_lfs/bufcache.h
1.3 sbin/fsck_lfs/lfs.h
1.8 sbin/fsck_lfs/lfs.c (via patch)
1.8 sbin/fsck_lfs/pass3.c (via patch)
1.18 sbin/fsck_lfs/pass0.c (via patch)
1.18 sbin/fsck_lfs/utilities.c (via patch)
1.7 sbin/fsck_lfs/segwrite.c
1.19 sbin/fsck_lfs/setup.c (via patch)
1.3 sbin/newfs_lfs/Makefile
0 sbin/newfs_lfs/lfs.c (yes, remove it)
1.1 sbin/newfs_lfs/make_lfs.c
1.15 sbin/newfs_lfs/newfs.c (via patch)
Various minor LFS improvements.
Kernel:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this. Should fix PR #29045.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
Fixes PR #26680.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
cleaner:
* Adapt lfs_cleanerd to use the fcntl call to get the Ifile filehandle,
so it need not be in the namespace.
* Make lfs_cleanerd be more careful when there are very few available
segments.
* Make lfs_cleanerd less verbose when the filesystem is unmounted.
newfs_lfs, fsck_lfs, and regression:
* Extend the lfs library from fsck_lfs(8) so that it can be used with a
not-yet-existent LFS. Make newfs_lfs(8) use this library, so it can
create LFSs whose Ifile is larger than one segment. Addresses PR #11110.
* Make newfs_lfs(8) use strsuftoi64() for its arguments, a la newfs(8).
* Make fsck_lfs(8) respect the "file system is clean" flag.
* Don't let fsck_lfs(8) think it has dirty blocks when invoked with the
-n flag.
* Remove the Ifile from the filesystem namespace. The cleaner now uses
a fcntl call on the root inode to find the Ifile filehandle. (As a
side-effect, addresses PR #29144.)
|
| 1.7.2.2 | 10-Nov-2005 |
skrll | Sync with HEAD. Here we go again...
|
| 1.7.2.1 | 04-Mar-2005 |
skrll | Sync with HEAD.
Hi Perry!
|
| 1.8.2.1 | 07-May-2005 |
tron | Apply patch (requested by perseant in ticket #242):
* fsck_lfs buffer cache fixes, including PR #29151
* Change fsck_lfs phase 0 message to reflect reality
* fsck_lfs: check phase 5 (cleanerinfo accounting) even on
roll-forward
* Keep better track of the free list during roll-forward, avoiding
a core dump
* Improve hash table use for fsck_lfs buffer and vnode cache
* Document fsck_lfs flag -f, and implement -q
* Add resize_lfs, including kernel support
* Add LFS to mountd's list of exportable filesystem types
* Make the LFS lkm work again [christos@]
* Add MP locking to the LFS kernel subsystem
* Fix pager_map deadlock in lfs_putpages()
* Avoid incomplete file extension that looks like "partial
truncation" to fsck
* Use lfs_malloc for cleaner malloc, since the cleaner often runs
in low-memory conditions.
* Use splay trees, not hash table, to track page allocation for
write.
* Fix mkdir panic on full fs
* Fix page accounting leak by counting differently.
* Use rightly named structure for lfs_getattr [skrll@]
* Cosmetic changes for readability.
|
| 1.9.2.1 | 21-Jun-2006 |
yamt | sync with head.
|
| 1.211 | 20-Oct-2025 |
perseant | * Generalize the partial-segment parser introduced for roll-forward,
using it to facilitate an in-kernel segment rewriter (cleaner), and a
mechanism to check whether a segment is in fact empty (only used
with DEBUG).
* Add these new fcntl calls:
- LFCNFILESTATS: For each inode given, report its number of direct
blocks, how many gaps (discontinuities) there are between direct
blocks, and how large the total gap distance is. This will be
useful for a coalescing agent.
- LFCNREWRITEFILE: For each inode given, rewrite its direct blocks,
effectively coalescing it into as compact a form as possible.
- LFCNSCRAMBLE: As above, except that it only rewrites every other
block. This causes the file to have many gaps that can be
measured with LFCNFILESTATS and addressed with LFCNREWRITEFILE,
for testing purposes.
- LFCNREWRITESEGS: Rewrite any live data in the given segments.
This is intended to simplify the cleaner API and facilitate an
in-kernel cleaner.
- LFCNCLEANERINFO: Get the most current CLEANERINFO data from the
kernel.
- LFCNSEGUSE: Retrieve segment usage data from the kernel.
* Vnodes marked IN_CLEANING now take a reference. Add a new "cleaner
lock", which must be taken by the cleaner before the segment lock,
and before marking nodes IN_CLEANING. This allows us to flush
vnodes, if necessary, before the cleaning segment is written, and
never to flush vnodes being cleaned. When the cleaner lock is
released, the vnodes are cleared of IN_CLEANING and the reference
dropped.
* Track a potential infinite loop in lfs_gatherblock.
* Pull "needs to flush" and "needs to wait for flush" into functions
instead of inlining their definitions.
|
| 1.210 | 17-Sep-2025 |
perseant | Add routines to check freelist consistency if compiled with DEBUG and
conditional on a kernel variable manipulated via sysctl.
Add checks before and after each routine that modifies the free list.
#if 0 a section of lfs_vfree() that was intended to keep the free list ordered
but instead corrupted it.
|
| 1.209 | 15-Sep-2025 |
perseant | If setting the head (or tail) of the inode free list to LFS_UNUSED_INUM, also
set the tail (resp. head) to LFS_UNUSED_INUM, as the list is now empty.
Add a check to ensure that lfs_valloc_fixed will always terminate, even
if the free list should contain a loop. Extend the ifile at the end if it
is empty, to match the assumption of lfs_valloc() that the free list is
never empty.
Needed for roll-forward.
|
| 1.208 | 28-Mar-2020 |
christos | Comment out some of the CTASSERTS for lint until I fix lint.
|
| 1.207 | 21-Mar-2020 |
riastradh | Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
|
| 1.206 | 21-Mar-2020 |
riastradh | CTASSERT lfs on-disk structure sizes.
|
| 1.205 | 23-Feb-2020 |
riastradh | Teach LFS_ORPHAN_NEXTFREE about lfs64.
|
| 1.204 | 10-Jan-2019 |
martin | branches: 1.204.4; 1.204.6;
Update comment (overlooked in r1.179).
From Jos� Luis Rodr�guez Garc�a in PR kern/53849.
|
| 1.203 | 26-Jul-2017 |
maya | branches: 1.203.2; 1.203.4;
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
|
| 1.202 | 05-Jun-2017 |
maya | Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
|
| 1.201 | 01-Apr-2017 |
maya | branches: 1.201.6;
switch lfs_dirops to condvar (from mtsleep)
|
| 1.200 | 01-Apr-2017 |
maya | switch lfs_sleepers to condvar (from mtsleep)
|
| 1.199 | 20-Jun-2016 |
dholland | branches: 1.199.2; 1.199.4;
Massedit u_int{8,16,32,64}_t to uint{8,16,32,64}_t. This effectively
merges ufs/dinode.h 1.25.
|
| 1.198 | 19-Jun-2016 |
dholland | we are actually synced with ufs/dinode.h 1.24 and ufs/dir.h 1.25.
|
| 1.197 | 26-Nov-2015 |
dholland | Update now-lying comment.
|
| 1.196 | 15-Oct-2015 |
dholland | For now bitflip the lfs64 magic number.
This will be unflipped when the format is finalized - right now I
still have pending changes to the superblock in mind (to reduce the
number of redundant fields) so anything created now is not future-
proof. However, the code's also nearing being ready for testing; so
I'm doing this before turning it on as a precaution.
|
| 1.195 | 15-Oct-2015 |
dholland | Move stuff from struct ulfsmount to struct lfs.
|
| 1.194 | 03-Oct-2015 |
dholland | Add lfs_checkword type for reading checksum data out of structures.
This is always uint32_t, but having a name for it both makes things
clearer and avoids confusion about whether it should be 32 or 64 bit.
Note: deployed in only one place (that was erroneously tagged
ondisk32) so far.
|
| 1.193 | 03-Oct-2015 |
dholland | Add an IINFO struct, which is like the FINFO struct but for the inode
blocks portion of the segment summary.
A segment summary block begins with a header (SEGSUM); the rest of the
block contains FINFO structures describing file blocks growing upward
from the bottom (after the header), and IINFO structures describing
inode blocks grown downward from the end of the block. (When they meet
the segment is full regardless of how many blocks might be left.)
IINFO contains just a block number, and until now this information was
handled by just using uint32_t*; switching to a structure will make
the code a lot easier to read, and also make it easier to have 32-bit
and 64-bit versions without making a mess.
This commit just adds the structures and accessors; they'll be
deployed into the code in subsequent commits.
|
| 1.192 | 21-Sep-2015 |
dholland | Oops, I forgot to make the atime in the 64-bit IFILE 64 bits.
Correct that. Incompatible change, but no LFS64 volumes can have been
created yet.
|
| 1.191 | 21-Sep-2015 |
dholland | Add 64-bit directory entry structures, and adjust accessors accordingly.
The LFS64 directory entry has a 64-bit inode number. This is stored as
two 32-bit values to avoid inducing 64-bit alignment requirements.
The exposed type for manipulating directory entries is now
LFS_DIRHEADER, following the same convention as e.g. IFILE and SEGUSE.
(But with LFS_ on it, because.)
|
| 1.190 | 20-Sep-2015 |
dholland | Clean up struct lfs_dirtemplate.
|
| 1.189 | 15-Sep-2015 |
dholland | Remove struct lfs_direct (no longer used) and update the big block
comment about directories.
|
| 1.188 | 15-Sep-2015 |
dholland | Add an accessor function for directory names.
|
| 1.187 | 15-Sep-2015 |
dholland | Move the header part of struct lfs_direct to its own structure.
(lfs_dirheader)
Take the opportunity to improve the directory generation code in
make_lfs.c. (Everything else was unaffected by virtue of using
accessor functions.)
|
| 1.186 | 15-Sep-2015 |
dholland | Add and use accessor functions for more of the directory entry fields.
|
| 1.185 | 01-Sep-2015 |
dholland | Add new accessors for the d_type and d_namlen fields of struct lfs_direct.
Napalm the old byteswap access logic for these.
|
| 1.184 | 01-Sep-2015 |
dholland | Comments on directories.
This includes a description of the struct direct byteswap horrors that
ought to be propagated to ufs/ufs.
|
| 1.183 | 01-Sep-2015 |
dholland | The ifile's inode number is constant. (it is always 1)
Therefore, storing the value in the superblock and reading it out
again is silly and offers the opportunity for it to become corrupted.
So, don't do that (most of the code already didn't) and use the
existing constant instead. Initialize new 32-bit superblocks with
the value for the sake of old userland programs, but don't keep the
value in the 64-bit superblock at all.
(approved by Margo Seltzer)
|
| 1.182 | 01-Sep-2015 |
dholland | Make the inode fields in the 64-bit superblock 64 bits wide.
Reasoning as before.
Note that I am not going through and checking for 64->32 truncations
in inode numbers; I'm sure there are quite a few, but that's a project
for later.
|
| 1.181 | 01-Sep-2015 |
dholland | Add byteswapping to the dinode accessors.
This prevents regressions in the ulfs code when switching to the new
accessors. Note that while adding byteswapping to the other accessors
is straightforward, I haven't done it yet; and that also is not enough
to make LFS_EI work, because there are places lying around that bypass
the accessors for one reason and another and all of them need to be
updated. That is going to have to wait for a later day as LFS_EI is
not on the critical path right now.
|
| 1.180 | 12-Aug-2015 |
dholland | Hack up dinode usage to be 64 vs. 32 as needed. Part 1.
(This part changes the native lfs code; the ufs-derived code already
has 64 vs. 32 logic, but as aspects of it are unsafe, and don't
entirely interoperate cleanly with the lfs 64/32 stuff, pass 2 will be
rehashing that.)
|
| 1.179 | 12-Aug-2015 |
dholland | Make the inode number in the 64-bit dinode 64 bits wide, like the
other lfs64 on-disk inode numbers; I've been doing that since this is
a new format and we may as well take the opportunity. This does assume
that more than 4 billion files on a single volume becomes desirable;
but for an average file size of 10K all that takes is a 40 TB volume,
and it's not that hard to make one of those these days if you want to
badly enough.
|
| 1.178 | 12-Aug-2015 |
dholland | Provide 32-bit and 64-bit versions of FINFO.
This also entailed sorting out part of struct segment, as that
contains a pointer into the current FINFO data.
|
| 1.177 | 12-Aug-2015 |
dholland | Make 32-bit and 64-bit versions of SEGSUM.
Also fix some of the FINFO handling as it's closely entangled.
|
| 1.176 | 12-Aug-2015 |
dholland | Add IFILE32 and IFILE64 structures for the on-disk ifile entries.
Add and use accessors. There are also a bunch of places that cast and
I hope I've found them all...
|
| 1.175 | 12-Aug-2015 |
dholland | Make 32-bit and 64-bit versions of CLEANERINFO.
XXX: while this is written to disk, it seems like much of it would
XXX: be better set up as a commpage shared with the cleaner.
|
| 1.174 | 12-Aug-2015 |
dholland | Widen several of the fields of BLOCK_INFO to 64 bits.
Keep the old BLOCK_INFO as BLOCK_INFO_70, and version the fcntls that
use it.
Note that BLOCK_INFO_70 has 64-bit padding issues so that it's
different on 32-bit and 64-bit machines. This has been fixed. However,
BLOCK_INFO also contains a pointer, so compat32 stuff for 32-on-64 is
still needed and doesn't currently exist.
|
| 1.173 | 12-Aug-2015 |
dholland | Fix assorted 64->32 truncations related to BLOCK_INFO.
Also make note of a cleaner limitation: it seems that when it goes to
coalesce discontiguous files, it mallocs an array with one BLOCK_INFO
for every block in the file. Therefore, with 64-bit LFS, on a 32-bit
platform it will be possible to have files large enough to overflow
the cleaner's address space. Currently these will be skipped and cause
warnings via syslog.
At some point someone should rewrite the logic to coalesce files to
use chunks of some reasonable size, as discontinuity between such
chunks is immaterial and mallocing this much space is silly and
fragile. Also, the kernel only accepts up to 65536 blocks at a time
for bmapv and markv, so processing more than this at once probably
isn't useful and may not even work currently. I don't want to change
this around just now as it's not entirely trivial.
|
| 1.172 | 02-Aug-2015 |
dholland | Pass the fs object to LFS_MAX_DADDR so it can check lfs_is64.
Remove some hackish intentional 64->32 truncations next to the checks
using LFS_MAX_DADDR, and tackle the problem they handled in bmap
instead.
The problem: the magic block pointer value UNWRITTEN has magic value
-2, and if it's not handled specifically, uint32 -> uint64 promotion
turns it into 4294967294, which then causes consternation and
monkeyhouse downstream.
What's here is still kind of a hack, but it's a step forward.
|
| 1.171 | 02-Aug-2015 |
dholland | Add a (draft) 64-bit superblock. Make things build again.
Add pieces of support for using both superblock types where
convenient, and specifically to the superblock accessors, but don't
actually enable it anywhere.
First substantive step on PR 50000.
|
| 1.170 | 02-Aug-2015 |
dholland | lfs_cleanint[] in the in-memory superblock needs to have 64-bit entries.
|
| 1.169 | 02-Aug-2015 |
dholland | Second batch of 64 -> 32 truncations in lfs, along with more minor
tidyups and corrections in passing.
|
| 1.168 | 02-Aug-2015 |
dholland | Fix assorted 64 -> 32 truncations in lfs. Also, some minor tidyups and
corrections in passing.
|
| 1.167 | 28-Jul-2015 |
dholland | Move struct salfs back inside libsa now that lfs_accessors.h is separate.
|
| 1.166 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.165 | 24-Jul-2015 |
dholland | More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)
The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.
|
| 1.164 | 24-Jul-2015 |
dholland | Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)
It also gets rid of cpp abuse in the form of fake structure member
macros.
Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)
XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.
|
| 1.163 | 24-Jul-2015 |
dholland | ulfs2_dinode, having never actually been used with lfs, doesn't have a
di_inumber field. Fix that. First preliminary step on PR 50000.
|
| 1.162 | 31-May-2015 |
hannken | Use VFS_PROTOS() for lfs.
Rename conflicting struct lfs field "lfs_start" to "lfs_s0addr".
No functional change.
|
| 1.161 | 28-Mar-2015 |
maxv | Remove the 'cred' argument from bread(). Remove a now unused var in
ffs_snapshot.c. Update the man page accordingly.
ok hannken@
|
| 1.160 | 28-Jul-2013 |
dholland | branches: 1.160.6;
Bring in a copy of ffs_quota2_mount() for reference.
Add stuff to struct lfs that it needs to initialize.
Clear these fields in mount as there's no on-disk support for quota2;
but this increases the chances of being able to add it (or something
like it) in the future.
|
| 1.159 | 28-Jul-2013 |
dholland | Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.158 | 28-Jul-2013 |
dholland | Add lfs_kernel.h for declarations that don't need to be exposed to userland.
lfs currently has the following headers:
lfs.h - on-disk structures and stuff needed for userlevel tools
lfs_inode.h - additional restricted materials for userlevel tools
that operate the fs (newfs_lfs, fsck_lfs, lfs_cleanerd)
lfs_kernel.h - stuff needed only in the kernel
and the following legacy headers that are expected to be mopped up and
folded into one of the above:
lfs_extern.h - function prototypes
ulfs_bswap.h - endian-independent support
ulfs_dinode.h - now contains very little
ulfs_dirhash.h - dirhash support
ulfs_extattr.h - extattr support
ulfs_extern.h - more function prototypes
ulfs_inode.h - assorted kernel-only declarations
ulfs_quota.h - quota support
ulfs_quota1.h - more quota support
ulfs_quota2.h - more quota support
ulfs_quotacommon.h - more quota support
ulfsmount.h - legacy copy of ufsmount material
|
| 1.157 | 28-Jun-2013 |
matt | branches: 1.157.2;
Remove duplicate define of LFS_MAXNAMLEN
|
| 1.156 | 23-Jun-2013 |
dholland | typo in comment
|
| 1.155 | 19-Jun-2013 |
dholland | Rename ambiguous macros:
MAXDIRSIZE -> UFS_MAXDIRSIZE or LFS_MAXDIRSIZE
NINDIR -> FFS_NINDIR, EXT2_NINDIR, LFS_NINDIR, or MFS_NINDIR
INOPB -> FFS_INOPB, LFS_INOPB
INOPF -> FFS_INOPF, LFS_INOPF
blksize -> ffs_blksize, ext2_blksize, or lfs_blksize
sblksize -> ffs_blksize
These are not the only ambiguously defined filesystem macros, of
course, there's a pile more. I may not have found all the ambiguous
definitions of blksize(), too, as there are a lot of other things
called 'blksize' in the system.
|
| 1.154 | 18-Jun-2013 |
christos | Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.
|
| 1.153 | 18-Jun-2013 |
dholland | Tuck away a bunch of symbols that don't need to be public.
|
| 1.152 | 09-Jun-2013 |
dholland | Move struct lfs_inode_ext to lfs_inode.h; it doesn't need to be public.
|
| 1.151 | 08-Jun-2013 |
dholland | Remove stale union and accessor macros.
|
| 1.150 | 08-Jun-2013 |
dholland | ulfs_dir.h has been emptied; remove it.
|
| 1.149 | 08-Jun-2013 |
dholland | Move a comment to lfs.h that belongs better there.
|
| 1.148 | 08-Jun-2013 |
dholland | As nearly all the content of ulfs_dir.h and ulfs_dinode.h has migrated
to lfs.h, propagate the copyright notices too.
|
| 1.147 | 08-Jun-2013 |
dholland | Move more symbols to lfs.h:
LFS_DIRBLKSIZ
LFS_DIRECTSIZ
LFS_DIRSIZ
LFS_OLDDIRFMT
LFS_NEWDIRFMT
LFS_IFTODT
LFS_DTTOIF
ULFS{,1,2}_MAXSYMLINKLEN
|
| 1.146 | 08-Jun-2013 |
dholland | DIRBLKSIZ -> LFS_DIRBLKSIZ
DIRECTSIZ -> LFS_DIRECTSIZ
DIRSIZ -> LFS_DIRSIZ
OLDDIRFMT -> LFS_OLDDIRFMT
NEWDIRFMT -> LFS_NEWDIRFMT
IFTODT -> LFS_IFTODT
DTTOIF -> LFS_DTTOIF
|
| 1.145 | 08-Jun-2013 |
dholland | Move stuff to lfs.h that's needed by userland:
LFS_DT_*
ULFS_ROOTINO
ULFS_WINO
struct lfs_direct
struct lfs_dirtemplate
struct lfs_odirtemplate
struct ulfs_args
Also fix FFS_MAXNAMLEN -> LFS_MAXNAMLEN in several places.
|
| 1.144 | 08-Jun-2013 |
dholland | Now move LFS_IFMT and friends from ulfs_dinode.h to lfs.h.
|
| 1.143 | 08-Jun-2013 |
dholland | Move the dinode (on-disk inode) structures to lfs.h, since they are
and will be obviously required by userland tools that need to read
the on-disk structures.
Also, DINODE{1,2}_SIZE -> LFS_DINODE{1,2}_SIZE.
|
| 1.142 | 08-Jun-2013 |
dholland | Split the definitions suitable for userland out of ulfs_inode.h into
lfs_inode.h. Since fsck_lfs, newfs_lfs, and lfs_cleanerd want to reuse
the inode structure for their own internal use, and some of them share
parts of the kernel code as well, the best way forward is to provide a
relatively sanitized header that doesn't bring in stray material.
Shuffle a few other definitions around so that lfs_inode.h depends
only on lfs.h.
Install lfs_inode.h into /usr/include.
|
| 1.141 | 06-Jun-2013 |
dholland | Fix some exposed symbols:
LOSTFOUNDINO -> LFS_LOSTFOUNDINO
struct ufid -> struct ulfs_ufid
|
| 1.140 | 06-Jun-2013 |
dholland | Cleanups to reduce symbol and header exposure:
- move struct ufid from ulfs_inode.h to lfs.h
- lfs.h needs sys/mount.h and sys/pool.h
- ulfs_quota2_subr.c needs lfs_inode.h
- remove ulfs_inode.h from lfs.h in favor of ulfs_dinode.h
- move ULFS_NDADDR, ULFS_NIADDR, ULFS_NXADDR from ulfs_dinode.h to lfs.h
- remove ulfs_dinode.h from lfs.h
- add lfs.h to ulfs_dinode.h
|
| 1.139 | 06-Jun-2013 |
dholland | Remove stray references to ext2fs, chfs, ffs, and mfs.
|
| 1.138 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.137 | 22-Jan-2013 |
dholland | Stuff UFS_ in front of a few of ufs's symbols to reduce namespace
pollution. Specifically:
ROOTINO -> UFS_ROOTINO
WINO -> UFS_WINO
NXADDR -> UFS_NXADDR
NDADDR -> UFS_NDADDR
NIADDR -> UFS_NIADDR
MAXSYMLINKLEN -> UFS_MAXSYMLINKLEN
MAXSYMLINKLEN_UFS[12] -> UFS[12]_MAXSYMLINKLEN (for consistency)
Sort out ext2fs's misuse of NDADDR and NIADDR; fortunately, these have
the same values in ext2fs and ffs.
No functional change intended.
|
| 1.136 | 16-Feb-2012 |
perseant | branches: 1.136.2;
Pass t_renamerace and t_rmdirrace tests.
Adapt dholland@'s fix to ufs_rename to fix PR kern/43582. Address several
other MP locking issues discovered during the course of investigating the
same problem.
Removed extraneous vn_lock() calls on the Ifile, since the Ifile writes
are controlled by the segment lock.
Fix PR kern/45982 by deemphasizing the estimate of how much metadata
will fill the empty space on disk when the disk is nearly empty
(t_renamerace crates a lot of inode blocks on a tiny empty disk).
|
| 1.135 | 02-Jan-2012 |
perseant | branches: 1.135.2;
* Remove PGO_RECLAIM during lfs_putpages()' call to genfs_putpages(),
to avoid a live lock in the latter when reclaiming a vnode with
dirty pages.
* Add a new segment flag, SEGM_RECLAIM, to note when a segment is
being written for vnode reclamation, and record which inode is being
reclaimed, to aid in forensic debugging.
* Add a new segment flag, SEGM_SINGLE, so that opportunistic writes
can write a single segment's worth of blocks and then stop, rather
than writing all the way up to the cleaner's reserved number of
segments.
* Add assert statements to check mutex ownership is the way it ought
to be, mostly in lfs_putpages; fix problems uncovered by this.
* Don't clear VU_DIROP until the inode actually makes its way to disk,
avoiding a problem where dirop inodes could become separated
(uncovered by a modified version of the "ckckp" forensic regression
test).
* Move the vfs_getopsbyname() call into lfs_writerd. Prepare code to
make lfs_writerd notice when there are no more LFSs, and exit losing
the reference, so that, in theory, the module can be unloaded. This
code is not enabled, since it causes a crash on exit.
* Set IN_MODIFIED on inodes flushed by lfs_flush_dirops. Really we
only need to set IN_MODIFIED if we are going to write them again
(e.g., to write pages); need to think about this more.
Finally, several changes to help avoid "no clean segments" panics:
* In lfs_bmapv, note when a vnode is loaded only to discover whether
its blocks are live, so it can immediately be recycled. Since the
cleaner will try to choose ~empty segments over full ones, this
prevents the cleaner from (1) filling the vnode cache with junk, and
(2) squeezing any unwritten writes to disk and running the fs out of
segments.
* Overestimate by half the amount of metadata that will be required
to fill the clean segments. This will make the disk appear smaller,
but should help avoid a "no clean segments" panic.
* Rearrange lfs_writerd. In particular, lfs_writerd now pays
attention to the number of clean segments available, and holds off
writing until there is room.
|
| 1.134 | 11-Jul-2011 |
hannken | branches: 1.134.2; 1.134.6;
Change VOP_BWRITE() to take a vnode as its first argument like all other
VOPs do. Layered file systems no longer have to modify bp->b_vp and run
into trouble when an async VOP_BWRITE() uses the wrong vnode.
- change all occurences of VOP_BWRITE(bp) to VOP_BWRITE(bp->b_vp, bp).
- remove layer_bwrite().
- welcome to 5.99.55
Adresses PR kern/38762 panic: vwakeup: neg numoutput
No objections from tech-kern@.
|
| 1.133 | 16-Feb-2010 |
mlelstv | Three changes in a single commit.
- drop the notion of frags (LFS fragments) vs fsb (FFS fragments)
The code uses a complicated unity function that just makes the
code difficult to understand.
- support larger sector sizes. Fix disk address computations
to use DEV_BSIZE in the kernel as required by device drivers
and to use sector sizes in userland.
- Fix several locking bugs in lfs_bio.c and lfs_subr.c.
|
| 1.132 | 05-Nov-2009 |
pooka | branches: 1.132.2;
... actually, define compat only for the kernel. Userlandia should
see only one version of the interfaces.
|
| 1.131 | 05-Nov-2009 |
pooka | Include compat/sys/time_types.h instead of compat/sys/time.h.
Fixes lint drama with interface name collisions.
|
| 1.130 | 05-Nov-2009 |
pooka | Include compat code by default.
|
| 1.129 | 29-Oct-2009 |
christos | PR/42246: NAKAJIMA Yoshihiro: provide COMPAT_50 for LFS
|
| 1.128 | 19-Jul-2009 |
dholland | typo in comment
|
| 1.127 | 16-May-2008 |
hannken | branches: 1.127.12;
Make sure all cached buffers with valid, not yet written data have been
run through copy-on-write. Call fscow_run() with valid data where possible.
The LP_UFSCOW hack is no longer needed to protect ffs_copyonwrite() against
endless recursion.
- Add a flag B_MODIFY to bread(), breada() and breadn(). If set the caller
intends to modify the buffer returned.
- Always run copy-on-write on buffers returned from ffs_balloc().
- Add new function ffs_getblk() that gets a buffer, assigns a new blkno,
may clear the buffer and runs copy-on-write. Process possible errors
from getblk() or fscow_run(). Part of PR kern/38664.
Welcome to 4.99.63
Reviewed by: YAMAMOTO Takashi <yamt@netbsd.org>
|
| 1.126 | 28-Apr-2008 |
martin | branches: 1.126.2;
Remove clause 3 and 4 from TNF licenses
|
| 1.125 | 15-Feb-2008 |
ad | branches: 1.125.6; 1.125.8; 1.125.10;
The buffer LOCKED flag need not be under the protection of bufcache_lock,
BUSY is enough.
|
| 1.124 | 03-Jan-2008 |
ad | Use pool_cache.
|
| 1.123 | 02-Jan-2008 |
ad | Merge vmlocking2 to head.
|
| 1.122 | 10-Oct-2007 |
ad | branches: 1.122.4; 1.122.6; 1.122.10;
Merge from vmlocking:
- Split vnode::v_flag into three fields, depending on field locking.
- simple_lock -> kmutex in a few places.
- Fix some simple locking problems.
|
| 1.121 | 08-Oct-2007 |
ad | Merge ffs locking & brelse changes from the vmlocking branch.
|
| 1.120 | 16-May-2007 |
perseant | branches: 1.120.6; 1.120.8; 1.120.10;
Change references to SEGM_W_DIROPS to SEGM_CKP, and replace the logic that
formerly used SEGM_W_DIROPS in lfs_segwrite() appropriately. This prevents
a problem in which processes could get stuck in "buffers" sleep forever.
|
| 1.119 | 17-Apr-2007 |
perseant | Install a new sysctl, vfs.lfs.ignore_lazy_sync, which causes LFS to ignore
the "smooth" syncer, as if vfs.sync.*delay = 0, but only for LFS. The
default is "on", i.e., ignore lazy sync.
Reduce the amount of polling/busy-waiting done by lfs_putpages(). To
accomplish this, copied genfs_putpages() and modified it to indicate which
page it was that caused it to return with EDEADLK. fsync()/fdatasync()
should no longer ever fail with EAGAIN, and should not consume huge
quantities of cpu.
Also, try to make dirops less likely to be written as the result of a
VOP_PUTPAGES(), while ensuring that they are written regularly.
|
| 1.118 | 15-Feb-2007 |
ad | branches: 1.118.2; 1.118.6; 1.118.8;
Replace some uses of lockmgr() / simplelocks.
|
| 1.117 | 28-Sep-2006 |
perseant | Use lockstatus instead of a homebrewed locking system to control
LFCNWRAPSTOP and LFCNWRAPGO.
Be less verbose about the various looping checks: use log() rather than
printf(), and only log anything if we are really looping ("count = 2" is
not an error condition).
Allow dirops sleeping on available space to be interruptible.
|
| 1.116 | 15-Sep-2006 |
perseant | branches: 1.116.2;
Don't remark a locked inode with IN_MODIFIED after writing it to disk,
if we ourselves hold the lock. This prevents e.g. mknod from hanging
indefinitely.
Also, always use the return value from VOP_ISLOCKED to determine whether
we hold the lock or someone else does, rather than looking into the lock
structure ourselves.
|
| 1.115 | 15-Sep-2006 |
yamt | merge yamt-pdpolicy branch.
- separate page replacement policy from the rest of kernel
- implement an alternative replacement policy
|
| 1.114 | 01-Sep-2006 |
perseant | branches: 1.114.2;
Changes to help the roll-forward agent, to wit:
* Mark being-deleted files in the Ifile so we can finish deleting them
at fs mount time.
* Flag the Ifile with "cleaner must clean" when writers are waiting for
the cleaner, rather than relying solely on the cleaner's estimation of
whether it should clean or not.
* Note partial segments written by a user agent (in particular,
fsck_lfs) so that repeated rolls forward don't interfere with one
another.
* Add a new fcntl, LFCNPASS, that allows the log to wrap exactly once,
for better testing of the validity of checkpoints.
* Keep track of the on-disk nlink count when cleaning, so that we don't
partially complete directory operations while cleaning.
* Ensure that every single Ifile inode write represents a consistent
view of the filesystem. In particular, the accounting for the segment
we are writing the inode into must be correct, and the accounting for
the segment that inode used to reside in must be correct. Rather than
just rewriting the inode if we wrote it wrong, rewrite the necessary
ifile blocks before writing the inode so we never write it wrong.
* Don't unmark any VDIROP vnodes if we haven't written them to disk,
avoiding yet another problem with the "wait for the cleaner" error
return from lfs_putpages().
Also, move the last callback to an aiodone call, so we no longer do any
memory management from interrupt context.
|
| 1.113 | 06-Aug-2006 |
martin | Fix size confusion with lfs_fhandle - and as it now turns out to be the same
as the lfs compat_30_fhandle, g/c the latter.
Add an alias for the LFCNIFILEFH fcntl, so that binaries compiled in the
meantime (with too large lfs_fhandle) continue to work.
This makes vfs_cleanerd work again after the kernel checks filehandle size
more strictly (problem reported by Kurt Schreiner on current-users).
|
| 1.112 | 31-Jul-2006 |
martin | Make filehandles opaque to userland
|
| 1.111 | 20-Jul-2006 |
perseant | Note partial segments that are written by the cleaner, to help out the
roll-forward agent.
|
| 1.110 | 13-Jul-2006 |
martin | Version the lfs_cleanerd internal fcntl() for filehandles too,
so old cleaners should work with newer kernels.
|
| 1.109 | 13-Jul-2006 |
martin | Fix alignement problems for fhandle_t, exposed by gcc4.1.
While touching all vptofh/fhtovp functions, get rid of VFS_MAXFIDSIZ,
version the getfh(2) syscall and explicitly pass the size available in
the filehandle from userland.
Discussed on tech-kern, with lots of help from yamt (thanks!).
|
| 1.108 | 24-Jun-2006 |
perseant | Change LFCNWRAP{STOP,GO} to make them more suitable for snapshotting; in
particular, the caller can now choose whether to wait for the condition
to be met, and if the caller of LFCNWRAPSTOP dies or otherwise closes
the descriptor, the filesystem is started again. Updated the ckckp
regression test to use the new semantics.
dump_lfs(8) now uses the fcntls to implement LFS-style snapshotting through
the -X flag, addressing PR#33457 albeit not using fss(4). Fixed a couple
other problems with dump_lfs that manifested themselves during testing.
|
| 1.107 | 14-May-2006 |
elad | branches: 1.107.4;
integrate kauth.
|
| 1.106 | 12-May-2006 |
perseant | Fixes to address the "vinvalbuf: dirty blocks" panic that can occur when
many inodes are cleaned at once. Make sure that we write all the pages
on vnodes that are being flushed, even if we don't think there's room;
drain v_numoutput before lfs_vflush() completes.
Also, don't allow a vnode that is in the process of being cleaned to be
chosen by getnewvnode(); this avoids a segment accounting panic in the case
that a large number of inodes are fed to lfs_markv() all at once.
|
| 1.105 | 04-May-2006 |
perseant | Introduce another per-filesystem parameter, lfs_resvseg, to separate the
notion of "how many segments are reserved for the cleaner" from that of
"how many segments are not counted in lfs_bfree". The default value
used for existing filesystems is the same as the previous implicit value
of (lfs_minfreeseg / 2 + 1), modulo some sanity checking.
Count pending dirops on a per-filesystem basis, since once we start
writing them we can't stop until we're done. This seems to help stave off
the "no clean segments" panic in the case of filling the filesystem with
directories and small files (e.g. simultaneously unpacking more copies of
pkgsrc than will fit).
|
| 1.104 | 30-Apr-2006 |
perseant | Postpone the segment accounting changes coming from truncation until the
inode that makes those changes valid is either written to disk by
lfs_writeinode() or discarded by lfs_vfree().
A couple of locking fixes are also included as well.
|
| 1.103 | 17-Apr-2006 |
perseant | Introduce two fcntl calls that freeze the filesystem right at the point
where segment 0 is being considered for writing. This allows for automated
checkpoint vailidity scanning, and could be used (in conjunction with the
existing LFCNREWIND) for e.g. snapshot dumps as well.
Include a regression test that does such scanning.
When writing the Ifile, loop through the dirty block list three times to
make sure that the checkpoint is always consistent (the first and second
times the Ifile blocks can cross a segment boundary; not so the third time
unless the segments are very small). Discovered by using the aforementioned
regression test.
|
| 1.102 | 13-Apr-2006 |
perseant | Make lfs_vref/lfs_vunref not need to know about VXLOCK and VFREEING
explicitly (especially since we didn't know about VFREEING at all before),
but notice the EBUSY return from vget() instead.
Fix some more MP locking protocol issues, most of which were pointed out by
Christian Ehrhardt this morning on tech-kern.
|
| 1.101 | 10-Apr-2006 |
perseant | Optimize the free list search a little more; in particular use words
instead of bytes for the index, and never search below fs->lfs_freehd.
Fix a bug in the previous version of the search (an erroneous assumption
that ino_t was signed).
Free the bitmap when we unmount the filesystem.
|
| 1.100 | 08-Apr-2006 |
perseant | Implement a somewhat finer-grained mechanism for paging LFS-backed pages.
The writer daemon, if it does not need to flush the whole filesystem,
now only writes the vnodes for which the pagedaemon has requested pageouts
(although it does not pay attention to the page ranges the pagedaemon
supplies).
|
| 1.99 | 08-Apr-2006 |
perseant | Keep the free list ordered. This solves a problem first pointed out to me
by Michel Oey, in which an aged LFS writes up to an extra Ifile block for
every file created; and paves the way for the truncation of the Ifile when
many files are deleted.
|
| 1.98 | 07-Apr-2006 |
perseant | Make the segment lock aware of LWPs. Fixes a (somewhat confusing)
"lockmgr: pid 3997, not exclusive lockholder 3997, unlocking" panic I
encountered while running blogbench on an LFS.
|
| 1.97 | 24-Mar-2006 |
perseant | Improvements to LFS's paging mechanism, to wit:
* Acknowledge that sometimes there are more dirty pages to be written to
disk than clean segments. When we reach the danger line,
lfs_gop_write() now returns EAGAIN. The caller of VOP_PUTPAGES(), if
it holds the segment lock, drops it and waits for the cleaner to make
room before continuing.
* Note and avoid a three-way deadlock in lfs_putpages (a writer holding
a page busy blocks on the cleaner while the cleaner blocks on the
segment lock while lfs_putpages blocks on the page).
|
| 1.96 | 17-Mar-2006 |
tls | From Konrad Schroeder, in response to strange df output on anoncvs.netbsd.org:
We were returning the wrong value for free space. Now we're not.
|
| 1.95 | 11-Dec-2005 |
christos | branches: 1.95.4; 1.95.6; 1.95.8; 1.95.10; 1.95.12;
merge ktrace-lwp.
|
| 1.94 | 13-Sep-2005 |
christos | split out lfs_itimes(). It is used in fsck_lfs.
|
| 1.93 | 12-Sep-2005 |
christos | Use nanotime() to update the time fields in filesystems. Convert the code
from macros to real functions. Original patch and review from chuq.
Note: ext2fs only keeps seconds in the on-disk inode, and msdosfs does not
have enough precision for all fields, so this is not very useful for those
two.
|
| 1.92 | 23-Aug-2005 |
christos | Don't overload MAXNAMLEN, use a separate constant for each filesystem type.
|
| 1.91 | 22-Aug-2005 |
yamt | whitespace.
|
| 1.90 | 22-Aug-2005 |
christos | change ino_t to u_int32_t for syscall compatibility.
|
| 1.89 | 31-Jul-2005 |
christos | Move extern kernel variable declarations, into a _KERNEL protected session
so that the don't pollute userland's namespace.
|
| 1.88 | 29-May-2005 |
christos | branches: 1.88.2;
- sprinkle const
- avoid shadow variables.
|
| 1.87 | 20-May-2005 |
perseant | Keep track of the number of segments reclaimed, since the cleaner doesn't
do this anymore (it hasn't for quite some time). Add a couple of conditional
debugging messages to indicate why segments are not cleaned, in the event
that lfs_segclean is used.
Make the LFCNSEGWAITALL fcntl work again.
|
| 1.86 | 23-Apr-2005 |
perseant | Provide a resize_lfs(8), including kernel and cleaner support. The current
implementation requires the fs to be mounted while resizing. Tested in both
directions, and everything appears to work happily, but ymmv.
|
| 1.85 | 19-Apr-2005 |
perseant | Keep per-inode, per-fs, and subsystem-wide counts of blocks allocated through
lfs_balloc(), and use that to estimate the number of dirty pages belonging
to LFS (subsystem or filesystem). This is almost certainly wrong for
the case of a large mmap()ed region, but the accounting is tighter than
what we had before, and performs much better in the typical case of pages
dirtied through write().
|
| 1.84 | 16-Apr-2005 |
perseant | Make userland compile again.
|
| 1.83 | 16-Apr-2005 |
perseant | Use splay trees, rather than a hash table, to manage the accounting of
blocks allocated through VOP_BALLOC() for pages to be written to disk.
This accounting no longer takes a noticeable fraction of the system CPU.
|
| 1.82 | 16-Apr-2005 |
perseant | Use lfs_malloc() to manage the blkiov arrays that the cleaner functions use,
since the cleaner is likely to operate in a low-memory condition.
|
| 1.81 | 14-Apr-2005 |
perseant | Tabify leading whitespace
|
| 1.80 | 14-Apr-2005 |
perseant | Consolidate the hash table we use to maintain the integrity of lfs_avail
into a single, system-wide table, rather than having a separate hash table
per inode. Significantly reduces the "system" cpu usage of your average
file write.
|
| 1.79 | 14-Apr-2005 |
perseant | Keep track of the highest block held by an LFS inode, so that we can
be assured that the last byte of a file is always allocated. Previously
a file extension could cause the filesystem to be flushed, writing an
inconsistent inode to disk. Although this condition would be corrected
the next time blocks were written to disk, an intervening crash would leave
the filesystem in an inconsistent state, leaving fsck_lfs to complain
of an inode "partially truncated".
|
| 1.78 | 01-Apr-2005 |
perseant | Protect various per-fs structures with fs->lfs_interlock simple_lock, to
improve behavior in the multiprocessor case. Add debugging segment-lock
assertion statements.
|
| 1.77 | 08-Mar-2005 |
perseant | branches: 1.77.2;
Straighten out the maze of ifdefs. Instead, consolidate all the debugging
stuff under '#ifdef DEBUG', and use sysctl knobs to turn on/off particular
parts of the debugging reporting (if DEBUG is enabled). Re-enable the LFS
statistics in sysctl, while I'm there. A bit of a rototill.
|
| 1.76 | 26-Feb-2005 |
perry | nuke trailing whitespace
|
| 1.75 | 26-Feb-2005 |
perseant | Various minor LFS improvements:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statvfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
|
| 1.74 | 14-Aug-2004 |
mycroft | branches: 1.74.4; 1.74.6;
Push atime/mtime updates even further -- into the reclaim path, so they happen
rarely in the normal case. (Note: This happens at reboot/shutdown time because
all file systems are unmounted.)
Also, for IN_MODIFY, use IN_ACCESSED, not IN_MODIFIED; otherwise "ls -l" of
your device node or FIFO would cause the time stamps to get written too
quickly.
|
| 1.73 | 14-Aug-2004 |
mycroft | Add a new flag, IN_MODIFY. This is like IN_UPDATE|IN_CHANGE, but unlike
setting those flags, it does not cause the inode to be written in the periodic
sync. This is used for writes to special files (devices and named pipes) and
FIFOs.
Do not preemptively sync updates to access times and modification times. They
are now updated in the inode only opportunistically, or when the file or device
is closed. (Really, it should be delayed beyond close, but this is enough to
help substantially with device nodes.)
And the most amusing part:
Trickle sync was broken on both FFS and ext2fs, in different ways. In FFS, the
periodic call to VFS_SYNC(MNT_LAZY) was still causing all file data to be
synced. In ext2fs, it was causing the metadata to *not* be synced. We now
only call VOP_UPDATE() on the node if we're doing MNT_LAZY. I've confirmed
that we do in fact trickle correctly now.
|
| 1.72 | 09-Mar-2004 |
yamt | branches: 1.72.4;
use correct segment size. this fixes memory corruption when using lfsv1.
|
| 1.71 | 28-Jan-2004 |
yamt | use bufmem instead of bufpages to make lfs a little less broken.
|
| 1.70 | 07-Sep-2003 |
yamt | - raise spl to bio in lfs_countlocked() rather than having callers to do so.
- buffer cache MP locks.
- assert B_CALL buffers are not on the free queue.
|
| 1.69 | 07-Aug-2003 |
agc | Move UCB-licensed code from 4-clause to 3-clause licence.
Patches provided by Joel Baker in PR 22364, verified by myself.
|
| 1.68 | 30-Jul-2003 |
yamt | using normal bufcache buffer for cluster buffer head.
|
| 1.67 | 12-Jul-2003 |
yamt | - protect global resource counts with lfs_subsys_lock.
- clean up scattered externs a little.
|
| 1.66 | 02-Jul-2003 |
yamt | use queue.h macros.
|
| 1.65 | 02-Jul-2003 |
yamt | - add a new functions, lfs_writer_enter/leave, and use them instead of
duplicated code fragments.
- add an assertion.
|
| 1.64 | 23-Apr-2003 |
perseant | branches: 1.64.2;
Make LFS work better (though still not "well") as an NFS-exported
filesystem (and other things that needed to be fixed before the tests
would complete), to wit:
* Include the fs ident in the filehandle; improve stale filehandle checks.
* Change definition of blksize() to use the on-dinode size instead of
the inode's i_size, so that fsck_lfs will work properly again.
* Use b_interlock in lfs_vtruncbuf.
* Postpone dirop reclamation until after the seglock has been released,
so that lfs_truncate is not called with the segment lock held.
* Don't loop in lfs_fsync(), just write everything and wait.
* Be more careful about the interlock/uobjlock in lfs_putpages: when we
lose this lock, we have to resynchronize dirtiness of pages in each
block.
* Be sure to always write indirect blocks and update metadata in
lfs_putpages; fixes a bug that caused blocks to be accounted to the
wrong segment.
|
| 1.63 | 09-Apr-2003 |
thorpej | Use PAGE_SIZE rather than NBPG.
|
| 1.62 | 02-Apr-2003 |
fvdl | Add support for UFS2. UFS2 is an enhanced FFS, adding support for
64 bit block pointers, extended attribute storage, and a few
other things.
This commit does not yet include the code to manipulate the extended
storage (for e.g. ACLs), this will be done later.
Originally written by Kirk McKusick and Network Associates Laboratories for
FreeBSD.
|
| 1.61 | 28-Mar-2003 |
perseant | Add a sleeper count, to prevent the cleaner from panicing the kernel
when the filesystem is unmounted, relocking the Ifile when its lock is
draining. (We can't use vfs_busy() since the process is sleeping for a
good long time.) Clean up / organize lfs.h, while I'm here.
In lfs_update_single, assert that disk addresses are either negative, or
are still positive when converted to int32_t, to prevent recurrence of a
negative/positive block problem.
|
| 1.60 | 21-Mar-2003 |
perseant | KNF (space after keywords).
|
| 1.59 | 21-Mar-2003 |
perseant | Use VONWORKLST as a heuristic for vnode emptiness, rather than exhaustively
checking the memq.
Take greater care not to dirty the Ifile vnode when unmounting the filesystem.
This should fix a "(vp->v_flag & VONWORKLST) == 0" assertion panic in vgonel
that could occur when unmounting.
Do not allow the Ifile to be mapped for writing.
|
| 1.58 | 15-Mar-2003 |
perseant | Add simple_lock protection for lfs_seglock and lfs_subsys_pages; these will
be expanded to cover other per-fs and subsystem-wide data as well.
Fix a case of IN_MODIFIED being set without updating lfs_uinodes, resulting
in a "lfs_uinodes < 0" panic.
Fix a deadlock in lfs_putpages arising from the need to busy all pages in a
block; unbusy any that had already been busied before starting over.
|
| 1.57 | 11-Mar-2003 |
perseant | - Get rid of unused #ifdefs LFS_NO_PAGEMOVE and LFS_MALLOC_SUMMARY (both
always true) and accompanying dead code.
- When constructing write clusters in lfs_writeseg, if the block we are
about to add is itself a cluster from GOP_WRITE, don't put a cluster
in a cluster, just write the GOP_WRITE cluster on its own. This seems
to represent a slight performance gain on my test machine.
- Charge someone's rusage for writes on LFSes. It's difficult to tell
who the "right" process to charge is; just charge whoever triggered
the write.
|
| 1.56 | 08-Mar-2003 |
perseant | Take away "#ifdef LFS_UBC".
|
| 1.55 | 08-Mar-2003 |
perseant | Add an lfs_strategy() that checks to make sure we're not trying to read
where the cleaner is trying to write, instead of tying up the "live"
buffers (or pages).
Fix a bug in the LFS_UBC case where oversized buffers would not be
checksummed correctly, causing uncleanable segments.
Make sure that wakeup(fs->lfs_iocount) is done if fs->lfs_iocount is 1
as well as 0, since we wait in some places for it to drop to 1.
Activate all pages that make it into lfs_gop_write without the segment
lock held, since they must have been dirtied very recently, even if
PG_DELWRI is not set.
|
| 1.54 | 02-Mar-2003 |
perseant | Account SEGUSE_ACTIVE correctly so that the automatic segment cleaning
actually happens.
Add a new fcntl call that will write the minimum necessary to checkpoint
(i.e., for on-disk directory structure to be consistent, not including
updates to file data) so that the cleaner can clean segments more quickly
without sacrificing three-way commit for cleaning.
|
| 1.53 | 27-Feb-2003 |
perseant | Do roundup and offset arithmetic in 64 bits, to allow >=2G files.
|
| 1.52 | 25-Feb-2003 |
perseant | Make fs-specific fcntl macros take three arguments (approved wrstuden).
Let LFS use fcntl for cleaner functions.
|
| 1.51 | 24-Feb-2003 |
perseant | Add lfs_ioctl vnode op, with ioctls to take over cleaner system call
functionality (not including segment clean, since that is now done
automatically as checkpoints happen).
|
| 1.50 | 23-Feb-2003 |
perseant | Fix a buffer overflow bug in the LFS_UBC case that manifested itself
either as a mysterious UVM error or as "panic: dirty bufs". Verify
maximum size in lfs_malloc.
Teach lfs_updatemeta and lfs_shellsort about oversized cluster blocks from
lfs_gop_write.
When unwiring pages in lfs_gop_write, deactivate them, under the theory
that the pagedaemon wanted to free them last we knew.
|
| 1.49 | 20-Feb-2003 |
perseant | Tabify, and fix some comment alignment problems.
|
| 1.48 | 19-Feb-2003 |
yamt | workaround for "another flush is..." infinity loop in writerd.
if we're writerd, sleep in lfs_flush until another writer goes away
instead of busy loop in writed.
|
| 1.47 | 18-Feb-2003 |
soren | Make libsa compile again.
|
| 1.46 | 17-Feb-2003 |
perseant | Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now
(there are still some details to work out) but expect that to go
away soon. To support these basic changes (creation of lfs_putpages,
lfs_gop_write, mods to lfs_balloc) several other changes were made, to
wit:
* Create a writer daemon kernel thread whose purpose is to handle page
writes for the pagedaemon, but which also takes over some of the
functions of lfs_check(). This thread is started the first time an
LFS is mounted.
* Add a "flags" parameter to GOP_SIZE. Current values are
GOP_SIZE_READ, meaning that the call should return the size of the
in-core version of the file, and GOP_SIZE_WRITE, meaning that it
should return the on-disk size. One of GOP_SIZE_READ or
GOP_SIZE_WRITE must be specified.
* Instead of using malloc(...M_WAITOK) for everything, reserve enough
resources to get by and use malloc(...M_NOWAIT), using the reserves if
necessary. Use the pool subsystem for structures small enough that
this is feasible. This also obsoletes LFS_THROTTLE.
And a few that are not strictly necessary:
* Moves the LFS inode extensions off onto a separately allocated
structure; getting closer to LFS as an LKM. "Welcome to 1.6O."
* Unified GOP_ALLOC between FFS and LFS.
* Update LFS copyright headers to correct values.
* Actually cast to unsigned in lfs_shellsort, like the comment says.
* Keep track of which segments were empty before the previous
checkpoint; any segments that pass two checkpoints both dirty and
empty can be summarily cleaned. Do this. Right now lfs_segclean
still works, but this should be turned into an effectless
compatibility syscall.
|
| 1.45 | 29-Jan-2003 |
yamt | don't use daddr_t for segment summary since it's an on-disk structure.
|
| 1.44 | 27-Jan-2003 |
yamt | make these compilable with lfs debug options.
(follow daddr_t change)
XXX maybe segment number should be 64bit.
|
| 1.43 | 24-Jan-2003 |
fvdl | Bump daddr_t to 64 bits. Replace it with int32_t in all places where
it was used on-disk, so that on-disk formats remain the same.
Remove ufs_daddr_t and ufs_lbn_t for the time being.
|
| 1.42 | 01-Dec-2002 |
matt | Add multiple inclusion protection for headers. Fix mismatched
variable declarations (missing const's) as needed.
|
| 1.41 | 06-Jul-2002 |
perseant | Deal with fragment size changes better. For each fragment that can
exist on an on-disk inode, we keep a record of its size in struct inode,
which is updated when we write the block to disk. The cleaner routines
thus have ready access to what size is the correct size for this block,
on disk.
Fixed a related bug: if a file with fragments is being cleaned
(fragments being cleaned) at the same time it is being extended beyond
NDADDR blocks, we could write a bogus FINFO record that has a frag in the
middle; when it was cleaned this would give back bogus file data. Don't
write the indirect blocks in this case, since there is no need.
lfs_fragextend and lfs_truncate no longer require the seglock, but instead
take a shared lock, which the seglock locks exclusively.
|
| 1.40 | 16-Jun-2002 |
perseant | For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.39 | 14-May-2002 |
perseant | branches: 1.39.2; 1.39.4;
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes
I found while making sure there weren't any new ones.
* Make the write clusters keep track of the buffers whose blocks they contain.
This should make it possible to (1) write clusters using a page mapping
instead of malloc, if desired, and (2) schedule blocks for rewriting
(somewhere else) if a write error occurs. Code is present to use
pagemove() to construct the clusters but that is untested and will go away
anyway in favor of page mapping.
* DEBUG now keeps a log of Ifile writes, so that any lingering instances of
the "dirty bufs" problem can be properly debugged.
* Keep track of whether the Ifile has been dirtied by various routines that
can be called by lfs_segwrite, and loop on that until it is clean, for
a checkpoint. Checkpoints need to be squeaky clean.
* Warn the user (once) if the Ifile grows larger than is reasonable for their
buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can
grow.
* If an inode is not found in a disk block, try rereading the block, under
the assumption that the block was copied to a cluster and then freed.
* Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
|
| 1.38 | 23-Nov-2001 |
chs | add spaces for KNF. confirmed to produce identical objects.
|
| 1.37 | 13-Jul-2001 |
perseant | Merge the short-lived perseant-lfsv2 branch into the trunk.
Kernels and tools understand both v1 and v2 filesystems; newfs_lfs
generates v2 by default. Changes for the v2 layout include:
- Segments of non-PO2 size and arbitrary block offset, so these can be
matched to convenient physical characteristics of the partition (e.g.,
stripe or track size and offset).
- Address by fragment instead of by disk sector, paving the way for
non-512-byte-sector devices. In theory fragments can be as large
as you like, though in reality they must be smaller than MAXBSIZE in size.
- Use serial number and filesystem identifier to ensure that roll-forward
doesn't get old data and think it's new. Roll-forward is enabled for
v2 filesystems, though not for v1 filesystems by default.
- The inode free list is now a tailq, paving the way for undelete (undelete
is not yet implemented, but can be without further non-backwards-compatible
changes to disk structures).
- Inode atime information is kept in the Ifile, instead of on the inode;
that is, the inode is never written *just* because atime was changed.
Because of this the inodes remain near the file data on the disk, rather
than wandering all over as the disk is read repeatedly. This speeds up
repeated reads by a small but noticeable amount.
Other changes of note include:
- The ifile written by newfs_lfs can now be of arbitrary length, it is no
longer restricted to a single indirect block.
- Fixed an old bug where ctime was changed every time a vnode was created.
I need to look more closely to make sure that the times are only updated
during write(2) and friends, not after-the-fact during a segment write,
and certainly not by the cleaner.
|
| 1.36 | 20-Dec-2000 |
cgd | branches: 1.36.2; 1.36.4; 1.36.6;
replace \<space(s)><newline> (wrong!) with \<newline>
|
| 1.35 | 17-Nov-2000 |
perseant | Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes.
(PR #11468). In the case of fragment allocation, check to see if enough
space is available before extending a fragment already scheduled for writing.
The locked_queue_* variables indicate the number of buffer headers and bytes,
respectively, that are unavailable to getnewbuf() because they are locked up
waiting for LFS to flush them; make sure that that is actually what we're
counting, i.e., never count malloced buffers, and always use b_bufsize instead
of b_bcount.
If DEBUG is defined, the periodic calls to lfs_countlocked will now complain
if either counter is incorrect. (In the future lfs_countlocked will not need
to be called at all if DEBUG is not defined.)
|
| 1.34 | 13-Nov-2000 |
perseant | Remove debugging code that accidentally went in with yesterday's commit.
|
| 1.33 | 12-Nov-2000 |
perseant | Do not needlessly dirty segment table blocks during lfs_segwrite,
preventing needless disk activity when the filesystem is idle. (PR #10979.)
|
| 1.32 | 13-Sep-2000 |
perseant | Cast back to int32_t in LFS_EST_BFREE and LFS_EST_RSVD macros, for
consistency with their arguments.
Change the debugging printf in lfs_reserve to match, and enclose it in
#ifdef DEBUG.
Tested on alpha, arm32, sparc.
|
| 1.31 | 09-Sep-2000 |
perseant | Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
|
| 1.30 | 09-Sep-2000 |
perseant | Change dlfs_dmeta and dlfs_avail to signed quantities, to prevent
underflow errors, visible in userland as impossibly high values
returned from df(1).
|
| 1.29 | 05-Jul-2000 |
perseant | Clean up accounting of lfs_uinodes (dirty but unwritten inodes).
Make lfs_uinodes a signed quantity for debugging purposes, and set it to
zero as fs mount time.
Enclose setting/clearing of the dirty flags (IN_MODIFIED, IN_ACCESSED,
IN_CLEANING) in macros, and use those macros everywhere. Make
LFS_ITIMES use these macros; updated the ITIMES macro in inode.h to know
about this. Make ufs_getattr use ITIMES instead of FFS_ITIMES.
|
| 1.28 | 04-Jul-2000 |
perseant | Fix errors observed while trying to fill the filesystem with yesterday's
fixes:
- Write copies of bfree and avail in the CLEANERINFO block, so the
cleaner doesn't have to guess which superblock has the current
information (if indeed any do).
- Tighten up accounting of lfs_avail (more needs to be done).
- When cleansing indirect blocks of UNWRITTEN, make sure not to mark
them clean, since they'll need to be rewritten later.
|
| 1.27 | 03-Jul-2000 |
perseant | Allow the number of free segments reserved for the cleaner to be
parametrized in the filesystem, defaulting to MIN_FREE_SEGS = 2 but set
to something more reasonable at newfs_lfs time.
Note the number of blocks that have been scheduled for writing but which
are not yet on disk in an inode extension, i_lfs_effnblks. Move
i_ffs_effnlink out of the ffs extension and onto the main inode, since
it's used all over the shared code and the lfs extension would clobber
it.
At inode write time, indirect blocks and inode-held blocks of inodes
that have i_lfs_effnblks != i_ffs_blocks are cleansed of UNWRITTEN disk
addresses, so that these never make it to disk.
|
| 1.26 | 27-Jun-2000 |
perseant | Fixes associated with filling an LFS:
Change the space computation to appear to change the size of the *disk*
rather than the *bytes used* when more segment summaries and inode
blocks are written. Try to estimate the amount of space that these will
take up when more files are written, so the disk size doesn't change too
much.
Regularize error returns from lfs_valloc, lfs_balloc, lfs_truncate: they
now fail entirely, rather than succeeding half-way and leaving the fs in
an inconsistent state.
Rewrite lfs_truncate, mostly stealing from ffs_truncate. The old
lfs_truncate had difficulty truncating a large file to a non-zero size
(indirect blocks were not handled appropriately).
Unmark VDIROP on fvp after ufs_remove, ufs_rmdir, so these can be
reclaimed immediately: this vnode would not be written to disk again
anyway if the removal succeeded, and if it failed, no directory
operation occurred.
ufs_makeinode and ufs_mkdir now remove IN_ADIROP on error.
|
| 1.25 | 06-Jun-2000 |
perseant | branches: 1.25.2;
Protect inode free list with seglock, instead of separate lock, so that
the head of the inode free list (on the superblock) always matches the
rest of the free list (in the ifile).
Protect lfs_fragextend with seglock, to prevent the segment byte count
fudging from making its way to disk.
Don't try to inactivate dirop vnodes that are still in the middle of
their dirop (may address PR#10285).
|
| 1.24 | 31-May-2000 |
perseant | update for IN_ACCESSED changes
|
| 1.23 | 27-May-2000 |
perseant | branches: 1.23.2;
Prevent dirops from getting around lfs_check and wedging the buffer cache.
All the dirop vnops now mark the inodes with a new flag, IN_ADIROP, which
is removed as soon as the dirop is done (as opposed to VDIROP which stays
until the file is written). To address one issue raised in PR#9357.
|
| 1.22 | 13-May-2000 |
perseant | Change the sementics of the last parameter from a boolean ("waitfor") to
a set of flags ("flags"). Two flags are defined, UPDATE_WAIT and
UPDATE_DIROP.
Under the old semantics, VOP_UPDATE would block if waitfor were set,
under the assumption that directory operations should be done
synchronously. At least LFS and FFS+softdep do not make this
assumption; FFS+softdep got around the problem by enclosing all relevant
calls to VOP_UPDATE in a "if(!DOINGSOFTDEP(vp))", while LFS simply
ignored waitfor, one of the reasons why NFS-serving an LFS filesystem
did not work properly.
Under the new semantics, the UPDATE_DIROP flag is a hint to the
fs-specific update routine that the call comes from a dirop routine, and
should be wait for, or not, accordingly.
Closes PR#8996.
|
| 1.21 | 05-May-2000 |
perseant | Change the way LFS does block accounting, from trying to infer from the
buffer cache flags, to marking the inode and/or indirect blocks with a
special disk address UNWRITTEN==-2 when a block is accounted for. (This
address is never written to disk, but only used in-core. This is essentially
the same method of block accounting as on the UBC branch, where the buffer
headers don't exist.) Make sure that truncation is handled properly,
especially in the case of holey files.
Fixes PR#9994.
|
| 1.20 | 19-Jan-2000 |
perseant | Changes to stabilize LFS. The first two of these should also apply to the
1.4 branch.
* Use a separate per-fs lock, instead of ufs_hashlock, to protect the Inode
free list. This seems to prevent the "lockmgr: %d, not exclusive lock holder
%d, unlocking" message I was mis-attributing last night to an unlocked vnode
being passed to vrele.
* Change calling semantics of lfs_ifind, to give better error reporting:
If fed a struct buf, it can report the block number of the offending inode
block as well as the inode number.
* Back out rev 1.10 of lfs_subr.c, since the replacement code was slightly
uglier while being functionally identical.
* Make lfs_vunref use the same free list convention as vrele/vput, so that
vget does not remove vnodes from a hash list they are not on.
|
| 1.19 | 15-Dec-1999 |
perseant | In lfs_bwrite, don't mark buffers dirty if lfs is mounted read-only.
(Previously buffers could be marked dirty by the cleaner, and possibly by
other means.)
Also check for softdep mount in vfs_shutdown before trying to bawrite
buffers, since other filesystems don't need it and lfs doesn't bawrite.
(This fragment reviewed by fvdl.)
Partially addresses PR#8964.
|
| 1.18 | 08-Dec-1999 |
simonb | Use an explicitly sized type (u_int32_t) for inode numbers in the super
block instead of ino_t. Reviewed by Konrad Schroder.
|
| 1.17 | 06-Nov-1999 |
perseant | branches: 1.17.2;
Address ufs_hashlock/ufs_ihashins protocol bug, discovered while doing a
post-mortem of a production machine. Also, take the active dirop
count off of the fs and make it global (since it is measuring a global
resource) and tie the threshold value LFS_MAXDIROP to desiredvnodes.
|
| 1.16 | 15-Jun-1999 |
perseant | branches: 1.16.2; 1.16.4; 1.16.6;
Minor changes to the segment live bytes calculation. In particular, fixed
a bug in fragment extension that could run the count negative. Also, don't
overcount for inodes, and don't count segment summaries. Thus, for empty
segments the live bytes count should now be exactly zero.
|
| 1.15 | 01-Jun-1999 |
perseant | Fixed lfs_update (and related functions) so that calls from lfs_fsync
will DTRT with vnodes marked VDIROP. In particular, the message
"flushing VDIROP" will no longer appear, and the filesystem will remain
stable in the event of a crash.
This was particularly a problem with NFS-exported LFSes, since fsync
was called on every file close.
|
| 1.14 | 25-Mar-1999 |
perseant | branches: 1.14.2; 1.14.4; 1.14.6;
clean up unused/required #ifdefs
|
| 1.13 | 17-Mar-1999 |
perseant | Move dlfs_pad to the end of struct dlfs (after the pad), for upward
compatibility.
|
| 1.12 | 17-Mar-1999 |
perseant | Fix pad on lfs.h so it is really 512 bytes, as advertized
|
| 1.11 | 10-Mar-1999 |
perseant | New sources should leave the LFS in a more-or-less working state. Changes
include:
- DIROP segregation is enabled, and greater care is taken
to make sure that a checkpoint completes. Fsck is not
needed to remount the filesystem.
- Several checks to make sure that the LFS subsystem does not
overuse various resources (memory, in particular).
- The cleaner routines, lfs_markv in particular, are completely
rewritten. A buffer overflow is removed. Greater care is taken
to ensure that inodes come from where lfs_cleanerd say they come
from (so we know nothing has changed since lfs_bmapv was called).
- Fragment allocation is fixed, so that writes beyond end-of-file
do the right thing.
|
| 1.10 | 11-Sep-1998 |
pk | PR#6032: define fixed sized on-disk superblock structure.
|
| 1.9 | 01-Mar-1998 |
fvdl | Merge with Lite2 + local changes
|
| 1.8 | 05-Dec-1996 |
is | Make the struct lfs 512 bytes long on 32bit machines whose compiler doesn't
align 32bit integers. Use explicit sized typing at some other places.
XXX This still won't fix lfs for 64bit machines, as we have some
assumptions about sizeof(pointer)=sizeof(u_int32_t) in here, and (if I
looked right) a misaligned u_int64_t. The right fix (to cite cgd) will
be to seperate on-disk-representation from in-core, but I don't have
the time (at the moment) to do this.
|
| 1.7 | 09-Feb-1996 |
christos | lfs prototypes
|
| 1.6 | 21-Dec-1994 |
mycroft | Add RCS ids where missing.
|
| 1.5 | 14-Dec-1994 |
mycroft | Sync with CSRG.
|
| 1.4 | 17-Nov-1994 |
mycroft | Round struct lfs to 512 bytes.
|
| 1.3 | 20-Oct-1994 |
cgd | update for new syscall args description mechanism, and deal safely
with wider types.
|
| 1.2 | 29-Jun-1994 |
cgd | New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
|
| 1.1 | 08-Jun-1994 |
mycroft | branches: 1.1.1;
Update to 4.4-Lite fs code, with local changes.
|
| 1.1.1.2 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite2
|
| 1.1.1.1 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite for reference
|
| 1.14.6.1 | 30-Nov-1999 |
itojun | bring in latest KAME (as of 19991130, KAME/NetBSD141) into kame branch
just for reference purposes.
This commit includes 1.4 -> 1.4.1 sync for kame branch.
The branch does not compile at all (due to the lack of ALTQ and some other
source code). Please do not try to modify the branch, this is just for
referenre purposes.
synchronization to latest KAME will take place on HEAD branch soon.
|
| 1.14.4.2 | 31-Aug-1999 |
perseant | Rudimentary support for LFS under UBC:
- LFS-specific VOP_BALLOC and VOP_PUTPAGES vnode ops.
- getblk VREG panic #ifdef'd out (can be reinstated when Ifile is
internalized and Ifile can be made another type from VREG)
- interface to VOP_PUTPAGES changed to pass all pager flags, not
just sync. FS putpages routines must know about the pager flags.
- new LFS magic disk address, -2 ("unwritten"), meaning accounted for
but not assigned to a fixed disk location (since LFS does these two
things separately, and the previous accounting method using buffer
headers no longer will work). Changed references to (foo == (daddr_t)-1)
to (foo < 0). Since disk drivers reject all addresses < 0, this should
not present a problem for other FSs.
|
| 1.14.4.1 | 21-Jun-1999 |
thorpej | Sync w/ -current.
|
| 1.14.2.4 | 20-Jan-2000 |
he | Pull up revision 1.20 (requested by perseant):
Files removed (through unlink, rmdir) are now really removed, though the
removal is postponed until the dirop is complete to ensure validity of
the filesystem through a crash. Use a separate per-fs lock, instead of
ufs_hashlock, to protect the inode free list. Change calling semantics
of lfs_ifind, to give better error reporting: If fed a struct buf, it
can report the block number of the offending inode block as well as the
inode number.
|
| 1.14.2.3 | 17-Dec-1999 |
he | Pull up revision 1.17 (requested by perseant):
Address locking protocol error for inode hash, and make the
maximum number of active dirops a global quantity.
|
| 1.14.2.2 | 17-Dec-1999 |
he | Pull up revision 1.15 (requested by perseant):
Avoid flushing vnodes involved in a dirop, making lfs' promise
of "no fsck needed, even in the event of a crash" closer to
reality.
|
| 1.14.2.1 | 25-Jun-1999 |
perry | pullup 1.15->1.16 (perseant)
|
| 1.16.6.1 | 27-Dec-1999 |
wrstuden | Pull up to last week's -current.
|
| 1.16.4.1 | 15-Nov-1999 |
fvdl | Sync with -current
|
| 1.16.2.3 | 05-Jan-2001 |
bouyer | Sync with HEAD
|
| 1.16.2.2 | 22-Nov-2000 |
bouyer | Sync with HEAD.
|
| 1.16.2.1 | 20-Nov-2000 |
bouyer | Update thorpej_scsipi to -current as of a month ago
|
| 1.17.2.2 | 06-Nov-1999 |
perseant | Address ufs_hashlock/ufs_ihashins protocol bug, discovered while doing a
post-mortem of a production machine. Also, take the active dirop
count off of the fs and make it global (since it is measuring a global
resource) and tie the threshold value LFS_MAXDIROP to desiredvnodes.
|
| 1.17.2.1 | 06-Nov-1999 |
perseant | file lfs.h was added on branch comdex-fall-1999 on 1999-11-06 20:33:06 +0000
|
| 1.23.2.1 | 22-Jun-2000 |
minoura | Sync w/ netbsd-1-5-base.
|
| 1.25.2.2 | 03-Feb-2001 |
he | Pull up revisions 1.33-1.35 (requested by perseant):
o Don't write anything if the filesystem is idle (PR#10979).
o Close up accounting holes in LFS' accounting of immediately-
available-space, number of clean segments, and amount of dirty
space taken up by metadata (PR#11468, PR#11470, PR#11534).
|
| 1.25.2.1 | 14-Sep-2000 |
perseant | Pull up recent LFS kernel changes (approved by thorpej):
ufs/ufs/inode.h, 1.20--1.22 (add i_lfs_effnblks extension ;
make ITIMES aware of LFS_ITIMES;
_LKM protection so userland progs
compile)
ufs/ufs/ufs_vnops.c, 1.69, 1.71 (remove IN_ADIROP;
use ITIMES instead of FFS_ITIMES)
ufs/ufs/ufs_readwrite.c, 1.27 (use lfs_reserve in lfs_write)
ufs/lfs/lfs.h, 1.26--1.32 (define LFS_EST_* macros ;
change MIN_FREE_SEGS to lfs_minfreesegs ;
add avail and bfree to CLEANERINFO ;
change lfs_uinodes to signed ;
change lfs_dmeta to signed ;
add whitespace to line up structure
members ;
explicit cast to int32_t in LFS_EST_*
macros)
ufs/lfs/lfs_alloc.c, back out 1.34.2.3 (pullups of 1.39, 1.40);
then pull up 1.38 (clean up on error)
1.39--1.43 (restore fvdl's ufs_hashlock fix ;
restore fvdl's ufs_hashlock fix ;
set i_lfs_effnblks ;
use UINO macros ;
add comments and fix long lines)
ufs/lfs/lfs_balloc.c, 1.19 (don't succeed halfway)
1.21--1.25 (use i_lfs_effnblks ;
fix i_lfs_effnblks computation and
quieten ;
fix i_ffs_blocks in unwritten fragment ;
remove useless debugging check ;
add comments and (c) 2000)
ufs/lfs/lfs_bio.c, 1.24--1.30 (cleanup and make lfs_flush_fs take
"struct lfs *" instead of "struct
mount *" ;
use lfs_minfreeseg instead of
MIN_FREE_SEGS ;
use UINO macros, and copy bfree/avail
to CLEANERINFO ;
add lfs_reserve function ;
1.28--1.30 fix printf formatting)
ufs/lfs/lfs_cksum.c, 1.13 (add (c) 2000)
ufs/lfs/lfs_debug.c, 1.11 (use btodb instead of DEV_BSIZE)
ufs/lfs/lfs_extern.h, 1.18, 1.20--1.21 (function prototype changes)
ufs/lfs/lfs_inode.c, 1.38 (rewrite lfs_truncate from
ffs_truncate)
1.40--1.44 (count written and unwritten blocks
seperately ;
use disk block units instead of bytes ;
remove unnecessary "mod" variable ;
correct B_DELWRI to avoid bawrite panic ;
use lfs_reserve)
ufs/lfs/lfs_segment.c, 1.52-1.59 (use lfs_dmeta to note used summaries ;
check for UNWRITTEN in indirect blocks ;
more debugging stuff inside #ifdef
DEBUG_LFS ;
use LK_CANRECURSE ;
don't drop dirty indirect blocks ;
use UINO macros ;
don't hose the free list ;
use btodb() instead of DEV_BSIZE ;
make it compile again (oops))
ufs/lfs/lfs_subr.c, 1.16--1.17 (check for locked inodes before
changing ;
use btodb() instead of DEV_BSIZE, (c)
2000)
ufs/lfs/lfs_syscalls.c, back out 1.41.4.2 (fvdl's ufs_hashlock fix);
then pull up 1.43 (use lfs_dmeta)
1.44--1.45 (restore fvdl's ufs_hashlock fix)
1.46--1.47 (fix lfs_avail leakage from sblock
segments ;
use UINO macros)
1.49 (bounds-check inode numbers in
lfs_markv)
ufs/lfs/lfs_vfsops.c, 1.53 (use LFS_EST_* macros in lfs_statfs)
1.56--1.58 (initialize lfs_minfreeseg, lfs_effnblk ;
initialize lfs_uinodes ;
initialize lfs_ravail)
ufs/lfs/lfs_vnops.c, 1.40 (remove VDIROP from removed files)
1.42--1.44 (move SET_ENDOP below the removal of
VDIROP ;
use UINO macros and add lfs_itimes
function ;
use lfs_reserve in dirops)
|
| 1.36.6.4 | 06-Sep-2002 |
jdolecek | sync kqueue branch with HEAD
|
| 1.36.6.3 | 23-Jun-2002 |
jdolecek | catch up with -current on kqueue branch
|
| 1.36.6.2 | 10-Jan-2002 |
thorpej | Sync kqueue branch with -current.
|
| 1.36.6.1 | 03-Aug-2001 |
lukem | update to -current
|
| 1.36.4.4 | 13-Jul-2001 |
perseant | Be more careful about when we update ctime/mtime. In particular, if we
are only writing indirect blocks, that doesn't count for mtime; and when
we first create a vnode, that certainly *does not* count for ctime
(a bug that's been there from the beginning).
This does not change the fact that mtime might still be set after write(2)
is "completed", but it does make the atime-in-the-ifile code have some
effect (noticeable less degradation of read time after an intervening
large write).
|
| 1.36.4.3 | 10-Jul-2001 |
perseant | Turn the free list into a tailq, with both head and tail kept on the ifile.
Update access times on the inode even if it does not get marked IN_ACCESS.
|
| 1.36.4.2 | 02-Jul-2001 |
perseant | Change disk addressing unit to be the fragment, instead of the disk sector.
All quantities in the superblock, inodes, indirect blocks, etc. refer now
to this abstract unit (called "fsb" as it is in FFS) instead of disk sectors;
as a consequence segment summary blocks have to be multiples of a fragment in
size. In v1 filesystems, compatibility code ensures that 1 fsb == 1 sector,
regardless of fragment size.
Fragments can now range in size between 512 and 32k; in the event that
LFS_LABELPAD (8k) is smaller than the disk address unit size, an extra
proto-superblock is kept at 8k from the beginning of the disk, to be used
*only* to locate the real superblocks. (Not all of the userland knows about
this yet.)
Almost all of this was done not by me, but by joff.
|
| 1.36.4.1 | 27-Jun-2001 |
perseant | Import of what I've been calling "LFSv2", that is, LFS with some features
added that require changes to the on-disk data structures. These include:
- 64-bit time in everything but inodes
- User-specified segment offset, and segment size no longer
restricted to PO2.
- Serial number on segment summaries in addition to timestamp, and
a new volume identifier, to make roll-forward feasible without
fear of finding old data and thinking it was new.
Although I think this version works at least as well as what's on the trunk,
we're not done yet; hence this commit is going in on a branch and not on
the trunk. Enhancements that are not here yet include fragment addressing,
like FFS does, instead of block addressing.
|
| 1.36.2.5 | 11-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.36.2.4 | 01-Aug-2002 |
nathanw | Catch up to -current.
|
| 1.36.2.3 | 20-Jun-2002 |
nathanw | Catch up to -current.
|
| 1.36.2.2 | 08-Jan-2002 |
nathanw | Catch up to -current.
|
| 1.36.2.1 | 24-Aug-2001 |
nathanw | Catch up with -current.
|
| 1.39.4.1 | 20-Jun-2002 |
lukem | Pull up revision 1.40 (requested by perseant in ticket #325):
For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.39.2.2 | 15-Jul-2002 |
gehenna | catch up with -current.
|
| 1.39.2.1 | 20-Jun-2002 |
gehenna | catch up with -current.
|
| 1.64.2.7 | 10-Nov-2005 |
skrll | Sync with HEAD. Here we go again...
|
| 1.64.2.6 | 08-Mar-2005 |
skrll | Sync with HEAD.
|
| 1.64.2.5 | 04-Mar-2005 |
skrll | Sync with HEAD.
Hi Perry!
|
| 1.64.2.4 | 21-Sep-2004 |
skrll | Fix the sync with head I botched.
|
| 1.64.2.3 | 18-Sep-2004 |
skrll | Sync with HEAD.
|
| 1.64.2.2 | 25-Aug-2004 |
skrll | Sync with HEAD.
|
| 1.64.2.1 | 03-Aug-2004 |
skrll | Sync with HEAD
|
| 1.72.4.1 | 10-May-2005 |
riz | Pull up the following revisions (requested by perseant in ticket #1281):
1.8 sys/ufs/lfs/TODO
1.75 sys/ufs/lfs/lfs.h (via patch)
1.74 sys/ufs/lfs/lfs_alloc.c (via patch)
1.49, 1.51 sys/ufs/lfs/lfs_balloc.c (1.51 via patch)
1.78 sys/ufs/lfs/lfs_bio.c
1.62 sys/ufs/lfs/lfs_extern.h (via patch)
1.156 sys/ufs/lfs/lfs_segment.c (via patch)
1.48 sys/ufs/lfs/lfs_subr.c
1.101 sys/ufs/lfs/lfs_syscalls.c
1.163 sys/ufs/lfs/lfs_vfsops.c (via patch)
1.134 sys/ufs/lfs/lfs_vnops.c (via patch)
1.61 sys/ufs/ufs/ufs_readwrite.c (via patch)
1.20 libexec/lfs_cleanerd/clean.h (via patch)
1.52 libexec/lfs_cleanerd/cleanerd.c (via patch)
1.41 libexec/lfs_cleanerd/library.c (via patch)
1.4 regress/sys/fs/lfs/newfs_fsck/Makefile
1.2 regress/sys/fs/lfs/newfs_fsck/mkfs_mount
1.2 regress/sys/fs/lfs/newfs_fsck/smallfiles
1.3 sbin/fsck_lfs/bufcache.c
1.3 sbin/fsck_lfs/bufcache.h
1.3 sbin/fsck_lfs/lfs.h
1.8 sbin/fsck_lfs/lfs.c (via patch)
1.8 sbin/fsck_lfs/pass3.c (via patch)
1.18 sbin/fsck_lfs/pass0.c (via patch)
1.18 sbin/fsck_lfs/utilities.c (via patch)
1.7 sbin/fsck_lfs/segwrite.c
1.19 sbin/fsck_lfs/setup.c (via patch)
1.3 sbin/newfs_lfs/Makefile
0 sbin/newfs_lfs/lfs.c (yes, remove it)
1.1 sbin/newfs_lfs/make_lfs.c
1.15 sbin/newfs_lfs/newfs.c (via patch)
Various minor LFS improvements.
Kernel:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this. Should fix PR #29045.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
Fixes PR #26680.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
cleaner:
* Adapt lfs_cleanerd to use the fcntl call to get the Ifile filehandle,
so it need not be in the namespace.
* Make lfs_cleanerd be more careful when there are very few available
segments.
* Make lfs_cleanerd less verbose when the filesystem is unmounted.
newfs_lfs, fsck_lfs, and regression:
* Extend the lfs library from fsck_lfs(8) so that it can be used with a
not-yet-existent LFS. Make newfs_lfs(8) use this library, so it can
create LFSs whose Ifile is larger than one segment. Addresses PR #11110.
* Make newfs_lfs(8) use strsuftoi64() for its arguments, a la newfs(8).
* Make fsck_lfs(8) respect the "file system is clean" flag.
* Don't let fsck_lfs(8) think it has dirty blocks when invoked with the
-n flag.
* Remove the Ifile from the filesystem namespace. The cleaner now uses
a fcntl call on the root inode to find the Ifile filehandle. (As a
side-effect, addresses PR #29144.)
|
| 1.74.6.1 | 19-Mar-2005 |
yamt | sync with head. xen and whitespace. xen part is not finished.
|
| 1.74.4.1 | 29-Apr-2005 |
kent | sync with -current
|
| 1.77.2.18 | 10-Aug-2006 |
tron | Apply patch (requested by fair in perseant #1457):
Bring LFS up to current, including a patch (1.95 lfs_alloc.c) that
should prevent the inode free list errors seen on the STABLE branch
subsequent to pullup ticket #1327.
|
| 1.77.2.17 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.93
sys/ufs/lfs/lfs.h: revision 1.106
sys/ufs/lfs/lfs_vfsops.c: revision 1.209
sys/ufs/lfs/lfs_vnops.c: revision 1.175
sys/ufs/lfs/lfs_segment.c: revision 1.178
Fixes to address the "vinvalbuf: dirty blocks" panic that can occur when
many inodes are cleaned at once. Make sure that we write all the pages
on vnodes that are being flushed, even if we don't think there's room;
drain v_numoutput before lfs_vflush() completes.
Also, don't allow a vnode that is in the process of being cleaned to be
chosen by getnewvnode(); this avoids a segment accounting panic in the case
that a large number of inodes are fed to lfs_markv() all at once.
|
| 1.77.2.16 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.92
sys/ufs/lfs/lfs.h: revision 1.105
sys/ufs/lfs/lfs_vfsops.c: revision 1.207
sys/ufs/lfs/lfs_subr.c: revision 1.59
sys/ufs/lfs/lfs_vnops.c: revision 1.173
sys/ufs/lfs/lfs_bio.c: revision 1.92
Introduce another per-filesystem parameter, lfs_resvseg, to separate the
notion of "how many segments are reserved for the cleaner" from that of
"how many segments are not counted in lfs_bfree". The default value
used for existing filesystems is the same as the previous implicit value
of (lfs_minfreeseg / 2 + 1), modulo some sanity checking.
Count pending dirops on a per-filesystem basis, since once we start
writing them we can't stop until we're done. This seems to help stave off
the "no clean segments" panic in the case of filling the filesystem with
directories and small files (e.g. simultaneously unpacking more copies of
pkgsrc than will fit).
|
| 1.77.2.15 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.104
sys/ufs/lfs/lfs_vfsops.c: revision 1.206
sys/ufs/lfs/lfs_vnops.c: revision 1.170
sys/ufs/lfs/lfs_extern.h: revision 1.80
sys/ufs/lfs/lfs_segment.c: revision 1.176
sys/ufs/lfs/lfs_inode.c: revision 1.103 via patch
sys/ufs/lfs/lfs_alloc.c: revision 1.90
Postpone the segment accounting changes coming from truncation until the
inode that makes those changes valid is either written to disk by
lfs_writeinode() or discarded by lfs_vfree().
A couple of locking fixes are also included as well.
|
| 1.77.2.14 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.103
sys/ufs/lfs/lfs_segment.c: revision 1.174
sys/ufs/lfs/lfs_vnops.c: revision 1.168
Introduce two fcntl calls that freeze the filesystem right at the point
where segment 0 is being considered for writing. This allows for automated
checkpoint vailidity scanning, and could be used (in conjunction with the
existing LFCNREWIND) for e.g. snapshot dumps as well.
Include a regression test that does such scanning.
When writing the Ifile, loop through the dirty block list three times to
make sure that the checkpoint is always consistent (the first and second
times the Ifile blocks can cross a segment boundary; not so the third time
unless the segments are very small). Discovered by using the aforementioned
regression test.
|
| 1.77.2.13 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.102
sys/ufs/lfs/lfs_segment.c: revision 1.173
sys/ufs/lfs/lfs_vnops.c: revision 1.167 via patch
sys/ufs/lfs/lfs_bio.c: revision 1.91
Make lfs_vref/lfs_vunref not need to know about VXLOCK and VFREEING
explicitly (especially since we didn't know about VFREEING at all before),
but notice the EBUSY return from vget() instead.
Fix some more MP locking protocol issues, most of which were pointed out by
Christian Ehrhardt this morning on tech-kern.
|
| 1.77.2.12 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.101
sys/ufs/lfs/lfs_vfsops.c: revision 1.202
sys/ufs/lfs/lfs_alloc.c: revision 1.88
Optimize the free list search a little more; in particular use words
instead of bytes for the index, and never search below fs->lfs_freehd.
Fix a bug in the previous version of the search (an erroneous assumption
that ino_t was signed).
Free the bitmap when we unmount the filesystem.
|
| 1.77.2.11 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.200
sys/ufs/lfs/lfs_vnops.c: revision 1.164
sys/ufs/lfs/lfs_inode.c: revision 1.101
sys/ufs/lfs/lfs_extern.h: revision 1.78
sys/ufs/lfs/lfs.h: revision 1.100
Implement a somewhat finer-grained mechanism for paging LFS-backed pages.
The writer daemon, if it does not need to flush the whole filesystem,
now only writes the vnodes for which the pagedaemon has requested pageouts
(although it does not pay attention to the page ranges the pagedaemon
supplies).
|
| 1.77.2.10 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.87
sys/ufs/lfs/lfs.h: revision 1.99
sys/ufs/lfs/lfs_vfsops.c: revision 1.199
sys/ufs/lfs/lfs_extern.h: revision 1.77 via patch
Keep the free list ordered. This solves a problem first pointed out to me
by Michel Oey, in which an aged LFS writes up to an extra Ifile block for
every file created; and paves the way for the truncation of the Ifile when
many files are deleted.
|
| 1.77.2.9 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_subr.c: revision 1.58
sys/ufs/lfs/lfs.h: revision 1.98
Make the segment lock aware of LWPs. Fixes a (somewhat confusing)
"lockmgr: pid 3997, not exclusive lockholder 3997, unlocking" panic I
encountered while running blogbench on an LFS.
|
| 1.77.2.8 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.158
sys/ufs/lfs/lfs_subr.c: revision 1.57
sys/ufs/lfs/lfs_segment.c: revision 1.171
sys/ufs/lfs/lfs.h: revision 1.97
sys/ufs/lfs/lfs_vfsops.c: revision 1.195
sys/ufs/lfs/lfs_extern.h: revision 1.76
Improvements to LFS's paging mechanism, to wit:
* Acknowledge that sometimes there are more dirty pages to be written to
disk than clean segments. When we reach the danger line,
lfs_gop_write() now returns EAGAIN. The caller of VOP_PUTPAGES(), if
it holds the segment lock, drops it and waits for the cleaner to make
room before continuing.
* Note and avoid a three-way deadlock in lfs_putpages (a writer holding
a page busy blocks on the cleaner while the cleaner blocks on the
segment lock while lfs_putpages blocks on the page).
|
| 1.77.2.7 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_segment.c: revision 1.170
sys/ufs/lfs/lfs.h: revision 1.96
sys/ufs/lfs/lfs_vfsops.c: revision 1.194
sys/ufs/lfs/lfs_syscalls.c: revision 1.109
From Konrad Schroeder, in response to strange df output on anoncvs.netbsd.org:
We were returning the wrong value for free space. Now we're not.
|
| 1.77.2.6 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.91
whitespace.
|
| 1.77.2.5 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.90
change ino_t to u_int32_t for syscall compatibility.
|
| 1.77.2.4 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.89
Move extern kernel variable declarations, into a _KERNEL protected session
so that the don't pollute userland's namespace.
|
| 1.77.2.3 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.152
sys/ufs/lfs/lfs_debug.c: revision 1.31
sys/ufs/lfs/lfs_subr.c: revision 1.53
sys/ufs/lfs/lfs_extern.h: revision 1.68
sys/ufs/lfs/lfs_inode.c: revision 1.96
sys/ufs/lfs/lfs_bio.c: revision 1.86
sys/ufs/lfs/lfs_alloc.c: revision 1.83
sys/ufs/lfs/lfs_vfsops.c: revision 1.181
sys/ufs/lfs/lfs.h: revision 1.88
sys/ufs/lfs/lfs_segment.c: revision 1.164
- sprinkle const
- avoid shadow variables.
|
| 1.77.2.2 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.180
sys/ufs/lfs/lfs_syscalls.c: revision 1.106
sys/ufs/lfs/lfs.h: revision 1.87
Keep track of the number of segments reclaimed, since the cleaner doesn't
do this anymore (it hasn't for quite some time). Add a couple of conditional
debugging messages to indicate why segments are not cleaned, in the event
that lfs_segclean is used.
Make the LFCNSEGWAITALL fcntl work again.
|
| 1.77.2.1 | 07-May-2005 |
tron | Apply patch (requested by perseant in ticket #242):
* fsck_lfs buffer cache fixes, including PR #29151
* Change fsck_lfs phase 0 message to reflect reality
* fsck_lfs: check phase 5 (cleanerinfo accounting) even on
roll-forward
* Keep better track of the free list during roll-forward, avoiding
a core dump
* Improve hash table use for fsck_lfs buffer and vnode cache
* Document fsck_lfs flag -f, and implement -q
* Add resize_lfs, including kernel support
* Add LFS to mountd's list of exportable filesystem types
* Make the LFS lkm work again [christos@]
* Add MP locking to the LFS kernel subsystem
* Fix pager_map deadlock in lfs_putpages()
* Avoid incomplete file extension that looks like "partial
truncation" to fsck
* Use lfs_malloc for cleaner malloc, since the cleaner often runs
in low-memory conditions.
* Use splay trees, not hash table, to track page allocation for
write.
* Fix mkdir panic on full fs
* Fix page accounting leak by counting differently.
* Use rightly named structure for lfs_getattr [skrll@]
* Cosmetic changes for readability.
|
| 1.88.2.6 | 27-Feb-2008 |
yamt | sync with head.
|
| 1.88.2.5 | 21-Jan-2008 |
yamt | sync with head
|
| 1.88.2.4 | 27-Oct-2007 |
yamt | sync with head.
|
| 1.88.2.3 | 26-Feb-2007 |
yamt | sync with head.
|
| 1.88.2.2 | 30-Dec-2006 |
yamt | sync with head.
|
| 1.88.2.1 | 21-Jun-2006 |
yamt | sync with head.
|
| 1.95.12.2 | 24-May-2006 |
tron | Merge 2006-05-24 NetBSD-current into the "peter-altq" branch.
|
| 1.95.12.1 | 28-Mar-2006 |
tron | Merge 2006-03-28 NetBSD-current into the "peter-altq" branch.
|
| 1.95.10.3 | 11-May-2006 |
elad | sync with head
|
| 1.95.10.2 | 19-Apr-2006 |
elad | sync with head.
|
| 1.95.10.1 | 08-Mar-2006 |
elad | Adapt to kernel authorization KPI.
|
| 1.95.8.7 | 03-Sep-2006 |
yamt | sync with head.
|
| 1.95.8.6 | 11-Aug-2006 |
yamt | sync with head
|
| 1.95.8.5 | 26-Jun-2006 |
yamt | sync with head.
|
| 1.95.8.4 | 24-May-2006 |
yamt | sync with head.
|
| 1.95.8.3 | 11-Apr-2006 |
yamt | sync with head
|
| 1.95.8.2 | 01-Apr-2006 |
yamt | sync with head.
|
| 1.95.8.1 | 05-Mar-2006 |
yamt | separate page replacement policy from the rest of kernel.
|
| 1.95.6.2 | 01-Jun-2006 |
kardel | Sync with head.
|
| 1.95.6.1 | 22-Apr-2006 |
simonb | Sync with head.
|
| 1.95.4.1 | 09-Sep-2006 |
rpaulo | sync with head
|
| 1.107.4.1 | 13-Jul-2006 |
gdamore | Merge from HEAD.
|
| 1.114.2.1 | 18-Nov-2006 |
ad | Sync with head.
|
| 1.116.2.1 | 22-Oct-2006 |
yamt | sync with head
|
| 1.118.8.1 | 11-Jul-2007 |
mjf | Sync with head.
|
| 1.118.6.4 | 24-Aug-2007 |
ad | Sync with buffer cache locking changes. See buf.h/vfs_bio.c for details.
Some minor portions are incomplete and needs to be verified as a whole.
|
| 1.118.6.3 | 17-Jun-2007 |
ad | - Increase the number of thread priorities from 128 to 256. How the space
is set up is to be revisited.
- Implement soft interrupts as kernel threads. A generic implementation
is provided, with hooks for fast-path MD code that can run the interrupt
threads over the top of other threads executing in the kernel.
- Split vnode::v_flag into three fields, depending on how the flag is
locked (by the interlock, by the vnode lock, by the file system).
- Miscellaneous locking fixes and improvements.
|
| 1.118.6.2 | 13-May-2007 |
ad | - Pass the error number and residual count to biodone(), and let it handle
setting error indicators. Prepare to eliminate B_ERROR.
- Add a flag argument to brelse() to be set into the buf's flags, instead
of doing it directly. Typically used to set B_INVAL.
- Add a "struct cpu_info *" argument to kthread_create(), to be used to
create bound threads. Change "bool mpsafe" to "int flags".
- Allow exit of LWPs in the IDL state when (l != curlwp).
- More locking fixes & conversion to the new API.
|
| 1.118.6.1 | 13-Mar-2007 |
ad | Pull in the initial set of changes for the vmlocking branch.
|
| 1.118.2.2 | 17-May-2007 |
yamt | sync with head.
|
| 1.118.2.1 | 07-May-2007 |
yamt | sync with head.
|
| 1.120.10.1 | 14-Oct-2007 |
yamt | sync with head.
|
| 1.120.8.3 | 23-Mar-2008 |
matt | sync with HEAD
|
| 1.120.8.2 | 09-Jan-2008 |
matt | sync with HEAD
|
| 1.120.8.1 | 06-Nov-2007 |
matt | sync with HEAD
|
| 1.120.6.1 | 26-Oct-2007 |
joerg | Sync with HEAD.
Follow the merge of pmap.c on i386 and amd64 and move
pmap_init_tmp_pgtbl into arch/x86/x86/pmap.c. Modify the ACPI wakeup
code to restore CR4 before jumping back into kernel space as the large
page option might cover that.
|
| 1.122.10.2 | 08-Jan-2008 |
bouyer | Sync with HEAD
|
| 1.122.10.1 | 02-Jan-2008 |
bouyer | Sync with HEAD
|
| 1.122.6.2 | 19-Dec-2007 |
ad | Use a global lfs_lock.
|
| 1.122.6.1 | 04-Dec-2007 |
ad | Pull the vmlocking changes into a new branch.
|
| 1.122.4.1 | 18-Feb-2008 |
mjf | Sync with HEAD.
|
| 1.125.10.4 | 11-Mar-2010 |
yamt | sync with head
|
| 1.125.10.3 | 19-Aug-2009 |
yamt | sync with head.
|
| 1.125.10.2 | 04-May-2009 |
yamt | sync with head.
|
| 1.125.10.1 | 16-May-2008 |
yamt | sync with head.
|
| 1.125.8.1 | 18-May-2008 |
yamt | sync with head.
|
| 1.125.6.1 | 02-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.126.2.1 | 23-Jun-2008 |
wrstuden | Sync w/ -current. 34 merge conflicts to follow.
|
| 1.127.12.1 | 23-Jul-2009 |
jym | Sync with HEAD.
|
| 1.132.2.1 | 30-Apr-2010 |
uebayasi | Sync with HEAD.
|
| 1.134.6.1 | 18-Feb-2012 |
mrg | merge to -current.
|
| 1.134.2.4 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.134.2.3 | 23-Jan-2013 |
yamt | sync with head
|
| 1.134.2.2 | 17-Apr-2012 |
yamt | sync with head
|
| 1.134.2.1 | 30-Nov-2011 |
yamt | make lfs another pager specific flag so that it won't be affected by
an nfs hack in genfs.
|
| 1.135.2.1 | 17-Mar-2012 |
bouyer | Pull up following revision(s) (requested by perseant in ticket #116):
sys/ufs/lfs/lfs_alloc.c: revision 1.112
tests/fs/vfs/t_rmdirrace.c: revision 1.9
tests/fs/vfs/t_renamerace.c: revision 1.25
sys/ufs/lfs/lfs_vnops.c: revision 1.240
sys/ufs/lfs/lfs_segment.c: revision 1.224
sys/ufs/lfs/lfs_bio.c: revision 1.122
sys/ufs/lfs/lfs_vfsops.c: revision 1.294
sbin/newfs_lfs/make_lfs.c: revision 1.19
sys/ufs/lfs/lfs.h: revision 1.136
Pass t_renamerace and t_rmdirrace tests.
Adapt dholland@'s fix to ufs_rename to fix PR kern/43582. Address several
other MP locking issues discovered during the course of investigating the
same problem.
Removed extraneous vn_lock() calls on the Ifile, since the Ifile writes
are controlled by the segment lock.
Fix PR kern/45982 by deemphasizing the estimate of how much metadata
will fill the empty space on disk when the disk is nearly empty
(t_renamerace crates a lot of inode blocks on a tiny empty disk).
|
| 1.136.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.136.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.136.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.136.2.1 | 25-Feb-2013 |
tls | resync with head
|
| 1.157.2.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.160.6.6 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.160.6.5 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.160.6.4 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.160.6.3 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.160.6.2 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.160.6.1 | 06-Apr-2015 |
skrll | Sync with HEAD
|
| 1.199.4.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.199.2.1 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.201.6.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.203.4.2 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.203.4.1 | 10-Jun-2019 |
christos | Sync with HEAD
|
| 1.203.2.1 | 18-Jan-2019 |
pgoyette | Synch with HEAD
|
| 1.204.6.1 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.204.4.1 | 17-Aug-2020 |
martin | Pull up following revision(s) (requested by riastradh in ticket #1050):
sys/ufs/lfs/lfs_subr.c: revision 1.101
sys/ufs/lfs/lfs_subr.c: revision 1.102
sys/ufs/lfs/lfs_inode.c: revision 1.158
sys/ufs/lfs/lfs_inode.h: revision 1.25
sys/ufs/lfs/lfs_balloc.c: revision 1.95
sys/ufs/lfs/lfs_pages.c: revision 1.21
sys/ufs/lfs/lfs_vnops.c: revision 1.330
sys/ufs/lfs/lfs_alloc.c: revision 1.140 (patch)
sys/ufs/lfs/lfs_alloc.c: revision 1.141 (patch)
lib/libp2k/p2k.c: revision 1.72
sys/ufs/lfs/lfs.h: revision 1.205
sys/ufs/lfs/lfs.h: revision 1.206
sys/ufs/lfs/lfs_segment.c: revision 1.284
sys/ufs/lfs/lfs.h: revision 1.207
sys/ufs/lfs/lfs_segment.c: revision 1.285
sys/ufs/lfs/lfs_debug.c: revision 1.55
sys/ufs/lfs/lfs_rename.c: revision 1.23
usr.sbin/dumplfs/dumplfs.c: revision 1.65
sys/ufs/lfs/lfs_vfsops.c: revision 1.371
sys/arch/i386/stand/efiboot/bootx64/Makefile: revision 1.3
sys/ufs/lfs/lfs_vfsops.c: revision 1.372
sys/ufs/lfs/lfs_vfsops.c: revision 1.373
sbin/fsck_lfs/pass1.c: revision 1.46
sys/ufs/lfs/lfs_vnops.c: revision 1.326
sys/ufs/lfs/lfs_vnops.c: revision 1.327
sys/ufs/lfs/lfs_vfsops.c: revision 1.375 (patch)
sys/ufs/lfs/lfs_vnops.c: revision 1.328
sys/ufs/lfs/lfs_subr.c: revision 1.98
sys/ufs/lfs/lfs_extern.h: revision 1.116
sys/ufs/lfs/lfs_vnops.c: revision 1.329
sys/ufs/lfs/lfs_subr.c: revision 1.99
sys/ufs/lfs/lfs_extern.h: revision 1.117
sys/ufs/lfs/lfs_accessors.h: revision 1.49
sys/ufs/lfs/lfs_extern.h: revision 1.118
sys/rump/fs/lib/liblfs/Makefile: revision 1.15
sys/ufs/lfs/lfs_bio.c: revision 1.146 (patch)
sys/ufs/lfs/lfs_bio.c: revision 1.147
sys/ufs/lfs/lfs_subr.c: revision 1.100
Fix kassert in lfs by initializing vp first.
Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
Teach lfs to transition ro<->rw.
Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
Assert lfs_writer where I think we can now prove it.
Serialize access to the splay tree with lfs_lock.
Change some cheap KDASSERT into KASSERT.
Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
Call lfs_orphan in lfs_rename while we're still in the dirop.
lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
Teach LFS_ORPHAN_NEXTFREE about lfs64.
Dust off the orphan detection code and try to make it work.
Fix !DIAGNOSTIC compile
Fix userland references to LFS_ORPHAN_NEXTFREE.
Forgot to grep for these or do a full distribution build, oops!
Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
OR into bp->b_cflags; don't overwrite.
CTASSERT lfs on-disk structure sizes.
Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
Fix clang build after packed lfs64 accessor change.
Suppress spurious address-of-packed error in rump lfs too.
|
| 1.53 | 20-Oct-2025 |
perseant | * Generalize the partial-segment parser introduced for roll-forward,
using it to facilitate an in-kernel segment rewriter (cleaner), and a
mechanism to check whether a segment is in fact empty (only used
with DEBUG).
* Add these new fcntl calls:
- LFCNFILESTATS: For each inode given, report its number of direct
blocks, how many gaps (discontinuities) there are between direct
blocks, and how large the total gap distance is. This will be
useful for a coalescing agent.
- LFCNREWRITEFILE: For each inode given, rewrite its direct blocks,
effectively coalescing it into as compact a form as possible.
- LFCNSCRAMBLE: As above, except that it only rewrites every other
block. This causes the file to have many gaps that can be
measured with LFCNFILESTATS and addressed with LFCNREWRITEFILE,
for testing purposes.
- LFCNREWRITESEGS: Rewrite any live data in the given segments.
This is intended to simplify the cleaner API and facilitate an
in-kernel cleaner.
- LFCNCLEANERINFO: Get the most current CLEANERINFO data from the
kernel.
- LFCNSEGUSE: Retrieve segment usage data from the kernel.
* Vnodes marked IN_CLEANING now take a reference. Add a new "cleaner
lock", which must be taken by the cleaner before the segment lock,
and before marking nodes IN_CLEANING. This allows us to flush
vnodes, if necessary, before the cleaning segment is written, and
never to flush vnodes being cleaned. When the cleaner lock is
released, the vnodes are cleared of IN_CLEANING and the reference
dropped.
* Track a potential infinite loop in lfs_gatherblock.
* Pull "needs to flush" and "needs to wait for flush" into functions
instead of inlining their definitions.
|
| 1.52 | 15-Sep-2025 |
perseant | If setting the head (or tail) of the inode free list to LFS_UNUSED_INUM, also
set the tail (resp. head) to LFS_UNUSED_INUM, as the list is now empty.
Add a check to ensure that lfs_valloc_fixed will always terminate, even
if the free list should contain a loop. Extend the ifile at the end if it
is empty, to match the assumption of lfs_valloc() that the free list is
never empty.
Needed for roll-forward.
|
| 1.51 | 24-Apr-2022 |
rillig | lfs: fix lint warning about empty declaration
|
| 1.50 | 07-Sep-2020 |
riastradh | Suppress -Waddress-of-packed-member just for lfs_accessors.h.
We can remove -Wno-error=address-of-packed-member from various
makefiles now.
|
| 1.49 | 21-Mar-2020 |
riastradh | Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
|
| 1.48 | 10-Jun-2017 |
maya | branches: 1.48.4; 1.48.8; 1.48.12;
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.47 | 12-Jan-2017 |
christos | branches: 1.47.8;
fix sign confusion
|
| 1.46 | 20-Jun-2016 |
dholland | branches: 1.46.2;
Massedit u_int{8,16,32,64}_t to uint{8,16,32,64}_t. This effectively
merges ufs/dinode.h 1.25.
|
| 1.45 | 19-Jun-2016 |
dholland | we are actually synced with ufs/dinode.h 1.24 and ufs/dir.h 1.25.
|
| 1.44 | 19-Feb-2016 |
riastradh | Explicitly cast between char and unsigned char here.
|
| 1.43 | 19-Feb-2016 |
riastradh | Various housekeeping.
- Include <ufs/lfs/lfs.h> for union lfs_dinode &c.
- Include <string.h> or <sys/systm.h> for memcpy.
- Avoid signedness mismatch in lfs dino accessor for `rdev'.
- Avoid shadowing global `index'.
|
| 1.42 | 10-Jan-2016 |
christos | there is no reason to use __unused here.
|
| 1.41 | 10-Jan-2016 |
dholland | Fix two functions that were accidentally "static __unused" instead of
"static __unused inline". Oops; but probably not actually harmful.
|
| 1.40 | 19-Oct-2015 |
dholland | fix stupid typo in the 64-bit branch of the d_namlen accessor
|
| 1.39 | 19-Oct-2015 |
dholland | improve some panic messages
|
| 1.38 | 15-Oct-2015 |
dholland | Remove stray #define of lfs_magic
(the last of the fake superblock structure field macros)
|
| 1.37 | 10-Oct-2015 |
dholland | Add byteswapping to the inode block-pointer accessors.
|
| 1.36 | 03-Oct-2015 |
dholland | Drop an explicit sign-extension in fsck that shouldn't be needed any
more.
|
| 1.35 | 03-Oct-2015 |
dholland | Use IINFO in lfs_writeinode().
(both the kernel and the userland copies)
|
| 1.34 | 03-Oct-2015 |
dholland | Add an IINFO struct, which is like the FINFO struct but for the inode
blocks portion of the segment summary.
A segment summary block begins with a header (SEGSUM); the rest of the
block contains FINFO structures describing file blocks growing upward
from the bottom (after the header), and IINFO structures describing
inode blocks grown downward from the end of the block. (When they meet
the segment is full regardless of how many blocks might be left.)
IINFO contains just a block number, and until now this information was
handled by just using uint32_t*; switching to a structure will make
the code a lot easier to read, and also make it easier to have 32-bit
and 64-bit versions without making a mess.
This commit just adds the structures and accessors; they'll be
deployed into the code in subsequent commits.
|
| 1.33 | 21-Sep-2015 |
dholland | branches: 1.33.2;
Fix some assorted 32-bit assumptions not yet otherwise handled.
Also apply patch to fix the overt problem in PR 50246: newfs was
calculating ifpb wrong for volumes with non-default block sizes.
|
| 1.32 | 21-Sep-2015 |
dholland | Oops, I forgot to make the atime in the 64-bit IFILE 64 bits.
Correct that. Incompatible change, but no LFS64 volumes can have been
created yet.
|
| 1.31 | 21-Sep-2015 |
dholland | Add 64-bit directory entry structures, and adjust accessors accordingly.
The LFS64 directory entry has a 64-bit inode number. This is stored as
two 32-bit values to avoid inducing 64-bit alignment requirements.
The exposed type for manipulating directory entries is now
LFS_DIRHEADER, following the same convention as e.g. IFILE and SEGUSE.
(But with LFS_ on it, because.)
|
| 1.30 | 21-Sep-2015 |
dholland | Oops; LFS_DIRECTSIZ() is going to need the fs as an argument.
Also, it turns out that dirhash needs a compile-time-constant version
of LFS_DIRECTSIZ(LFS_MAXNAMLEN+1), independent of 64-vs-32, so create
LFS_MAXDIRENTRYSIZE for this. Sigh.
|
| 1.29 | 20-Sep-2015 |
dholland | Clean up struct lfs_dirtemplate.
|
| 1.28 | 20-Sep-2015 |
dholland | Fix glaringly stupid overflow/sizing bug in -r1.25. The part I don't
get is how it passed testing...
|
| 1.27 | 15-Sep-2015 |
dholland | Pass around struct lfs_dirheader instead of struct lfs_direct.
|
| 1.26 | 15-Sep-2015 |
dholland | Add an accessor function for directory names.
|
| 1.25 | 15-Sep-2015 |
dholland | Add a function lfs_copydirname() to copy directory names in place; use
it in place of (variously) memcpy and strlcpy. (The latter isn't even
correct; was probably changed blindly from strncpy at some point.)
The new function zeroes the padding in the directory entry instead of
leaving trash behind.
|
| 1.24 | 15-Sep-2015 |
dholland | Move the header part of struct lfs_direct to its own structure.
(lfs_dirheader)
Take the opportunity to improve the directory generation code in
make_lfs.c. (Everything else was unaffected by virtue of using
accessor functions.)
|
| 1.23 | 15-Sep-2015 |
dholland | Add and use accessor functions for more of the directory entry fields.
|
| 1.22 | 01-Sep-2015 |
dholland | Add new accessors for the d_type and d_namlen fields of struct lfs_direct.
Napalm the old byteswap access logic for these.
|
| 1.21 | 01-Sep-2015 |
dholland | Fix up indirect block handling in truncate to be 32/64 clean.
|
| 1.20 | 01-Sep-2015 |
dholland | Tidy the MAXSYMLINKLEN macros.
|
| 1.19 | 01-Sep-2015 |
dholland | The ifile's inode number is constant. (it is always 1)
Therefore, storing the value in the superblock and reading it out
again is silly and offers the opportunity for it to become corrupted.
So, don't do that (most of the code already didn't) and use the
existing constant instead. Initialize new 32-bit superblocks with
the value for the sake of old userland programs, but don't keep the
value in the 64-bit superblock at all.
(approved by Margo Seltzer)
|
| 1.18 | 01-Sep-2015 |
dholland | Make the inode fields in the 64-bit superblock 64 bits wide.
Reasoning as before.
Note that I am not going through and checking for 64->32 truncations
in inode numbers; I'm sure there are quite a few, but that's a project
for later.
|
| 1.17 | 01-Sep-2015 |
dholland | Add byteswapping to the dinode accessors.
This prevents regressions in the ulfs code when switching to the new
accessors. Note that while adding byteswapping to the other accessors
is straightforward, I haven't done it yet; and that also is not enough
to make LFS_EI work, because there are places lying around that bypass
the accessors for one reason and another and all of them need to be
updated. That is going to have to wait for a later day as LFS_EI is
not on the critical path right now.
|
| 1.16 | 01-Sep-2015 |
dholland | Use the lfs dinode accessors in place of the ufs-derived ones.
(Mostly.)
The ufs-derived ones are fake structure member macros, which are gross
and not very safe. Also, it seems that a lot of places in the lfs code
were using the ffsv1 branch of them unconditionally, and this way it's
guaranteed all those places have been updated.
Found while doing this: for non-devices, have getattr produce NODEV
in the rdev field instead of leaking the address of the first direct
block.
|
| 1.15 | 29-Aug-2015 |
mlelstv | Fix IFILE pointer calculation when scanning freelist.
|
| 1.14 | 19-Aug-2015 |
dholland | Part two of dinodes; use the same union everywhere.
(previously the ufs-derived code had things set up slightly different)
Remove a bunch of associated mess.
|
| 1.13 | 12-Aug-2015 |
dholland | Hack up dinode usage to be 64 vs. 32 as needed. Part 1.
(This part changes the native lfs code; the ufs-derived code already
has 64 vs. 32 logic, but as aspects of it are unsafe, and don't
entirely interoperate cleanly with the lfs 64/32 stuff, pass 2 will be
rehashing that.)
|
| 1.12 | 12-Aug-2015 |
dholland | Provide 32-bit and 64-bit versions of FINFO.
This also entailed sorting out part of struct segment, as that
contains a pointer into the current FINFO data.
|
| 1.11 | 12-Aug-2015 |
dholland | Make 32-bit and 64-bit versions of SEGSUM.
Also fix some of the FINFO handling as it's closely entangled.
|
| 1.10 | 12-Aug-2015 |
dholland | Add IFILE32 and IFILE64 structures for the on-disk ifile entries.
Add and use accessors. There are also a bunch of places that cast and
I hope I've found them all...
|
| 1.9 | 12-Aug-2015 |
dholland | Make 32-bit and 64-bit versions of CLEANERINFO.
XXX: while this is written to disk, it seems like much of it would
XXX: be better set up as a commpage shared with the cleaner.
|
| 1.8 | 02-Aug-2015 |
dholland | Pass the fs object to LFS_MAX_DADDR so it can check lfs_is64.
Remove some hackish intentional 64->32 truncations next to the checks
using LFS_MAX_DADDR, and tackle the problem they handled in bmap
instead.
The problem: the magic block pointer value UNWRITTEN has magic value
-2, and if it's not handled specifically, uint32 -> uint64 promotion
turns it into 4294967294, which then causes consternation and
monkeyhouse downstream.
What's here is still kind of a hack, but it's a step forward.
|
| 1.7 | 02-Aug-2015 |
dholland | Add a (draft) 64-bit superblock. Make things build again.
Add pieces of support for using both superblock types where
convenient, and specifically to the superblock accessors, but don't
actually enable it anywhere.
First substantive step on PR 50000.
|
| 1.6 | 02-Aug-2015 |
dholland | Use accessor functions for the version field of the lfs superblock.
I thought at first maybe the cases that test the version should be
rolled into the accessors, but on the whole I think the conclusion on
that is no.
|
| 1.5 | 02-Aug-2015 |
dholland | Second batch of 64 -> 32 truncations in lfs, along with more minor
tidyups and corrections in passing.
|
| 1.4 | 02-Aug-2015 |
dholland | Fix assorted 64 -> 32 truncations in lfs. Also, some minor tidyups and
corrections in passing.
|
| 1.3 | 02-Aug-2015 |
dholland | Allow superblock accessors that widen 32-bit disk fields to 64-bit
memory values.
|
| 1.2 | 28-Jul-2015 |
dholland | Use lfs_accessors.h in conjunction with the cleaner's struct clfs.
Remove previous hacks.
|
| 1.1 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.33.2.7 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.33.2.6 | 05-Feb-2017 |
skrll | Sync with HEAD
|
| 1.33.2.5 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.33.2.4 | 19-Mar-2016 |
skrll | Sync with HEAD
|
| 1.33.2.3 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.33.2.2 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.33.2.1 | 21-Sep-2015 |
skrll | file lfs_accessors.h was added on branch nick-nhusb on 2015-09-22 12:06:17 +0000
|
| 1.46.2.1 | 20-Mar-2017 |
pgoyette | Sync with HEAD
|
| 1.47.8.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.48.12.1 | 17-Aug-2020 |
martin | Pull up following revision(s) (requested by riastradh in ticket #1050):
sys/ufs/lfs/lfs_subr.c: revision 1.101
sys/ufs/lfs/lfs_subr.c: revision 1.102
sys/ufs/lfs/lfs_inode.c: revision 1.158
sys/ufs/lfs/lfs_inode.h: revision 1.25
sys/ufs/lfs/lfs_balloc.c: revision 1.95
sys/ufs/lfs/lfs_pages.c: revision 1.21
sys/ufs/lfs/lfs_vnops.c: revision 1.330
sys/ufs/lfs/lfs_alloc.c: revision 1.140 (patch)
sys/ufs/lfs/lfs_alloc.c: revision 1.141 (patch)
lib/libp2k/p2k.c: revision 1.72
sys/ufs/lfs/lfs.h: revision 1.205
sys/ufs/lfs/lfs.h: revision 1.206
sys/ufs/lfs/lfs_segment.c: revision 1.284
sys/ufs/lfs/lfs.h: revision 1.207
sys/ufs/lfs/lfs_segment.c: revision 1.285
sys/ufs/lfs/lfs_debug.c: revision 1.55
sys/ufs/lfs/lfs_rename.c: revision 1.23
usr.sbin/dumplfs/dumplfs.c: revision 1.65
sys/ufs/lfs/lfs_vfsops.c: revision 1.371
sys/arch/i386/stand/efiboot/bootx64/Makefile: revision 1.3
sys/ufs/lfs/lfs_vfsops.c: revision 1.372
sys/ufs/lfs/lfs_vfsops.c: revision 1.373
sbin/fsck_lfs/pass1.c: revision 1.46
sys/ufs/lfs/lfs_vnops.c: revision 1.326
sys/ufs/lfs/lfs_vnops.c: revision 1.327
sys/ufs/lfs/lfs_vfsops.c: revision 1.375 (patch)
sys/ufs/lfs/lfs_vnops.c: revision 1.328
sys/ufs/lfs/lfs_subr.c: revision 1.98
sys/ufs/lfs/lfs_extern.h: revision 1.116
sys/ufs/lfs/lfs_vnops.c: revision 1.329
sys/ufs/lfs/lfs_subr.c: revision 1.99
sys/ufs/lfs/lfs_extern.h: revision 1.117
sys/ufs/lfs/lfs_accessors.h: revision 1.49
sys/ufs/lfs/lfs_extern.h: revision 1.118
sys/rump/fs/lib/liblfs/Makefile: revision 1.15
sys/ufs/lfs/lfs_bio.c: revision 1.146 (patch)
sys/ufs/lfs/lfs_bio.c: revision 1.147
sys/ufs/lfs/lfs_subr.c: revision 1.100
Fix kassert in lfs by initializing vp first.
Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
Teach lfs to transition ro<->rw.
Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
Assert lfs_writer where I think we can now prove it.
Serialize access to the splay tree with lfs_lock.
Change some cheap KDASSERT into KASSERT.
Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
Call lfs_orphan in lfs_rename while we're still in the dirop.
lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
Teach LFS_ORPHAN_NEXTFREE about lfs64.
Dust off the orphan detection code and try to make it work.
Fix !DIAGNOSTIC compile
Fix userland references to LFS_ORPHAN_NEXTFREE.
Forgot to grep for these or do a full distribution build, oops!
Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
OR into bp->b_cflags; don't overwrite.
CTASSERT lfs on-disk structure sizes.
Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
Fix clang build after packed lfs64 accessor change.
Suppress spurious address-of-packed error in rump lfs too.
|
| 1.48.8.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.48.4.2 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.48.4.1 | 10-Jun-2017 |
jdolecek | file lfs_accessors.h was added on branch tls-maxphys on 2017-12-03 11:39:22 +0000
|
| 1.145 | 21-Sep-2025 |
christos | lfs_freelist_prev is unused
|
| 1.144 | 17-Sep-2025 |
perseant | Add routines to check freelist consistency if compiled with DEBUG and
conditional on a kernel variable manipulated via sysctl.
Add checks before and after each routine that modifies the free list.
#if 0 a section of lfs_vfree() that was intended to keep the free list ordered
but instead corrupted it.
|
| 1.143 | 15-Sep-2025 |
perseant | Initialize nextfree, to placate gcc.
|
| 1.142 | 15-Sep-2025 |
perseant | If setting the head (or tail) of the inode free list to LFS_UNUSED_INUM, also
set the tail (resp. head) to LFS_UNUSED_INUM, as the list is now empty.
Add a check to ensure that lfs_valloc_fixed will always terminate, even
if the free list should contain a loop. Extend the ifile at the end if it
is empty, to match the assumption of lfs_valloc() that the free list is
never empty.
Needed for roll-forward.
|
| 1.141 | 23-Feb-2020 |
riastradh | Dust off the orphan detection code and try to make it work.
|
| 1.140 | 23-Feb-2020 |
riastradh | Teach LFS_ORPHAN_NEXTFREE about lfs64.
|
| 1.139 | 22-Feb-2020 |
kamil | Avoid undefined behavior in *_BITMAP_FREE() macros
left shift of 1 by 31 places cannot be represented in type 'int'
|
| 1.138 | 17-Jan-2020 |
ad | VFS_VGET(), VFS_ROOT(), VFS_FHTOVP(): give them a "int lktype" argument, to
allow us to get shared locks (or no lock) on the returned vnode. Matches
FreeBSD.
|
| 1.137 | 19-Aug-2017 |
maya | branches: 1.137.4; 1.137.8; 1.137.10;
Consistently use {,UN}MARK_VNODE macros rather than function calls.
|
| 1.136 | 10-Jun-2017 |
maya | Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.135 | 13-Mar-2017 |
maya | branches: 1.135.6;
Fill in some XXXs with the exact action described in them. match
lfs_valloc behaviour.
|
| 1.134 | 13-Mar-2017 |
riastradh | #if DIAGNOSTIC panic ---> KASSERT
Replace some #if DEBUG by this too. DEBUG is only for expensive
assertions; these are not.
|
| 1.133 | 07-Aug-2016 |
dholland | branches: 1.133.2;
Remove unused <sys/tree.h>
|
| 1.132 | 07-Aug-2016 |
dholland | Comments
|
| 1.131 | 10-Oct-2015 |
dholland | branches: 1.131.2;
Fix minor bitrot in #if 0 or otherwise disabled code.
|
| 1.130 | 13-Sep-2015 |
dholland | Fix wrong code in lfs_valloc_fixed(). It was overwriting the inode
number it was supposed to be allocating with the head of the inode
freelist, then applying the wrong test to that result. Net result:
unless the freelist was empty (in which case it would always fail),
it would in general drop a bunch of entries from the freelist.
This code seems to have been broken when the first version of lfsv2
was imported onto the perseant-lfsv2 branch in -r1.47.2.1, and
remained broken since, in spite of having been moved to lfs_rfw.c and
back and rearranged quite a bit in the meantime.
Sigh.
Found by Coverity in a rather confusing way as CID 1316545.
|
| 1.129 | 01-Sep-2015 |
dholland | Use the lfs dinode accessors in place of the ufs-derived ones.
(Mostly.)
The ufs-derived ones are fake structure member macros, which are gross
and not very safe. Also, it seems that a lot of places in the lfs code
were using the ffsv1 branch of them unconditionally, and this way it's
guaranteed all those places have been updated.
Found while doing this: for non-devices, have getattr produce NODEV
in the rdev field instead of leaking the address of the first direct
block.
|
| 1.128 | 29-Aug-2015 |
mlelstv | Fix IFILE pointer calculation when scanning freelist.
|
| 1.127 | 12-Aug-2015 |
dholland | Hack up dinode usage to be 64 vs. 32 as needed. Part 1.
(This part changes the native lfs code; the ufs-derived code already
has 64 vs. 32 logic, but as aspects of it are unsafe, and don't
entirely interoperate cleanly with the lfs 64/32 stuff, pass 2 will be
rehashing that.)
|
| 1.126 | 12-Aug-2015 |
dholland | Add IFILE32 and IFILE64 structures for the on-disk ifile entries.
Add and use accessors. There are also a bunch of places that cast and
I hope I've found them all...
|
| 1.125 | 02-Aug-2015 |
dholland | Use accessor functions for the version field of the lfs superblock.
I thought at first maybe the cases that test the version should be
rolled into the accessors, but on the whole I think the conclusion on
that is no.
|
| 1.124 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.123 | 24-Jul-2015 |
dholland | More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)
The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.
|
| 1.122 | 24-Jul-2015 |
dholland | Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)
It also gets rid of cpp abuse in the form of fake structure member
macros.
Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)
XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.
|
| 1.121 | 16-Jul-2015 |
dholland | Don't cast the return value of malloc.
|
| 1.120 | 31-May-2015 |
hannken | Change lfs from hash table to vcache.
- Change lfs_valloc() to return an inode number and version instead of
a vnode and move lfs_ialloc() and lfs_vcreate() to new lfs_init_vnode().
- Add lfs_valloc_fixed() to allocate a known inode, used by kernel
roll forward.
- Remove lfs_*ref(), these functions cannot coexist with vcache and
their commented behaviour is far away from their implementation.
- Add the cleaner lwp and blockinfo to struct ulfsmount so lfs_loadvnode()
may use hints from the cleaner.
- Remove vnode locks from ulfs_lookup() like we did with ufs_lookup().
|
| 1.119 | 28-Jul-2013 |
dholland | branches: 1.119.6;
Add more of the bits for supporting quotas.
|
| 1.118 | 28-Jul-2013 |
dholland | Add lfs_kernel.h for declarations that don't need to be exposed to userland.
lfs currently has the following headers:
lfs.h - on-disk structures and stuff needed for userlevel tools
lfs_inode.h - additional restricted materials for userlevel tools
that operate the fs (newfs_lfs, fsck_lfs, lfs_cleanerd)
lfs_kernel.h - stuff needed only in the kernel
and the following legacy headers that are expected to be mopped up and
folded into one of the above:
lfs_extern.h - function prototypes
ulfs_bswap.h - endian-independent support
ulfs_dinode.h - now contains very little
ulfs_dirhash.h - dirhash support
ulfs_extattr.h - extattr support
ulfs_extern.h - more function prototypes
ulfs_inode.h - assorted kernel-only declarations
ulfs_quota.h - quota support
ulfs_quota1.h - more quota support
ulfs_quota2.h - more quota support
ulfs_quotacommon.h - more quota support
ulfsmount.h - legacy copy of ufsmount material
|
| 1.117 | 18-Jun-2013 |
christos | branches: 1.117.2;
Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.
|
| 1.116 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.115 | 06-Jun-2013 |
dholland | Split lfs from ufs step 3: rearrange config stuff.
Add new options:
LFS_EI
LFS_DIRHASH
LFS_EXTATTR
LFS_EXTATTR_AUTOSTART
LFS_QUOTA
LFS_QUOTA2
and update code referring to the corresponding FFS and UFS config
symbols to use the LFS versions. Disable the one extant reference
to APPLE_UFS in the ulfs files. Use opt_lfs.h only, not opt_ffs.h.
|
| 1.114 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.113 | 22-Jan-2013 |
dholland | Stuff UFS_ in front of a few of ufs's symbols to reduce namespace
pollution. Specifically:
ROOTINO -> UFS_ROOTINO
WINO -> UFS_WINO
NXADDR -> UFS_NXADDR
NDADDR -> UFS_NDADDR
NIADDR -> UFS_NIADDR
MAXSYMLINKLEN -> UFS_MAXSYMLINKLEN
MAXSYMLINKLEN_UFS[12] -> UFS[12]_MAXSYMLINKLEN (for consistency)
Sort out ext2fs's misuse of NDADDR and NIADDR; fortunately, these have
the same values in ext2fs and ffs.
No functional change intended.
|
| 1.112 | 16-Feb-2012 |
perseant | branches: 1.112.2;
Pass t_renamerace and t_rmdirrace tests.
Adapt dholland@'s fix to ufs_rename to fix PR kern/43582. Address several
other MP locking issues discovered during the course of investigating the
same problem.
Removed extraneous vn_lock() calls on the Ifile, since the Ifile writes
are controlled by the segment lock.
Fix PR kern/45982 by deemphasizing the estimate of how much metadata
will fill the empty space on disk when the disk is nearly empty
(t_renamerace crates a lot of inode blocks on a tiny empty disk).
|
| 1.111 | 12-Jun-2011 |
rmind | branches: 1.111.2; 1.111.6; 1.111.8;
Welcome to 5.99.53! Merge rmind-uvmplock branch:
- Reorganize locking in UVM and provide extra serialisation for pmap(9).
New lock order: [vmpage-owner-lock] -> pmap-lock.
- Simplify locking in some pmap(9) modules by removing P->V locking.
- Use lock object on vmobjlock (and thus vnode_t::v_interlock) to share
the locks amongst UVM objects where necessary (tmpfs, layerfs, unionfs).
- Rewrite and optimise x86 TLB shootdown code, make it simpler and cleaner.
Add TLBSTATS option for x86 to collect statistics about TLB shootdowns.
- Unify /dev/mem et al in MI code and provide required locking (removes
kernel-lock on some ports). Also, avoid cache-aliasing issues.
Thanks to Andrew Doran and Joerg Sonnenberger, as their initial patches
formed the core changes of this branch.
|
| 1.110 | 24-Jun-2010 |
hannken | branches: 1.110.6;
Clean up vnode lock operations pass 2:
VOP_UNLOCK(vp, flags) -> VOP_UNLOCK(vp): Remove the unneeded flags argument.
Welcome to 5.99.32.
Discussed on tech-kern.
|
| 1.109 | 08-Jan-2010 |
pooka | branches: 1.109.2; 1.109.4;
The VATTR_NULL/VREF/VHOLD/HOLDRELE() macros lost their will to live
years ago when the kernel was modified to not alter ABI based on
DIAGNOSTIC, and now just call the respective function interfaces
(in lowercase). Plenty of mix'n match upper/lowercase has creeped
into the tree since then. Nuke the macros and convert all callsites
to lowercase.
no functional change
|
| 1.108 | 13-Sep-2009 |
tsutsui | Move declaration of ufs_hashlock into <ufs/ufs_extern.h> from each c source.
|
| 1.107 | 28-Apr-2008 |
martin | Remove clause 3 and 4 from TNF licenses
|
| 1.106 | 30-Jan-2008 |
ad | branches: 1.106.6; 1.106.8; 1.106.10;
Replace struct lock on vnodes with a simpler lock object built on
krwlock_t. This is a step towards removing lockmgr and simplifying
vnode locking. Discussed on tech-kern.
|
| 1.105 | 02-Jan-2008 |
ad | Merge vmlocking2 to head.
|
| 1.104 | 12-Dec-2007 |
he | Fix a use of lfs_truncate() inside an #ifdef notyet (so no resulting change);
lfs_truncate() has lost its lwp argument.
|
| 1.103 | 10-Oct-2007 |
ad | branches: 1.103.4; 1.103.6; 1.103.8; 1.103.10;
Merge from vmlocking:
- Split vnode::v_flag into three fields, depending on field locking.
- simple_lock -> kmutex in a few places.
- Fix some simple locking problems.
|
| 1.102 | 08-Oct-2007 |
ad | Merge ffs locking & brelse changes from the vmlocking branch.
|
| 1.101 | 10-Jul-2007 |
hannken | branches: 1.101.6; 1.101.8; 1.101.10;
Move `struct dquot' and its supporting functions from quota.h to ufs_quota.c.
- Make quota-internal functions static.
- Clean up declarations in quota.h and ufs_extern.h. quota.h now has the
description of quota criterions, on-disk structure, user-kernel interface and
declaration of init/done functions. All ufs quota related function
prototypes go to ufs_extern.h.
- New functions ufsquota_init() and ufsquota_free() create or destroy the
quota fields of `struct inode'.
- chkdq() and chkiq() always update the quota fields of `struct inode' first.
- Only ufs_access() explicitely calls getinoquota().
No objections on tech-kern@
|
| 1.100 | 15-Feb-2007 |
ad | branches: 1.100.6; 1.100.8;
Replace some uses of lockmgr() / simplelocks.
|
| 1.99 | 16-Nov-2006 |
christos | __unused removal on arguments; approved by core.
|
| 1.98 | 12-Oct-2006 |
christos | - sprinkle __unused on function decls.
- fix a couple of unused bugs
- no more -Wno-unused for i386
|
| 1.97 | 01-Sep-2006 |
perseant | branches: 1.97.2; 1.97.4;
Changes to help the roll-forward agent, to wit:
* Mark being-deleted files in the Ifile so we can finish deleting them
at fs mount time.
* Flag the Ifile with "cleaner must clean" when writers are waiting for
the cleaner, rather than relying solely on the cleaner's estimation of
whether it should clean or not.
* Note partial segments written by a user agent (in particular,
fsck_lfs) so that repeated rolls forward don't interfere with one
another.
* Add a new fcntl, LFCNPASS, that allows the log to wrap exactly once,
for better testing of the validity of checkpoints.
* Keep track of the on-disk nlink count when cleaning, so that we don't
partially complete directory operations while cleaning.
* Ensure that every single Ifile inode write represents a consistent
view of the filesystem. In particular, the accounting for the segment
we are writing the inode into must be correct, and the accounting for
the segment that inode used to reside in must be correct. Rather than
just rewriting the inode if we wrote it wrong, rewrite the necessary
ifile blocks before writing the inode so we never write it wrong.
* Don't unmark any VDIROP vnodes if we haven't written them to disk,
avoiding yet another problem with the "wait for the cleaner" error
return from lfs_putpages().
Also, move the last callback to an aiodone call, so we no longer do any
memory management from interrupt context.
|
| 1.96 | 20-Jul-2006 |
perseant | Separate the (non-working) LFS kernel roll-forward code into its own file,
lfs_rfw.c.
|
| 1.95 | 06-Jul-2006 |
perseant | Protect lfs_order_freelist() with the segment lock.
|
| 1.94 | 14-May-2006 |
elad | branches: 1.94.4;
integrate kauth.
|
| 1.93 | 12-May-2006 |
perseant | Fixes to address the "vinvalbuf: dirty blocks" panic that can occur when
many inodes are cleaned at once. Make sure that we write all the pages
on vnodes that are being flushed, even if we don't think there's room;
drain v_numoutput before lfs_vflush() completes.
Also, don't allow a vnode that is in the process of being cleaned to be
chosen by getnewvnode(); this avoids a segment accounting panic in the case
that a large number of inodes are fed to lfs_markv() all at once.
|
| 1.92 | 04-May-2006 |
perseant | Introduce another per-filesystem parameter, lfs_resvseg, to separate the
notion of "how many segments are reserved for the cleaner" from that of
"how many segments are not counted in lfs_bfree". The default value
used for existing filesystems is the same as the previous implicit value
of (lfs_minfreeseg / 2 + 1), modulo some sanity checking.
Count pending dirops on a per-filesystem basis, since once we start
writing them we can't stop until we're done. This seems to help stave off
the "no clean segments" panic in the case of filling the filesystem with
directories and small files (e.g. simultaneously unpacking more copies of
pkgsrc than will fit).
|
| 1.91 | 30-Apr-2006 |
perseant | Add an explicit list initialization that was missing from my last commit.
|
| 1.90 | 30-Apr-2006 |
perseant | Postpone the segment accounting changes coming from truncation until the
inode that makes those changes valid is either written to disk by
lfs_writeinode() or discarded by lfs_vfree().
A couple of locking fixes are also included as well.
|
| 1.89 | 22-Apr-2006 |
perseant | Fix a fencepost error in the bitmap handling in extend_ifile(), and another
in lfs_freelist_prev().
|
| 1.88 | 10-Apr-2006 |
perseant | Optimize the free list search a little more; in particular use words
instead of bytes for the index, and never search below fs->lfs_freehd.
Fix a bug in the previous version of the search (an erroneous assumption
that ino_t was signed).
Free the bitmap when we unmount the filesystem.
|
| 1.87 | 08-Apr-2006 |
perseant | Keep the free list ordered. This solves a problem first pointed out to me
by Michel Oey, in which an aged LFS writes up to an extra Ifile block for
every file created; and paves the way for the truncation of the Ifile when
many files are deleted.
|
| 1.86 | 11-Dec-2005 |
christos | branches: 1.86.4; 1.86.6; 1.86.8; 1.86.10; 1.86.12;
merge ktrace-lwp.
|
| 1.85 | 02-Nov-2005 |
yamt | merge yamt-vop branch. remove following VOPs.
VOP_BLKATOFF
VOP_VALLOC
VOP_BALLOC
VOP_REALLOCBLKS
VOP_VFREE
VOP_TRUNCATE
VOP_UPDATE
|
| 1.84 | 19-Aug-2005 |
christos | branches: 1.84.2;
64 bit inode changes.
|
| 1.83 | 29-May-2005 |
christos | branches: 1.83.2;
- sprinkle const
- avoid shadow variables.
|
| 1.82 | 19-Apr-2005 |
perseant | Keep per-inode, per-fs, and subsystem-wide counts of blocks allocated through
lfs_balloc(), and use that to estimate the number of dirty pages belonging
to LFS (subsystem or filesystem). This is almost certainly wrong for
the case of a large mmap()ed region, but the accounting is tighter than
what we had before, and performs much better in the typical case of pages
dirtied through write().
|
| 1.81 | 16-Apr-2005 |
perseant | Use splay trees, rather than a hash table, to manage the accounting of
blocks allocated through VOP_BALLOC() for pages to be written to disk.
This accounting no longer takes a noticeable fraction of the system CPU.
|
| 1.80 | 14-Apr-2005 |
perseant | Consolidate the hash table we use to maintain the integrity of lfs_avail
into a single, system-wide table, rather than having a separate hash table
per inode. Significantly reduces the "system" cpu usage of your average
file write.
|
| 1.79 | 14-Apr-2005 |
perseant | Keep track of the highest block held by an LFS inode, so that we can
be assured that the last byte of a file is always allocated. Previously
a file extension could cause the filesystem to be flushed, writing an
inconsistent inode to disk. Although this condition would be corrected
the next time blocks were written to disk, an intervening crash would leave
the filesystem in an inconsistent state, leaving fsck_lfs to complain
of an inode "partially truncated".
|
| 1.78 | 01-Apr-2005 |
perseant | Protect various per-fs structures with fs->lfs_interlock simple_lock, to
improve behavior in the multiprocessor case. Add debugging segment-lock
assertion statements.
|
| 1.77 | 23-Mar-2005 |
perseant | Make LFS dirops get their vnode first, before incrementing the dirop count,
to prevent a deadlock trying to call VOP_PUTPAGES() on a VDIROP vnode.
This can happen when a stacked filesystem is mounted on top of an LFS: an
LFS dirop needs to get a vnode, which is available from the upper layer.
The corresponding lower layer vnode, however, is VDIROP, so the upper layer
can't be cleaned out since its VOP_PUTPAGES() is passed through to the lower
layer, which waits for dirops to drain before it can proceed. Deadlock.
Tweak ufs_makeinode() and ufs_mkdir() to pass the a_vpp argument through
to VOP_VALLOC().
Partially addresses PR # 26043, though it probably does not completely fix
the problem described there.
|
| 1.76 | 08-Mar-2005 |
perseant | branches: 1.76.2;
Straighten out the maze of ifdefs. Instead, consolidate all the debugging
stuff under '#ifdef DEBUG', and use sysctl knobs to turn on/off particular
parts of the debugging reporting (if DEBUG is enabled). Re-enable the LFS
statistics in sysctl, while I'm there. A bit of a rototill.
|
| 1.75 | 26-Feb-2005 |
perry | nuke trailing whitespace
|
| 1.74 | 26-Feb-2005 |
perseant | Various minor LFS improvements:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statvfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
|
| 1.73 | 14-Aug-2004 |
mycroft | branches: 1.73.4; 1.73.6;
Add a new flag, IN_MODIFY. This is like IN_UPDATE|IN_CHANGE, but unlike
setting those flags, it does not cause the inode to be written in the periodic
sync. This is used for writes to special files (devices and named pipes) and
FIFOs.
Do not preemptively sync updates to access times and modification times. They
are now updated in the inode only opportunistically, or when the file or device
is closed. (Really, it should be delayed beyond close, but this is enough to
help substantially with device nodes.)
And the most amusing part:
Trickle sync was broken on both FFS and ext2fs, in different ways. In FFS, the
periodic call to VFS_SYNC(MNT_LAZY) was still causing all file data to be
synced. In ext2fs, it was causing the metadata to *not* be synced. We now
only call VOP_UPDATE() on the node if we're doing MNT_LAZY. I've confirmed
that we do in fact trickle correctly now.
|
| 1.72 | 23-Sep-2003 |
yamt | branches: 1.72.4;
cleanup IN_ADIROP/VDIROP handling a little.
|
| 1.71 | 07-Aug-2003 |
agc | Move UCB-licensed code from 4-clause to 3-clause licence.
Patches provided by Joel Baker in PR 22364, verified by myself.
|
| 1.70 | 12-Jul-2003 |
yamt | - protect global resource counts with lfs_subsys_lock.
- clean up scattered externs a little.
|
| 1.69 | 29-Jun-2003 |
fvdl | branches: 1.69.2;
Back out the lwp/ktrace changes. They contained a lot of colateral damage,
and need to be examined and discussed more.
|
| 1.68 | 29-Jun-2003 |
thorpej | Undo part of the ktrace/lwp changes. In particular:
* Remove the "lwp *" argument that was added to vget(). Turns out
that nothing actually used it!
* Remove the "lwp *" arguments that were added to VFS_ROOT(), VFS_VGET(),
and VFS_FHTOVP(); all they did was pass it to vget() (which, as noted
above, didn't use it).
* Remove all of the "lwp *" arguments to internal functions that were added
just to appease the above.
|
| 1.67 | 28-Jun-2003 |
darrenr | Pass lwp pointers throughtout the kernel, as required, so that the lwpid can
be inserted into ktrace records. The general change has been to replace
"struct proc *" with "struct lwp *" in various function prototypes, pass
the lwp through and use l_proc to get the process pointer when needed.
Bump the kernel rev up to 1.6V
|
| 1.66 | 02-Apr-2003 |
fvdl | Add support for UFS2. UFS2 is an enhanced FFS, adding support for
64 bit block pointers, extended attribute storage, and a few
other things.
This commit does not yet include the code to manipulate the extended
storage (for e.g. ACLs), this will be done later.
Originally written by Kirk McKusick and Network Associates Laboratories for
FreeBSD.
|
| 1.65 | 15-Mar-2003 |
perseant | Add simple_lock protection for lfs_seglock and lfs_subsys_pages; these will
be expanded to cover other per-fs and subsystem-wide data as well.
Fix a case of IN_MODIFIED being set without updating lfs_uinodes, resulting
in a "lfs_uinodes < 0" panic.
Fix a deadlock in lfs_putpages arising from the need to busy all pages in a
block; unbusy any that had already been busied before starting over.
|
| 1.64 | 20-Feb-2003 |
perseant | Tabify, and fix some comment alignment problems.
|
| 1.63 | 17-Feb-2003 |
perseant | Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now
(there are still some details to work out) but expect that to go
away soon. To support these basic changes (creation of lfs_putpages,
lfs_gop_write, mods to lfs_balloc) several other changes were made, to
wit:
* Create a writer daemon kernel thread whose purpose is to handle page
writes for the pagedaemon, but which also takes over some of the
functions of lfs_check(). This thread is started the first time an
LFS is mounted.
* Add a "flags" parameter to GOP_SIZE. Current values are
GOP_SIZE_READ, meaning that the call should return the size of the
in-core version of the file, and GOP_SIZE_WRITE, meaning that it
should return the on-disk size. One of GOP_SIZE_READ or
GOP_SIZE_WRITE must be specified.
* Instead of using malloc(...M_WAITOK) for everything, reserve enough
resources to get by and use malloc(...M_NOWAIT), using the reserves if
necessary. Use the pool subsystem for structures small enough that
this is feasible. This also obsoletes LFS_THROTTLE.
And a few that are not strictly necessary:
* Moves the LFS inode extensions off onto a separately allocated
structure; getting closer to LFS as an LKM. "Welcome to 1.6O."
* Unified GOP_ALLOC between FFS and LFS.
* Update LFS copyright headers to correct values.
* Actually cast to unsigned in lfs_shellsort, like the comment says.
* Keep track of which segments were empty before the previous
checkpoint; any segments that pass two checkpoints both dirty and
empty can be summarily cleaned. Do this. Right now lfs_segclean
still works, but this should be turned into an effectless
compatibility syscall.
|
| 1.62 | 27-Jan-2003 |
yamt | make these compilable with lfs debug options.
(follow daddr_t change)
XXX maybe segment number should be 64bit.
|
| 1.61 | 25-Jan-2003 |
tron | Use PRId64 instead of hard coding "%lld" to fix build problems under
LP64 ports.
|
| 1.60 | 25-Jan-2003 |
tron | Fix printf() format strings problems caused by "daddr_t" change.
|
| 1.59 | 24-Jan-2003 |
fvdl | Bump daddr_t to 64 bits. Replace it with int32_t in all places where
it was used on-disk, so that on-disk formats remain the same.
Remove ufs_daddr_t and ufs_lbn_t for the time being.
|
| 1.58 | 08-Jan-2003 |
yamt | use lfs_unmark_vnode instead of duplicated code fragments.
|
| 1.57 | 24-Nov-2002 |
yamt | make sure i_lfs_fragsize is initialized.
fix panic "lfs_writefile: more than one fragment!"
PR 18974.
|
| 1.56 | 14-May-2002 |
perseant | Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes
I found while making sure there weren't any new ones.
* Make the write clusters keep track of the buffers whose blocks they contain.
This should make it possible to (1) write clusters using a page mapping
instead of malloc, if desired, and (2) schedule blocks for rewriting
(somewhere else) if a write error occurs. Code is present to use
pagemove() to construct the clusters but that is untested and will go away
anyway in favor of page mapping.
* DEBUG now keeps a log of Ifile writes, so that any lingering instances of
the "dirty bufs" problem can be properly debugged.
* Keep track of whether the Ifile has been dirtied by various routines that
can be called by lfs_segwrite, and loop on that until it is clean, for
a checkpoint. Checkpoints need to be squeaky clean.
* Warn the user (once) if the Ifile grows larger than is reasonable for their
buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can
grow.
* If an inode is not found in a disk block, try rereading the block, under
the assumption that the block was copied to a cluster and then freed.
* Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
|
| 1.55 | 04-Feb-2002 |
perseant | Correct free list tail pointer, when adding blocks of new inodes to v2
filesystems. Should fix PR #14408.
|
| 1.54 | 18-Dec-2001 |
chs | use the new compatibility routines to allow mmap() to work
(in the same non-coherent fashion that it worked pre-UBC)
until someone has time to do it the right way.
|
| 1.53 | 23-Nov-2001 |
chs | add spaces for KNF. confirmed to produce identical objects.
|
| 1.52 | 08-Nov-2001 |
lukem | add RCSID
|
| 1.51 | 14-Oct-2001 |
chs | branches: 1.51.2;
initialize the vnode's copy of the size in lfs_ialloc().
|
| 1.50 | 28-Sep-2001 |
chs | don't depend on other headers to include sys/proc.h for us.
|
| 1.49 | 15-Sep-2001 |
chs | a whole bunch of changes to improve performance and robustness under load:
- remove special treatment of pager_map mappings in pmaps. this is
required now, since I've removed the globals that expose the address range.
pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's
no longer any need to special-case it.
- eliminate struct uvm_vnode by moving its fields into struct vnode.
- rewrite the pageout path. the pager is now responsible for handling the
high-level requests instead of only getting control after a bunch of work
has already been done on its behalf. this will allow us to UBCify LFS,
which needs tighter control over its pages than other filesystems do.
writing a page to disk no longer requires making it read-only, which
allows us to write wired pages without causing all kinds of havoc.
- use a new PG_PAGEOUT flag to indicate that a page should be freed
on behalf of the pagedaemon when it's unlocked. this flag is very similar
to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the
pageout fails due to eg. an indirect-block buffer being locked.
this allows us to remove the "version" field from struct vm_page,
and together with shrinking "loan_count" from 32 bits to 16,
struct vm_page is now 4 bytes smaller.
- no longer use PG_RELEASED for swap-backed pages. if the page is busy
because it's being paged out, we can't release the swap slot to be
reallocated until that write is complete, but unlike with vnodes we
don't keep a count of in-progress writes so there's no good way to
know when the write is done. instead, when we need to free a busy
swap-backed page, just sleep until we can get it busy ourselves.
- implement a fast-path for extending writes which allows us to avoid
zeroing new pages. this substantially reduces cpu usage.
- encapsulate the data used by the genfs code in a struct genfs_node,
which must be the first element of the filesystem-specific vnode data
for filesystems which use genfs_{get,put}pages().
- eliminate many of the UVM pagerops, since they aren't needed anymore
now that the pager "put" operation is a higher-level operation.
- enhance the genfs code to allow NFS to use the genfs_{get,put}pages
instead of a modified copy.
- clean up struct vnode by removing all the fields that used to be used by
the vfs_cluster.c code (which we don't use anymore with UBC).
- remove kmem_object and mb_object since they were useless.
instead of allocating pages to these objects, we now just allocate
pages with no object. such pages are mapped in the kernel until they
are freed, so we can use the mapping to find the page to free it.
this allows us to remove splvm() protection in several places.
The sum of all these changes improves write throughput on my
decstation 5000/200 to within 1% of the rate of NetBSD 1.5
and reduces the elapsed time for "make release" of a NetBSD 1.5
source tree on my 128MB pc to 10% less than a 1.5 kernel took.
|
| 1.48 | 13-Jul-2001 |
perseant | branches: 1.48.2;
Merge the short-lived perseant-lfsv2 branch into the trunk.
Kernels and tools understand both v1 and v2 filesystems; newfs_lfs
generates v2 by default. Changes for the v2 layout include:
- Segments of non-PO2 size and arbitrary block offset, so these can be
matched to convenient physical characteristics of the partition (e.g.,
stripe or track size and offset).
- Address by fragment instead of by disk sector, paving the way for
non-512-byte-sector devices. In theory fragments can be as large
as you like, though in reality they must be smaller than MAXBSIZE in size.
- Use serial number and filesystem identifier to ensure that roll-forward
doesn't get old data and think it's new. Roll-forward is enabled for
v2 filesystems, though not for v1 filesystems by default.
- The inode free list is now a tailq, paving the way for undelete (undelete
is not yet implemented, but can be without further non-backwards-compatible
changes to disk structures).
- Inode atime information is kept in the Ifile, instead of on the inode;
that is, the inode is never written *just* because atime was changed.
Because of this the inodes remain near the file data on the disk, rather
than wandering all over as the disk is read repeatedly. This speeds up
repeated reads by a small but noticeable amount.
Other changes of note include:
- The ifile written by newfs_lfs can now be of arbitrary length, it is no
longer restricted to a single indirect block.
- Fixed an old bug where ctime was changed every time a vnode was created.
I need to look more closely to make sure that the times are only updated
during write(2) and friends, not after-the-fact during a segment write,
and certainly not by the cleaner.
|
| 1.47 | 30-May-2001 |
mrg | branches: 1.47.2; 1.47.4;
use _KERNEL_OPT
|
| 1.46 | 03-Dec-2000 |
perseant | branches: 1.46.2;
Get rid of some old unnecessary code that cleared B_NEEDCOMMIT from buffers in
lfs_writeseg (possibly after they had been freed).
If MALLOCLOG is defined, make lfs_newbuf and lfs_freebuf pass along the
caller's file and line to _malloc and _free.
|
| 1.45 | 27-Nov-2000 |
chs | Initial integration of the Unified Buffer Cache project.
|
| 1.44 | 27-Nov-2000 |
perseant | If LFS_DO_ROLLFORWARD is defined, roll forward from the older checkpoint
on mount, through the newer checkpoint and on through any newer
partial-segments that may have been written but not checkpointed because
of an intervening crash.
LFS_DO_ROLLFORWARD is not defined by default.
|
| 1.43 | 09-Sep-2000 |
perseant | Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
|
| 1.42 | 05-Jul-2000 |
perseant | Clean up accounting of lfs_uinodes (dirty but unwritten inodes).
Make lfs_uinodes a signed quantity for debugging purposes, and set it to
zero as fs mount time.
Enclose setting/clearing of the dirty flags (IN_MODIFIED, IN_ACCESSED,
IN_CLEANING) in macros, and use those macros everywhere. Make
LFS_ITIMES use these macros; updated the ITIMES macro in inode.h to know
about this. Make ufs_getattr use ITIMES instead of FFS_ITIMES.
|
| 1.41 | 03-Jul-2000 |
perseant | i_lfs_effnblks fixes. Put debugging printfs under #ifdef DEBUG_LFS.
|
| 1.40 | 30-Jun-2000 |
fvdl | Rearrange code around getnewvnode as was already done for ffs, to avoid
locking against oneself because getnewvnode recycles a softdep-using vnode.
|
| 1.39 | 28-Jun-2000 |
mrg | remove include of <vm/vm.h> and <uvm/uvm_extern.h>
|
| 1.38 | 27-Jun-2000 |
perseant | Fixes associated with filling an LFS:
Change the space computation to appear to change the size of the *disk*
rather than the *bytes used* when more segment summaries and inode
blocks are written. Try to estimate the amount of space that these will
take up when more files are written, so the disk size doesn't change too
much.
Regularize error returns from lfs_valloc, lfs_balloc, lfs_truncate: they
now fail entirely, rather than succeeding half-way and leaving the fs in
an inconsistent state.
Rewrite lfs_truncate, mostly stealing from ffs_truncate. The old
lfs_truncate had difficulty truncating a large file to a non-zero size
(indirect blocks were not handled appropriately).
Unmark VDIROP on fvp after ufs_remove, ufs_rmdir, so these can be
reclaimed immediately: this vnode would not be written to disk again
anyway if the removal succeeded, and if it failed, no directory
operation occurred.
ufs_makeinode and ufs_mkdir now remove IN_ADIROP on error.
|
| 1.37 | 22-Jun-2000 |
perseant | fix my own typo, grr....
|
| 1.36 | 22-Jun-2000 |
perseant | Read i_ffs_gen from the version number in the Ifile during lfs_valloc,
instead of keeping it always == 1. (The ifile version number is
increased on vfree.) May address PR #7213, but I haven't been able to
test thoroughly enough to say for sure.
|
| 1.35 | 22-Jun-2000 |
perseant | Update lfs_vunref for the fact that now a vnode can be locked with no
references (locked for VOP_INACTIVE at the end of vrele) and it's okay.
Check the return value of lfs_vref where appropriate.
Fixes PR #s 10285 and 10352.
|
| 1.34 | 06-Jun-2000 |
perseant | branches: 1.34.2;
Protect inode free list with seglock, instead of separate lock, so that
the head of the inode free list (on the superblock) always matches the
rest of the free list (in the ifile).
Protect lfs_fragextend with seglock, to prevent the segment byte count
fudging from making its way to disk.
Don't try to inactivate dirop vnodes that are still in the middle of
their dirop (may address PR#10285).
|
| 1.33 | 31-May-2000 |
perseant | update for IN_ACCESSED changes
|
| 1.32 | 27-May-2000 |
perseant | branches: 1.32.2;
Prevent dirops from getting around lfs_check and wedging the buffer cache.
All the dirop vnops now mark the inodes with a new flag, IN_ADIROP, which
is removed as soon as the dirop is done (as opposed to VDIROP which stays
until the file is written). To address one issue raised in PR#9357.
|
| 1.31 | 19-Jan-2000 |
perseant | Changes to stabilize LFS. The first two of these should also apply to the
1.4 branch.
* Use a separate per-fs lock, instead of ufs_hashlock, to protect the Inode
free list. This seems to prevent the "lockmgr: %d, not exclusive lock holder
%d, unlocking" message I was mis-attributing last night to an unlocked vnode
being passed to vrele.
* Change calling semantics of lfs_ifind, to give better error reporting:
If fed a struct buf, it can report the block number of the offending inode
block as well as the inode number.
* Back out rev 1.10 of lfs_subr.c, since the replacement code was slightly
uglier while being functionally identical.
* Make lfs_vunref use the same free list convention as vrele/vput, so that
vget does not remove vnodes from a hash list they are not on.
|
| 1.30 | 15-Dec-1999 |
perseant | Fix error returns on lfs vnops so that locks and reference counts are
preserved. Handle dirop accounting in lfs_vfree for this case as well.
May address PR#8823.
|
| 1.29 | 15-Nov-1999 |
fvdl | Add Kirk McKusick's soft updates code to the trunk. Not enabled by
default, as the copyright on the main file (ffs_softdep.c) is such
that is has been put into gnusrc. options SOFTDEP will pull this
in. This code also contains the trickle syncer.
Bump version number to 1.4O
|
| 1.28 | 12-Nov-1999 |
perseant | Back out my patch of the 8th (to address unreferenced inode problem).
Apparently this needs more thought.
|
| 1.27 | 09-Nov-1999 |
perseant | If ifile blocks were written before dirops were complete, and then the
system crashed, inodes could be allocated that were not referenced. (Though
not a serious problem, it evidences itself in phase 4 of fsck_lfs.) Fix
this by marking if_daddr with UNASSIGNED before the inodes are actually
written; at mount time the ifile is checked for UNASSIGNED entries and
any that are found are linked back into the free list. (The latter
functionality should move into the roll-forward agent when it materializes.)
|
| 1.26 | 06-Nov-1999 |
perseant | branches: 1.26.2;
Address ufs_hashlock/ufs_ihashins protocol bug, discovered while doing a
post-mortem of a production machine. Also, take the active dirop
count off of the fs and make it global (since it is measuring a global
resource) and tie the threshold value LFS_MAXDIROP to desiredvnodes.
|
| 1.25 | 03-Sep-1999 |
perseant | branches: 1.25.2; 1.25.4; 1.25.6;
Make changes that will allow an LFS filesystem to be used as the root
filesystem. In particular,
- Fix mknod deadlock, described in PR 8172.
- Enable lfs_mountroot.
- Make lfs_writevnodes treat filesystems mounted on lfs device nodes properly,
by flushing that device rather than trying to add blocks to the device inode.
This, in combination with lfs boot blocks, will allow operation of an all-lfs
system.
|
| 1.24 | 08-Jul-1999 |
wrstuden | Modify file systems to deal with struct lock in struct vnode. All leaf
fs's other than nfs use genfs_lock() for locking.
Modify lookup routines to set PDIRUNLOCK when they unlock the parrent.
|
| 1.23 | 17-Jun-1999 |
tls | squash some compiler warnings on debug printfs by casting to int
|
| 1.22 | 15-Jun-1999 |
perseant | Minor changes to the segment live bytes calculation. In particular, fixed
a bug in fragment extension that could run the count negative. Also, don't
overcount for inodes, and don't count segment summaries. Thus, for empty
segments the live bytes count should now be exactly zero.
|
| 1.21 | 16-Apr-1999 |
perseant | Other half of the ufs_hashlock locking fix (oops)
|
| 1.20 | 16-Apr-1999 |
perseant | Fix locking panic on ufs_hashlock
|
| 1.19 | 11-Apr-1999 |
perseant | Take out the `#ifdef USE_UFSHASH'; use ufs_hashlock to lock the inode free
list instead of free_lock.
|
| 1.18 | 24-Mar-1999 |
mrg | branches: 1.18.2;
completely remove Mach VM support. all that is left is the all the
header files as UVM still uses (most of) these.
|
| 1.17 | 10-Mar-1999 |
perseant | New sources should leave the LFS in a more-or-less working state. Changes
include:
- DIROP segregation is enabled, and greater care is taken
to make sure that a checkpoint completes. Fsck is not
needed to remount the filesystem.
- Several checks to make sure that the LFS subsystem does not
overuse various resources (memory, in particular).
- The cleaner routines, lfs_markv in particular, are completely
rewritten. A buffer overflow is removed. Greater care is taken
to ensure that inodes come from where lfs_cleanerd say they come
from (so we know nothing has changed since lfs_bmapv was called).
- Fragment allocation is fixed, so that writes beyond end-of-file
do the right thing.
|
| 1.16 | 23-Oct-1998 |
thorpej | Use DINODE_SIZE rather than sizeof(struct dinode).
|
| 1.15 | 01-Sep-1998 |
thorpej | Use the pool allocator and the "nointr" pool page allocator for LFS inodes.
|
| 1.14 | 24-Jun-1998 |
sommerfe | Always include fifos; "not an option any more".
|
| 1.13 | 09-Jun-1998 |
scottr | Protect various config(8)-generated files from inclusion while
building LKMs. Fixes PR 5557.
|
| 1.12 | 08-Jun-1998 |
scottr | Use the newly-defined opt_quota.h.
|
| 1.11 | 01-Mar-1998 |
fvdl | Merge with Lite2 + local changes
|
| 1.10 | 07-Feb-1998 |
chs | add UVM stuff.
|
| 1.9 | 04-Jul-1997 |
drochner | Don't cast 64bit (off_t) file sizes to vm_offset_t (32bit on many
architectures), truncate them intelligently instead.
The truncation is done centralized in vnode_pager.c.
This prevents from wrap-over effects when parts of large (>2^32 byte) files
are mmapped.
Don't allow to mmap above the numerical range of vm_offset_t.
This is considered a temporary solution until the vm system handles the
object sizes/offsets more cleanly.
|
| 1.8 | 11-Jun-1997 |
bouyer | Add support for ext2fs, this needed a few modifications to ufs/ufs/inode.h:
- added an "union inode_ext" to struct inode, for the per-fs extentions.
For now only ext2fs uses it.
- i_din is now an union:
union {
struct dinode ffs_din; /* 128 bytes of the on-disk dinode. */
struct ext2fs_dinode e2fs_din; /* 128 bytes of the on-disk dinode. */
} i_din
Added a lot of #define i_ffs_* and i_e2fs_* to access the fields.
- Added two macros: FFS_ITIMES and EXT2FS_ITIMES. ITIMES calls the rigth
macro, depending on the time of the inode. ITIMES is used where necessary,
FFS_ITIMES and EXT2FS_ITIMES in other places.
|
| 1.7 | 10-Mar-1997 |
mycroft | Just increment the generation count. Using the time is bogus and defeats
fsirand(8).
|
| 1.6 | 12-Oct-1996 |
christos | branches: 1.6.6;
revert previous kprintf changes
|
| 1.5 | 10-Oct-1996 |
christos | printf -> kprintf, sprintf -> ksprintf
|
| 1.4 | 25-Mar-1996 |
pk | Appease gcc: unused variables if !QUOTA
|
| 1.3 | 09-Feb-1996 |
christos | lfs prototypes
|
| 1.2 | 29-Jun-1994 |
cgd | New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
|
| 1.1 | 08-Jun-1994 |
mycroft | branches: 1.1.1;
Update to 4.4-Lite fs code, with local changes.
|
| 1.1.1.2 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite2
|
| 1.1.1.1 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite for reference
|
| 1.6.6.1 | 12-Mar-1997 |
is | Merge in changes from Trunk
|
| 1.18.2.8 | 20-Jan-2000 |
he | Pull up revision 1.31 (via patch, requested by perseant):
Files removed (through unlink, rmdir) are now really removed, though the
removal is postponed until the dirop is complete to ensure validity of
the filesystem through a crash. Use a separate per-fs lock, instead of
ufs_hashlock, to protect the inode free list. Change calling semantics
of lfs_ifind, to give better error reporting: If fed a struct buf, it
can report the block number of the offending inode block as well as the
inode number.
|
| 1.18.2.7 | 15-Jan-2000 |
he | Pull up revision 1.30 (requested by perseant):
Fix error returns on lfs vnops so that locks and reference counts
are preserved. Handle dirop accounting in lfs_vfree for this
case as well. Addresses PR#8823.
|
| 1.18.2.6 | 15-Jan-2000 |
he | Pull up revision 1.25 (requested by perseant):
Address problems related to using an LFS filesystem as the root
filesystem, including mknod hangs. Fixes PR#8172 and PR#9072.
|
| 1.18.2.5 | 17-Dec-1999 |
he | Pull up revision 1.26 (requested by perseant):
Address locking protocol error for inode hash, and make the
maximum number of active dirops a global quantity.
|
| 1.18.2.4 | 03-Sep-1999 |
he | Pull up revision 1.23:
Fix a printf format bug that gives compiler warnings/errors on
64-bit platforms, fixing PR#8241. (perseant)
|
| 1.18.2.3 | 25-Jun-1999 |
perry | pullup 1.21->1.22 (perseant)
|
| 1.18.2.2 | 16-Apr-1999 |
perseant | branches: 1.18.2.2.2; 1.18.2.2.4;
Pull up src/sys/ufs/lfs: lfs_alloc.c 1.19->1.21.
This fixes another locking problem, this time a lock on ufs_hashlock in
lfs_vfree. The lock could be held by a process calling getnewvnode, and
then attempted again by lfs_vfree. This works around that, not attempting
to get the lock if curproc already holds it.
|
| 1.18.2.1 | 13-Apr-1999 |
perseant | Pull-up of changes made to the trunk on Sunday [1.18->1.19], to wit:
Take out the `#ifdef USE_UFSHASH'; use ufs_hashlock to lock the inode free
list instead of free_lock.
Fix inode reporting in lfs_statfs (the meaning of f_files and f_ffree was
reversed).
Fix "lfs_ifind: dinode xxx not found" panic. When inodes were freed, then
immediately reloaded, their dinodes were located in an inode block which
was not on disk at the advertized location, nor in the cache (although it
would be flushed to disk next segment write). Fix this by using getblk()
instead of lfs_newbuf() for inode blocks.
Better checking for held inode locks in lfs_fastvget, for a number of
error conditions. Also change the default setting of lfs_clean_vnhead to
0, which seems to make the locking problems go away (although this is
difficult to test as I can't reliably reproduce them).
Make sure that the wakeup occurs for vnodes that lfs_update might be
sleeping on (nodes which are not marked IN_MODIFIED/IN_CLEANING, but which
have dirty buffers), by marking them with the appropriate flag if
dirtybuffers were added while the write was in progress.
Fix block counting during file truncation, if not truncating to zero.
Disallow threshold-initiated cache flush when dirops are active. Also,
make SET_ENDOP use lfs_check instead of inlining most of it.
Improve the debugging printfs in the cleaner syscalls (in particular, make
it obvious that they're coming from lfs).
Check the superblock version field, and refuse to mount the filesystem if
the version number is higher than we know about. This allows, e.g.,
changes in the format of the ifile, segment size restrictions and
boundaries, etc., which would not affect existing fields in the
superblock, but which would drastically affect the filesystem, to be
smoothly integrated at a later date.
|
| 1.18.2.2.4.1 | 30-Nov-1999 |
itojun | bring in latest KAME (as of 19991130, KAME/NetBSD141) into kame branch
just for reference purposes.
This commit includes 1.4 -> 1.4.1 sync for kame branch.
The branch does not compile at all (due to the lack of ALTQ and some other
source code). Please do not try to modify the branch, this is just for
referenre purposes.
synchronization to latest KAME will take place on HEAD branch soon.
|
| 1.18.2.2.2.4 | 31-Aug-1999 |
perseant | Rudimentary support for LFS under UBC:
- LFS-specific VOP_BALLOC and VOP_PUTPAGES vnode ops.
- getblk VREG panic #ifdef'd out (can be reinstated when Ifile is
internalized and Ifile can be made another type from VREG)
- interface to VOP_PUTPAGES changed to pass all pager flags, not
just sync. FS putpages routines must know about the pager flags.
- new LFS magic disk address, -2 ("unwritten"), meaning accounted for
but not assigned to a fixed disk location (since LFS does these two
things separately, and the previous accounting method using buffer
headers no longer will work). Changed references to (foo == (daddr_t)-1)
to (foo < 0). Since disk drivers reject all addresses < 0, this should
not present a problem for other FSs.
|
| 1.18.2.2.2.3 | 02-Aug-1999 |
thorpej | Update from trunk.
|
| 1.18.2.2.2.2 | 11-Jul-1999 |
chs | remove uvm_vnp_uncache(), it's no longer needed.
|
| 1.18.2.2.2.1 | 21-Jun-1999 |
thorpej | Sync w/ -current.
|
| 1.25.6.1 | 27-Dec-1999 |
wrstuden | Pull up to last week's -current.
|
| 1.25.4.2 | 15-Nov-1999 |
fvdl | Sync with -current
|
| 1.25.4.1 | 19-Oct-1999 |
fvdl | Bring in Kirk McKusick's FFS softdep code on a branch.
|
| 1.25.2.2 | 08-Dec-2000 |
bouyer | Sync with HEAD.
|
| 1.25.2.1 | 20-Nov-2000 |
bouyer | Update thorpej_scsipi to -current as of a month ago
|
| 1.26.2.2 | 06-Nov-1999 |
perseant | Address ufs_hashlock/ufs_ihashins protocol bug, discovered while doing a
post-mortem of a production machine. Also, take the active dirop
count off of the fs and make it global (since it is measuring a global
resource) and tie the threshold value LFS_MAXDIROP to desiredvnodes.
|
| 1.26.2.1 | 06-Nov-1999 |
perseant | file lfs_alloc.c was added on branch comdex-fall-1999 on 1999-11-06 20:33:06 +0000
|
| 1.32.2.1 | 22-Jun-2000 |
minoura | Sync w/ netbsd-1-5-base.
|
| 1.34.2.4 | 14-Sep-2000 |
perseant | Pull up recent LFS kernel changes (approved by thorpej):
ufs/ufs/inode.h, 1.20--1.22 (add i_lfs_effnblks extension ;
make ITIMES aware of LFS_ITIMES;
_LKM protection so userland progs
compile)
ufs/ufs/ufs_vnops.c, 1.69, 1.71 (remove IN_ADIROP;
use ITIMES instead of FFS_ITIMES)
ufs/ufs/ufs_readwrite.c, 1.27 (use lfs_reserve in lfs_write)
ufs/lfs/lfs.h, 1.26--1.32 (define LFS_EST_* macros ;
change MIN_FREE_SEGS to lfs_minfreesegs ;
add avail and bfree to CLEANERINFO ;
change lfs_uinodes to signed ;
change lfs_dmeta to signed ;
add whitespace to line up structure
members ;
explicit cast to int32_t in LFS_EST_*
macros)
ufs/lfs/lfs_alloc.c, back out 1.34.2.3 (pullups of 1.39, 1.40);
then pull up 1.38 (clean up on error)
1.39--1.43 (restore fvdl's ufs_hashlock fix ;
restore fvdl's ufs_hashlock fix ;
set i_lfs_effnblks ;
use UINO macros ;
add comments and fix long lines)
ufs/lfs/lfs_balloc.c, 1.19 (don't succeed halfway)
1.21--1.25 (use i_lfs_effnblks ;
fix i_lfs_effnblks computation and
quieten ;
fix i_ffs_blocks in unwritten fragment ;
remove useless debugging check ;
add comments and (c) 2000)
ufs/lfs/lfs_bio.c, 1.24--1.30 (cleanup and make lfs_flush_fs take
"struct lfs *" instead of "struct
mount *" ;
use lfs_minfreeseg instead of
MIN_FREE_SEGS ;
use UINO macros, and copy bfree/avail
to CLEANERINFO ;
add lfs_reserve function ;
1.28--1.30 fix printf formatting)
ufs/lfs/lfs_cksum.c, 1.13 (add (c) 2000)
ufs/lfs/lfs_debug.c, 1.11 (use btodb instead of DEV_BSIZE)
ufs/lfs/lfs_extern.h, 1.18, 1.20--1.21 (function prototype changes)
ufs/lfs/lfs_inode.c, 1.38 (rewrite lfs_truncate from
ffs_truncate)
1.40--1.44 (count written and unwritten blocks
seperately ;
use disk block units instead of bytes ;
remove unnecessary "mod" variable ;
correct B_DELWRI to avoid bawrite panic ;
use lfs_reserve)
ufs/lfs/lfs_segment.c, 1.52-1.59 (use lfs_dmeta to note used summaries ;
check for UNWRITTEN in indirect blocks ;
more debugging stuff inside #ifdef
DEBUG_LFS ;
use LK_CANRECURSE ;
don't drop dirty indirect blocks ;
use UINO macros ;
don't hose the free list ;
use btodb() instead of DEV_BSIZE ;
make it compile again (oops))
ufs/lfs/lfs_subr.c, 1.16--1.17 (check for locked inodes before
changing ;
use btodb() instead of DEV_BSIZE, (c)
2000)
ufs/lfs/lfs_syscalls.c, back out 1.41.4.2 (fvdl's ufs_hashlock fix);
then pull up 1.43 (use lfs_dmeta)
1.44--1.45 (restore fvdl's ufs_hashlock fix)
1.46--1.47 (fix lfs_avail leakage from sblock
segments ;
use UINO macros)
1.49 (bounds-check inode numbers in
lfs_markv)
ufs/lfs/lfs_vfsops.c, 1.53 (use LFS_EST_* macros in lfs_statfs)
1.56--1.58 (initialize lfs_minfreeseg, lfs_effnblk ;
initialize lfs_uinodes ;
initialize lfs_ravail)
ufs/lfs/lfs_vnops.c, 1.40 (remove VDIROP from removed files)
1.42--1.44 (move SET_ENDOP below the removal of
VDIROP ;
use UINO macros and add lfs_itimes
function ;
use lfs_reserve in dirops)
|
| 1.34.2.3 | 03-Jul-2000 |
fvdl | pullup the fixes from the trunk to not hold ufs_hashlock across
getnewvnode()
|
| 1.34.2.2 | 28-Jun-2000 |
perseant | pull up i_ffs_gen patch from trunk
|
| 1.34.2.1 | 22-Jun-2000 |
perseant | Pull up lfs_vunref fix from the trunk.
|
| 1.46.2.11 | 08-Jan-2003 |
thorpej | Sync with HEAD.
|
| 1.46.2.10 | 11-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.46.2.9 | 20-Jun-2002 |
nathanw | Catch up to -current.
|
| 1.46.2.8 | 28-Feb-2002 |
nathanw | Catch up to -current.
|
| 1.46.2.7 | 08-Jan-2002 |
nathanw | Catch up to -current.
|
| 1.46.2.6 | 14-Nov-2001 |
nathanw | Catch up to -current.
|
| 1.46.2.5 | 22-Oct-2001 |
nathanw | Catch up to -current.
|
| 1.46.2.4 | 08-Oct-2001 |
nathanw | Catch up to -current.
|
| 1.46.2.3 | 21-Sep-2001 |
nathanw | Catch up to -current.
|
| 1.46.2.2 | 24-Aug-2001 |
nathanw | Catch up with -current.
|
| 1.46.2.1 | 21-Jun-2001 |
nathanw | Catch up to -current.
|
| 1.47.4.4 | 23-Jun-2002 |
jdolecek | catch up with -current on kqueue branch
|
| 1.47.4.3 | 11-Feb-2002 |
jdolecek | Sync w/ -current.
|
| 1.47.4.2 | 10-Jan-2002 |
thorpej | Sync kqueue branch with -current.
|
| 1.47.4.1 | 03-Aug-2001 |
lukem | update to -current
|
| 1.47.2.6 | 13-Jul-2001 |
perseant | Be more careful about when we update ctime/mtime. In particular, if we
are only writing indirect blocks, that doesn't count for mtime; and when
we first create a vnode, that certainly *does not* count for ctime
(a bug that's been there from the beginning).
This does not change the fact that mtime might still be set after write(2)
is "completed", but it does make the atime-in-the-ifile code have some
effect (noticeable less degradation of read time after an intervening
large write).
|
| 1.47.2.5 | 10-Jul-2001 |
perseant | Turn the free list into a tailq, with both head and tail kept on the ifile.
Update access times on the inode even if it does not get marked IN_ACCESS.
|
| 1.47.2.4 | 02-Jul-2001 |
perseant | Change disk addressing unit to be the fragment, instead of the disk sector.
All quantities in the superblock, inodes, indirect blocks, etc. refer now
to this abstract unit (called "fsb" as it is in FFS) instead of disk sectors;
as a consequence segment summary blocks have to be multiples of a fragment in
size. In v1 filesystems, compatibility code ensures that 1 fsb == 1 sector,
regardless of fragment size.
Fragments can now range in size between 512 and 32k; in the event that
LFS_LABELPAD (8k) is smaller than the disk address unit size, an extra
proto-superblock is kept at 8k from the beginning of the disk, to be used
*only* to locate the real superblocks. (Not all of the userland knows about
this yet.)
Almost all of this was done not by me, but by joff.
|
| 1.47.2.3 | 29-Jun-2001 |
perseant | Update the Ifile copy of the free list head in lfs_vfree, so inode numbers
actually get reused.
|
| 1.47.2.2 | 29-Jun-2001 |
perseant | Get rid of __P(), protoizing where it had not already been done
|
| 1.47.2.1 | 27-Jun-2001 |
perseant | Import of what I've been calling "LFSv2", that is, LFS with some features
added that require changes to the on-disk data structures. These include:
- 64-bit time in everything but inodes
- User-specified segment offset, and segment size no longer
restricted to PO2.
- Serial number on segment summaries in addition to timestamp, and
a new volume identifier, to make roll-forward feasible without
fear of finding old data and thinking it was new.
Although I think this version works at least as well as what's on the trunk,
we're not done yet; hence this commit is going in on a branch and not on
the trunk. Enhancements that are not here yet include fragment addressing,
like FFS does, instead of block addressing.
|
| 1.48.2.1 | 01-Oct-2001 |
fvdl | Catch up with -current.
|
| 1.51.2.1 | 12-Nov-2001 |
thorpej | Sync the thorpej-mips-cache branch with -current.
|
| 1.69.2.10 | 10-Nov-2005 |
skrll | Sync with HEAD. Here we go again...
|
| 1.69.2.9 | 01-Apr-2005 |
skrll | Sync with HEAD.
|
| 1.69.2.8 | 08-Mar-2005 |
skrll | Sync with HEAD.
|
| 1.69.2.7 | 04-Mar-2005 |
skrll | Sync with HEAD.
Hi Perry!
|
| 1.69.2.6 | 21-Sep-2004 |
skrll | Fix the sync with head I botched.
|
| 1.69.2.5 | 18-Sep-2004 |
skrll | Sync with HEAD.
|
| 1.69.2.4 | 25-Aug-2004 |
skrll | Sync with HEAD.
|
| 1.69.2.3 | 24-Aug-2004 |
skrll | Undo part of the ktrace/lwp changes. In particular:
* Remove the "lwp *" argument that was added to vget(). Turns out
that nothing actually used it!
* Remove the "lwp *" arguments that were added to VFS_ROOT(), VFS_VGET(),
and VFS_FHTOVP(); all they did was pass it to vget() (which, as noted
above, didn't use it).
* Remove all of the "lwp *" arguments to internal functions that were added
just to appease the above.
|
| 1.69.2.2 | 03-Aug-2004 |
skrll | Sync with HEAD
|
| 1.69.2.1 | 02-Jul-2003 |
darrenr | Apply the aborted ktrace-lwp changes to a specific branch. This is just for
others to review, I'm concerned that patch fuziness may have resulted in some
errant code being generated but I'll look at that later by comparing the diff
from the base to the branch with the file I attempt to apply to it. This will,
at the very least, put the changes in a better context for others to review
them and attempt to tinker with removing passing of 'struct lwp' through
the kernel.
|
| 1.72.4.1 | 10-May-2005 |
riz | Pull up the following revisions (requested by perseant in ticket #1281):
1.8 sys/ufs/lfs/TODO
1.75 sys/ufs/lfs/lfs.h (via patch)
1.74 sys/ufs/lfs/lfs_alloc.c (via patch)
1.49, 1.51 sys/ufs/lfs/lfs_balloc.c (1.51 via patch)
1.78 sys/ufs/lfs/lfs_bio.c
1.62 sys/ufs/lfs/lfs_extern.h (via patch)
1.156 sys/ufs/lfs/lfs_segment.c (via patch)
1.48 sys/ufs/lfs/lfs_subr.c
1.101 sys/ufs/lfs/lfs_syscalls.c
1.163 sys/ufs/lfs/lfs_vfsops.c (via patch)
1.134 sys/ufs/lfs/lfs_vnops.c (via patch)
1.61 sys/ufs/ufs/ufs_readwrite.c (via patch)
1.20 libexec/lfs_cleanerd/clean.h (via patch)
1.52 libexec/lfs_cleanerd/cleanerd.c (via patch)
1.41 libexec/lfs_cleanerd/library.c (via patch)
1.4 regress/sys/fs/lfs/newfs_fsck/Makefile
1.2 regress/sys/fs/lfs/newfs_fsck/mkfs_mount
1.2 regress/sys/fs/lfs/newfs_fsck/smallfiles
1.3 sbin/fsck_lfs/bufcache.c
1.3 sbin/fsck_lfs/bufcache.h
1.3 sbin/fsck_lfs/lfs.h
1.8 sbin/fsck_lfs/lfs.c (via patch)
1.8 sbin/fsck_lfs/pass3.c (via patch)
1.18 sbin/fsck_lfs/pass0.c (via patch)
1.18 sbin/fsck_lfs/utilities.c (via patch)
1.7 sbin/fsck_lfs/segwrite.c
1.19 sbin/fsck_lfs/setup.c (via patch)
1.3 sbin/newfs_lfs/Makefile
0 sbin/newfs_lfs/lfs.c (yes, remove it)
1.1 sbin/newfs_lfs/make_lfs.c
1.15 sbin/newfs_lfs/newfs.c (via patch)
Various minor LFS improvements.
Kernel:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this. Should fix PR #29045.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
Fixes PR #26680.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
cleaner:
* Adapt lfs_cleanerd to use the fcntl call to get the Ifile filehandle,
so it need not be in the namespace.
* Make lfs_cleanerd be more careful when there are very few available
segments.
* Make lfs_cleanerd less verbose when the filesystem is unmounted.
newfs_lfs, fsck_lfs, and regression:
* Extend the lfs library from fsck_lfs(8) so that it can be used with a
not-yet-existent LFS. Make newfs_lfs(8) use this library, so it can
create LFSs whose Ifile is larger than one segment. Addresses PR #11110.
* Make newfs_lfs(8) use strsuftoi64() for its arguments, a la newfs(8).
* Make fsck_lfs(8) respect the "file system is clean" flag.
* Don't let fsck_lfs(8) think it has dirty blocks when invoked with the
-n flag.
* Remove the Ifile from the filesystem namespace. The cleaner now uses
a fcntl call on the root inode to find the Ifile filehandle. (As a
side-effect, addresses PR #29144.)
|
| 1.73.6.2 | 26-Mar-2005 |
yamt | sync with head.
|
| 1.73.6.1 | 19-Mar-2005 |
yamt | sync with head. xen and whitespace. xen part is not finished.
|
| 1.73.4.1 | 29-Apr-2005 |
kent | sync with -current
|
| 1.76.2.12 | 10-Aug-2006 |
tron | Apply patch (requested by fair in perseant #1457):
Bring LFS up to current, including a patch (1.95 lfs_alloc.c) that
should prevent the inode free list errors seen on the STABLE branch
subsequent to pullup ticket #1327.
|
| 1.76.2.11 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.93
sys/ufs/lfs/lfs.h: revision 1.106
sys/ufs/lfs/lfs_vfsops.c: revision 1.209
sys/ufs/lfs/lfs_vnops.c: revision 1.175
sys/ufs/lfs/lfs_segment.c: revision 1.178
Fixes to address the "vinvalbuf: dirty blocks" panic that can occur when
many inodes are cleaned at once. Make sure that we write all the pages
on vnodes that are being flushed, even if we don't think there's room;
drain v_numoutput before lfs_vflush() completes.
Also, don't allow a vnode that is in the process of being cleaned to be
chosen by getnewvnode(); this avoids a segment accounting panic in the case
that a large number of inodes are fed to lfs_markv() all at once.
|
| 1.76.2.10 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.92
sys/ufs/lfs/lfs.h: revision 1.105
sys/ufs/lfs/lfs_vfsops.c: revision 1.207
sys/ufs/lfs/lfs_subr.c: revision 1.59
sys/ufs/lfs/lfs_vnops.c: revision 1.173
sys/ufs/lfs/lfs_bio.c: revision 1.92
Introduce another per-filesystem parameter, lfs_resvseg, to separate the
notion of "how many segments are reserved for the cleaner" from that of
"how many segments are not counted in lfs_bfree". The default value
used for existing filesystems is the same as the previous implicit value
of (lfs_minfreeseg / 2 + 1), modulo some sanity checking.
Count pending dirops on a per-filesystem basis, since once we start
writing them we can't stop until we're done. This seems to help stave off
the "no clean segments" panic in the case of filling the filesystem with
directories and small files (e.g. simultaneously unpacking more copies of
pkgsrc than will fit).
|
| 1.76.2.9 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.91
Add an explicit list initialization that was missing from my last commit.
|
| 1.76.2.8 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.104
sys/ufs/lfs/lfs_vfsops.c: revision 1.206
sys/ufs/lfs/lfs_vnops.c: revision 1.170
sys/ufs/lfs/lfs_extern.h: revision 1.80
sys/ufs/lfs/lfs_segment.c: revision 1.176
sys/ufs/lfs/lfs_inode.c: revision 1.103 via patch
sys/ufs/lfs/lfs_alloc.c: revision 1.90
Postpone the segment accounting changes coming from truncation until the
inode that makes those changes valid is either written to disk by
lfs_writeinode() or discarded by lfs_vfree().
A couple of locking fixes are also included as well.
|
| 1.76.2.7 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.89
Fix a fencepost error in the bitmap handling in extend_ifile(), and another
in lfs_freelist_prev().
|
| 1.76.2.6 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.101
sys/ufs/lfs/lfs_vfsops.c: revision 1.202
sys/ufs/lfs/lfs_alloc.c: revision 1.88
Optimize the free list search a little more; in particular use words
instead of bytes for the index, and never search below fs->lfs_freehd.
Fix a bug in the previous version of the search (an erroneous assumption
that ino_t was signed).
Free the bitmap when we unmount the filesystem.
|
| 1.76.2.5 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.87
sys/ufs/lfs/lfs.h: revision 1.99
sys/ufs/lfs/lfs_vfsops.c: revision 1.199
sys/ufs/lfs/lfs_extern.h: revision 1.77 via patch
Keep the free list ordered. This solves a problem first pointed out to me
by Michel Oey, in which an aged LFS writes up to an extra Ifile block for
every file created; and paves the way for the truncation of the Ifile when
many files are deleted.
|
| 1.76.2.4 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.153
sys/ufs/lfs/lfs_debug.c: revision 1.32
sys/ufs/lfs/lfs_alloc.c: revision 1.84
sys/ufs/lfs/lfs_vfsops.c: revision 1.185
sys/ufs/lfs/lfs_segment.c: revision 1.165
64 bit inode changes.
|
| 1.76.2.3 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.152
sys/ufs/lfs/lfs_debug.c: revision 1.31
sys/ufs/lfs/lfs_subr.c: revision 1.53
sys/ufs/lfs/lfs_extern.h: revision 1.68
sys/ufs/lfs/lfs_inode.c: revision 1.96
sys/ufs/lfs/lfs_bio.c: revision 1.86
sys/ufs/lfs/lfs_alloc.c: revision 1.83
sys/ufs/lfs/lfs_vfsops.c: revision 1.181
sys/ufs/lfs/lfs.h: revision 1.88
sys/ufs/lfs/lfs_segment.c: revision 1.164
- sprinkle const
- avoid shadow variables.
|
| 1.76.2.2 | 07-May-2005 |
tron | Apply patch (requested by perseant in ticket #242):
* fsck_lfs buffer cache fixes, including PR #29151
* Change fsck_lfs phase 0 message to reflect reality
* fsck_lfs: check phase 5 (cleanerinfo accounting) even on
roll-forward
* Keep better track of the free list during roll-forward, avoiding
a core dump
* Improve hash table use for fsck_lfs buffer and vnode cache
* Document fsck_lfs flag -f, and implement -q
* Add resize_lfs, including kernel support
* Add LFS to mountd's list of exportable filesystem types
* Make the LFS lkm work again [christos@]
* Add MP locking to the LFS kernel subsystem
* Fix pager_map deadlock in lfs_putpages()
* Avoid incomplete file extension that looks like "partial
truncation" to fsck
* Use lfs_malloc for cleaner malloc, since the cleaner often runs
in low-memory conditions.
* Use splay trees, not hash table, to track page allocation for
write.
* Fix mkdir panic on full fs
* Fix page accounting leak by counting differently.
* Use rightly named structure for lfs_getattr [skrll@]
* Cosmetic changes for readability.
|
| 1.76.2.1 | 30-Mar-2005 |
tron | Pull up revision 1.77 (requested by perseant in ticket #74):
Make LFS dirops get their vnode first, before incrementing the dirop
count, to prevent a deadlock trying to call VOP_PUTPAGES() on a VDIROP
vnode. This can happen when a stacked filesystem is mounted on top of an
LFS: an LFS dirop needs to get a vnode, which is available from the upper
layer. The corresponding lower layer vnode, however, is VDIROP, so the
upper layer can't be cleaned out since its VOP_PUTPAGES() is passed
through to the lower layer, which waits for dirops to drain before it can
proceed. Deadlock.
Tweak ufs_makeinode() and ufs_mkdir() to pass the a_vpp argument through
to VOP_VALLOC().
Partially addresses PR # 26043, though it probably does not completely fix
the problem described there.
|
| 1.83.2.7 | 04-Feb-2008 |
yamt | sync with head.
|
| 1.83.2.6 | 21-Jan-2008 |
yamt | sync with head
|
| 1.83.2.5 | 27-Oct-2007 |
yamt | sync with head.
|
| 1.83.2.4 | 03-Sep-2007 |
yamt | sync with head.
|
| 1.83.2.3 | 26-Feb-2007 |
yamt | sync with head.
|
| 1.83.2.2 | 30-Dec-2006 |
yamt | sync with head.
|
| 1.83.2.1 | 21-Jun-2006 |
yamt | sync with head.
|
| 1.84.2.2 | 29-Oct-2005 |
yamt | use lfs_* directly rather than via ufs_ops.
suggested by Chuck Silvers.
|
| 1.84.2.1 | 20-Oct-2005 |
yamt | adapt ufs.
|
| 1.86.12.1 | 24-May-2006 |
tron | Merge 2006-05-24 NetBSD-current into the "peter-altq" branch.
|
| 1.86.10.4 | 11-May-2006 |
elad | sync with head
|
| 1.86.10.3 | 06-May-2006 |
christos | - Move kauth_cred_t declaration to <sys/types.h>
- Cleanup struct ucred; forward declarations that are unused.
- Don't include <sys/kauth.h> in any header, but include it in the c files
that need it.
Approved by core.
|
| 1.86.10.2 | 19-Apr-2006 |
elad | sync with head.
|
| 1.86.10.1 | 08-Mar-2006 |
elad | Adapt to kernel authorization KPI.
|
| 1.86.8.4 | 03-Sep-2006 |
yamt | sync with head.
|
| 1.86.8.3 | 11-Aug-2006 |
yamt | sync with head
|
| 1.86.8.2 | 24-May-2006 |
yamt | sync with head.
|
| 1.86.8.1 | 11-Apr-2006 |
yamt | sync with head
|
| 1.86.6.2 | 01-Jun-2006 |
kardel | Sync with head.
|
| 1.86.6.1 | 22-Apr-2006 |
simonb | Sync with head.
|
| 1.86.4.1 | 09-Sep-2006 |
rpaulo | sync with head
|
| 1.94.4.1 | 13-Jul-2006 |
gdamore | Merge from HEAD.
|
| 1.97.4.2 | 10-Dec-2006 |
yamt | sync with head.
|
| 1.97.4.1 | 22-Oct-2006 |
yamt | sync with head
|
| 1.97.2.1 | 18-Nov-2006 |
ad | Sync with head.
|
| 1.100.8.1 | 11-Jul-2007 |
mjf | Sync with head.
|
| 1.100.6.4 | 15-Jul-2007 |
ad | Sync with head.
|
| 1.100.6.3 | 17-Jun-2007 |
ad | - Increase the number of thread priorities from 128 to 256. How the space
is set up is to be revisited.
- Implement soft interrupts as kernel threads. A generic implementation
is provided, with hooks for fast-path MD code that can run the interrupt
threads over the top of other threads executing in the kernel.
- Split vnode::v_flag into three fields, depending on how the flag is
locked (by the interlock, by the vnode lock, by the file system).
- Miscellaneous locking fixes and improvements.
|
| 1.100.6.2 | 13-May-2007 |
ad | - Pass the error number and residual count to biodone(), and let it handle
setting error indicators. Prepare to eliminate B_ERROR.
- Add a flag argument to brelse() to be set into the buf's flags, instead
of doing it directly. Typically used to set B_INVAL.
- Add a "struct cpu_info *" argument to kthread_create(), to be used to
create bound threads. Change "bool mpsafe" to "int flags".
- Allow exit of LWPs in the IDL state when (l != curlwp).
- More locking fixes & conversion to the new API.
|
| 1.100.6.1 | 13-Mar-2007 |
ad | Pull in the initial set of changes for the vmlocking branch.
|
| 1.101.10.1 | 14-Oct-2007 |
yamt | sync with head.
|
| 1.101.8.3 | 23-Mar-2008 |
matt | sync with HEAD
|
| 1.101.8.2 | 09-Jan-2008 |
matt | sync with HEAD
|
| 1.101.8.1 | 06-Nov-2007 |
matt | sync with HEAD
|
| 1.101.6.1 | 26-Oct-2007 |
joerg | Sync with HEAD.
Follow the merge of pmap.c on i386 and amd64 and move
pmap_init_tmp_pgtbl into arch/x86/x86/pmap.c. Modify the ACPI wakeup
code to restore CR4 before jumping back into kernel space as the large
page option might cover that.
|
| 1.103.10.2 | 02-Jan-2008 |
bouyer | Sync with HEAD
|
| 1.103.10.1 | 13-Dec-2007 |
bouyer | Sync with HEAD
|
| 1.103.8.1 | 13-Dec-2007 |
yamt | sync with head.
|
| 1.103.6.4 | 26-Dec-2007 |
ad | Sync with head.
|
| 1.103.6.3 | 19-Dec-2007 |
ad | Use a global lfs_lock.
|
| 1.103.6.2 | 19-Dec-2007 |
ad | Get lfs mostly working.
|
| 1.103.6.1 | 04-Dec-2007 |
ad | Pull the vmlocking changes into a new branch.
|
| 1.103.4.1 | 18-Feb-2008 |
mjf | Sync with HEAD.
|
| 1.106.10.4 | 11-Aug-2010 |
yamt | sync with head.
|
| 1.106.10.3 | 11-Mar-2010 |
yamt | sync with head
|
| 1.106.10.2 | 16-Sep-2009 |
yamt | sync with head
|
| 1.106.10.1 | 16-May-2008 |
yamt | sync with head.
|
| 1.106.8.1 | 18-May-2008 |
yamt | sync with head.
|
| 1.106.6.1 | 02-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.109.4.2 | 03-Jul-2010 |
rmind | sync with head
|
| 1.109.4.1 | 16-Mar-2010 |
rmind | Change struct uvm_object::vmobjlock to be dynamically allocated with
mutex_obj_alloc(). It allows us to share the locks among UVM objects.
|
| 1.109.2.1 | 17-Aug-2010 |
uebayasi | Sync with HEAD.
|
| 1.110.6.1 | 23-Jun-2011 |
cherry | Catchup with rmind-uvmplock merge.
|
| 1.111.8.1 | 17-Mar-2012 |
bouyer | Pull up following revision(s) (requested by perseant in ticket #116):
sys/ufs/lfs/lfs_alloc.c: revision 1.112
tests/fs/vfs/t_rmdirrace.c: revision 1.9
tests/fs/vfs/t_renamerace.c: revision 1.25
sys/ufs/lfs/lfs_vnops.c: revision 1.240
sys/ufs/lfs/lfs_segment.c: revision 1.224
sys/ufs/lfs/lfs_bio.c: revision 1.122
sys/ufs/lfs/lfs_vfsops.c: revision 1.294
sbin/newfs_lfs/make_lfs.c: revision 1.19
sys/ufs/lfs/lfs.h: revision 1.136
Pass t_renamerace and t_rmdirrace tests.
Adapt dholland@'s fix to ufs_rename to fix PR kern/43582. Address several
other MP locking issues discovered during the course of investigating the
same problem.
Removed extraneous vn_lock() calls on the Ifile, since the Ifile writes
are controlled by the segment lock.
Fix PR kern/45982 by deemphasizing the estimate of how much metadata
will fill the empty space on disk when the disk is nearly empty
(t_renamerace crates a lot of inode blocks on a tiny empty disk).
|
| 1.111.6.1 | 18-Feb-2012 |
mrg | merge to -current.
|
| 1.111.2.3 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.111.2.2 | 23-Jan-2013 |
yamt | sync with head
|
| 1.111.2.1 | 17-Apr-2012 |
yamt | sync with head
|
| 1.112.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.112.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.112.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.112.2.1 | 25-Feb-2013 |
tls | resync with head
|
| 1.117.2.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.119.6.5 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.119.6.4 | 05-Oct-2016 |
skrll | Sync with HEAD
|
| 1.119.6.3 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.119.6.2 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.119.6.1 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.131.2.1 | 20-Mar-2017 |
pgoyette | Sync with HEAD
|
| 1.133.2.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.135.6.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.137.10.2 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.137.10.1 | 17-Jan-2020 |
ad | Sync with head.
|
| 1.137.8.1 | 17-Aug-2020 |
martin | Pull up following revision(s) (requested by riastradh in ticket #1050):
sys/ufs/lfs/lfs_subr.c: revision 1.101
sys/ufs/lfs/lfs_subr.c: revision 1.102
sys/ufs/lfs/lfs_inode.c: revision 1.158
sys/ufs/lfs/lfs_inode.h: revision 1.25
sys/ufs/lfs/lfs_balloc.c: revision 1.95
sys/ufs/lfs/lfs_pages.c: revision 1.21
sys/ufs/lfs/lfs_vnops.c: revision 1.330
sys/ufs/lfs/lfs_alloc.c: revision 1.140 (patch)
sys/ufs/lfs/lfs_alloc.c: revision 1.141 (patch)
lib/libp2k/p2k.c: revision 1.72
sys/ufs/lfs/lfs.h: revision 1.205
sys/ufs/lfs/lfs.h: revision 1.206
sys/ufs/lfs/lfs_segment.c: revision 1.284
sys/ufs/lfs/lfs.h: revision 1.207
sys/ufs/lfs/lfs_segment.c: revision 1.285
sys/ufs/lfs/lfs_debug.c: revision 1.55
sys/ufs/lfs/lfs_rename.c: revision 1.23
usr.sbin/dumplfs/dumplfs.c: revision 1.65
sys/ufs/lfs/lfs_vfsops.c: revision 1.371
sys/arch/i386/stand/efiboot/bootx64/Makefile: revision 1.3
sys/ufs/lfs/lfs_vfsops.c: revision 1.372
sys/ufs/lfs/lfs_vfsops.c: revision 1.373
sbin/fsck_lfs/pass1.c: revision 1.46
sys/ufs/lfs/lfs_vnops.c: revision 1.326
sys/ufs/lfs/lfs_vnops.c: revision 1.327
sys/ufs/lfs/lfs_vfsops.c: revision 1.375 (patch)
sys/ufs/lfs/lfs_vnops.c: revision 1.328
sys/ufs/lfs/lfs_subr.c: revision 1.98
sys/ufs/lfs/lfs_extern.h: revision 1.116
sys/ufs/lfs/lfs_vnops.c: revision 1.329
sys/ufs/lfs/lfs_subr.c: revision 1.99
sys/ufs/lfs/lfs_extern.h: revision 1.117
sys/ufs/lfs/lfs_accessors.h: revision 1.49
sys/ufs/lfs/lfs_extern.h: revision 1.118
sys/rump/fs/lib/liblfs/Makefile: revision 1.15
sys/ufs/lfs/lfs_bio.c: revision 1.146 (patch)
sys/ufs/lfs/lfs_bio.c: revision 1.147
sys/ufs/lfs/lfs_subr.c: revision 1.100
Fix kassert in lfs by initializing vp first.
Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
Teach lfs to transition ro<->rw.
Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
Assert lfs_writer where I think we can now prove it.
Serialize access to the splay tree with lfs_lock.
Change some cheap KDASSERT into KASSERT.
Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
Call lfs_orphan in lfs_rename while we're still in the dirop.
lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
Teach LFS_ORPHAN_NEXTFREE about lfs64.
Dust off the orphan detection code and try to make it work.
Fix !DIAGNOSTIC compile
Fix userland references to LFS_ORPHAN_NEXTFREE.
Forgot to grep for these or do a full distribution build, oops!
Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
OR into bp->b_cflags; don't overwrite.
CTASSERT lfs on-disk structure sizes.
Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
Fix clang build after packed lfs64 accessor change.
Suppress spurious address-of-packed error in rump lfs too.
|
| 1.137.4.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.97 | 20-Oct-2025 |
perseant | * Generalize the partial-segment parser introduced for roll-forward,
using it to facilitate an in-kernel segment rewriter (cleaner), and a
mechanism to check whether a segment is in fact empty (only used
with DEBUG).
* Add these new fcntl calls:
- LFCNFILESTATS: For each inode given, report its number of direct
blocks, how many gaps (discontinuities) there are between direct
blocks, and how large the total gap distance is. This will be
useful for a coalescing agent.
- LFCNREWRITEFILE: For each inode given, rewrite its direct blocks,
effectively coalescing it into as compact a form as possible.
- LFCNSCRAMBLE: As above, except that it only rewrites every other
block. This causes the file to have many gaps that can be
measured with LFCNFILESTATS and addressed with LFCNREWRITEFILE,
for testing purposes.
- LFCNREWRITESEGS: Rewrite any live data in the given segments.
This is intended to simplify the cleaner API and facilitate an
in-kernel cleaner.
- LFCNCLEANERINFO: Get the most current CLEANERINFO data from the
kernel.
- LFCNSEGUSE: Retrieve segment usage data from the kernel.
* Vnodes marked IN_CLEANING now take a reference. Add a new "cleaner
lock", which must be taken by the cleaner before the segment lock,
and before marking nodes IN_CLEANING. This allows us to flush
vnodes, if necessary, before the cleaning segment is written, and
never to flush vnodes being cleaned. When the cleaner lock is
released, the vnodes are cleared of IN_CLEANING and the reference
dropped.
* Track a potential infinite loop in lfs_gatherblock.
* Pull "needs to flush" and "needs to wait for flush" into functions
instead of inlining their definitions.
|
| 1.96 | 05-Sep-2020 |
riastradh | Round of uvm.h cleanup.
The poorly named uvm.h is generally supposed to be for uvm-internal
users only.
- Narrow it to files that actually need it -- mostly files that need
to query whether curlwp is the pagedaemon, which should maybe be
exposed by an external header.
- Use uvm_extern.h where feasible and uvm_*.h for things not exposed
by it. We should split up uvm_extern.h but this will serve for now
to reduce the uvm.h dependencies.
- Use uvm_stat.h and #ifdef UVMHIST uvm.h for files that use
UVMHIST(ubchist), since ubchist is declared in uvm.h but the
reference evaporates if UVMHIST is not defined, so we reduce header
file dependencies.
- Make uvm_device.h and uvm_swap.h independently includable while
here.
ok chs@
|
| 1.95 | 23-Feb-2020 |
riastradh | Serialize access to the splay tree with lfs_lock.
|
| 1.94 | 10-Jun-2017 |
maya | branches: 1.94.6; 1.94.10; 1.94.12;
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.93 | 08-Jun-2017 |
chs | move some buffer cache internals declarations from buf.h to vfs_bio.c.
this is needed to avoid name conflicts with ZFS and also
makes it clearer that other code shouldn't be messing with these.
remove the LFS debug code that poked around in bufqueues and
remove the BQ_EMPTY bufqueue since nothing uses it anymore.
provide a function to let LFS and wapbl read the value of nbuf for now.
|
| 1.92 | 06-Apr-2017 |
maya | branches: 1.92.6;
Provide a LFS_ENTER_LOG (__nothing) in the !DEBUG case.
so I can drop lots of #ifdef DEBUG around this macro. NFCI
|
| 1.91 | 07-Aug-2016 |
dholland | branches: 1.91.2;
Fix stupid thinko.
|
| 1.90 | 07-Aug-2016 |
dholland | comments
|
| 1.89 | 07-Aug-2016 |
dholland | use static properly
|
| 1.88 | 10-Oct-2015 |
dholland | branches: 1.88.2;
Use accessors for some more indirect block manipulations.
|
| 1.87 | 01-Sep-2015 |
dholland | Use the lfs dinode accessors in place of the ufs-derived ones.
(Mostly.)
The ufs-derived ones are fake structure member macros, which are gross
and not very safe. Also, it seems that a lot of places in the lfs code
were using the ffsv1 branch of them unconditionally, and this way it's
guaranteed all those places have been updated.
Found while doing this: for non-devices, have getattr produce NODEV
in the rdev field instead of leaking the address of the first direct
block.
|
| 1.86 | 02-Aug-2015 |
dholland | Pass the fs object to LFS_MAX_DADDR so it can check lfs_is64.
Remove some hackish intentional 64->32 truncations next to the checks
using LFS_MAX_DADDR, and tackle the problem they handled in bmap
instead.
The problem: the magic block pointer value UNWRITTEN has magic value
-2, and if it's not handled specifically, uint32 -> uint64 promotion
turns it into 4294967294, which then causes consternation and
monkeyhouse downstream.
What's here is still kind of a hack, but it's a step forward.
|
| 1.85 | 02-Aug-2015 |
dholland | Fix assorted 64 -> 32 truncations in lfs. Also, some minor tidyups and
corrections in passing.
|
| 1.84 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.83 | 24-Jul-2015 |
dholland | More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)
The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.
|
| 1.82 | 24-Jul-2015 |
dholland | Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)
It also gets rid of cpp abuse in the form of fake structure member
macros.
Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)
XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.
|
| 1.81 | 28-Mar-2015 |
maxv | Remove the 'cred' argument from bread(). Remove a now unused var in
ffs_snapshot.c. Update the man page accordingly.
ok hannken@
|
| 1.80 | 28-Jul-2013 |
dholland | branches: 1.80.6;
Add more of the bits for supporting quotas.
|
| 1.79 | 28-Jul-2013 |
dholland | Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.78 | 28-Jul-2013 |
dholland | Add lfs_kernel.h for declarations that don't need to be exposed to userland.
lfs currently has the following headers:
lfs.h - on-disk structures and stuff needed for userlevel tools
lfs_inode.h - additional restricted materials for userlevel tools
that operate the fs (newfs_lfs, fsck_lfs, lfs_cleanerd)
lfs_kernel.h - stuff needed only in the kernel
and the following legacy headers that are expected to be mopped up and
folded into one of the above:
lfs_extern.h - function prototypes
ulfs_bswap.h - endian-independent support
ulfs_dinode.h - now contains very little
ulfs_dirhash.h - dirhash support
ulfs_extattr.h - extattr support
ulfs_extern.h - more function prototypes
ulfs_inode.h - assorted kernel-only declarations
ulfs_quota.h - quota support
ulfs_quota1.h - more quota support
ulfs_quota2.h - more quota support
ulfs_quotacommon.h - more quota support
ulfsmount.h - legacy copy of ufsmount material
|
| 1.77 | 18-Jun-2013 |
christos | branches: 1.77.2;
Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.
|
| 1.76 | 06-Jun-2013 |
dholland | Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.75 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.74 | 06-Jun-2013 |
dholland | Split lfs from ufs step 3: rearrange config stuff.
Add new options:
LFS_EI
LFS_DIRHASH
LFS_EXTATTR
LFS_EXTATTR_AUTOSTART
LFS_QUOTA
LFS_QUOTA2
and update code referring to the corresponding FFS and UFS config
symbols to use the LFS versions. Disable the one extant reference
to APPLE_UFS in the ulfs files. Use opt_lfs.h only, not opt_ffs.h.
|
| 1.73 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.72 | 22-Jan-2013 |
dholland | Stuff UFS_ in front of a few of ufs's symbols to reduce namespace
pollution. Specifically:
ROOTINO -> UFS_ROOTINO
WINO -> UFS_WINO
NXADDR -> UFS_NXADDR
NDADDR -> UFS_NDADDR
NIADDR -> UFS_NIADDR
MAXSYMLINKLEN -> UFS_MAXSYMLINKLEN
MAXSYMLINKLEN_UFS[12] -> UFS[12]_MAXSYMLINKLEN (for consistency)
Sort out ext2fs's misuse of NDADDR and NIADDR; fortunately, these have
the same values in ext2fs and ffs.
No functional change intended.
|
| 1.71 | 20-Dec-2012 |
hannken | Change bread() and breadn() to never return a buffer on
error and modify all callers to not brelse() on error.
Welcome to 6.99.16
PR kern/46282 (6.0_BETA crash: msdosfs_bmap -> pcbmap -> bread -> bio_doread)
|
| 1.70 | 11-Jul-2011 |
hannken | branches: 1.70.2; 1.70.12;
Change VOP_BWRITE() to take a vnode as its first argument like all other
VOPs do. Layered file systems no longer have to modify bp->b_vp and run
into trouble when an async VOP_BWRITE() uses the wrong vnode.
- change all occurences of VOP_BWRITE(bp) to VOP_BWRITE(bp->b_vp, bp).
- remove layer_bwrite().
- welcome to 5.99.55
Adresses PR kern/38762 panic: vwakeup: neg numoutput
No objections from tech-kern@.
|
| 1.69 | 16-Feb-2010 |
mlelstv | Three changes in a single commit.
- drop the notion of frags (LFS fragments) vs fsb (FFS fragments)
The code uses a complicated unity function that just makes the
code difficult to understand.
- support larger sector sizes. Fix disk address computations
to use DEV_BSIZE in the kernel as required by device drivers
and to use sector sizes in userland.
- Fix several locking bugs in lfs_bio.c and lfs_subr.c.
|
| 1.68 | 18-Mar-2009 |
cegger | branches: 1.68.2;
bzero -> memset
|
| 1.67 | 16-May-2008 |
hannken | branches: 1.67.6; 1.67.12;
Make sure all cached buffers with valid, not yet written data have been
run through copy-on-write. Call fscow_run() with valid data where possible.
The LP_UFSCOW hack is no longer needed to protect ffs_copyonwrite() against
endless recursion.
- Add a flag B_MODIFY to bread(), breada() and breadn(). If set the caller
intends to modify the buffer returned.
- Always run copy-on-write on buffers returned from ffs_balloc().
- Add new function ffs_getblk() that gets a buffer, assigns a new blkno,
may clear the buffer and runs copy-on-write. Process possible errors
from getblk() or fscow_run(). Part of PR kern/38664.
Welcome to 4.99.63
Reviewed by: YAMAMOTO Takashi <yamt@netbsd.org>
|
| 1.66 | 28-Apr-2008 |
martin | branches: 1.66.2;
Remove clause 3 and 4 from TNF licenses
|
| 1.65 | 15-Feb-2008 |
ad | branches: 1.65.6; 1.65.8; 1.65.10;
The buffer LOCKED flag need not be under the protection of bufcache_lock,
BUSY is enough.
|
| 1.64 | 02-Jan-2008 |
ad | Merge vmlocking2 to head.
|
| 1.63 | 08-Oct-2007 |
ad | branches: 1.63.4; 1.63.6; 1.63.10;
Merge ffs locking & brelse changes from the vmlocking branch.
|
| 1.62 | 15-Feb-2007 |
ad | branches: 1.62.6; 1.62.18; 1.62.20; 1.62.22;
Replace some uses of lockmgr() / simplelocks.
|
| 1.61 | 14-May-2006 |
elad | integrate kauth.
|
| 1.60 | 07-Apr-2006 |
perseant | Several minor bug fixes:
* Correct (weak) segment lock assertions in lfs_fragextend and lfs_putpages.
* Keep IN_MODIFIED set if we run out of avail in lfs_putpages.
* Don't try to (re)write buffers on a VBLK vnode; fixes a panic I found
while running with an LFS root.
* Raise priority of LFCNSEGWAIT to PVFS; PUSER is way too low for
something the pagedaemon is relying on.
|
| 1.59 | 24-Dec-2005 |
perry | branches: 1.59.4; 1.59.6; 1.59.8; 1.59.10; 1.59.12;
Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.
|
| 1.58 | 11-Dec-2005 |
christos | merge ktrace-lwp.
|
| 1.57 | 02-Nov-2005 |
yamt | merge yamt-vop branch. remove following VOPs.
VOP_BLKATOFF
VOP_VALLOC
VOP_BALLOC
VOP_REALLOCBLKS
VOP_VFREE
VOP_TRUNCATE
VOP_UPDATE
|
| 1.56 | 19-Apr-2005 |
perseant | branches: 1.56.2; 1.56.4;
Keep per-inode, per-fs, and subsystem-wide counts of blocks allocated through
lfs_balloc(), and use that to estimate the number of dirty pages belonging
to LFS (subsystem or filesystem). This is almost certainly wrong for
the case of a large mmap()ed region, but the accounting is tighter than
what we had before, and performs much better in the typical case of pages
dirtied through write().
|
| 1.55 | 16-Apr-2005 |
perseant | Use splay trees, rather than a hash table, to manage the accounting of
blocks allocated through VOP_BALLOC() for pages to be written to disk.
This accounting no longer takes a noticeable fraction of the system CPU.
|
| 1.54 | 14-Apr-2005 |
perseant | Tabify leading whitespace
|
| 1.53 | 14-Apr-2005 |
perseant | Consolidate the hash table we use to maintain the integrity of lfs_avail
into a single, system-wide table, rather than having a separate hash table
per inode. Significantly reduces the "system" cpu usage of your average
file write.
|
| 1.52 | 01-Apr-2005 |
perseant | Protect various per-fs structures with fs->lfs_interlock simple_lock, to
improve behavior in the multiprocessor case. Add debugging segment-lock
assertion statements.
|
| 1.51 | 02-Mar-2005 |
perseant | branches: 1.51.2;
Put the ISSPACE() check where it belongs. This allows rewriting a file
on a full filesystem while still returning ENOSPC on an attempt to allocate
new blocks.
|
| 1.50 | 26-Feb-2005 |
perry | nuke trailing whitespace
|
| 1.49 | 26-Feb-2005 |
perseant | Various minor LFS improvements:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statvfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
|
| 1.48 | 25-Jan-2004 |
hannken | branches: 1.48.6; 1.48.8; 1.48.10;
Make VOP_STRATEGY(bp) a real VOP as discussed on tech-kern.
VOP_STRATEGY(bp) is replaced by one of two new functions:
- VOP_STRATEGY(vp, bp) Call the strategy routine of vp for bp.
- DEV_STRATEGY(bp) Call the d_strategy routine of bp->b_dev for bp.
DEV_STRATEGY(bp) is used only for block-to-block device situations.
|
| 1.47 | 30-Dec-2003 |
pk | Replace the traditional buffer memory management -- based on fixed per buffer
virtual memory reservation and a private pool of memory pages -- by a scheme
based on memory pools.
This allows better utilization of memory because buffers can now be allocated
with a granularity finer than the system's native page size (useful for
filesystems with e.g. 1k or 2k fragment sizes). It also avoids fragmentation
of virtual to physical memory mappings (due to the former fixed virtual
address reservation) resulting in better utilization of MMU resources on some
platforms. Finally, the scheme is more flexible by allowing run-time decisions
on the amount of memory to be used for buffers.
On the other hand, the effectiveness of the LRU queue for buffer recycling
may be somewhat reduced compared to the traditional method since, due to the
nature of the pool based memory allocation, the actual least recently used
buffer may release its memory to a pool different from the one needed by a
newly allocated buffer. However, this effect will kick in only if the
system is under memory pressure.
|
| 1.46 | 29-Oct-2003 |
mycroft | Adjust to remove bogus initializer.
|
| 1.45 | 25-Oct-2003 |
christos | Fix uninitialized variable warnings.
|
| 1.44 | 04-Sep-2003 |
yamt | don't call LFS_DEBUG_COUNTLOCKED after bread().
lfs_countlocked doesn't count buffers that isn't on the freelist.
|
| 1.43 | 07-Aug-2003 |
agc | Move UCB-licensed code from 4-clause to 3-clause licence.
Patches provided by Joel Baker in PR 22364, verified by myself.
|
| 1.42 | 18-May-2003 |
yamt | branches: 1.42.2;
make is_sequential a callback in order to achieve better lfs write clustering.
since lfs always rewrite blocks into the new segment,
current on-disk place of the block doesn't affect to write clustering.
ok'ed by Konrad Schroder.
|
| 1.41 | 29-Apr-2003 |
yamt | add an assertion.
|
| 1.40 | 02-Apr-2003 |
fvdl | Add support for UFS2. UFS2 is an enhanced FFS, adding support for
64 bit block pointers, extended attribute storage, and a few
other things.
This commit does not yet include the code to manipulate the extended
storage (for e.g. ACLs), this will be done later.
Originally written by Kirk McKusick and Network Associates Laboratories for
FreeBSD.
|
| 1.39 | 15-Mar-2003 |
perseant | Add simple_lock protection for lfs_seglock and lfs_subsys_pages; these will
be expanded to cover other per-fs and subsystem-wide data as well.
Fix a case of IN_MODIFIED being set without updating lfs_uinodes, resulting
in a "lfs_uinodes < 0" panic.
Fix a deadlock in lfs_putpages arising from the need to busy all pages in a
block; unbusy any that had already been busied before starting over.
|
| 1.38 | 28-Feb-2003 |
perseant | Fix a clrbuf() on an uninitialized pointer.
|
| 1.37 | 20-Feb-2003 |
perseant | Tabify, and fix some comment alignment problems.
|
| 1.36 | 17-Feb-2003 |
perseant | Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now
(there are still some details to work out) but expect that to go
away soon. To support these basic changes (creation of lfs_putpages,
lfs_gop_write, mods to lfs_balloc) several other changes were made, to
wit:
* Create a writer daemon kernel thread whose purpose is to handle page
writes for the pagedaemon, but which also takes over some of the
functions of lfs_check(). This thread is started the first time an
LFS is mounted.
* Add a "flags" parameter to GOP_SIZE. Current values are
GOP_SIZE_READ, meaning that the call should return the size of the
in-core version of the file, and GOP_SIZE_WRITE, meaning that it
should return the on-disk size. One of GOP_SIZE_READ or
GOP_SIZE_WRITE must be specified.
* Instead of using malloc(...M_WAITOK) for everything, reserve enough
resources to get by and use malloc(...M_NOWAIT), using the reserves if
necessary. Use the pool subsystem for structures small enough that
this is feasible. This also obsoletes LFS_THROTTLE.
And a few that are not strictly necessary:
* Moves the LFS inode extensions off onto a separately allocated
structure; getting closer to LFS as an LKM. "Welcome to 1.6O."
* Unified GOP_ALLOC between FFS and LFS.
* Update LFS copyright headers to correct values.
* Actually cast to unsigned in lfs_shellsort, like the comment says.
* Keep track of which segments were empty before the previous
checkpoint; any segments that pass two checkpoints both dirty and
empty can be summarily cleaned. Do this. Right now lfs_segclean
still works, but this should be turned into an effectless
compatibility syscall.
|
| 1.35 | 24-Jan-2003 |
fvdl | Bump daddr_t to 64 bits. Replace it with int32_t in all places where
it was used on-disk, so that on-disk formats remain the same.
Remove ufs_daddr_t and ufs_lbn_t for the time being.
|
| 1.34 | 11-Dec-2002 |
yamt | take care of B_CLRBUF in lfs_balloc.
otherwise you'll see uninitialized blocks.
|
| 1.33 | 06-Jul-2002 |
perseant | Deal with fragment size changes better. For each fragment that can
exist on an on-disk inode, we keep a record of its size in struct inode,
which is updated when we write the block to disk. The cleaner routines
thus have ready access to what size is the correct size for this block,
on disk.
Fixed a related bug: if a file with fragments is being cleaned
(fragments being cleaned) at the same time it is being extended beyond
NDADDR blocks, we could write a bogus FINFO record that has a frag in the
middle; when it was cleaned this would give back bogus file data. Don't
write the indirect blocks in this case, since there is no need.
lfs_fragextend and lfs_truncate no longer require the seglock, but instead
take a shared lock, which the seglock locks exclusively.
|
| 1.32 | 14-May-2002 |
perseant | branches: 1.32.2;
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes
I found while making sure there weren't any new ones.
* Make the write clusters keep track of the buffers whose blocks they contain.
This should make it possible to (1) write clusters using a page mapping
instead of malloc, if desired, and (2) schedule blocks for rewriting
(somewhere else) if a write error occurs. Code is present to use
pagemove() to construct the clusters but that is untested and will go away
anyway in favor of page mapping.
* DEBUG now keeps a log of Ifile writes, so that any lingering instances of
the "dirty bufs" problem can be properly debugged.
* Keep track of whether the Ifile has been dirtied by various routines that
can be called by lfs_segwrite, and loop on that until it is clean, for
a checkpoint. Checkpoints need to be squeaky clean.
* Warn the user (once) if the Ifile grows larger than is reasonable for their
buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can
grow.
* If an inode is not found in a disk block, try rereading the block, under
the assumption that the block was copied to a cluster and then freed.
* Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
|
| 1.31 | 23-Nov-2001 |
chs | add spaces for KNF. confirmed to produce identical objects.
|
| 1.30 | 08-Nov-2001 |
lukem | add RCSID
|
| 1.29 | 13-Jul-2001 |
perseant | branches: 1.29.4;
Merge the short-lived perseant-lfsv2 branch into the trunk.
Kernels and tools understand both v1 and v2 filesystems; newfs_lfs
generates v2 by default. Changes for the v2 layout include:
- Segments of non-PO2 size and arbitrary block offset, so these can be
matched to convenient physical characteristics of the partition (e.g.,
stripe or track size and offset).
- Address by fragment instead of by disk sector, paving the way for
non-512-byte-sector devices. In theory fragments can be as large
as you like, though in reality they must be smaller than MAXBSIZE in size.
- Use serial number and filesystem identifier to ensure that roll-forward
doesn't get old data and think it's new. Roll-forward is enabled for
v2 filesystems, though not for v1 filesystems by default.
- The inode free list is now a tailq, paving the way for undelete (undelete
is not yet implemented, but can be without further non-backwards-compatible
changes to disk structures).
- Inode atime information is kept in the Ifile, instead of on the inode;
that is, the inode is never written *just* because atime was changed.
Because of this the inodes remain near the file data on the disk, rather
than wandering all over as the disk is read repeatedly. This speeds up
repeated reads by a small but noticeable amount.
Other changes of note include:
- The ifile written by newfs_lfs can now be of arbitrary length, it is no
longer restricted to a single indirect block.
- Fixed an old bug where ctime was changed every time a vnode was created.
I need to look more closely to make sure that the times are only updated
during write(2) and friends, not after-the-fact during a segment write,
and certainly not by the cleaner.
|
| 1.28 | 30-May-2001 |
mrg | branches: 1.28.2; 1.28.4;
use _KERNEL_OPT
|
| 1.27 | 21-Nov-2000 |
perseant | branches: 1.27.2;
More locked_queue_* and lfs_avail accounting fixes from Jesse Off
<joff@gci-net.com>. Remove a specious btodb() in lfs_fragextend, and
count blocks shrunk or removed by VOP_TRUNCATE in lfs_avail.
|
| 1.26 | 17-Nov-2000 |
perseant | Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes.
(PR #11468). In the case of fragment allocation, check to see if enough
space is available before extending a fragment already scheduled for writing.
The locked_queue_* variables indicate the number of buffer headers and bytes,
respectively, that are unavailable to getnewbuf() because they are locked up
waiting for LFS to flush them; make sure that that is actually what we're
counting, i.e., never count malloced buffers, and always use b_bufsize instead
of b_bcount.
If DEBUG is defined, the periodic calls to lfs_countlocked will now complain
if either counter is incorrect. (In the future lfs_countlocked will not need
to be called at all if DEBUG is not defined.)
|
| 1.25 | 09-Sep-2000 |
perseant | Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
|
| 1.24 | 04-Jul-2000 |
perseant | Fix errors observed while trying to fill the filesystem with yesterday's
fixes:
- Write copies of bfree and avail in the CLEANERINFO block, so the
cleaner doesn't have to guess which superblock has the current
information (if indeed any do).
- Tighten up accounting of lfs_avail (more needs to be done).
- When cleansing indirect blocks of UNWRITTEN, make sure not to mark
them clean, since they'll need to be rewritten later.
|
| 1.23 | 03-Jul-2000 |
perseant | Fix i_ffs_blocks in fragment extension case where fragment has not yet
been written to disk.
|
| 1.22 | 03-Jul-2000 |
perseant | i_lfs_effnblks fixes. Put debugging printfs under #ifdef DEBUG_LFS.
|
| 1.21 | 03-Jul-2000 |
perseant | Allow the number of free segments reserved for the cleaner to be
parametrized in the filesystem, defaulting to MIN_FREE_SEGS = 2 but set
to something more reasonable at newfs_lfs time.
Note the number of blocks that have been scheduled for writing but which
are not yet on disk in an inode extension, i_lfs_effnblks. Move
i_ffs_effnlink out of the ffs extension and onto the main inode, since
it's used all over the shared code and the lfs extension would clobber
it.
At inode write time, indirect blocks and inode-held blocks of inodes
that have i_lfs_effnblks != i_ffs_blocks are cleansed of UNWRITTEN disk
addresses, so that these never make it to disk.
|
| 1.20 | 28-Jun-2000 |
mrg | remove include of <vm/vm.h> and <uvm/uvm_extern.h>
|
| 1.19 | 27-Jun-2000 |
perseant | Fixes associated with filling an LFS:
Change the space computation to appear to change the size of the *disk*
rather than the *bytes used* when more segment summaries and inode
blocks are written. Try to estimate the amount of space that these will
take up when more files are written, so the disk size doesn't change too
much.
Regularize error returns from lfs_valloc, lfs_balloc, lfs_truncate: they
now fail entirely, rather than succeeding half-way and leaving the fs in
an inconsistent state.
Rewrite lfs_truncate, mostly stealing from ffs_truncate. The old
lfs_truncate had difficulty truncating a large file to a non-zero size
(indirect blocks were not handled appropriately).
Unmark VDIROP on fvp after ufs_remove, ufs_rmdir, so these can be
reclaimed immediately: this vnode would not be written to disk again
anyway if the removal succeeded, and if it failed, no directory
operation occurred.
ufs_makeinode and ufs_mkdir now remove IN_ADIROP on error.
|
| 1.18 | 06-Jun-2000 |
perseant | branches: 1.18.2;
Protect inode free list with seglock, instead of separate lock, so that
the head of the inode free list (on the superblock) always matches the
rest of the free list (in the ifile).
Protect lfs_fragextend with seglock, to prevent the segment byte count
fudging from making its way to disk.
Don't try to inactivate dirop vnodes that are still in the middle of
their dirop (may address PR#10285).
|
| 1.17 | 30-May-2000 |
perseant | Don't try to "correct" accounting for fragments being extended but which
have never been written to disk.
|
| 1.16 | 05-May-2000 |
perseant | branches: 1.16.2;
Change the way LFS does block accounting, from trying to infer from the
buffer cache flags, to marking the inode and/or indirect blocks with a
special disk address UNWRITTEN==-2 when a block is accounted for. (This
address is never written to disk, but only used in-core. This is essentially
the same method of block accounting as on the UBC branch, where the buffer
headers don't exist.) Make sure that truncation is handled properly,
especially in the case of holey files.
Fixes PR#9994.
|
| 1.15 | 23-Apr-2000 |
perseant | Fix problems outlined in PR#9926:
- lfs_truncate extends the file if called with length > i_ffs_size;
- lfs_truncate errors out if called with length < 0;
- lfs_balloc block accounting corrected for the case of blocks read
into the cache before they exist on disk;
- mp->mnt_stat.f_iosize is initialized in lfs_mountfs.
|
| 1.14 | 15-Nov-1999 |
fvdl | Add Kirk McKusick's soft updates code to the trunk. Not enabled by
default, as the copyright on the main file (ffs_softdep.c) is such
that is has been put into gnusrc. options SOFTDEP will pull this
in. This code also contains the trickle syncer.
Bump version number to 1.4O
|
| 1.13 | 15-Jun-1999 |
perseant | branches: 1.13.2; 1.13.4; 1.13.8;
Minor changes to the segment live bytes calculation. In particular, fixed
a bug in fragment extension that could run the count negative. Also, don't
overcount for inodes, and don't count segment summaries. Thus, for empty
segments the live bytes count should now be exactly zero.
|
| 1.12 | 24-Mar-1999 |
mrg | branches: 1.12.2; 1.12.4; 1.12.6;
completely remove Mach VM support. all that is left is the all the
header files as UVM still uses (most of) these.
|
| 1.11 | 10-Mar-1999 |
perseant | New sources should leave the LFS in a more-or-less working state. Changes
include:
- DIROP segregation is enabled, and greater care is taken
to make sure that a checkpoint completes. Fsck is not
needed to remount the filesystem.
- Several checks to make sure that the LFS subsystem does not
overuse various resources (memory, in particular).
- The cleaner routines, lfs_markv in particular, are completely
rewritten. A buffer overflow is removed. Greater care is taken
to ensure that inodes come from where lfs_cleanerd say they come
from (so we know nothing has changed since lfs_bmapv was called).
- Fragment allocation is fixed, so that writes beyond end-of-file
do the right thing.
|
| 1.10 | 09-Nov-1998 |
mycroft | GC the B_CACHE bit.
|
| 1.9 | 09-Jun-1998 |
scottr | Protect various config(8)-generated files from inclusion while
building LKMs. Fixes PR 5557.
|
| 1.8 | 08-Jun-1998 |
scottr | Use the newly-defined opt_quota.h.
|
| 1.7 | 03-Mar-1998 |
drochner | Don't cast the quad_t file size to u_long, this can cause overflows.
|
| 1.6 | 03-Mar-1998 |
fvdl | Make this compile again with UVM
|
| 1.5 | 01-Mar-1998 |
fvdl | Merge with Lite2 + local changes
|
| 1.4 | 11-Jun-1997 |
bouyer | Add support for ext2fs, this needed a few modifications to ufs/ufs/inode.h:
- added an "union inode_ext" to struct inode, for the per-fs extentions.
For now only ext2fs uses it.
- i_din is now an union:
union {
struct dinode ffs_din; /* 128 bytes of the on-disk dinode. */
struct ext2fs_dinode e2fs_din; /* 128 bytes of the on-disk dinode. */
} i_din
Added a lot of #define i_ffs_* and i_e2fs_* to access the fields.
- Added two macros: FFS_ITIMES and EXT2FS_ITIMES. ITIMES calls the rigth
macro, depending on the time of the inode. ITIMES is used where necessary,
FFS_ITIMES and EXT2FS_ITIMES in other places.
|
| 1.3 | 09-Feb-1996 |
christos | lfs prototypes
|
| 1.2 | 29-Jun-1994 |
cgd | New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
|
| 1.1 | 08-Jun-1994 |
mycroft | branches: 1.1.1;
Update to 4.4-Lite fs code, with local changes.
|
| 1.1.1.2 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite2
|
| 1.1.1.1 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite for reference
|
| 1.12.6.1 | 30-Nov-1999 |
itojun | bring in latest KAME (as of 19991130, KAME/NetBSD141) into kame branch
just for reference purposes.
This commit includes 1.4 -> 1.4.1 sync for kame branch.
The branch does not compile at all (due to the lack of ALTQ and some other
source code). Please do not try to modify the branch, this is just for
referenre purposes.
synchronization to latest KAME will take place on HEAD branch soon.
|
| 1.12.4.3 | 31-Aug-1999 |
perseant | Rudimentary support for LFS under UBC:
- LFS-specific VOP_BALLOC and VOP_PUTPAGES vnode ops.
- getblk VREG panic #ifdef'd out (can be reinstated when Ifile is
internalized and Ifile can be made another type from VREG)
- interface to VOP_PUTPAGES changed to pass all pager flags, not
just sync. FS putpages routines must know about the pager flags.
- new LFS magic disk address, -2 ("unwritten"), meaning accounted for
but not assigned to a fixed disk location (since LFS does these two
things separately, and the previous accounting method using buffer
headers no longer will work). Changed references to (foo == (daddr_t)-1)
to (foo < 0). Since disk drivers reject all addresses < 0, this should
not present a problem for other FSs.
|
| 1.12.4.2 | 04-Jul-1999 |
chs | a couple steps towards supporting UBC.
|
| 1.12.4.1 | 21-Jun-1999 |
thorpej | Sync w/ -current.
|
| 1.12.2.1 | 25-Jun-1999 |
perry | pullup 1.12->1.13 (perseant)
|
| 1.13.8.1 | 27-Dec-1999 |
wrstuden | Pull up to last week's -current.
|
| 1.13.4.1 | 19-Oct-1999 |
fvdl | Bring in Kirk McKusick's FFS softdep code on a branch.
|
| 1.13.2.2 | 22-Nov-2000 |
bouyer | Sync with HEAD.
|
| 1.13.2.1 | 20-Nov-2000 |
bouyer | Update thorpej_scsipi to -current as of a month ago
|
| 1.16.2.1 | 22-Jun-2000 |
minoura | Sync w/ netbsd-1-5-base.
|
| 1.18.2.2 | 03-Feb-2001 |
he | Pull up revisions 1.26-1.27 (requested by perseant):
o Close up accounting holes in LFS' accounting of immediately-
available-space, number of clean segments, and amount of dirty
space taken up by metadata (PR#11468, PR#11470, PR#11534).
|
| 1.18.2.1 | 14-Sep-2000 |
perseant | Pull up recent LFS kernel changes (approved by thorpej):
ufs/ufs/inode.h, 1.20--1.22 (add i_lfs_effnblks extension ;
make ITIMES aware of LFS_ITIMES;
_LKM protection so userland progs
compile)
ufs/ufs/ufs_vnops.c, 1.69, 1.71 (remove IN_ADIROP;
use ITIMES instead of FFS_ITIMES)
ufs/ufs/ufs_readwrite.c, 1.27 (use lfs_reserve in lfs_write)
ufs/lfs/lfs.h, 1.26--1.32 (define LFS_EST_* macros ;
change MIN_FREE_SEGS to lfs_minfreesegs ;
add avail and bfree to CLEANERINFO ;
change lfs_uinodes to signed ;
change lfs_dmeta to signed ;
add whitespace to line up structure
members ;
explicit cast to int32_t in LFS_EST_*
macros)
ufs/lfs/lfs_alloc.c, back out 1.34.2.3 (pullups of 1.39, 1.40);
then pull up 1.38 (clean up on error)
1.39--1.43 (restore fvdl's ufs_hashlock fix ;
restore fvdl's ufs_hashlock fix ;
set i_lfs_effnblks ;
use UINO macros ;
add comments and fix long lines)
ufs/lfs/lfs_balloc.c, 1.19 (don't succeed halfway)
1.21--1.25 (use i_lfs_effnblks ;
fix i_lfs_effnblks computation and
quieten ;
fix i_ffs_blocks in unwritten fragment ;
remove useless debugging check ;
add comments and (c) 2000)
ufs/lfs/lfs_bio.c, 1.24--1.30 (cleanup and make lfs_flush_fs take
"struct lfs *" instead of "struct
mount *" ;
use lfs_minfreeseg instead of
MIN_FREE_SEGS ;
use UINO macros, and copy bfree/avail
to CLEANERINFO ;
add lfs_reserve function ;
1.28--1.30 fix printf formatting)
ufs/lfs/lfs_cksum.c, 1.13 (add (c) 2000)
ufs/lfs/lfs_debug.c, 1.11 (use btodb instead of DEV_BSIZE)
ufs/lfs/lfs_extern.h, 1.18, 1.20--1.21 (function prototype changes)
ufs/lfs/lfs_inode.c, 1.38 (rewrite lfs_truncate from
ffs_truncate)
1.40--1.44 (count written and unwritten blocks
seperately ;
use disk block units instead of bytes ;
remove unnecessary "mod" variable ;
correct B_DELWRI to avoid bawrite panic ;
use lfs_reserve)
ufs/lfs/lfs_segment.c, 1.52-1.59 (use lfs_dmeta to note used summaries ;
check for UNWRITTEN in indirect blocks ;
more debugging stuff inside #ifdef
DEBUG_LFS ;
use LK_CANRECURSE ;
don't drop dirty indirect blocks ;
use UINO macros ;
don't hose the free list ;
use btodb() instead of DEV_BSIZE ;
make it compile again (oops))
ufs/lfs/lfs_subr.c, 1.16--1.17 (check for locked inodes before
changing ;
use btodb() instead of DEV_BSIZE, (c)
2000)
ufs/lfs/lfs_syscalls.c, back out 1.41.4.2 (fvdl's ufs_hashlock fix);
then pull up 1.43 (use lfs_dmeta)
1.44--1.45 (restore fvdl's ufs_hashlock fix)
1.46--1.47 (fix lfs_avail leakage from sblock
segments ;
use UINO macros)
1.49 (bounds-check inode numbers in
lfs_markv)
ufs/lfs/lfs_vfsops.c, 1.53 (use LFS_EST_* macros in lfs_statfs)
1.56--1.58 (initialize lfs_minfreeseg, lfs_effnblk ;
initialize lfs_uinodes ;
initialize lfs_ravail)
ufs/lfs/lfs_vnops.c, 1.40 (remove VDIROP from removed files)
1.42--1.44 (move SET_ENDOP below the removal of
VDIROP ;
use UINO macros and add lfs_itimes
function ;
use lfs_reserve in dirops)
|
| 1.27.2.8 | 11-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.27.2.7 | 01-Aug-2002 |
nathanw | Catch up to -current.
|
| 1.27.2.6 | 20-Jun-2002 |
nathanw | Catch up to -current.
|
| 1.27.2.5 | 08-Jan-2002 |
nathanw | Catch up to -current.
|
| 1.27.2.4 | 14-Nov-2001 |
nathanw | Catch up to -current.
|
| 1.27.2.3 | 24-Aug-2001 |
nathanw | Catch up with -current.
|
| 1.27.2.2 | 21-Jun-2001 |
nathanw | Catch up to -current.
|
| 1.27.2.1 | 05-Mar-2001 |
nathanw | Initial commit of scheduler activations and lightweight process support.
|
| 1.28.4.4 | 06-Sep-2002 |
jdolecek | sync kqueue branch with HEAD
|
| 1.28.4.3 | 23-Jun-2002 |
jdolecek | catch up with -current on kqueue branch
|
| 1.28.4.2 | 10-Jan-2002 |
thorpej | Sync kqueue branch with -current.
|
| 1.28.4.1 | 03-Aug-2001 |
lukem | update to -current
|
| 1.28.2.2 | 02-Jul-2001 |
perseant | Change disk addressing unit to be the fragment, instead of the disk sector.
All quantities in the superblock, inodes, indirect blocks, etc. refer now
to this abstract unit (called "fsb" as it is in FFS) instead of disk sectors;
as a consequence segment summary blocks have to be multiples of a fragment in
size. In v1 filesystems, compatibility code ensures that 1 fsb == 1 sector,
regardless of fragment size.
Fragments can now range in size between 512 and 32k; in the event that
LFS_LABELPAD (8k) is smaller than the disk address unit size, an extra
proto-superblock is kept at 8k from the beginning of the disk, to be used
*only* to locate the real superblocks. (Not all of the userland knows about
this yet.)
Almost all of this was done not by me, but by joff.
|
| 1.28.2.1 | 29-Jun-2001 |
perseant | Get rid of __P(), protoizing where it had not already been done
|
| 1.29.4.1 | 12-Nov-2001 |
thorpej | Sync the thorpej-mips-cache branch with -current.
|
| 1.32.2.1 | 15-Jul-2002 |
gehenna | catch up with -current.
|
| 1.42.2.5 | 10-Nov-2005 |
skrll | Sync with HEAD. Here we go again...
|
| 1.42.2.4 | 04-Mar-2005 |
skrll | Sync with HEAD.
Hi Perry!
|
| 1.42.2.3 | 21-Sep-2004 |
skrll | Fix the sync with head I botched.
|
| 1.42.2.2 | 18-Sep-2004 |
skrll | Sync with HEAD.
|
| 1.42.2.1 | 03-Aug-2004 |
skrll | Sync with HEAD
|
| 1.48.10.1 | 19-Mar-2005 |
yamt | sync with head. xen and whitespace. xen part is not finished.
|
| 1.48.8.1 | 29-Apr-2005 |
kent | sync with -current
|
| 1.48.6.1 | 10-May-2005 |
riz | Pull up the following revisions (requested by perseant in ticket #1281):
1.8 sys/ufs/lfs/TODO
1.75 sys/ufs/lfs/lfs.h (via patch)
1.74 sys/ufs/lfs/lfs_alloc.c (via patch)
1.49, 1.51 sys/ufs/lfs/lfs_balloc.c (1.51 via patch)
1.78 sys/ufs/lfs/lfs_bio.c
1.62 sys/ufs/lfs/lfs_extern.h (via patch)
1.156 sys/ufs/lfs/lfs_segment.c (via patch)
1.48 sys/ufs/lfs/lfs_subr.c
1.101 sys/ufs/lfs/lfs_syscalls.c
1.163 sys/ufs/lfs/lfs_vfsops.c (via patch)
1.134 sys/ufs/lfs/lfs_vnops.c (via patch)
1.61 sys/ufs/ufs/ufs_readwrite.c (via patch)
1.20 libexec/lfs_cleanerd/clean.h (via patch)
1.52 libexec/lfs_cleanerd/cleanerd.c (via patch)
1.41 libexec/lfs_cleanerd/library.c (via patch)
1.4 regress/sys/fs/lfs/newfs_fsck/Makefile
1.2 regress/sys/fs/lfs/newfs_fsck/mkfs_mount
1.2 regress/sys/fs/lfs/newfs_fsck/smallfiles
1.3 sbin/fsck_lfs/bufcache.c
1.3 sbin/fsck_lfs/bufcache.h
1.3 sbin/fsck_lfs/lfs.h
1.8 sbin/fsck_lfs/lfs.c (via patch)
1.8 sbin/fsck_lfs/pass3.c (via patch)
1.18 sbin/fsck_lfs/pass0.c (via patch)
1.18 sbin/fsck_lfs/utilities.c (via patch)
1.7 sbin/fsck_lfs/segwrite.c
1.19 sbin/fsck_lfs/setup.c (via patch)
1.3 sbin/newfs_lfs/Makefile
0 sbin/newfs_lfs/lfs.c (yes, remove it)
1.1 sbin/newfs_lfs/make_lfs.c
1.15 sbin/newfs_lfs/newfs.c (via patch)
Various minor LFS improvements.
Kernel:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this. Should fix PR #29045.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
Fixes PR #26680.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
cleaner:
* Adapt lfs_cleanerd to use the fcntl call to get the Ifile filehandle,
so it need not be in the namespace.
* Make lfs_cleanerd be more careful when there are very few available
segments.
* Make lfs_cleanerd less verbose when the filesystem is unmounted.
newfs_lfs, fsck_lfs, and regression:
* Extend the lfs library from fsck_lfs(8) so that it can be used with a
not-yet-existent LFS. Make newfs_lfs(8) use this library, so it can
create LFSs whose Ifile is larger than one segment. Addresses PR #11110.
* Make newfs_lfs(8) use strsuftoi64() for its arguments, a la newfs(8).
* Make fsck_lfs(8) respect the "file system is clean" flag.
* Don't let fsck_lfs(8) think it has dirty blocks when invoked with the
-n flag.
* Remove the Ifile from the filesystem namespace. The cleaner now uses
a fcntl call on the root inode to find the Ifile filehandle. (As a
side-effect, addresses PR #29144.)
|
| 1.51.2.2 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_balloc.c: revision 1.60
sys/ufs/lfs/lfs_syscalls.c: revision 1.111
sys/ufs/lfs/lfs_segment.c: revision 1.172
sys/ufs/lfs/lfs_vnops.c: revision 1.163
Several minor bug fixes:
* Correct (weak) segment lock assertions in lfs_fragextend and lfs_putpages.
* Keep IN_MODIFIED set if we run out of avail in lfs_putpages.
* Don't try to (re)write buffers on a VBLK vnode; fixes a panic I found
while running with an LFS root.
* Raise priority of LFCNSEGWAIT to PVFS; PUSER is way too low for
something the pagedaemon is relying on.
|
| 1.51.2.1 | 07-May-2005 |
tron | Apply patch (requested by perseant in ticket #242):
* fsck_lfs buffer cache fixes, including PR #29151
* Change fsck_lfs phase 0 message to reflect reality
* fsck_lfs: check phase 5 (cleanerinfo accounting) even on
roll-forward
* Keep better track of the free list during roll-forward, avoiding
a core dump
* Improve hash table use for fsck_lfs buffer and vnode cache
* Document fsck_lfs flag -f, and implement -q
* Add resize_lfs, including kernel support
* Add LFS to mountd's list of exportable filesystem types
* Make the LFS lkm work again [christos@]
* Add MP locking to the LFS kernel subsystem
* Fix pager_map deadlock in lfs_putpages()
* Avoid incomplete file extension that looks like "partial
truncation" to fsck
* Use lfs_malloc for cleaner malloc, since the cleaner often runs
in low-memory conditions.
* Use splay trees, not hash table, to track page allocation for
write.
* Fix mkdir panic on full fs
* Fix page accounting leak by counting differently.
* Use rightly named structure for lfs_getattr [skrll@]
* Cosmetic changes for readability.
|
| 1.56.4.1 | 20-Oct-2005 |
yamt | adapt ufs.
|
| 1.56.2.5 | 27-Feb-2008 |
yamt | sync with head.
|
| 1.56.2.4 | 21-Jan-2008 |
yamt | sync with head
|
| 1.56.2.3 | 27-Oct-2007 |
yamt | sync with head.
|
| 1.56.2.2 | 26-Feb-2007 |
yamt | sync with head.
|
| 1.56.2.1 | 21-Jun-2006 |
yamt | sync with head.
|
| 1.59.12.1 | 24-May-2006 |
tron | Merge 2006-05-24 NetBSD-current into the "peter-altq" branch.
|
| 1.59.10.3 | 06-May-2006 |
christos | - Move kauth_cred_t declaration to <sys/types.h>
- Cleanup struct ucred; forward declarations that are unused.
- Don't include <sys/kauth.h> in any header, but include it in the c files
that need it.
Approved by core.
|
| 1.59.10.2 | 19-Apr-2006 |
elad | sync with head.
|
| 1.59.10.1 | 08-Mar-2006 |
elad | Adapt to kernel authorization KPI.
|
| 1.59.8.2 | 24-May-2006 |
yamt | sync with head.
|
| 1.59.8.1 | 11-Apr-2006 |
yamt | sync with head
|
| 1.59.6.2 | 01-Jun-2006 |
kardel | Sync with head.
|
| 1.59.6.1 | 22-Apr-2006 |
simonb | Sync with head.
|
| 1.59.4.1 | 09-Sep-2006 |
rpaulo | sync with head
|
| 1.62.22.1 | 14-Oct-2007 |
yamt | sync with head.
|
| 1.62.20.3 | 23-Mar-2008 |
matt | sync with HEAD
|
| 1.62.20.2 | 09-Jan-2008 |
matt | sync with HEAD
|
| 1.62.20.1 | 06-Nov-2007 |
matt | sync with HEAD
|
| 1.62.18.1 | 26-Oct-2007 |
joerg | Sync with HEAD.
Follow the merge of pmap.c on i386 and amd64 and move
pmap_init_tmp_pgtbl into arch/x86/x86/pmap.c. Modify the ACPI wakeup
code to restore CR4 before jumping back into kernel space as the large
page option might cover that.
|
| 1.62.6.3 | 24-Aug-2007 |
ad | Sync with buffer cache locking changes. See buf.h/vfs_bio.c for details.
Some minor portions are incomplete and needs to be verified as a whole.
|
| 1.62.6.2 | 13-May-2007 |
ad | - Pass the error number and residual count to biodone(), and let it handle
setting error indicators. Prepare to eliminate B_ERROR.
- Add a flag argument to brelse() to be set into the buf's flags, instead
of doing it directly. Typically used to set B_INVAL.
- Add a "struct cpu_info *" argument to kthread_create(), to be used to
create bound threads. Change "bool mpsafe" to "int flags".
- Allow exit of LWPs in the IDL state when (l != curlwp).
- More locking fixes & conversion to the new API.
|
| 1.62.6.1 | 13-Mar-2007 |
ad | Pull in the initial set of changes for the vmlocking branch.
|
| 1.63.10.1 | 02-Jan-2008 |
bouyer | Sync with HEAD
|
| 1.63.6.2 | 19-Dec-2007 |
ad | Use a global lfs_lock.
|
| 1.63.6.1 | 04-Dec-2007 |
ad | Pull the vmlocking changes into a new branch.
|
| 1.63.4.1 | 18-Feb-2008 |
mjf | Sync with HEAD.
|
| 1.65.10.3 | 11-Mar-2010 |
yamt | sync with head
|
| 1.65.10.2 | 04-May-2009 |
yamt | sync with head.
|
| 1.65.10.1 | 16-May-2008 |
yamt | sync with head.
|
| 1.65.8.1 | 18-May-2008 |
yamt | sync with head.
|
| 1.65.6.1 | 02-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.66.2.1 | 23-Jun-2008 |
wrstuden | Sync w/ -current. 34 merge conflicts to follow.
|
| 1.67.12.1 | 13-May-2009 |
jym | Sync with HEAD.
Commit is split, to avoid a "too many arguments" protocol error.
|
| 1.67.6.1 | 28-Apr-2009 |
skrll | Sync with HEAD.
|
| 1.68.2.1 | 30-Apr-2010 |
uebayasi | Sync with HEAD.
|
| 1.70.12.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.70.12.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.70.12.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.70.12.1 | 25-Feb-2013 |
tls | resync with head
|
| 1.70.2.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.70.2.1 | 23-Jan-2013 |
yamt | sync with head
|
| 1.77.2.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.80.6.5 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.80.6.4 | 05-Oct-2016 |
skrll | Sync with HEAD
|
| 1.80.6.3 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.80.6.2 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.80.6.1 | 06-Apr-2015 |
skrll | Sync with HEAD
|
| 1.88.2.1 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.91.2.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.92.6.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.94.12.1 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.94.10.1 | 17-Aug-2020 |
martin | Pull up following revision(s) (requested by riastradh in ticket #1050):
sys/ufs/lfs/lfs_subr.c: revision 1.101
sys/ufs/lfs/lfs_subr.c: revision 1.102
sys/ufs/lfs/lfs_inode.c: revision 1.158
sys/ufs/lfs/lfs_inode.h: revision 1.25
sys/ufs/lfs/lfs_balloc.c: revision 1.95
sys/ufs/lfs/lfs_pages.c: revision 1.21
sys/ufs/lfs/lfs_vnops.c: revision 1.330
sys/ufs/lfs/lfs_alloc.c: revision 1.140 (patch)
sys/ufs/lfs/lfs_alloc.c: revision 1.141 (patch)
lib/libp2k/p2k.c: revision 1.72
sys/ufs/lfs/lfs.h: revision 1.205
sys/ufs/lfs/lfs.h: revision 1.206
sys/ufs/lfs/lfs_segment.c: revision 1.284
sys/ufs/lfs/lfs.h: revision 1.207
sys/ufs/lfs/lfs_segment.c: revision 1.285
sys/ufs/lfs/lfs_debug.c: revision 1.55
sys/ufs/lfs/lfs_rename.c: revision 1.23
usr.sbin/dumplfs/dumplfs.c: revision 1.65
sys/ufs/lfs/lfs_vfsops.c: revision 1.371
sys/arch/i386/stand/efiboot/bootx64/Makefile: revision 1.3
sys/ufs/lfs/lfs_vfsops.c: revision 1.372
sys/ufs/lfs/lfs_vfsops.c: revision 1.373
sbin/fsck_lfs/pass1.c: revision 1.46
sys/ufs/lfs/lfs_vnops.c: revision 1.326
sys/ufs/lfs/lfs_vnops.c: revision 1.327
sys/ufs/lfs/lfs_vfsops.c: revision 1.375 (patch)
sys/ufs/lfs/lfs_vnops.c: revision 1.328
sys/ufs/lfs/lfs_subr.c: revision 1.98
sys/ufs/lfs/lfs_extern.h: revision 1.116
sys/ufs/lfs/lfs_vnops.c: revision 1.329
sys/ufs/lfs/lfs_subr.c: revision 1.99
sys/ufs/lfs/lfs_extern.h: revision 1.117
sys/ufs/lfs/lfs_accessors.h: revision 1.49
sys/ufs/lfs/lfs_extern.h: revision 1.118
sys/rump/fs/lib/liblfs/Makefile: revision 1.15
sys/ufs/lfs/lfs_bio.c: revision 1.146 (patch)
sys/ufs/lfs/lfs_bio.c: revision 1.147
sys/ufs/lfs/lfs_subr.c: revision 1.100
Fix kassert in lfs by initializing vp first.
Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
Teach lfs to transition ro<->rw.
Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
Assert lfs_writer where I think we can now prove it.
Serialize access to the splay tree with lfs_lock.
Change some cheap KDASSERT into KASSERT.
Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
Call lfs_orphan in lfs_rename while we're still in the dirop.
lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
Teach LFS_ORPHAN_NEXTFREE about lfs64.
Dust off the orphan detection code and try to make it work.
Fix !DIAGNOSTIC compile
Fix userland references to LFS_ORPHAN_NEXTFREE.
Forgot to grep for these or do a full distribution build, oops!
Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
OR into bp->b_cflags; don't overwrite.
CTASSERT lfs on-disk structure sizes.
Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
Fix clang build after packed lfs64 accessor change.
Suppress spurious address-of-packed error in rump lfs too.
|
| 1.94.6.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.151 | 20-Oct-2025 |
perseant | * Generalize the partial-segment parser introduced for roll-forward,
using it to facilitate an in-kernel segment rewriter (cleaner), and a
mechanism to check whether a segment is in fact empty (only used
with DEBUG).
* Add these new fcntl calls:
- LFCNFILESTATS: For each inode given, report its number of direct
blocks, how many gaps (discontinuities) there are between direct
blocks, and how large the total gap distance is. This will be
useful for a coalescing agent.
- LFCNREWRITEFILE: For each inode given, rewrite its direct blocks,
effectively coalescing it into as compact a form as possible.
- LFCNSCRAMBLE: As above, except that it only rewrites every other
block. This causes the file to have many gaps that can be
measured with LFCNFILESTATS and addressed with LFCNREWRITEFILE,
for testing purposes.
- LFCNREWRITESEGS: Rewrite any live data in the given segments.
This is intended to simplify the cleaner API and facilitate an
in-kernel cleaner.
- LFCNCLEANERINFO: Get the most current CLEANERINFO data from the
kernel.
- LFCNSEGUSE: Retrieve segment usage data from the kernel.
* Vnodes marked IN_CLEANING now take a reference. Add a new "cleaner
lock", which must be taken by the cleaner before the segment lock,
and before marking nodes IN_CLEANING. This allows us to flush
vnodes, if necessary, before the cleaning segment is written, and
never to flush vnodes being cleaned. When the cleaner lock is
released, the vnodes are cleared of IN_CLEANING and the reference
dropped.
* Track a potential infinite loop in lfs_gatherblock.
* Pull "needs to flush" and "needs to wait for flush" into functions
instead of inlining their definitions.
|
| 1.150 | 15-Sep-2025 |
perseant | If we don't have enough space, flush with checkpoint: the Ifile might be
clogging up the buffer cache.
Rewrite the logic in lfs_flush() so that the requested filesystem is always
flushed, regardless of whether only_onefs is set.
Use LFS_WAIT_BYTES and LFS_WAIT_BUFS as the thresholds when determining
whether to wait for resources, rather than their _MAX_ counterparts.
|
| 1.149 | 05-Sep-2020 |
riastradh | Round of uvm.h cleanup.
The poorly named uvm.h is generally supposed to be for uvm-internal
users only.
- Narrow it to files that actually need it -- mostly files that need
to query whether curlwp is the pagedaemon, which should maybe be
exposed by an external header.
- Use uvm_extern.h where feasible and uvm_*.h for things not exposed
by it. We should split up uvm_extern.h but this will serve for now
to reduce the uvm.h dependencies.
- Use uvm_stat.h and #ifdef UVMHIST uvm.h for files that use
UVMHIST(ubchist), since ubchist is declared in uvm.h but the
reference evaporates if UVMHIST is not defined, so we reduce header
file dependencies.
- Make uvm_device.h and uvm_swap.h independently includable while
here.
ok chs@
|
| 1.148 | 11-Jun-2020 |
ad | uvm_availmem(): give it a boolean argument to specify whether a recent
cached value will do, or if the very latest total must be fetched. It can
be called thousands of times a second and fetching the totals impacts not
only the calling LWP but other CPUs doing unrelated activity in the VM
system.
|
| 1.147 | 14-Mar-2020 |
ad | OR into bp->b_cflags; don't overwrite.
|
| 1.146 | 23-Feb-2020 |
riastradh | Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
|
| 1.145 | 18-Feb-2020 |
chs | remove the aiodoned thread. I originally added this to provide a thread context
for doing page cache iodone work, but since then biodone() has changed to
hand off all iodone work to a softint thread, so we no longer need the
special-purpose aiodoned thread.
|
| 1.144 | 31-Dec-2019 |
ad | branches: 1.144.2;
Rename uvm_free() -> uvm_availmem().
|
| 1.143 | 21-Dec-2019 |
ad | uvmexp.free -> uvm_free()
|
| 1.142 | 09-Jun-2018 |
zafer | branches: 1.142.2; 1.142.6;
Add missing b_cflags and b_oflags.
Ok dholland@
Addresses PR kern/42342 by Yoshihiro Nakajima
|
| 1.141 | 10-Jun-2017 |
maya | branches: 1.141.4;
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.140 | 08-Jun-2017 |
chs | move some buffer cache internals declarations from buf.h to vfs_bio.c.
this is needed to avoid name conflicts with ZFS and also
makes it clearer that other code shouldn't be messing with these.
remove the LFS debug code that poked around in bufqueues and
remove the BQ_EMPTY bufqueue since nothing uses it anymore.
provide a function to let LFS and wapbl read the value of nbuf for now.
|
| 1.139 | 17-Apr-2017 |
hannken | branches: 1.139.4;
Remove unused argument "nextp" from vfs_busy() and vfs_unbusy().
Remove argument "keepref" from vfs_unbusy() and add vfs_ref() where needed.
|
| 1.138 | 13-Apr-2017 |
hannken | Switch lfs_flush() and lfs_writerd() to mountlist iterator.
|
| 1.137 | 01-Apr-2017 |
maya | Switch lfs_writer_daemon to use condvar instead of mtsleep.
track thread existence with struct lwp instead of pid + lid,
it's more useful from ddb.
|
| 1.136 | 13-Mar-2017 |
riastradh | #if DIAGNOSTIC panic ---> KASSERT
Replace some #if DEBUG by this too. DEBUG is only for expensive
assertions; these are not.
|
| 1.135 | 03-Oct-2015 |
hannken | branches: 1.135.2; 1.135.4;
Remove dubious vhold()/holdrele() from lfs_reserve().
The vnodes are always referenced on entry.
If we changed ulfs_remove() and ulfs_rmdir() to return the locked dvp
the vnodes were always locked on entry.
Remove an outdated comment from lfs_reserveavail(), unlocking/relocking
the vnode was removed in rev 1.49.
|
| 1.134 | 12-Aug-2015 |
dholland | Hack up dinode usage to be 64 vs. 32 as needed. Part 1.
(This part changes the native lfs code; the ufs-derived code already
has 64 vs. 32 logic, but as aspects of it are unsafe, and don't
entirely interoperate cleanly with the lfs 64/32 stuff, pass 2 will be
rehashing that.)
|
| 1.133 | 02-Aug-2015 |
dholland | Fix assorted 64 -> 32 truncations in lfs. Also, some minor tidyups and
corrections in passing.
|
| 1.132 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.131 | 25-Jul-2015 |
martin | Use accessors in DEBUG and DIAGNOSTIC code as well
|
| 1.130 | 24-Jul-2015 |
dholland | More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)
The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.
|
| 1.129 | 24-Jul-2015 |
dholland | Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)
It also gets rid of cpp abuse in the form of fake structure member
macros.
Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)
XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.
|
| 1.128 | 27-Nov-2013 |
christos | branches: 1.128.6;
Change the queue.3 *_END(&head) macros to NULL. Since we don't have CIRCLEQ
anymore, all the macros expand to NULL anyway, so this improves readability.
Requested by rmind@
|
| 1.127 | 23-Nov-2013 |
christos | change the mountlist CIRCLEQ into a TAILQ
|
| 1.126 | 28-Jul-2013 |
dholland | Add lfs_kernel.h for declarations that don't need to be exposed to userland.
lfs currently has the following headers:
lfs.h - on-disk structures and stuff needed for userlevel tools
lfs_inode.h - additional restricted materials for userlevel tools
that operate the fs (newfs_lfs, fsck_lfs, lfs_cleanerd)
lfs_kernel.h - stuff needed only in the kernel
and the following legacy headers that are expected to be mopped up and
folded into one of the above:
lfs_extern.h - function prototypes
ulfs_bswap.h - endian-independent support
ulfs_dinode.h - now contains very little
ulfs_dirhash.h - dirhash support
ulfs_extattr.h - extattr support
ulfs_extern.h - more function prototypes
ulfs_inode.h - assorted kernel-only declarations
ulfs_quota.h - quota support
ulfs_quota1.h - more quota support
ulfs_quota2.h - more quota support
ulfs_quotacommon.h - more quota support
ulfsmount.h - legacy copy of ufsmount material
|
| 1.125 | 18-Jun-2013 |
christos | branches: 1.125.2;
Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.
|
| 1.124 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.123 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.122 | 16-Feb-2012 |
perseant | branches: 1.122.2;
Pass t_renamerace and t_rmdirrace tests.
Adapt dholland@'s fix to ufs_rename to fix PR kern/43582. Address several
other MP locking issues discovered during the course of investigating the
same problem.
Removed extraneous vn_lock() calls on the Ifile, since the Ifile writes
are controlled by the segment lock.
Fix PR kern/45982 by deemphasizing the estimate of how much metadata
will fill the empty space on disk when the disk is nearly empty
(t_renamerace crates a lot of inode blocks on a tiny empty disk).
|
| 1.121 | 02-Jan-2012 |
perseant | branches: 1.121.2;
* Remove PGO_RECLAIM during lfs_putpages()' call to genfs_putpages(),
to avoid a live lock in the latter when reclaiming a vnode with
dirty pages.
* Add a new segment flag, SEGM_RECLAIM, to note when a segment is
being written for vnode reclamation, and record which inode is being
reclaimed, to aid in forensic debugging.
* Add a new segment flag, SEGM_SINGLE, so that opportunistic writes
can write a single segment's worth of blocks and then stop, rather
than writing all the way up to the cleaner's reserved number of
segments.
* Add assert statements to check mutex ownership is the way it ought
to be, mostly in lfs_putpages; fix problems uncovered by this.
* Don't clear VU_DIROP until the inode actually makes its way to disk,
avoiding a problem where dirop inodes could become separated
(uncovered by a modified version of the "ckckp" forensic regression
test).
* Move the vfs_getopsbyname() call into lfs_writerd. Prepare code to
make lfs_writerd notice when there are no more LFSs, and exit losing
the reference, so that, in theory, the module can be unloaded. This
code is not enabled, since it causes a crash on exit.
* Set IN_MODIFIED on inodes flushed by lfs_flush_dirops. Really we
only need to set IN_MODIFIED if we are going to write them again
(e.g., to write pages); need to think about this more.
Finally, several changes to help avoid "no clean segments" panics:
* In lfs_bmapv, note when a vnode is loaded only to discover whether
its blocks are live, so it can immediately be recycled. Since the
cleaner will try to choose ~empty segments over full ones, this
prevents the cleaner from (1) filling the vnode cache with junk, and
(2) squeezing any unwritten writes to disk and running the fs out of
segments.
* Overestimate by half the amount of metadata that will be required
to fill the clean segments. This will make the disk appear smaller,
but should help avoid a "no clean segments" panic.
* Rearrange lfs_writerd. In particular, lfs_writerd now pays
attention to the number of clean segments available, and holds off
writing until there is room.
|
| 1.120 | 11-Jul-2011 |
hannken | branches: 1.120.2; 1.120.6;
Change VOP_BWRITE() to take a vnode as its first argument like all other
VOPs do. Layered file systems no longer have to modify bp->b_vp and run
into trouble when an async VOP_BWRITE() uses the wrong vnode.
- change all occurences of VOP_BWRITE(bp) to VOP_BWRITE(bp->b_vp, bp).
- remove layer_bwrite().
- welcome to 5.99.55
Adresses PR kern/38762 panic: vwakeup: neg numoutput
No objections from tech-kern@.
|
| 1.119 | 12-Jun-2011 |
rmind | Welcome to 5.99.53! Merge rmind-uvmplock branch:
- Reorganize locking in UVM and provide extra serialisation for pmap(9).
New lock order: [vmpage-owner-lock] -> pmap-lock.
- Simplify locking in some pmap(9) modules by removing P->V locking.
- Use lock object on vmobjlock (and thus vnode_t::v_interlock) to share
the locks amongst UVM objects where necessary (tmpfs, layerfs, unionfs).
- Rewrite and optimise x86 TLB shootdown code, make it simpler and cleaner.
Add TLBSTATS option for x86 to collect statistics about TLB shootdowns.
- Unify /dev/mem et al in MI code and provide required locking (removes
kernel-lock on some ports). Also, avoid cache-aliasing issues.
Thanks to Andrew Doran and Joerg Sonnenberger, as their initial patches
formed the core changes of this branch.
|
| 1.118 | 24-Jun-2010 |
hannken | branches: 1.118.6;
Clean up vnode lock operations pass 2:
VOP_UNLOCK(vp, flags) -> VOP_UNLOCK(vp): Remove the unneeded flags argument.
Welcome to 5.99.32.
Discussed on tech-kern.
|
| 1.117 | 16-Feb-2010 |
mlelstv | branches: 1.117.2;
Three changes in a single commit.
- drop the notion of frags (LFS fragments) vs fsb (FFS fragments)
The code uses a complicated unity function that just makes the
code difficult to understand.
- support larger sector sizes. Fix disk address computations
to use DEV_BSIZE in the kernel as required by device drivers
and to use sector sizes in userland.
- Fix several locking bugs in lfs_bio.c and lfs_subr.c.
|
| 1.116 | 08-Jan-2010 |
pooka | branches: 1.116.2;
The VATTR_NULL/VREF/VHOLD/HOLDRELE() macros lost their will to live
years ago when the kernel was modified to not alter ABI based on
DIAGNOSTIC, and now just call the respective function interfaces
(in lowercase). Plenty of mix'n match upper/lowercase has creeped
into the tree since then. Nuke the macros and convert all callsites
to lowercase.
no functional change
|
| 1.115 | 07-Dec-2009 |
eeh | Fix some more hangs and deadlocks.
|
| 1.114 | 06-May-2008 |
ad | branches: 1.114.18;
PR kern/38141 lookup/vfs_busy acquire rwlock recursively
Simplify the mount locking. Remove all the crud to deal with recursion on
the mount lock, and crud to deal with unmount as another weirdo lock.
Hopefully this will once and for all fix the deadlocks with this. With this
commit there are two locks on each mount:
- krwlock_t mnt_unmounting. This is used to prevent unmount across critical
sections like getnewvnode(). It's only ever read locked with rw_tryenter(),
and is only ever write locked in dounmount(). A write hold can't be taken
on this lock if the current LWP could hold a vnode lock.
- kmutex_t mnt_updating. This is taken by threads updating the mount, for
example when going r/o -> r/w, and is only present to serialize updates.
In order to take this lock, a read hold must first be taken on
mnt_unmounting, and the two need to be held across the operation.
One effect of this change: previously if an unmount failed, we would make a
half hearted attempt to back out of it gracefully, but that was unlikely to
work in a lot of cases. Now while an unmount that will be aborted is in
progress, new file operations within the mount will fail instead of being
delayed. That is unlikely to be a problem though, because if the admin
requests unmount of a file system then s(he) has made a decision to deny
access to the resource.
|
| 1.113 | 30-Apr-2008 |
ad | PR kern/38135 vfs_busy/vfs_trybusy confusion
The previous fix worked, but it opened a window where mounts could have
disappeared from mountlist while the caller was traversing it using
vfs_trybusy(). Fix that.
|
| 1.112 | 29-Apr-2008 |
ad | kern/38135 vfs_busy/vfs_trybusy confusion
The symptom was that sometimes file systems would occasionally not appear
in output from 'df' or 'mount' if the system was busy. Resolution:
- Make mount locks work somewhat like vm_map locks.
- vfs_trybusy() now only fails if the mount is gone, or if someone is
unmounting the file system. Simple contention on mnt_lock doesn't
cause it to fail.
- vfs_busy() will wait even if the file system is being unmounted.
|
| 1.111 | 28-Apr-2008 |
martin | Remove clause 3 and 4 from TNF licenses
|
| 1.110 | 20-Feb-2008 |
matt | branches: 1.110.6; 1.110.8; 1.110.10;
Merge all the *different* definitions of bufqueues into one common one.
|
| 1.109 | 15-Feb-2008 |
ad | The buffer LOCKED flag need not be under the protection of bufcache_lock,
BUSY is enough.
|
| 1.108 | 30-Jan-2008 |
ad | PR kern/37706 (forced unmount of file systems is unsafe):
- Do reference counting for 'struct mount'. Each vnode associated with a
mount takes a reference, and in turn the mount takes a reference to the
vfsops.
- Now that mounts are reference counted, replace the overcomplicated mount
locking inherited from 4.4BSD with a recursable rwlock.
|
| 1.107 | 02-Jan-2008 |
ad | Merge vmlocking2 to head.
|
| 1.106 | 11-Oct-2007 |
ad | branches: 1.106.4; 1.106.6; 1.106.10;
Remove LOCK_ASSERT(!simple_lock_held(&foo));
|
| 1.105 | 10-Oct-2007 |
ad | Merge from vmlocking:
- Split vnode::v_flag into three fields, depending on field locking.
- simple_lock -> kmutex in a few places.
- Fix some simple locking problems.
|
| 1.104 | 08-Oct-2007 |
ad | Merge ffs locking & brelse changes from the vmlocking branch.
|
| 1.103 | 29-Jul-2007 |
ad | branches: 1.103.4; 1.103.6; 1.103.8; 1.103.10;
It's not a good idea for device drivers to modify b_flags, as they don't
need to understand the locking around that field. Instead of setting
B_ERROR, set b_error instead. b_error is 'owned' by whoever completes
the I/O request.
|
| 1.102 | 17-Jul-2007 |
christos | branches: 1.102.2;
eliminate MFSNAMELEN
|
| 1.101 | 16-May-2007 |
perseant | Change references to SEGM_W_DIROPS to SEGM_CKP, and replace the logic that
formerly used SEGM_W_DIROPS in lfs_segwrite() appropriately. This prevents
a problem in which processes could get stuck in "buffers" sleep forever.
|
| 1.100 | 18-Apr-2007 |
perseant | Add/change a couple of comments about locking restrictions.
|
| 1.99 | 17-Apr-2007 |
perseant | Install a new sysctl, vfs.lfs.ignore_lazy_sync, which causes LFS to ignore
the "smooth" syncer, as if vfs.sync.*delay = 0, but only for LFS. The
default is "on", i.e., ignore lazy sync.
Reduce the amount of polling/busy-waiting done by lfs_putpages(). To
accomplish this, copied genfs_putpages() and modified it to indicate which
page it was that caused it to return with EDEADLK. fsync()/fdatasync()
should no longer ever fail with EAGAIN, and should not consume huge
quantities of cpu.
Also, try to make dirops less likely to be written as the result of a
VOP_PUTPAGES(), while ensuring that they are written regularly.
|
| 1.98 | 16-Nov-2006 |
christos | branches: 1.98.2; 1.98.4; 1.98.8; 1.98.10; 1.98.16;
__unused removal on arguments; approved by core.
|
| 1.97 | 12-Oct-2006 |
christos | - sprinkle __unused on function decls.
- fix a couple of unused bugs
- no more -Wno-unused for i386
|
| 1.96 | 04-Oct-2006 |
christos | fix empty if
|
| 1.95 | 15-Sep-2006 |
yamt | branches: 1.95.2;
merge yamt-pdpolicy branch.
- separate page replacement policy from the rest of kernel
- implement an alternative replacement policy
|
| 1.94 | 29-Jun-2006 |
perseant | branches: 1.94.4;
Don't wake up the cleaner if the filesystem is unwrappable, and fix the
compatibility fcntls.
Also includes one-line fixes for an MP locking bug and a zero-length FINFO
problem that manifested during testing.
|
| 1.93 | 14-May-2006 |
elad | branches: 1.93.4;
integrate kauth.
|
| 1.92 | 04-May-2006 |
perseant | Introduce another per-filesystem parameter, lfs_resvseg, to separate the
notion of "how many segments are reserved for the cleaner" from that of
"how many segments are not counted in lfs_bfree". The default value
used for existing filesystems is the same as the previous implicit value
of (lfs_minfreeseg / 2 + 1), modulo some sanity checking.
Count pending dirops on a per-filesystem basis, since once we start
writing them we can't stop until we're done. This seems to help stave off
the "no clean segments" panic in the case of filling the filesystem with
directories and small files (e.g. simultaneously unpacking more copies of
pkgsrc than will fit).
|
| 1.91 | 13-Apr-2006 |
perseant | Make lfs_vref/lfs_vunref not need to know about VXLOCK and VFREEING
explicitly (especially since we didn't know about VFREEING at all before),
but notice the EBUSY return from vget() instead.
Fix some more MP locking protocol issues, most of which were pointed out by
Christian Ehrhardt this morning on tech-kern.
|
| 1.90 | 05-Mar-2006 |
christos | branches: 1.90.2; 1.90.4;
cleanup more SET/CLR/ISSET lossage
|
| 1.89 | 06-Jan-2006 |
yamt | branches: 1.89.2; 1.89.4; 1.89.6;
initialize necessary members of struct buf. PR/32462 from Reinoud Zandijk.
|
| 1.88 | 04-Jan-2006 |
yamt | - add simple functions to allocate/free a buffer for i/o.
- make bufpool static.
|
| 1.87 | 11-Dec-2005 |
christos | branches: 1.87.2;
merge ktrace-lwp.
|
| 1.86 | 29-May-2005 |
christos | branches: 1.86.2;
- sprinkle const
- avoid shadow variables.
|
| 1.85 | 23-Apr-2005 |
perseant | Provide a resize_lfs(8), including kernel and cleaner support. The current
implementation requires the fs to be mounted while resizing. Tested in both
directions, and everything appears to work happily, but ymmv.
|
| 1.84 | 19-Apr-2005 |
perseant | Keep per-inode, per-fs, and subsystem-wide counts of blocks allocated through
lfs_balloc(), and use that to estimate the number of dirty pages belonging
to LFS (subsystem or filesystem). This is almost certainly wrong for
the case of a large mmap()ed region, but the accounting is tighter than
what we had before, and performs much better in the typical case of pages
dirtied through write().
|
| 1.83 | 06-Apr-2005 |
perseant | Fix some locking issues that appeared with the simple_lock work.
Address a "pager_map" deadlock in lfs_putpages().
|
| 1.82 | 01-Apr-2005 |
perseant | Protect various per-fs structures with fs->lfs_interlock simple_lock, to
improve behavior in the multiprocessor case. Add debugging segment-lock
assertion statements.
|
| 1.81 | 09-Mar-2005 |
perseant | branches: 1.81.2;
Be more careful about handling of flags to lfs_flush, to ensure that
the lfs_writing mutex is respected.
|
| 1.80 | 08-Mar-2005 |
perseant | Straighten out the maze of ifdefs. Instead, consolidate all the debugging
stuff under '#ifdef DEBUG', and use sysctl knobs to turn on/off particular
parts of the debugging reporting (if DEBUG is enabled). Re-enable the LFS
statistics in sysctl, while I'm there. A bit of a rototill.
|
| 1.79 | 26-Feb-2005 |
perry | nuke trailing whitespace
|
| 1.78 | 26-Feb-2005 |
perseant | Various minor LFS improvements:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statvfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
|
| 1.77 | 28-Jan-2004 |
yamt | branches: 1.77.6; 1.77.8; 1.77.10;
use bufmem instead of bufpages to make lfs a little less broken.
|
| 1.76 | 04-Dec-2003 |
yamt | use b_private rather than b_saveaddr.
XXX LFS_USE_B_INVAL
|
| 1.75 | 03-Oct-2003 |
yamt | assertions.
|
| 1.74 | 23-Sep-2003 |
yamt | remove unnecessary externs of lfs_do_flush.
|
| 1.73 | 07-Sep-2003 |
yamt | - raise spl to bio in lfs_countlocked() rather than having callers to do so.
- buffer cache MP locks.
- assert B_CALL buffers are not on the free queue.
|
| 1.72 | 07-Aug-2003 |
agc | Move UCB-licensed code from 4-clause to 3-clause licence.
Patches provided by Joel Baker in PR 22364, verified by myself.
|
| 1.71 | 12-Jul-2003 |
yamt | - protect global resource counts with lfs_subsys_lock.
- clean up scattered externs a little.
|
| 1.70 | 02-Jul-2003 |
yamt | a comment.
|
| 1.69 | 02-Jul-2003 |
yamt | use queue.h macros.
|
| 1.68 | 02-Jul-2003 |
yamt | use VFSTOUFS macro.
|
| 1.67 | 02-Jul-2003 |
yamt | - add a new functions, lfs_writer_enter/leave, and use them instead of
duplicated code fragments.
- add an assertion.
|
| 1.66 | 27-Apr-2003 |
perseant | branches: 1.66.2;
Don't change update time on block write; lets e.g. "tar xp" work properly.
|
| 1.65 | 02-Apr-2003 |
fvdl | Add support for UFS2. UFS2 is an enhanced FFS, adding support for
64 bit block pointers, extended attribute storage, and a few
other things.
This commit does not yet include the code to manipulate the extended
storage (for e.g. ACLs), this will be done later.
Originally written by Kirk McKusick and Network Associates Laboratories for
FreeBSD.
|
| 1.64 | 15-Mar-2003 |
perseant | Add simple_lock protection for lfs_seglock and lfs_subsys_pages; these will
be expanded to cover other per-fs and subsystem-wide data as well.
Fix a case of IN_MODIFIED being set without updating lfs_uinodes, resulting
in a "lfs_uinodes < 0" panic.
Fix a deadlock in lfs_putpages arising from the need to busy all pages in a
block; unbusy any that had already been busied before starting over.
|
| 1.63 | 02-Mar-2003 |
perseant | Account SEGUSE_ACTIVE correctly so that the automatic segment cleaning
actually happens.
Add a new fcntl call that will write the minimum necessary to checkpoint
(i.e., for on-disk directory structure to be consistent, not including
updates to file data) so that the cleaner can clean segments more quickly
without sacrificing three-way commit for cleaning.
|
| 1.62 | 25-Feb-2003 |
thorpej | Add a new BUF_INIT() macro which initializes b_dep and b_interlock, and
use it. This fixes a few places where either b_dep or b_interlock were
not properly initialized.
|
| 1.61 | 20-Feb-2003 |
perseant | Tabify, and fix some comment alignment problems.
|
| 1.60 | 19-Feb-2003 |
yamt | workaround for "another flush is..." infinity loop in writerd.
if we're writerd, sleep in lfs_flush until another writer goes away
instead of busy loop in writed.
|
| 1.59 | 19-Feb-2003 |
yamt | init b_interlock.
|
| 1.58 | 17-Feb-2003 |
perseant | Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now
(there are still some details to work out) but expect that to go
away soon. To support these basic changes (creation of lfs_putpages,
lfs_gop_write, mods to lfs_balloc) several other changes were made, to
wit:
* Create a writer daemon kernel thread whose purpose is to handle page
writes for the pagedaemon, but which also takes over some of the
functions of lfs_check(). This thread is started the first time an
LFS is mounted.
* Add a "flags" parameter to GOP_SIZE. Current values are
GOP_SIZE_READ, meaning that the call should return the size of the
in-core version of the file, and GOP_SIZE_WRITE, meaning that it
should return the on-disk size. One of GOP_SIZE_READ or
GOP_SIZE_WRITE must be specified.
* Instead of using malloc(...M_WAITOK) for everything, reserve enough
resources to get by and use malloc(...M_NOWAIT), using the reserves if
necessary. Use the pool subsystem for structures small enough that
this is feasible. This also obsoletes LFS_THROTTLE.
And a few that are not strictly necessary:
* Moves the LFS inode extensions off onto a separately allocated
structure; getting closer to LFS as an LKM. "Welcome to 1.6O."
* Unified GOP_ALLOC between FFS and LFS.
* Update LFS copyright headers to correct values.
* Actually cast to unsigned in lfs_shellsort, like the comment says.
* Keep track of which segments were empty before the previous
checkpoint; any segments that pass two checkpoints both dirty and
empty can be summarily cleaned. Do this. Right now lfs_segclean
still works, but this should be turned into an effectless
compatibility syscall.
|
| 1.57 | 05-Feb-2003 |
pk | Make the buffer cache code MP-safe.
|
| 1.56 | 24-Jan-2003 |
fvdl | Bump daddr_t to 64 bits. Replace it with int32_t in all places where
it was used on-disk, so that on-disk formats remain the same.
Remove ufs_daddr_t and ufs_lbn_t for the time being.
|
| 1.55 | 30-Dec-2002 |
yamt | comment and assertions
|
| 1.54 | 30-Dec-2002 |
yamt | move check of lfs_unlockvp from lfs_reserveavail to lfs_reserve
because lfs_reservebuf needs same check as well.
|
| 1.53 | 29-Dec-2002 |
yamt | fix vref/vunref mismatch.
|
| 1.52 | 28-Dec-2002 |
yamt | - in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't
try to reclaim them.
(workaround for deadlock noted in the comment in lfs_reserveavail)
- in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
|
| 1.51 | 26-Dec-2002 |
yamt | - in lfs_reserve, reserve locked buffer count as well.
- don't wait for locking buf in lfs_bwrite_ext to avoid deadlocks.
- skip lfs_reserve when we're doing dirop.
reserve more (for lfs_truncate) in set_dirop instead.
this mostly solves PR 18972. (and hopefully PR 19196)
|
| 1.50 | 22-Dec-2002 |
yamt | add a XXX comment. (description of possible deadlock)
|
| 1.49 | 17-Dec-2002 |
yamt | #if 0 out vnode unlock/lock in lfs_reserve for now and add a comment about it.
deadlock is better than corruption (or panic), IMO.
|
| 1.48 | 14-Dec-2002 |
yamt | - in lfs_bwrite_ext, if we're cleaner,
mark inode IN_CLEANING rather then IN_MODIFIED.
otherwise cleaned (indirect) blocks belongs to the inode isn't written
until next sync.
- add assertions.
|
| 1.47 | 27-Nov-2002 |
yamt | more XXX comment.
|
| 1.46 | 24-Nov-2002 |
yamt | add a XXX comment to lfs_reserve.
* it isn't safe to unlock vp here
* because we're passing data using inode from namei.
* (eg. i_offset)
|
| 1.45 | 24-Nov-2002 |
yamt | lfs_reserve shouldn't block for lfs_unlockvp.
otherwise cleaner deadlocks.
PR 19134.
|
| 1.44 | 20-Jun-2002 |
perseant | Fix miscalculation in lfs_fits found by Trevin Beattie <trevin@xmission.com>.
Change some of the variable names from "nb", "db" to "fsb" to reflect their
calling conventions.
|
| 1.43 | 14-May-2002 |
perseant | branches: 1.43.2;
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes
I found while making sure there weren't any new ones.
* Make the write clusters keep track of the buffers whose blocks they contain.
This should make it possible to (1) write clusters using a page mapping
instead of malloc, if desired, and (2) schedule blocks for rewriting
(somewhere else) if a write error occurs. Code is present to use
pagemove() to construct the clusters but that is untested and will go away
anyway in favor of page mapping.
* DEBUG now keeps a log of Ifile writes, so that any lingering instances of
the "dirty bufs" problem can be properly debugged.
* Keep track of whether the Ifile has been dirtied by various routines that
can be called by lfs_segwrite, and loop on that until it is clean, for
a checkpoint. Checkpoints need to be squeaky clean.
* Warn the user (once) if the Ifile grows larger than is reasonable for their
buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can
grow.
* If an inode is not found in a disk block, try rereading the block, under
the assumption that the block was copied to a cluster and then freed.
* Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
|
| 1.42 | 12-May-2002 |
matt | Eliminate commons.
|
| 1.41 | 11-Feb-2002 |
perseant | Include the space taken by inodes in the count made by lfs_check();
make VOP_SETATTR call lfs_check. This prevents large numbers of inode
changes (say, at the end of tar(1)) from filling the buffer cache.
|
| 1.40 | 23-Nov-2001 |
chs | add spaces for KNF. confirmed to produce identical objects.
|
| 1.39 | 08-Nov-2001 |
lukem | add RCSID
|
| 1.38 | 06-Nov-2001 |
simonb | Remove some variables that are set but never used.
|
| 1.37 | 26-Oct-2001 |
lukem | remove #include <ufs/ufs/quota.h> where it was just to appease
<ufs/ufs/inode.h>, since the latter now includes the former. leave the former
in source that obviously uses specific bits of it (for completeness.)
|
| 1.36 | 13-Jul-2001 |
perseant | branches: 1.36.4;
Merge the short-lived perseant-lfsv2 branch into the trunk.
Kernels and tools understand both v1 and v2 filesystems; newfs_lfs
generates v2 by default. Changes for the v2 layout include:
- Segments of non-PO2 size and arbitrary block offset, so these can be
matched to convenient physical characteristics of the partition (e.g.,
stripe or track size and offset).
- Address by fragment instead of by disk sector, paving the way for
non-512-byte-sector devices. In theory fragments can be as large
as you like, though in reality they must be smaller than MAXBSIZE in size.
- Use serial number and filesystem identifier to ensure that roll-forward
doesn't get old data and think it's new. Roll-forward is enabled for
v2 filesystems, though not for v1 filesystems by default.
- The inode free list is now a tailq, paving the way for undelete (undelete
is not yet implemented, but can be without further non-backwards-compatible
changes to disk structures).
- Inode atime information is kept in the Ifile, instead of on the inode;
that is, the inode is never written *just* because atime was changed.
Because of this the inodes remain near the file data on the disk, rather
than wandering all over as the disk is read repeatedly. This speeds up
repeated reads by a small but noticeable amount.
Other changes of note include:
- The ifile written by newfs_lfs can now be of arbitrary length, it is no
longer restricted to a single indirect block.
- Fixed an old bug where ctime was changed every time a vnode was created.
I need to look more closely to make sure that the times are only updated
during write(2) and friends, not after-the-fact during a segment write,
and certainly not by the cleaner.
|
| 1.35 | 03-Dec-2000 |
perseant | branches: 1.35.2; 1.35.4; 1.35.6;
Fix typo in 'malloc' for non-MALLOCLOG case
|
| 1.34 | 03-Dec-2000 |
perseant | Get rid of some old unnecessary code that cleared B_NEEDCOMMIT from buffers in
lfs_writeseg (possibly after they had been freed).
If MALLOCLOG is defined, make lfs_newbuf and lfs_freebuf pass along the
caller's file and line to _malloc and _free.
|
| 1.33 | 27-Nov-2000 |
perseant | If LFS_DO_ROLLFORWARD is defined, roll forward from the older checkpoint
on mount, through the newer checkpoint and on through any newer
partial-segments that may have been written but not checkpointed because
of an intervening crash.
LFS_DO_ROLLFORWARD is not defined by default.
|
| 1.32 | 17-Nov-2000 |
perseant | Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes.
(PR #11468). In the case of fragment allocation, check to see if enough
space is available before extending a fragment already scheduled for writing.
The locked_queue_* variables indicate the number of buffer headers and bytes,
respectively, that are unavailable to getnewbuf() because they are locked up
waiting for LFS to flush them; make sure that that is actually what we're
counting, i.e., never count malloced buffers, and always use b_bufsize instead
of b_bcount.
If DEBUG is defined, the periodic calls to lfs_countlocked will now complain
if either counter is incorrect. (In the future lfs_countlocked will not need
to be called at all if DEBUG is not defined.)
|
| 1.31 | 12-Nov-2000 |
perseant | Do not needlessly dirty segment table blocks during lfs_segwrite,
preventing needless disk activity when the filesystem is idle. (PR #10979.)
|
| 1.30 | 13-Sep-2000 |
perseant | Cast back to int32_t in LFS_EST_BFREE and LFS_EST_RSVD macros, for
consistency with their arguments.
Change the debugging printf in lfs_reserve to match, and enclose it in
#ifdef DEBUG.
Tested on alpha, arm32, sparc.
|
| 1.29 | 12-Sep-2000 |
perseant | Make this file compile on the alpha as well (use %ld and cast to long,
instead of %qd with no cast).
|
| 1.28 | 10-Sep-2000 |
augustss | Make this file compile again.
|
| 1.27 | 09-Sep-2000 |
perseant | Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
|
| 1.26 | 05-Jul-2000 |
perseant | Clean up accounting of lfs_uinodes (dirty but unwritten inodes).
Make lfs_uinodes a signed quantity for debugging purposes, and set it to
zero as fs mount time.
Enclose setting/clearing of the dirty flags (IN_MODIFIED, IN_ACCESSED,
IN_CLEANING) in macros, and use those macros everywhere. Make
LFS_ITIMES use these macros; updated the ITIMES macro in inode.h to know
about this. Make ufs_getattr use ITIMES instead of FFS_ITIMES.
|
| 1.25 | 03-Jul-2000 |
perseant | Allow the number of free segments reserved for the cleaner to be
parametrized in the filesystem, defaulting to MIN_FREE_SEGS = 2 but set
to something more reasonable at newfs_lfs time.
Note the number of blocks that have been scheduled for writing but which
are not yet on disk in an inode extension, i_lfs_effnblks. Move
i_ffs_effnlink out of the ffs extension and onto the main inode, since
it's used all over the shared code and the lfs extension would clobber
it.
At inode write time, indirect blocks and inode-held blocks of inodes
that have i_lfs_effnblks != i_ffs_blocks are cleansed of UNWRITTEN disk
addresses, so that these never make it to disk.
|
| 1.24 | 27-Jun-2000 |
perseant | Fixes associated with filling an LFS:
Change the space computation to appear to change the size of the *disk*
rather than the *bytes used* when more segment summaries and inode
blocks are written. Try to estimate the amount of space that these will
take up when more files are written, so the disk size doesn't change too
much.
Regularize error returns from lfs_valloc, lfs_balloc, lfs_truncate: they
now fail entirely, rather than succeeding half-way and leaving the fs in
an inconsistent state.
Rewrite lfs_truncate, mostly stealing from ffs_truncate. The old
lfs_truncate had difficulty truncating a large file to a non-zero size
(indirect blocks were not handled appropriately).
Unmark VDIROP on fvp after ufs_remove, ufs_rmdir, so these can be
reclaimed immediately: this vnode would not be written to disk again
anyway if the removal succeeded, and if it failed, no directory
operation occurred.
ufs_makeinode and ufs_mkdir now remove IN_ADIROP on error.
|
| 1.23 | 06-Jun-2000 |
perseant | branches: 1.23.2;
Protect inode free list with seglock, instead of separate lock, so that
the head of the inode free list (on the superblock) always matches the
rest of the free list (in the ifile).
Protect lfs_fragextend with seglock, to prevent the segment byte count
fudging from making its way to disk.
Don't try to inactivate dirop vnodes that are still in the middle of
their dirop (may address PR#10285).
|
| 1.22 | 31-May-2000 |
fredb | Make this build. (Balance parenthesis.
|
| 1.21 | 31-May-2000 |
perseant | update for IN_ACCESSED changes
|
| 1.20 | 27-May-2000 |
perseant | branches: 1.20.2;
Prevent dirops from getting around lfs_check and wedging the buffer cache.
All the dirop vnops now mark the inodes with a new flag, IN_ADIROP, which
is removed as soon as the dirop is done (as opposed to VDIROP which stays
until the file is written). To address one issue raised in PR#9357.
|
| 1.19 | 19-May-2000 |
thorpej | NULL != 0
|
| 1.18 | 05-May-2000 |
perseant | Change the way LFS does block accounting, from trying to infer from the
buffer cache flags, to marking the inode and/or indirect blocks with a
special disk address UNWRITTEN==-2 when a block is accounted for. (This
address is never written to disk, but only used in-core. This is essentially
the same method of block accounting as on the UBC branch, where the buffer
headers don't exist.) Make sure that truncation is handled properly,
especially in the case of holey files.
Fixes PR#9994.
|
| 1.17 | 30-Mar-2000 |
augustss | Remove register declarations.
|
| 1.16 | 15-Dec-1999 |
perseant | In lfs_bwrite, don't mark buffers dirty if lfs is mounted read-only.
(Previously buffers could be marked dirty by the cleaner, and possibly by
other means.)
Also check for softdep mount in vfs_shutdown before trying to bawrite
buffers, since other filesystems don't need it and lfs doesn't bawrite.
(This fragment reviewed by fvdl.)
Partially addresses PR#8964.
|
| 1.15 | 04-Dec-1999 |
ragge | CL* discarding.
|
| 1.14 | 23-Nov-1999 |
fvdl | Be more careful to block bio interrupts for some data structures. There
were at least a few missed cases where vp->v_{clean,dirty}blkhd were
unprotected since the softdep/trickle sync merge.
|
| 1.13 | 06-Nov-1999 |
perseant | branches: 1.13.2;
Address ufs_hashlock/ufs_ihashins protocol bug, discovered while doing a
post-mortem of a production machine. Also, take the active dirop
count off of the fs and make it global (since it is measuring a global
resource) and tie the threshold value LFS_MAXDIROP to desiredvnodes.
|
| 1.12 | 21-Oct-1999 |
perseant | Under degenerate access patterns (e.g. `bonnie' benchmark) lfs_check could
fail, because the particular block being requested was always in the cache
(although other routines that cannot afford to call lfs_check have in the
meantime stuffed the cache full of dirty blocks). Partially addresses PR 8383.
|
| 1.11 | 01-Jun-1999 |
perseant | branches: 1.11.2; 1.11.4; 1.11.6;
Fixed lfs_update (and related functions) so that calls from lfs_fsync
will DTRT with vnodes marked VDIROP. In particular, the message
"flushing VDIROP" will no longer appear, and the filesystem will remain
stable in the event of a crash.
This was particularly a problem with NFS-exported LFSes, since fsync
was called on every file close.
|
| 1.10 | 12-Apr-1999 |
perseant | Disallow threshold-initiated cache flush when dirops are active. Also, make
SET_ENDOP use lfs_check instead of inlining most of it.
|
| 1.9 | 25-Mar-1999 |
perseant | branches: 1.9.2;
Fixes to make dirops and lfs_vflush play together well. In particular,
if we are short on vnodes, lfs_vflush from another process can grab a
vnode that lfs_markv has already processed but not yet written; but
lfs_markv holds the seglock. When lfs_vflush gets around to writing it,
the context for copyin is gone. So, now lfs_markv calls copyin itself,
rather than having lfs_writeseg do it.
|
| 1.8 | 25-Mar-1999 |
perseant | clean up unused/required #ifdefs
|
| 1.7 | 10-Mar-1999 |
perseant | New sources should leave the LFS in a more-or-less working state. Changes
include:
- DIROP segregation is enabled, and greater care is taken
to make sure that a checkpoint completes. Fsck is not
needed to remount the filesystem.
- Several checks to make sure that the LFS subsystem does not
overuse various resources (memory, in particular).
- The cleaner routines, lfs_markv in particular, are completely
rewritten. A buffer overflow is removed. Greater care is taken
to ensure that inodes come from where lfs_cleanerd say they come
from (so we know nothing has changed since lfs_bmapv was called).
- Fragment allocation is fixed, so that writes beyond end-of-file
do the right thing.
|
| 1.6 | 01-Mar-1998 |
fvdl | Merge with Lite2 + local changes
|
| 1.5 | 09-Feb-1996 |
christos | lfs prototypes
|
| 1.4 | 18-Jun-1995 |
cgd | don't assume the f_fsnamelen is nul-truncated or longer than MFSNAMELEN
|
| 1.3 | 18-Jan-1995 |
mycroft | Turn mountlist into a CIRCLEQ, and handle setting and checking of MNT_ROOTFS
differently.
|
| 1.2 | 29-Jun-1994 |
cgd | New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
|
| 1.1 | 08-Jun-1994 |
mycroft | branches: 1.1.1;
Update to 4.4-Lite fs code, with local changes.
|
| 1.1.1.2 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite2
|
| 1.1.1.1 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite for reference
|
| 1.9.2.4 | 17-Dec-1999 |
he | Pull up revision 1.13 (requested by perseant):
Address locking protocol error for inode hash, and make the
maximum number of active dirops a global quantity.
|
| 1.9.2.3 | 17-Dec-1999 |
he | Pull up revision 1.11 (requested by perseant):
Avoid flushing vnodes involved in a dirop, making lfs' promise
of "no fsck needed, even in the event of a crash" closer to
reality.
|
| 1.9.2.2 | 26-Oct-1999 |
he | Pull up revision 1.12 (requested by perseant):
Fix LFS buffer starvation under degenerate access patterns.
|
| 1.9.2.1 | 13-Apr-1999 |
perseant | branches: 1.9.2.1.2;
Pull-up of changes made to the trunk on Sunday [1.9->1.10], to wit:
Take out the `#ifdef USE_UFSHASH'; use ufs_hashlock to lock the inode free
list instead of free_lock.
Fix inode reporting in lfs_statfs (the meaning of f_files and f_ffree was
reversed).
Fix "lfs_ifind: dinode xxx not found" panic. When inodes were freed, then
immediately reloaded, their dinodes were located in an inode block which
was not on disk at the advertized location, nor in the cache (although it
would be flushed to disk next segment write). Fix this by using getblk()
instead of lfs_newbuf() for inode blocks.
Better checking for held inode locks in lfs_fastvget, for a number of
error conditions. Also change the default setting of lfs_clean_vnhead to
0, which seems to make the locking problems go away (although this is
difficult to test as I can't reliably reproduce them).
Make sure that the wakeup occurs for vnodes that lfs_update might be
sleeping on (nodes which are not marked IN_MODIFIED/IN_CLEANING, but which
have dirty buffers), by marking them with the appropriate flag if
dirtybuffers were added while the write was in progress.
Fix block counting during file truncation, if not truncating to zero.
Disallow threshold-initiated cache flush when dirops are active. Also,
make SET_ENDOP use lfs_check instead of inlining most of it.
Improve the debugging printfs in the cleaner syscalls (in particular, make
it obvious that they're coming from lfs).
Check the superblock version field, and refuse to mount the filesystem if
the version number is higher than we know about. This allows, e.g.,
changes in the format of the ifile, segment size restrictions and
boundaries, etc., which would not affect existing fields in the
superblock, but which would drastically affect the filesystem, to be
smoothly integrated at a later date.
|
| 1.9.2.1.2.2 | 31-Aug-1999 |
perseant | Rudimentary support for LFS under UBC:
- LFS-specific VOP_BALLOC and VOP_PUTPAGES vnode ops.
- getblk VREG panic #ifdef'd out (can be reinstated when Ifile is
internalized and Ifile can be made another type from VREG)
- interface to VOP_PUTPAGES changed to pass all pager flags, not
just sync. FS putpages routines must know about the pager flags.
- new LFS magic disk address, -2 ("unwritten"), meaning accounted for
but not assigned to a fixed disk location (since LFS does these two
things separately, and the previous accounting method using buffer
headers no longer will work). Changed references to (foo == (daddr_t)-1)
to (foo < 0). Since disk drivers reject all addresses < 0, this should
not present a problem for other FSs.
|
| 1.9.2.1.2.1 | 21-Jun-1999 |
thorpej | Sync w/ -current.
|
| 1.11.6.2 | 27-Dec-1999 |
wrstuden | Pull up to last week's -current.
|
| 1.11.6.1 | 21-Dec-1999 |
wrstuden | Initial commit of recent changes to make DEV_BSIZE go away.
Runs on i386, needs work on other arch's. Main kernel routines should be
fine, but a number of the stand programs need help.
cd, fd, ccd, wd, and sd have been updated. sd has been tested with non-512
byte block devices. vnd, raidframe, and lfs need work.
Non 2**n block support is automatic for LKM's and conditional for kernels
on "options NON_PO2_BLOCKS".
|
| 1.11.4.1 | 15-Nov-1999 |
fvdl | Sync with -current
|
| 1.11.2.3 | 08-Dec-2000 |
bouyer | Sync with HEAD.
|
| 1.11.2.2 | 22-Nov-2000 |
bouyer | Sync with HEAD.
|
| 1.11.2.1 | 20-Nov-2000 |
bouyer | Update thorpej_scsipi to -current as of a month ago
|
| 1.13.2.2 | 06-Nov-1999 |
perseant | Address ufs_hashlock/ufs_ihashins protocol bug, discovered while doing a
post-mortem of a production machine. Also, take the active dirop
count off of the fs and make it global (since it is measuring a global
resource) and tie the threshold value LFS_MAXDIROP to desiredvnodes.
|
| 1.13.2.1 | 06-Nov-1999 |
perseant | file lfs_bio.c was added on branch comdex-fall-1999 on 1999-11-06 20:33:06 +0000
|
| 1.20.2.1 | 22-Jun-2000 |
minoura | Sync w/ netbsd-1-5-base.
|
| 1.23.2.2 | 03-Feb-2001 |
he | Pull up revisions 1.31-1.32 (requested by perseant):
o Don't write anything if the filesystem is idle (PR#10979).
o Close up accounting holes in LFS' accounting of immediately-
available-space, number of clean segments, and amount of dirty
space taken up by metadata (PR#11468, PR#11470, PR#11534).
|
| 1.23.2.1 | 14-Sep-2000 |
perseant | Pull up recent LFS kernel changes (approved by thorpej):
ufs/ufs/inode.h, 1.20--1.22 (add i_lfs_effnblks extension ;
make ITIMES aware of LFS_ITIMES;
_LKM protection so userland progs
compile)
ufs/ufs/ufs_vnops.c, 1.69, 1.71 (remove IN_ADIROP;
use ITIMES instead of FFS_ITIMES)
ufs/ufs/ufs_readwrite.c, 1.27 (use lfs_reserve in lfs_write)
ufs/lfs/lfs.h, 1.26--1.32 (define LFS_EST_* macros ;
change MIN_FREE_SEGS to lfs_minfreesegs ;
add avail and bfree to CLEANERINFO ;
change lfs_uinodes to signed ;
change lfs_dmeta to signed ;
add whitespace to line up structure
members ;
explicit cast to int32_t in LFS_EST_*
macros)
ufs/lfs/lfs_alloc.c, back out 1.34.2.3 (pullups of 1.39, 1.40);
then pull up 1.38 (clean up on error)
1.39--1.43 (restore fvdl's ufs_hashlock fix ;
restore fvdl's ufs_hashlock fix ;
set i_lfs_effnblks ;
use UINO macros ;
add comments and fix long lines)
ufs/lfs/lfs_balloc.c, 1.19 (don't succeed halfway)
1.21--1.25 (use i_lfs_effnblks ;
fix i_lfs_effnblks computation and
quieten ;
fix i_ffs_blocks in unwritten fragment ;
remove useless debugging check ;
add comments and (c) 2000)
ufs/lfs/lfs_bio.c, 1.24--1.30 (cleanup and make lfs_flush_fs take
"struct lfs *" instead of "struct
mount *" ;
use lfs_minfreeseg instead of
MIN_FREE_SEGS ;
use UINO macros, and copy bfree/avail
to CLEANERINFO ;
add lfs_reserve function ;
1.28--1.30 fix printf formatting)
ufs/lfs/lfs_cksum.c, 1.13 (add (c) 2000)
ufs/lfs/lfs_debug.c, 1.11 (use btodb instead of DEV_BSIZE)
ufs/lfs/lfs_extern.h, 1.18, 1.20--1.21 (function prototype changes)
ufs/lfs/lfs_inode.c, 1.38 (rewrite lfs_truncate from
ffs_truncate)
1.40--1.44 (count written and unwritten blocks
seperately ;
use disk block units instead of bytes ;
remove unnecessary "mod" variable ;
correct B_DELWRI to avoid bawrite panic ;
use lfs_reserve)
ufs/lfs/lfs_segment.c, 1.52-1.59 (use lfs_dmeta to note used summaries ;
check for UNWRITTEN in indirect blocks ;
more debugging stuff inside #ifdef
DEBUG_LFS ;
use LK_CANRECURSE ;
don't drop dirty indirect blocks ;
use UINO macros ;
don't hose the free list ;
use btodb() instead of DEV_BSIZE ;
make it compile again (oops))
ufs/lfs/lfs_subr.c, 1.16--1.17 (check for locked inodes before
changing ;
use btodb() instead of DEV_BSIZE, (c)
2000)
ufs/lfs/lfs_syscalls.c, back out 1.41.4.2 (fvdl's ufs_hashlock fix);
then pull up 1.43 (use lfs_dmeta)
1.44--1.45 (restore fvdl's ufs_hashlock fix)
1.46--1.47 (fix lfs_avail leakage from sblock
segments ;
use UINO macros)
1.49 (bounds-check inode numbers in
lfs_markv)
ufs/lfs/lfs_vfsops.c, 1.53 (use LFS_EST_* macros in lfs_statfs)
1.56--1.58 (initialize lfs_minfreeseg, lfs_effnblk ;
initialize lfs_uinodes ;
initialize lfs_ravail)
ufs/lfs/lfs_vnops.c, 1.40 (remove VDIROP from removed files)
1.42--1.44 (move SET_ENDOP below the removal of
VDIROP ;
use UINO macros and add lfs_itimes
function ;
use lfs_reserve in dirops)
|
| 1.35.6.5 | 06-Sep-2002 |
jdolecek | sync kqueue branch with HEAD
|
| 1.35.6.4 | 23-Jun-2002 |
jdolecek | catch up with -current on kqueue branch
|
| 1.35.6.3 | 16-Mar-2002 |
jdolecek | Catch up with -current.
|
| 1.35.6.2 | 10-Jan-2002 |
thorpej | Sync kqueue branch with -current.
|
| 1.35.6.1 | 03-Aug-2001 |
lukem | update to -current
|
| 1.35.4.4 | 13-Jul-2001 |
perseant | Be more careful about when we update ctime/mtime. In particular, if we
are only writing indirect blocks, that doesn't count for mtime; and when
we first create a vnode, that certainly *does not* count for ctime
(a bug that's been there from the beginning).
This does not change the fact that mtime might still be set after write(2)
is "completed", but it does make the atime-in-the-ifile code have some
effect (noticeable less degradation of read time after an intervening
large write).
|
| 1.35.4.3 | 02-Jul-2001 |
perseant | Change disk addressing unit to be the fragment, instead of the disk sector.
All quantities in the superblock, inodes, indirect blocks, etc. refer now
to this abstract unit (called "fsb" as it is in FFS) instead of disk sectors;
as a consequence segment summary blocks have to be multiples of a fragment in
size. In v1 filesystems, compatibility code ensures that 1 fsb == 1 sector,
regardless of fragment size.
Fragments can now range in size between 512 and 32k; in the event that
LFS_LABELPAD (8k) is smaller than the disk address unit size, an extra
proto-superblock is kept at 8k from the beginning of the disk, to be used
*only* to locate the real superblocks. (Not all of the userland knows about
this yet.)
Almost all of this was done not by me, but by joff.
|
| 1.35.4.2 | 29-Jun-2001 |
perseant | Get rid of __P(), protoizing where it had not already been done
|
| 1.35.4.1 | 27-Jun-2001 |
perseant | Import of what I've been calling "LFSv2", that is, LFS with some features
added that require changes to the on-disk data structures. These include:
- 64-bit time in everything but inodes
- User-specified segment offset, and segment size no longer
restricted to PO2.
- Serial number on segment summaries in addition to timestamp, and
a new volume identifier, to make roll-forward feasible without
fear of finding old data and thinking it was new.
Although I think this version works at least as well as what's on the trunk,
we're not done yet; hence this commit is going in on a branch and not on
the trunk. Enhancements that are not here yet include fragment addressing,
like FFS does, instead of block addressing.
|
| 1.35.2.13 | 03-Jan-2003 |
thorpej | Sync with HEAD.
|
| 1.35.2.12 | 29-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.35.2.11 | 19-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.35.2.10 | 11-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.35.2.9 | 01-Aug-2002 |
nathanw | Catch up to -current.
|
| 1.35.2.8 | 15-Jul-2002 |
nathanw | Whitespace.
|
| 1.35.2.7 | 24-Jun-2002 |
nathanw | Curproc->curlwp renaming.
Change uses of "curproc->l_proc" back to "curproc", which is more like the
original use. Bare uses of "curproc" are now "curlwp".
"curproc" is now #defined in proc.h as ((curlwp) ? (curlwp)->l_proc) : NULL)
so that it is always safe to reference curproc (*de*referencing curproc
is another story, but that's always been true).
|
| 1.35.2.6 | 20-Jun-2002 |
nathanw | Catch up to -current.
|
| 1.35.2.5 | 28-Feb-2002 |
nathanw | Catch up to -current.
|
| 1.35.2.4 | 08-Jan-2002 |
nathanw | Catch up to -current.
|
| 1.35.2.3 | 14-Nov-2001 |
nathanw | Catch up to -current.
|
| 1.35.2.2 | 24-Aug-2001 |
nathanw | Catch up with -current.
|
| 1.35.2.1 | 05-Mar-2001 |
nathanw | Initial commit of scheduler activations and lightweight process support.
|
| 1.36.4.1 | 12-Nov-2001 |
thorpej | Sync the thorpej-mips-cache branch with -current.
|
| 1.43.2.1 | 15-Jul-2002 |
gehenna | catch up with -current.
|
| 1.66.2.7 | 10-Nov-2005 |
skrll | Sync with HEAD. Here we go again...
|
| 1.66.2.6 | 01-Apr-2005 |
skrll | Sync with HEAD.
|
| 1.66.2.5 | 08-Mar-2005 |
skrll | Sync with HEAD.
|
| 1.66.2.4 | 04-Mar-2005 |
skrll | Sync with HEAD.
Hi Perry!
|
| 1.66.2.3 | 21-Sep-2004 |
skrll | Fix the sync with head I botched.
|
| 1.66.2.2 | 18-Sep-2004 |
skrll | Sync with HEAD.
|
| 1.66.2.1 | 03-Aug-2004 |
skrll | Sync with HEAD
|
| 1.77.10.1 | 19-Mar-2005 |
yamt | sync with head. xen and whitespace. xen part is not finished.
|
| 1.77.8.1 | 29-Apr-2005 |
kent | sync with -current
|
| 1.77.6.1 | 10-May-2005 |
riz | Pull up the following revisions (requested by perseant in ticket #1281):
1.8 sys/ufs/lfs/TODO
1.75 sys/ufs/lfs/lfs.h (via patch)
1.74 sys/ufs/lfs/lfs_alloc.c (via patch)
1.49, 1.51 sys/ufs/lfs/lfs_balloc.c (1.51 via patch)
1.78 sys/ufs/lfs/lfs_bio.c
1.62 sys/ufs/lfs/lfs_extern.h (via patch)
1.156 sys/ufs/lfs/lfs_segment.c (via patch)
1.48 sys/ufs/lfs/lfs_subr.c
1.101 sys/ufs/lfs/lfs_syscalls.c
1.163 sys/ufs/lfs/lfs_vfsops.c (via patch)
1.134 sys/ufs/lfs/lfs_vnops.c (via patch)
1.61 sys/ufs/ufs/ufs_readwrite.c (via patch)
1.20 libexec/lfs_cleanerd/clean.h (via patch)
1.52 libexec/lfs_cleanerd/cleanerd.c (via patch)
1.41 libexec/lfs_cleanerd/library.c (via patch)
1.4 regress/sys/fs/lfs/newfs_fsck/Makefile
1.2 regress/sys/fs/lfs/newfs_fsck/mkfs_mount
1.2 regress/sys/fs/lfs/newfs_fsck/smallfiles
1.3 sbin/fsck_lfs/bufcache.c
1.3 sbin/fsck_lfs/bufcache.h
1.3 sbin/fsck_lfs/lfs.h
1.8 sbin/fsck_lfs/lfs.c (via patch)
1.8 sbin/fsck_lfs/pass3.c (via patch)
1.18 sbin/fsck_lfs/pass0.c (via patch)
1.18 sbin/fsck_lfs/utilities.c (via patch)
1.7 sbin/fsck_lfs/segwrite.c
1.19 sbin/fsck_lfs/setup.c (via patch)
1.3 sbin/newfs_lfs/Makefile
0 sbin/newfs_lfs/lfs.c (yes, remove it)
1.1 sbin/newfs_lfs/make_lfs.c
1.15 sbin/newfs_lfs/newfs.c (via patch)
Various minor LFS improvements.
Kernel:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this. Should fix PR #29045.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
Fixes PR #26680.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
cleaner:
* Adapt lfs_cleanerd to use the fcntl call to get the Ifile filehandle,
so it need not be in the namespace.
* Make lfs_cleanerd be more careful when there are very few available
segments.
* Make lfs_cleanerd less verbose when the filesystem is unmounted.
newfs_lfs, fsck_lfs, and regression:
* Extend the lfs library from fsck_lfs(8) so that it can be used with a
not-yet-existent LFS. Make newfs_lfs(8) use this library, so it can
create LFSs whose Ifile is larger than one segment. Addresses PR #11110.
* Make newfs_lfs(8) use strsuftoi64() for its arguments, a la newfs(8).
* Make fsck_lfs(8) respect the "file system is clean" flag.
* Don't let fsck_lfs(8) think it has dirty blocks when invoked with the
-n flag.
* Remove the Ifile from the filesystem namespace. The cleaner now uses
a fcntl call on the root inode to find the Ifile filehandle. (As a
side-effect, addresses PR #29144.)
|
| 1.81.2.5 | 10-Aug-2006 |
tron | Apply patch (requested by fair in perseant #1457):
Bring LFS up to current, including a patch (1.95 lfs_alloc.c) that
should prevent the inode free list errors seen on the STABLE branch
subsequent to pullup ticket #1327.
|
| 1.81.2.4 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.92
sys/ufs/lfs/lfs.h: revision 1.105
sys/ufs/lfs/lfs_vfsops.c: revision 1.207
sys/ufs/lfs/lfs_subr.c: revision 1.59
sys/ufs/lfs/lfs_vnops.c: revision 1.173
sys/ufs/lfs/lfs_bio.c: revision 1.92
Introduce another per-filesystem parameter, lfs_resvseg, to separate the
notion of "how many segments are reserved for the cleaner" from that of
"how many segments are not counted in lfs_bfree". The default value
used for existing filesystems is the same as the previous implicit value
of (lfs_minfreeseg / 2 + 1), modulo some sanity checking.
Count pending dirops on a per-filesystem basis, since once we start
writing them we can't stop until we're done. This seems to help stave off
the "no clean segments" panic in the case of filling the filesystem with
directories and small files (e.g. simultaneously unpacking more copies of
pkgsrc than will fit).
|
| 1.81.2.3 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.102
sys/ufs/lfs/lfs_segment.c: revision 1.173
sys/ufs/lfs/lfs_vnops.c: revision 1.167 via patch
sys/ufs/lfs/lfs_bio.c: revision 1.91
Make lfs_vref/lfs_vunref not need to know about VXLOCK and VFREEING
explicitly (especially since we didn't know about VFREEING at all before),
but notice the EBUSY return from vget() instead.
Fix some more MP locking protocol issues, most of which were pointed out by
Christian Ehrhardt this morning on tech-kern.
|
| 1.81.2.2 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.152
sys/ufs/lfs/lfs_debug.c: revision 1.31
sys/ufs/lfs/lfs_subr.c: revision 1.53
sys/ufs/lfs/lfs_extern.h: revision 1.68
sys/ufs/lfs/lfs_inode.c: revision 1.96
sys/ufs/lfs/lfs_bio.c: revision 1.86
sys/ufs/lfs/lfs_alloc.c: revision 1.83
sys/ufs/lfs/lfs_vfsops.c: revision 1.181
sys/ufs/lfs/lfs.h: revision 1.88
sys/ufs/lfs/lfs_segment.c: revision 1.164
- sprinkle const
- avoid shadow variables.
|
| 1.81.2.1 | 07-May-2005 |
tron | Apply patch (requested by perseant in ticket #242):
* fsck_lfs buffer cache fixes, including PR #29151
* Change fsck_lfs phase 0 message to reflect reality
* fsck_lfs: check phase 5 (cleanerinfo accounting) even on
roll-forward
* Keep better track of the free list during roll-forward, avoiding
a core dump
* Improve hash table use for fsck_lfs buffer and vnode cache
* Document fsck_lfs flag -f, and implement -q
* Add resize_lfs, including kernel support
* Add LFS to mountd's list of exportable filesystem types
* Make the LFS lkm work again [christos@]
* Add MP locking to the LFS kernel subsystem
* Fix pager_map deadlock in lfs_putpages()
* Avoid incomplete file extension that looks like "partial
truncation" to fsck
* Use lfs_malloc for cleaner malloc, since the cleaner often runs
in low-memory conditions.
* Use splay trees, not hash table, to track page allocation for
write.
* Fix mkdir panic on full fs
* Fix page accounting leak by counting differently.
* Use rightly named structure for lfs_getattr [skrll@]
* Cosmetic changes for readability.
|
| 1.86.2.7 | 27-Feb-2008 |
yamt | sync with head.
|
| 1.86.2.6 | 04-Feb-2008 |
yamt | sync with head.
|
| 1.86.2.5 | 21-Jan-2008 |
yamt | sync with head
|
| 1.86.2.4 | 27-Oct-2007 |
yamt | sync with head.
|
| 1.86.2.3 | 03-Sep-2007 |
yamt | sync with head.
|
| 1.86.2.2 | 30-Dec-2006 |
yamt | sync with head.
|
| 1.86.2.1 | 21-Jun-2006 |
yamt | sync with head.
|
| 1.87.2.1 | 15-Jan-2006 |
yamt | sync with head.
|
| 1.89.6.4 | 11-Aug-2006 |
yamt | sync with head
|
| 1.89.6.3 | 24-May-2006 |
yamt | sync with head.
|
| 1.89.6.2 | 13-Mar-2006 |
yamt | sync with head.
|
| 1.89.6.1 | 05-Mar-2006 |
yamt | separate page replacement policy from the rest of kernel.
|
| 1.89.4.2 | 01-Jun-2006 |
kardel | Sync with head.
|
| 1.89.4.1 | 22-Apr-2006 |
simonb | Sync with head.
|
| 1.89.2.1 | 09-Sep-2006 |
rpaulo | sync with head
|
| 1.90.4.1 | 24-May-2006 |
tron | Merge 2006-05-24 NetBSD-current into the "peter-altq" branch.
|
| 1.90.2.3 | 11-May-2006 |
elad | sync with head
|
| 1.90.2.2 | 06-May-2006 |
christos | - Move kauth_cred_t declaration to <sys/types.h>
- Cleanup struct ucred; forward declarations that are unused.
- Don't include <sys/kauth.h> in any header, but include it in the c files
that need it.
Approved by core.
|
| 1.90.2.1 | 19-Apr-2006 |
elad | sync with head.
|
| 1.93.4.1 | 13-Jul-2006 |
gdamore | Merge from HEAD.
|
| 1.94.4.1 | 18-Nov-2006 |
ad | Sync with head.
|
| 1.95.2.2 | 10-Dec-2006 |
yamt | sync with head.
|
| 1.95.2.1 | 22-Oct-2006 |
yamt | sync with head
|
| 1.98.16.1 | 03-Sep-2007 |
wrstuden | Sync w/ NetBSD-4-RC_1
|
| 1.98.10.1 | 11-Jul-2007 |
mjf | Sync with head.
|
| 1.98.8.7 | 24-Aug-2007 |
ad | Sync with buffer cache locking changes. See buf.h/vfs_bio.c for details.
Some minor portions are incomplete and needs to be verified as a whole.
|
| 1.98.8.6 | 20-Aug-2007 |
ad | Sync with HEAD.
|
| 1.98.8.5 | 19-Aug-2007 |
ad | - Back out the biodone() changes.
- Eliminate B_ERROR (from HEAD).
|
| 1.98.8.4 | 23-Jun-2007 |
ad | - Lock v_cleanblkhd, v_dirtyblkhd, v_numoutput with the vnode's interlock.
Get rid of global_v_numoutput_lock. Partially incomplete as the buffer
cache locking doesn't work very well and needs an overhaul.
- Some changes to try and make softdep MP safe. Untested.
|
| 1.98.8.3 | 08-Jun-2007 |
ad | Sync with head.
|
| 1.98.8.2 | 13-May-2007 |
ad | - Pass the error number and residual count to biodone(), and let it handle
setting error indicators. Prepare to eliminate B_ERROR.
- Add a flag argument to brelse() to be set into the buf's flags, instead
of doing it directly. Typically used to set B_INVAL.
- Add a "struct cpu_info *" argument to kthread_create(), to be used to
create bound threads. Change "bool mpsafe" to "int flags".
- Allow exit of LWPs in the IDL state when (l != curlwp).
- More locking fixes & conversion to the new API.
|
| 1.98.8.1 | 13-Mar-2007 |
ad | Pull in the initial set of changes for the vmlocking branch.
|
| 1.98.4.2 | 17-May-2007 |
yamt | sync with head.
|
| 1.98.4.1 | 07-May-2007 |
yamt | sync with head.
|
| 1.98.2.1 | 05-Jun-2007 |
bouyer | Pull up following revision(s) (requested by perseant in ticket #703):
sys/miscfs/genfs/genfs.h 1.21
sys/miscfs/genfs/genfs_vnops.c 1.151
sys/ufs/lfs/lfs.h 1.119, 1.120
sys/ufs/lfs/lfs_bio.c 1.99-101
sys/ufs/lfs/lfs_extern.h 1.89
sys/ufs/lfs/lfs_inode.c 1.108, 1.109
sys/ufs/lfs/lfs_segment.c 1.197, 1.199, 1.200
sys/ufs/lfs/lfs_subr.c 1.69, 1.70
sys/ufs/lfs/lfs_syscalls.c 1.119
sys/ufs/lfs/lfs_vfsops.c 1.234, 1.235
sys/ufs/lfs/lfs_vnops.c 1.195, 1.196, 1.200, 1.202-206
Reduce busy waiting in lfs_putpages(), and other LFS improvements.
|
| 1.102.2.1 | 15-Aug-2007 |
skrll | Sync with HEAD.
|
| 1.103.10.2 | 29-Jul-2007 |
ad | It's not a good idea for device drivers to modify b_flags, as they don't
need to understand the locking around that field. Instead of setting
B_ERROR, set b_error instead. b_error is 'owned' by whoever completes
the I/O request.
|
| 1.103.10.1 | 29-Jul-2007 |
ad | file lfs_bio.c was added on branch matt-mips64 on 2007-07-29 13:31:15 +0000
|
| 1.103.8.1 | 14-Oct-2007 |
yamt | sync with head.
|
| 1.103.6.3 | 23-Mar-2008 |
matt | sync with HEAD
|
| 1.103.6.2 | 09-Jan-2008 |
matt | sync with HEAD
|
| 1.103.6.1 | 06-Nov-2007 |
matt | sync with HEAD
|
| 1.103.4.1 | 26-Oct-2007 |
joerg | Sync with HEAD.
Follow the merge of pmap.c on i386 and amd64 and move
pmap_init_tmp_pgtbl into arch/x86/x86/pmap.c. Modify the ACPI wakeup
code to restore CR4 before jumping back into kernel space as the large
page option might cover that.
|
| 1.106.10.1 | 02-Jan-2008 |
bouyer | Sync with HEAD
|
| 1.106.6.5 | 19-Dec-2007 |
ad | Use a global lfs_lock.
|
| 1.106.6.4 | 19-Dec-2007 |
ad | Fix some more problems w/lfs on this branch.
|
| 1.106.6.3 | 19-Dec-2007 |
ad | Get lfs mostly working.
|
| 1.106.6.2 | 08-Dec-2007 |
ad | Minor locking fixes.
|
| 1.106.6.1 | 04-Dec-2007 |
ad | Pull the vmlocking changes into a new branch.
|
| 1.106.4.1 | 18-Feb-2008 |
mjf | Sync with HEAD.
|
| 1.110.10.3 | 11-Aug-2010 |
yamt | sync with head.
|
| 1.110.10.2 | 11-Mar-2010 |
yamt | sync with head
|
| 1.110.10.1 | 16-May-2008 |
yamt | sync with head.
|
| 1.110.8.1 | 18-May-2008 |
yamt | sync with head.
|
| 1.110.6.1 | 02-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.114.18.1 | 19-Dec-2013 |
matt | Adapt to new uvm_estimatepageable arguments
|
| 1.116.2.2 | 17-Aug-2010 |
uebayasi | Sync with HEAD.
|
| 1.116.2.1 | 30-Apr-2010 |
uebayasi | Sync with HEAD.
|
| 1.117.2.2 | 03-Jul-2010 |
rmind | sync with head
|
| 1.117.2.1 | 16-Mar-2010 |
rmind | Change struct uvm_object::vmobjlock to be dynamically allocated with
mutex_obj_alloc(). It allows us to share the locks among UVM objects.
|
| 1.118.6.1 | 23-Jun-2011 |
cherry | Catchup with rmind-uvmplock merge.
|
| 1.120.6.1 | 18-Feb-2012 |
mrg | merge to -current.
|
| 1.120.2.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.120.2.1 | 17-Apr-2012 |
yamt | sync with head
|
| 1.121.2.1 | 17-Mar-2012 |
bouyer | Pull up following revision(s) (requested by perseant in ticket #116):
sys/ufs/lfs/lfs_alloc.c: revision 1.112
tests/fs/vfs/t_rmdirrace.c: revision 1.9
tests/fs/vfs/t_renamerace.c: revision 1.25
sys/ufs/lfs/lfs_vnops.c: revision 1.240
sys/ufs/lfs/lfs_segment.c: revision 1.224
sys/ufs/lfs/lfs_bio.c: revision 1.122
sys/ufs/lfs/lfs_vfsops.c: revision 1.294
sbin/newfs_lfs/make_lfs.c: revision 1.19
sys/ufs/lfs/lfs.h: revision 1.136
Pass t_renamerace and t_rmdirrace tests.
Adapt dholland@'s fix to ufs_rename to fix PR kern/43582. Address several
other MP locking issues discovered during the course of investigating the
same problem.
Removed extraneous vn_lock() calls on the Ifile, since the Ifile writes
are controlled by the segment lock.
Fix PR kern/45982 by deemphasizing the estimate of how much metadata
will fill the empty space on disk when the disk is nearly empty
(t_renamerace crates a lot of inode blocks on a tiny empty disk).
|
| 1.122.2.3 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.122.2.2 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.122.2.1 | 23-Jun-2013 |
tls | resync from head
|
| 1.125.2.2 | 18-May-2014 |
rmind | sync with head
|
| 1.125.2.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.128.6.3 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.128.6.2 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.128.6.1 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.135.4.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.135.2.2 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.135.2.1 | 20-Mar-2017 |
pgoyette | Sync with HEAD
|
| 1.139.4.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.141.4.1 | 25-Jun-2018 |
pgoyette | Sync with HEAD
|
| 1.142.6.1 | 17-Aug-2020 |
martin | Pull up following revision(s) (requested by riastradh in ticket #1050):
sys/ufs/lfs/lfs_subr.c: revision 1.101
sys/ufs/lfs/lfs_subr.c: revision 1.102
sys/ufs/lfs/lfs_inode.c: revision 1.158
sys/ufs/lfs/lfs_inode.h: revision 1.25
sys/ufs/lfs/lfs_balloc.c: revision 1.95
sys/ufs/lfs/lfs_pages.c: revision 1.21
sys/ufs/lfs/lfs_vnops.c: revision 1.330
sys/ufs/lfs/lfs_alloc.c: revision 1.140 (patch)
sys/ufs/lfs/lfs_alloc.c: revision 1.141 (patch)
lib/libp2k/p2k.c: revision 1.72
sys/ufs/lfs/lfs.h: revision 1.205
sys/ufs/lfs/lfs.h: revision 1.206
sys/ufs/lfs/lfs_segment.c: revision 1.284
sys/ufs/lfs/lfs.h: revision 1.207
sys/ufs/lfs/lfs_segment.c: revision 1.285
sys/ufs/lfs/lfs_debug.c: revision 1.55
sys/ufs/lfs/lfs_rename.c: revision 1.23
usr.sbin/dumplfs/dumplfs.c: revision 1.65
sys/ufs/lfs/lfs_vfsops.c: revision 1.371
sys/arch/i386/stand/efiboot/bootx64/Makefile: revision 1.3
sys/ufs/lfs/lfs_vfsops.c: revision 1.372
sys/ufs/lfs/lfs_vfsops.c: revision 1.373
sbin/fsck_lfs/pass1.c: revision 1.46
sys/ufs/lfs/lfs_vnops.c: revision 1.326
sys/ufs/lfs/lfs_vnops.c: revision 1.327
sys/ufs/lfs/lfs_vfsops.c: revision 1.375 (patch)
sys/ufs/lfs/lfs_vnops.c: revision 1.328
sys/ufs/lfs/lfs_subr.c: revision 1.98
sys/ufs/lfs/lfs_extern.h: revision 1.116
sys/ufs/lfs/lfs_vnops.c: revision 1.329
sys/ufs/lfs/lfs_subr.c: revision 1.99
sys/ufs/lfs/lfs_extern.h: revision 1.117
sys/ufs/lfs/lfs_accessors.h: revision 1.49
sys/ufs/lfs/lfs_extern.h: revision 1.118
sys/rump/fs/lib/liblfs/Makefile: revision 1.15
sys/ufs/lfs/lfs_bio.c: revision 1.146 (patch)
sys/ufs/lfs/lfs_bio.c: revision 1.147
sys/ufs/lfs/lfs_subr.c: revision 1.100
Fix kassert in lfs by initializing vp first.
Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
Teach lfs to transition ro<->rw.
Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
Assert lfs_writer where I think we can now prove it.
Serialize access to the splay tree with lfs_lock.
Change some cheap KDASSERT into KASSERT.
Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
Call lfs_orphan in lfs_rename while we're still in the dirop.
lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
Teach LFS_ORPHAN_NEXTFREE about lfs64.
Dust off the orphan detection code and try to make it work.
Fix !DIAGNOSTIC compile
Fix userland references to LFS_ORPHAN_NEXTFREE.
Forgot to grep for these or do a full distribution build, oops!
Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
OR into bp->b_cflags; don't overwrite.
CTASSERT lfs on-disk structure sizes.
Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
Fix clang build after packed lfs64 accessor change.
Suppress spurious address-of-packed error in rump lfs too.
|
| 1.142.2.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.144.2.1 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.30 | 02-Aug-2015 |
dholland | Add a (draft) 64-bit superblock. Make things build again.
Add pieces of support for using both superblock types where
convenient, and specifically to the superblock accessors, but don't
actually enable it anywhere.
First substantive step on PR 50000.
|
| 1.29 | 06-Jun-2013 |
dholland | branches: 1.29.10;
Cleanups and hacks to make lfs userland stuff build:
- lfs_cksum.c doesn't actually need ulfs_inode.h any more.
- neither does lfs_itimes.c.
- add hacks to fsck_lfs to make it compile.
- add hacks to newfs_lfs to make it compile.
- fix warning in ulfs_quota.c when quotas are fully disabled
(as I guess is happening with the rumpity version)
XXX: This commit adds -I${NETBSDSRCDIR}/sys to the Makefiles for
XXX: fsck_lfs, newfs_lfs, and lfs_cleanerd. This needs to be cleaned
XXX: up ASAP; but I consider this less problematic in the short term
XXX: than spewing ulfs_*.h into /usr/include.
|
| 1.28 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.27 | 28-Apr-2008 |
martin | branches: 1.27.34; 1.27.44;
Remove clause 3 and 4 from TNF licenses
|
| 1.26 | 11-Dec-2005 |
christos | branches: 1.26.70; 1.26.72; 1.26.74;
merge ktrace-lwp.
|
| 1.25 | 26-Feb-2005 |
perry | nuke trailing whitespace
|
| 1.24 | 09-Mar-2004 |
yamt | branches: 1.24.8; 1.24.10;
calculate data checksum inline.
|
| 1.23 | 07-Aug-2003 |
agc | Move UCB-licensed code from 4-clause to 3-clause licence.
Patches provided by Joel Baker in PR 22364, verified by myself.
|
| 1.22 | 20-Feb-2003 |
perseant | branches: 1.22.2;
Tabify, and fix some comment alignment problems.
|
| 1.21 | 17-Feb-2003 |
perseant | Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now
(there are still some details to work out) but expect that to go
away soon. To support these basic changes (creation of lfs_putpages,
lfs_gop_write, mods to lfs_balloc) several other changes were made, to
wit:
* Create a writer daemon kernel thread whose purpose is to handle page
writes for the pagedaemon, but which also takes over some of the
functions of lfs_check(). This thread is started the first time an
LFS is mounted.
* Add a "flags" parameter to GOP_SIZE. Current values are
GOP_SIZE_READ, meaning that the call should return the size of the
in-core version of the file, and GOP_SIZE_WRITE, meaning that it
should return the on-disk size. One of GOP_SIZE_READ or
GOP_SIZE_WRITE must be specified.
* Instead of using malloc(...M_WAITOK) for everything, reserve enough
resources to get by and use malloc(...M_NOWAIT), using the reserves if
necessary. Use the pool subsystem for structures small enough that
this is feasible. This also obsoletes LFS_THROTTLE.
And a few that are not strictly necessary:
* Moves the LFS inode extensions off onto a separately allocated
structure; getting closer to LFS as an LKM. "Welcome to 1.6O."
* Unified GOP_ALLOC between FFS and LFS.
* Update LFS copyright headers to correct values.
* Actually cast to unsigned in lfs_shellsort, like the comment says.
* Keep track of which segments were empty before the previous
checkpoint; any segments that pass two checkpoints both dirty and
empty can be summarily cleaned. Do this. Right now lfs_segclean
still works, but this should be turned into an effectless
compatibility syscall.
|
| 1.20 | 16-Jun-2002 |
perseant | For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.19 | 15-Nov-2001 |
lukem | branches: 1.19.8; 1.19.10;
don't need <sys/types.h> when including <sys/param.h>
|
| 1.18 | 08-Nov-2001 |
lukem | add RCSID
|
| 1.17 | 26-Oct-2001 |
lukem | remove #include <ufs/ufs/quota.h> where it was just to appease
<ufs/ufs/inode.h>, since the latter now includes the former. leave the former
in source that obviously uses specific bits of it (for completeness.)
|
| 1.16 | 13-Jul-2001 |
perseant | branches: 1.16.4;
Merge the short-lived perseant-lfsv2 branch into the trunk.
Kernels and tools understand both v1 and v2 filesystems; newfs_lfs
generates v2 by default. Changes for the v2 layout include:
- Segments of non-PO2 size and arbitrary block offset, so these can be
matched to convenient physical characteristics of the partition (e.g.,
stripe or track size and offset).
- Address by fragment instead of by disk sector, paving the way for
non-512-byte-sector devices. In theory fragments can be as large
as you like, though in reality they must be smaller than MAXBSIZE in size.
- Use serial number and filesystem identifier to ensure that roll-forward
doesn't get old data and think it's new. Roll-forward is enabled for
v2 filesystems, though not for v1 filesystems by default.
- The inode free list is now a tailq, paving the way for undelete (undelete
is not yet implemented, but can be without further non-backwards-compatible
changes to disk structures).
- Inode atime information is kept in the Ifile, instead of on the inode;
that is, the inode is never written *just* because atime was changed.
Because of this the inodes remain near the file data on the disk, rather
than wandering all over as the disk is read repeatedly. This speeds up
repeated reads by a small but noticeable amount.
Other changes of note include:
- The ifile written by newfs_lfs can now be of arbitrary length, it is no
longer restricted to a single indirect block.
- Fixed an old bug where ctime was changed every time a vnode was created.
I need to look more closely to make sure that the times are only updated
during write(2) and friends, not after-the-fact during a segment write,
and certainly not by the cleaner.
|
| 1.15 | 04-Feb-2001 |
christos | branches: 1.15.2; 1.15.4; 1.15.6;
don't include lfs_extern.h; ufs/inode.h does too.
|
| 1.14 | 25-Nov-2000 |
perseant | Use u_int32_t instead of u_long to compute LFS checksums, since the
checksum is stored in a u_int32_t.
|
| 1.13 | 09-Sep-2000 |
perseant | Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
|
| 1.12 | 30-Mar-2000 |
augustss | branches: 1.12.4;
Remove register declarations.
|
| 1.11 | 25-Mar-1999 |
perseant | branches: 1.11.8;
Change lfs_sb_cksum to use offsetof() instead of an inlined version.
Fix lfs_vref/lfs_vunredf to ignore VXLOCKed vnodes that are also being
flushed.
Improve the debugging messages somewhat.
|
| 1.10 | 10-Mar-1999 |
perseant | New sources should leave the LFS in a more-or-less working state. Changes
include:
- DIROP segregation is enabled, and greater care is taken
to make sure that a checkpoint completes. Fsck is not
needed to remount the filesystem.
- Several checks to make sure that the LFS subsystem does not
overuse various resources (memory, in particular).
- The cleaner routines, lfs_markv in particular, are completely
rewritten. A buffer overflow is removed. Greater care is taken
to ensure that inodes come from where lfs_cleanerd say they come
from (so we know nothing has changed since lfs_bmapv was called).
- Fragment allocation is fixed, so that writes beyond end-of-file
do the right thing.
|
| 1.9 | 11-Sep-1998 |
pk | PR#6032: define fixed sized on-disk superblock structure.
|
| 1.8 | 01-Mar-1998 |
fvdl | Merge with Lite2 + local changes
|
| 1.7 | 15-Sep-1997 |
lukem | prototype lfs_cksum ifndef KERNEL
|
| 1.6 | 16-Feb-1996 |
christos | branches: 1.6.12;
Protect include in lfs_cksum.c so that it can be used by userland programs.
|
| 1.5 | 09-Feb-1996 |
christos | lfs prototypes
|
| 1.4 | 14-Dec-1994 |
mycroft | Sync with CSRG.
|
| 1.3 | 20-Sep-1994 |
cgd | c syntax
|
| 1.2 | 29-Jun-1994 |
cgd | New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
|
| 1.1 | 08-Jun-1994 |
mycroft | branches: 1.1.1;
Update to 4.4-Lite fs code, with local changes.
|
| 1.1.1.2 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite2
|
| 1.1.1.1 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite for reference
|
| 1.6.12.1 | 22-Sep-1997 |
thorpej | Update marc-pcmcia branch from trunk.
|
| 1.11.8.3 | 11-Feb-2001 |
bouyer | Sync with HEAD.
|
| 1.11.8.2 | 08-Dec-2000 |
bouyer | Sync with HEAD.
|
| 1.11.8.1 | 20-Nov-2000 |
bouyer | Update thorpej_scsipi to -current as of a month ago
|
| 1.12.4.1 | 14-Sep-2000 |
perseant | Pull up recent LFS kernel changes (approved by thorpej):
ufs/ufs/inode.h, 1.20--1.22 (add i_lfs_effnblks extension ;
make ITIMES aware of LFS_ITIMES;
_LKM protection so userland progs
compile)
ufs/ufs/ufs_vnops.c, 1.69, 1.71 (remove IN_ADIROP;
use ITIMES instead of FFS_ITIMES)
ufs/ufs/ufs_readwrite.c, 1.27 (use lfs_reserve in lfs_write)
ufs/lfs/lfs.h, 1.26--1.32 (define LFS_EST_* macros ;
change MIN_FREE_SEGS to lfs_minfreesegs ;
add avail and bfree to CLEANERINFO ;
change lfs_uinodes to signed ;
change lfs_dmeta to signed ;
add whitespace to line up structure
members ;
explicit cast to int32_t in LFS_EST_*
macros)
ufs/lfs/lfs_alloc.c, back out 1.34.2.3 (pullups of 1.39, 1.40);
then pull up 1.38 (clean up on error)
1.39--1.43 (restore fvdl's ufs_hashlock fix ;
restore fvdl's ufs_hashlock fix ;
set i_lfs_effnblks ;
use UINO macros ;
add comments and fix long lines)
ufs/lfs/lfs_balloc.c, 1.19 (don't succeed halfway)
1.21--1.25 (use i_lfs_effnblks ;
fix i_lfs_effnblks computation and
quieten ;
fix i_ffs_blocks in unwritten fragment ;
remove useless debugging check ;
add comments and (c) 2000)
ufs/lfs/lfs_bio.c, 1.24--1.30 (cleanup and make lfs_flush_fs take
"struct lfs *" instead of "struct
mount *" ;
use lfs_minfreeseg instead of
MIN_FREE_SEGS ;
use UINO macros, and copy bfree/avail
to CLEANERINFO ;
add lfs_reserve function ;
1.28--1.30 fix printf formatting)
ufs/lfs/lfs_cksum.c, 1.13 (add (c) 2000)
ufs/lfs/lfs_debug.c, 1.11 (use btodb instead of DEV_BSIZE)
ufs/lfs/lfs_extern.h, 1.18, 1.20--1.21 (function prototype changes)
ufs/lfs/lfs_inode.c, 1.38 (rewrite lfs_truncate from
ffs_truncate)
1.40--1.44 (count written and unwritten blocks
seperately ;
use disk block units instead of bytes ;
remove unnecessary "mod" variable ;
correct B_DELWRI to avoid bawrite panic ;
use lfs_reserve)
ufs/lfs/lfs_segment.c, 1.52-1.59 (use lfs_dmeta to note used summaries ;
check for UNWRITTEN in indirect blocks ;
more debugging stuff inside #ifdef
DEBUG_LFS ;
use LK_CANRECURSE ;
don't drop dirty indirect blocks ;
use UINO macros ;
don't hose the free list ;
use btodb() instead of DEV_BSIZE ;
make it compile again (oops))
ufs/lfs/lfs_subr.c, 1.16--1.17 (check for locked inodes before
changing ;
use btodb() instead of DEV_BSIZE, (c)
2000)
ufs/lfs/lfs_syscalls.c, back out 1.41.4.2 (fvdl's ufs_hashlock fix);
then pull up 1.43 (use lfs_dmeta)
1.44--1.45 (restore fvdl's ufs_hashlock fix)
1.46--1.47 (fix lfs_avail leakage from sblock
segments ;
use UINO macros)
1.49 (bounds-check inode numbers in
lfs_markv)
ufs/lfs/lfs_vfsops.c, 1.53 (use LFS_EST_* macros in lfs_statfs)
1.56--1.58 (initialize lfs_minfreeseg, lfs_effnblk ;
initialize lfs_uinodes ;
initialize lfs_ravail)
ufs/lfs/lfs_vnops.c, 1.40 (remove VDIROP from removed files)
1.42--1.44 (move SET_ENDOP below the removal of
VDIROP ;
use UINO macros and add lfs_itimes
function ;
use lfs_reserve in dirops)
|
| 1.15.6.3 | 23-Jun-2002 |
jdolecek | catch up with -current on kqueue branch
|
| 1.15.6.2 | 10-Jan-2002 |
thorpej | Sync kqueue branch with -current.
|
| 1.15.6.1 | 03-Aug-2001 |
lukem | update to -current
|
| 1.15.4.1 | 29-Jun-2001 |
perseant | Get rid of __P(), protoizing where it had not already been done
|
| 1.15.2.4 | 20-Jun-2002 |
nathanw | Catch up to -current.
|
| 1.15.2.3 | 08-Jan-2002 |
nathanw | Catch up to -current.
|
| 1.15.2.2 | 14-Nov-2001 |
nathanw | Catch up to -current.
|
| 1.15.2.1 | 24-Aug-2001 |
nathanw | Catch up with -current.
|
| 1.16.4.1 | 12-Nov-2001 |
thorpej | Sync the thorpej-mips-cache branch with -current.
|
| 1.19.10.1 | 20-Jun-2002 |
lukem | Pull up revision 1.20 (requested by perseant in ticket #325):
For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.19.8.1 | 20-Jun-2002 |
gehenna | catch up with -current.
|
| 1.22.2.4 | 04-Mar-2005 |
skrll | Sync with HEAD.
Hi Perry!
|
| 1.22.2.3 | 21-Sep-2004 |
skrll | Fix the sync with head I botched.
|
| 1.22.2.2 | 18-Sep-2004 |
skrll | Sync with HEAD.
|
| 1.22.2.1 | 03-Aug-2004 |
skrll | Sync with HEAD
|
| 1.24.10.1 | 19-Mar-2005 |
yamt | sync with head. xen and whitespace. xen part is not finished.
|
| 1.24.8.1 | 29-Apr-2005 |
kent | sync with -current
|
| 1.26.74.1 | 16-May-2008 |
yamt | sync with head.
|
| 1.26.72.1 | 18-May-2008 |
yamt | sync with head.
|
| 1.26.70.1 | 02-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.27.44.2 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.27.44.1 | 23-Jun-2013 |
tls | resync from head
|
| 1.27.34.1 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.29.10.1 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.55 | 23-Feb-2020 |
riastradh | Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
|
| 1.54 | 01-Sep-2015 |
dholland | branches: 1.54.18; 1.54.22; 1.54.24;
The ifile's inode number is constant. (it is always 1)
Therefore, storing the value in the superblock and reading it out
again is silly and offers the opportunity for it to become corrupted.
So, don't do that (most of the code already didn't) and use the
existing constant instead. Initialize new 32-bit superblocks with
the value for the sake of old userland programs, but don't keep the
value in the 64-bit superblock at all.
(approved by Margo Seltzer)
|
| 1.53 | 01-Sep-2015 |
dholland | Make the inode fields in the 64-bit superblock 64 bits wide.
Reasoning as before.
Note that I am not going through and checking for 64->32 truncations
in inode numbers; I'm sure there are quite a few, but that's a project
for later.
|
| 1.52 | 12-Aug-2015 |
dholland | Hack up dinode usage to be 64 vs. 32 as needed. Part 1.
(This part changes the native lfs code; the ufs-derived code already
has 64 vs. 32 logic, but as aspects of it are unsafe, and don't
entirely interoperate cleanly with the lfs 64/32 stuff, pass 2 will be
rehashing that.)
|
| 1.51 | 12-Aug-2015 |
dholland | Provide 32-bit and 64-bit versions of FINFO.
This also entailed sorting out part of struct segment, as that
contains a pointer into the current FINFO data.
|
| 1.50 | 02-Aug-2015 |
dholland | Add a (draft) 64-bit superblock. Make things build again.
Add pieces of support for using both superblock types where
convenient, and specifically to the superblock accessors, but don't
actually enable it anywhere.
First substantive step on PR 50000.
|
| 1.49 | 02-Aug-2015 |
dholland | Use accessor functions for the version field of the lfs superblock.
I thought at first maybe the cases that test the version should be
rolled into the accessors, but on the whole I think the conclusion on
that is no.
|
| 1.48 | 02-Aug-2015 |
dholland | Second batch of 64 -> 32 truncations in lfs, along with more minor
tidyups and corrections in passing.
|
| 1.47 | 02-Aug-2015 |
dholland | Fix assorted 64 -> 32 truncations in lfs. Also, some minor tidyups and
corrections in passing.
|
| 1.46 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.45 | 25-Jul-2015 |
hannken | Another lfs superblock accessor (inside #ifdef 0).
|
| 1.44 | 25-Jul-2015 |
martin | Use accessors in DEBUG and DIAGNOSTIC code as well
|
| 1.43 | 18-Jun-2013 |
christos | branches: 1.43.10;
Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.
|
| 1.42 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.41 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.40 | 22-Jan-2013 |
dholland | Stuff UFS_ in front of a few of ufs's symbols to reduce namespace
pollution. Specifically:
ROOTINO -> UFS_ROOTINO
WINO -> UFS_WINO
NXADDR -> UFS_NXADDR
NDADDR -> UFS_NDADDR
NIADDR -> UFS_NIADDR
MAXSYMLINKLEN -> UFS_MAXSYMLINKLEN
MAXSYMLINKLEN_UFS[12] -> UFS[12]_MAXSYMLINKLEN (for consistency)
Sort out ext2fs's misuse of NDADDR and NIADDR; fortunately, these have
the same values in ext2fs and ffs.
No functional change intended.
|
| 1.39 | 17-Jul-2011 |
joerg | branches: 1.39.2; 1.39.12;
Retire varargs.h support. Move machine/stdarg.h logic into MI
sys/stdarg.h and expect compiler to provide proper builtins, defaulting
to the GCC interface. lint still has a special fallback.
Reduce abuse of _BSD_VA_LIST_ by defining __va_list by default and
derive va_list as required by standards.
|
| 1.38 | 19-Jul-2009 |
dholland | minor knf
|
| 1.37 | 28-Apr-2008 |
martin | branches: 1.37.14;
Remove clause 3 and 4 from TNF licenses
|
| 1.36 | 02-Jan-2008 |
ad | branches: 1.36.6; 1.36.8; 1.36.10;
Merge vmlocking2 to head.
|
| 1.35 | 12-Dec-2007 |
lukem | Move __KERNEL_RCSID() so that it's always available if this file is
compiled, even if DEBUG isn't defined.
(This matches the behaviour of various other source files that
provide functions only if DEBUG is enabled.)
|
| 1.34 | 22-Jul-2007 |
christos | branches: 1.34.6; 1.34.12; 1.34.14; 1.34.16; 1.34.18; 1.34.22;
make this compile again
|
| 1.33 | 11-Dec-2005 |
christos | branches: 1.33.30; 1.33.40;
merge ktrace-lwp.
|
| 1.32 | 19-Aug-2005 |
christos | 64 bit inode changes.
|
| 1.31 | 29-May-2005 |
christos | branches: 1.31.2;
- sprinkle const
- avoid shadow variables.
|
| 1.30 | 01-Apr-2005 |
perseant | Protect various per-fs structures with fs->lfs_interlock simple_lock, to
improve behavior in the multiprocessor case. Add debugging segment-lock
assertion statements.
|
| 1.29 | 26-Mar-2005 |
christos | make this compile again :-(
|
| 1.28 | 26-Mar-2005 |
christos | Use vlog(9). Open-coding vlog here breaks lkm's because including
<sys/kprintf.h> includes opt_multiprocessor.h. One could argue
that the lock stuff should just move to subr_prf.c since nothing
else uses it.
|
| 1.27 | 08-Mar-2005 |
simonb | branches: 1.27.2;
Tab Police.
|
| 1.26 | 08-Mar-2005 |
perseant | Straighten out the maze of ifdefs. Instead, consolidate all the debugging
stuff under '#ifdef DEBUG', and use sysctl knobs to turn on/off particular
parts of the debugging reporting (if DEBUG is enabled). Re-enable the LFS
statistics in sysctl, while I'm there. A bit of a rototill.
|
| 1.25 | 26-Feb-2005 |
perry | nuke trailing whitespace
|
| 1.24 | 30-Oct-2003 |
simonb | branches: 1.24.8; 1.24.10;
Remove some assigned-to but otherwise unused variables.
|
| 1.23 | 07-Aug-2003 |
agc | Move UCB-licensed code from 4-clause to 3-clause licence.
Patches provided by Joel Baker in PR 22364, verified by myself.
|
| 1.22 | 02-Apr-2003 |
fvdl | branches: 1.22.2;
Add support for UFS2. UFS2 is an enhanced FFS, adding support for
64 bit block pointers, extended attribute storage, and a few
other things.
This commit does not yet include the code to manipulate the extended
storage (for e.g. ACLs), this will be done later.
Originally written by Kirk McKusick and Network Associates Laboratories for
FreeBSD.
|
| 1.21 | 20-Feb-2003 |
perseant | Tabify, and fix some comment alignment problems.
|
| 1.20 | 17-Feb-2003 |
perseant | Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now
(there are still some details to work out) but expect that to go
away soon. To support these basic changes (creation of lfs_putpages,
lfs_gop_write, mods to lfs_balloc) several other changes were made, to
wit:
* Create a writer daemon kernel thread whose purpose is to handle page
writes for the pagedaemon, but which also takes over some of the
functions of lfs_check(). This thread is started the first time an
LFS is mounted.
* Add a "flags" parameter to GOP_SIZE. Current values are
GOP_SIZE_READ, meaning that the call should return the size of the
in-core version of the file, and GOP_SIZE_WRITE, meaning that it
should return the on-disk size. One of GOP_SIZE_READ or
GOP_SIZE_WRITE must be specified.
* Instead of using malloc(...M_WAITOK) for everything, reserve enough
resources to get by and use malloc(...M_NOWAIT), using the reserves if
necessary. Use the pool subsystem for structures small enough that
this is feasible. This also obsoletes LFS_THROTTLE.
And a few that are not strictly necessary:
* Moves the LFS inode extensions off onto a separately allocated
structure; getting closer to LFS as an LKM. "Welcome to 1.6O."
* Unified GOP_ALLOC between FFS and LFS.
* Update LFS copyright headers to correct values.
* Actually cast to unsigned in lfs_shellsort, like the comment says.
* Keep track of which segments were empty before the previous
checkpoint; any segments that pass two checkpoints both dirty and
empty can be summarily cleaned. Do this. Right now lfs_segclean
still works, but this should be turned into an effectless
compatibility syscall.
|
| 1.19 | 29-Jan-2003 |
yamt | don't use daddr_t for segment summary since it's an on-disk structure.
|
| 1.18 | 25-Jan-2003 |
kleink | Fix further printf format warnings for DEBUG, in the wake of daddr_t
having changed.
|
| 1.17 | 24-Jan-2003 |
fvdl | Bump daddr_t to 64 bits. Replace it with int32_t in all places where
it was used on-disk, so that on-disk formats remain the same.
Remove ufs_daddr_t and ufs_lbn_t for the time being.
|
| 1.16 | 14-May-2002 |
perseant | Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes
I found while making sure there weren't any new ones.
* Make the write clusters keep track of the buffers whose blocks they contain.
This should make it possible to (1) write clusters using a page mapping
instead of malloc, if desired, and (2) schedule blocks for rewriting
(somewhere else) if a write error occurs. Code is present to use
pagemove() to construct the clusters but that is untested and will go away
anyway in favor of page mapping.
* DEBUG now keeps a log of Ifile writes, so that any lingering instances of
the "dirty bufs" problem can be properly debugged.
* Keep track of whether the Ifile has been dirtied by various routines that
can be called by lfs_segwrite, and loop on that until it is clean, for
a checkpoint. Checkpoints need to be squeaky clean.
* Warn the user (once) if the Ifile grows larger than is reasonable for their
buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can
grow.
* If an inode is not found in a disk block, try rereading the block, under
the assumption that the block was copied to a cluster and then freed.
* Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
|
| 1.15 | 23-Nov-2001 |
chs | add spaces for KNF. confirmed to produce identical objects.
|
| 1.14 | 08-Nov-2001 |
lukem | add RCSID
|
| 1.13 | 26-Oct-2001 |
lukem | remove #include <ufs/ufs/quota.h> where it was just to appease
<ufs/ufs/inode.h>, since the latter now includes the former. leave the former
in source that obviously uses specific bits of it (for completeness.)
|
| 1.12 | 13-Jul-2001 |
perseant | branches: 1.12.4;
Merge the short-lived perseant-lfsv2 branch into the trunk.
Kernels and tools understand both v1 and v2 filesystems; newfs_lfs
generates v2 by default. Changes for the v2 layout include:
- Segments of non-PO2 size and arbitrary block offset, so these can be
matched to convenient physical characteristics of the partition (e.g.,
stripe or track size and offset).
- Address by fragment instead of by disk sector, paving the way for
non-512-byte-sector devices. In theory fragments can be as large
as you like, though in reality they must be smaller than MAXBSIZE in size.
- Use serial number and filesystem identifier to ensure that roll-forward
doesn't get old data and think it's new. Roll-forward is enabled for
v2 filesystems, though not for v1 filesystems by default.
- The inode free list is now a tailq, paving the way for undelete (undelete
is not yet implemented, but can be without further non-backwards-compatible
changes to disk structures).
- Inode atime information is kept in the Ifile, instead of on the inode;
that is, the inode is never written *just* because atime was changed.
Because of this the inodes remain near the file data on the disk, rather
than wandering all over as the disk is read repeatedly. This speeds up
repeated reads by a small but noticeable amount.
Other changes of note include:
- The ifile written by newfs_lfs can now be of arbitrary length, it is no
longer restricted to a single indirect block.
- Fixed an old bug where ctime was changed every time a vnode was created.
I need to look more closely to make sure that the times are only updated
during write(2) and friends, not after-the-fact during a segment write,
and certainly not by the cleaner.
|
| 1.11 | 09-Sep-2000 |
perseant | branches: 1.11.2; 1.11.4; 1.11.6;
Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
|
| 1.10 | 23-Apr-2000 |
perseant | branches: 1.10.4;
Fix problems outlined in PR#9926:
- lfs_truncate extends the file if called with length > i_ffs_size;
- lfs_truncate errors out if called with length < 0;
- lfs_balloc block accounting corrected for the case of blocks read
into the cache before they exist on disk;
- mp->mnt_stat.f_iosize is initialized in lfs_mountfs.
|
| 1.9 | 10-Mar-1999 |
perseant | branches: 1.9.8; 1.9.14;
New sources should leave the LFS in a more-or-less working state. Changes
include:
- DIROP segregation is enabled, and greater care is taken
to make sure that a checkpoint completes. Fsck is not
needed to remount the filesystem.
- Several checks to make sure that the LFS subsystem does not
overuse various resources (memory, in particular).
- The cleaner routines, lfs_markv in particular, are completely
rewritten. A buffer overflow is removed. Greater care is taken
to ensure that inodes come from where lfs_cleanerd say they come
from (so we know nothing has changed since lfs_bmapv was called).
- Fragment allocation is fixed, so that writes beyond end-of-file
do the right thing.
|
| 1.8 | 01-Mar-1998 |
fvdl | Merge with Lite2 + local changes
|
| 1.7 | 15-Nov-1996 |
cgd | cast int64_t-sized types to "long long" before printing them with %qd.
gcc thinks that the 'q' modifier describes a "long long", and so -Wformat
whines when printing with 'q' on the alpha, since int64_t-sized types are
done with variations on "long" rather than "long long".
|
| 1.6 | 12-Oct-1996 |
christos | revert previous kprintf changes
|
| 1.5 | 10-Oct-1996 |
christos | printf -> kprintf, sprintf -> ksprintf
|
| 1.4 | 17-Mar-1996 |
christos | Fix printf format strings
|
| 1.3 | 12-Feb-1996 |
christos | di_size is a quad and needs %qu not %lu
|
| 1.2 | 29-Jun-1994 |
cgd | New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
|
| 1.1 | 08-Jun-1994 |
mycroft | branches: 1.1.1;
Update to 4.4-Lite fs code, with local changes.
|
| 1.1.1.1 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite for reference
|
| 1.9.14.1 | 21-Dec-1999 |
wrstuden | Initial commit of recent changes to make DEV_BSIZE go away.
Runs on i386, needs work on other arch's. Main kernel routines should be
fine, but a number of the stand programs need help.
cd, fd, ccd, wd, and sd have been updated. sd has been tested with non-512
byte block devices. vnd, raidframe, and lfs need work.
Non 2**n block support is automatic for LKM's and conditional for kernels
on "options NON_PO2_BLOCKS".
|
| 1.9.8.1 | 20-Nov-2000 |
bouyer | Update thorpej_scsipi to -current as of a month ago
|
| 1.10.4.1 | 14-Sep-2000 |
perseant | Pull up recent LFS kernel changes (approved by thorpej):
ufs/ufs/inode.h, 1.20--1.22 (add i_lfs_effnblks extension ;
make ITIMES aware of LFS_ITIMES;
_LKM protection so userland progs
compile)
ufs/ufs/ufs_vnops.c, 1.69, 1.71 (remove IN_ADIROP;
use ITIMES instead of FFS_ITIMES)
ufs/ufs/ufs_readwrite.c, 1.27 (use lfs_reserve in lfs_write)
ufs/lfs/lfs.h, 1.26--1.32 (define LFS_EST_* macros ;
change MIN_FREE_SEGS to lfs_minfreesegs ;
add avail and bfree to CLEANERINFO ;
change lfs_uinodes to signed ;
change lfs_dmeta to signed ;
add whitespace to line up structure
members ;
explicit cast to int32_t in LFS_EST_*
macros)
ufs/lfs/lfs_alloc.c, back out 1.34.2.3 (pullups of 1.39, 1.40);
then pull up 1.38 (clean up on error)
1.39--1.43 (restore fvdl's ufs_hashlock fix ;
restore fvdl's ufs_hashlock fix ;
set i_lfs_effnblks ;
use UINO macros ;
add comments and fix long lines)
ufs/lfs/lfs_balloc.c, 1.19 (don't succeed halfway)
1.21--1.25 (use i_lfs_effnblks ;
fix i_lfs_effnblks computation and
quieten ;
fix i_ffs_blocks in unwritten fragment ;
remove useless debugging check ;
add comments and (c) 2000)
ufs/lfs/lfs_bio.c, 1.24--1.30 (cleanup and make lfs_flush_fs take
"struct lfs *" instead of "struct
mount *" ;
use lfs_minfreeseg instead of
MIN_FREE_SEGS ;
use UINO macros, and copy bfree/avail
to CLEANERINFO ;
add lfs_reserve function ;
1.28--1.30 fix printf formatting)
ufs/lfs/lfs_cksum.c, 1.13 (add (c) 2000)
ufs/lfs/lfs_debug.c, 1.11 (use btodb instead of DEV_BSIZE)
ufs/lfs/lfs_extern.h, 1.18, 1.20--1.21 (function prototype changes)
ufs/lfs/lfs_inode.c, 1.38 (rewrite lfs_truncate from
ffs_truncate)
1.40--1.44 (count written and unwritten blocks
seperately ;
use disk block units instead of bytes ;
remove unnecessary "mod" variable ;
correct B_DELWRI to avoid bawrite panic ;
use lfs_reserve)
ufs/lfs/lfs_segment.c, 1.52-1.59 (use lfs_dmeta to note used summaries ;
check for UNWRITTEN in indirect blocks ;
more debugging stuff inside #ifdef
DEBUG_LFS ;
use LK_CANRECURSE ;
don't drop dirty indirect blocks ;
use UINO macros ;
don't hose the free list ;
use btodb() instead of DEV_BSIZE ;
make it compile again (oops))
ufs/lfs/lfs_subr.c, 1.16--1.17 (check for locked inodes before
changing ;
use btodb() instead of DEV_BSIZE, (c)
2000)
ufs/lfs/lfs_syscalls.c, back out 1.41.4.2 (fvdl's ufs_hashlock fix);
then pull up 1.43 (use lfs_dmeta)
1.44--1.45 (restore fvdl's ufs_hashlock fix)
1.46--1.47 (fix lfs_avail leakage from sblock
segments ;
use UINO macros)
1.49 (bounds-check inode numbers in
lfs_markv)
ufs/lfs/lfs_vfsops.c, 1.53 (use LFS_EST_* macros in lfs_statfs)
1.56--1.58 (initialize lfs_minfreeseg, lfs_effnblk ;
initialize lfs_uinodes ;
initialize lfs_ravail)
ufs/lfs/lfs_vnops.c, 1.40 (remove VDIROP from removed files)
1.42--1.44 (move SET_ENDOP below the removal of
VDIROP ;
use UINO macros and add lfs_itimes
function ;
use lfs_reserve in dirops)
|
| 1.11.6.3 | 23-Jun-2002 |
jdolecek | catch up with -current on kqueue branch
|
| 1.11.6.2 | 10-Jan-2002 |
thorpej | Sync kqueue branch with -current.
|
| 1.11.6.1 | 03-Aug-2001 |
lukem | update to -current
|
| 1.11.4.3 | 02-Jul-2001 |
perseant | Change disk addressing unit to be the fragment, instead of the disk sector.
All quantities in the superblock, inodes, indirect blocks, etc. refer now
to this abstract unit (called "fsb" as it is in FFS) instead of disk sectors;
as a consequence segment summary blocks have to be multiples of a fragment in
size. In v1 filesystems, compatibility code ensures that 1 fsb == 1 sector,
regardless of fragment size.
Fragments can now range in size between 512 and 32k; in the event that
LFS_LABELPAD (8k) is smaller than the disk address unit size, an extra
proto-superblock is kept at 8k from the beginning of the disk, to be used
*only* to locate the real superblocks. (Not all of the userland knows about
this yet.)
Almost all of this was done not by me, but by joff.
|
| 1.11.4.2 | 29-Jun-2001 |
perseant | Get rid of __P(), protoizing where it had not already been done
|
| 1.11.4.1 | 27-Jun-2001 |
perseant | Import of what I've been calling "LFSv2", that is, LFS with some features
added that require changes to the on-disk data structures. These include:
- 64-bit time in everything but inodes
- User-specified segment offset, and segment size no longer
restricted to PO2.
- Serial number on segment summaries in addition to timestamp, and
a new volume identifier, to make roll-forward feasible without
fear of finding old data and thinking it was new.
Although I think this version works at least as well as what's on the trunk,
we're not done yet; hence this commit is going in on a branch and not on
the trunk. Enhancements that are not here yet include fragment addressing,
like FFS does, instead of block addressing.
|
| 1.11.2.4 | 20-Jun-2002 |
nathanw | Catch up to -current.
|
| 1.11.2.3 | 08-Jan-2002 |
nathanw | Catch up to -current.
|
| 1.11.2.2 | 14-Nov-2001 |
nathanw | Catch up to -current.
|
| 1.11.2.1 | 24-Aug-2001 |
nathanw | Catch up with -current.
|
| 1.12.4.1 | 12-Nov-2001 |
thorpej | Sync the thorpej-mips-cache branch with -current.
|
| 1.22.2.7 | 10-Nov-2005 |
skrll | Sync with HEAD. Here we go again...
|
| 1.22.2.6 | 01-Apr-2005 |
skrll | Sync with HEAD.
|
| 1.22.2.5 | 08-Mar-2005 |
skrll | Sync with HEAD.
|
| 1.22.2.4 | 04-Mar-2005 |
skrll | Sync with HEAD.
Hi Perry!
|
| 1.22.2.3 | 21-Sep-2004 |
skrll | Fix the sync with head I botched.
|
| 1.22.2.2 | 18-Sep-2004 |
skrll | Sync with HEAD.
|
| 1.22.2.1 | 03-Aug-2004 |
skrll | Sync with HEAD
|
| 1.24.10.1 | 19-Mar-2005 |
yamt | sync with head. xen and whitespace. xen part is not finished.
|
| 1.24.8.1 | 29-Apr-2005 |
kent | sync with -current
|
| 1.27.2.3 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.153
sys/ufs/lfs/lfs_debug.c: revision 1.32
sys/ufs/lfs/lfs_alloc.c: revision 1.84
sys/ufs/lfs/lfs_vfsops.c: revision 1.185
sys/ufs/lfs/lfs_segment.c: revision 1.165
64 bit inode changes.
|
| 1.27.2.2 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.152
sys/ufs/lfs/lfs_debug.c: revision 1.31
sys/ufs/lfs/lfs_subr.c: revision 1.53
sys/ufs/lfs/lfs_extern.h: revision 1.68
sys/ufs/lfs/lfs_inode.c: revision 1.96
sys/ufs/lfs/lfs_bio.c: revision 1.86
sys/ufs/lfs/lfs_alloc.c: revision 1.83
sys/ufs/lfs/lfs_vfsops.c: revision 1.181
sys/ufs/lfs/lfs.h: revision 1.88
sys/ufs/lfs/lfs_segment.c: revision 1.164
- sprinkle const
- avoid shadow variables.
|
| 1.27.2.1 | 07-May-2005 |
tron | Apply patch (requested by perseant in ticket #242):
* fsck_lfs buffer cache fixes, including PR #29151
* Change fsck_lfs phase 0 message to reflect reality
* fsck_lfs: check phase 5 (cleanerinfo accounting) even on
roll-forward
* Keep better track of the free list during roll-forward, avoiding
a core dump
* Improve hash table use for fsck_lfs buffer and vnode cache
* Document fsck_lfs flag -f, and implement -q
* Add resize_lfs, including kernel support
* Add LFS to mountd's list of exportable filesystem types
* Make the LFS lkm work again [christos@]
* Add MP locking to the LFS kernel subsystem
* Fix pager_map deadlock in lfs_putpages()
* Avoid incomplete file extension that looks like "partial
truncation" to fsck
* Use lfs_malloc for cleaner malloc, since the cleaner often runs
in low-memory conditions.
* Use splay trees, not hash table, to track page allocation for
write.
* Fix mkdir panic on full fs
* Fix page accounting leak by counting differently.
* Use rightly named structure for lfs_getattr [skrll@]
* Cosmetic changes for readability.
|
| 1.31.2.3 | 21-Jan-2008 |
yamt | sync with head
|
| 1.31.2.2 | 03-Sep-2007 |
yamt | sync with head.
|
| 1.31.2.1 | 21-Jun-2006 |
yamt | sync with head.
|
| 1.33.40.1 | 15-Aug-2007 |
skrll | Sync with HEAD.
|
| 1.33.30.4 | 28-Aug-2007 |
yamt | make this compile with DEBUG.
|
| 1.33.30.3 | 20-Aug-2007 |
ad | Sync with HEAD.
|
| 1.33.30.2 | 15-Jul-2007 |
ad | Sync with head.
|
| 1.33.30.1 | 05-Apr-2007 |
ad | Compile fixes.
|
| 1.34.22.2 | 22-Jul-2007 |
christos | make this compile again
|
| 1.34.22.1 | 22-Jul-2007 |
christos | file lfs_debug.c was added on branch matt-mips64 on 2007-07-22 03:41:00 +0000
|
| 1.34.18.2 | 02-Jan-2008 |
bouyer | Sync with HEAD
|
| 1.34.18.1 | 13-Dec-2007 |
bouyer | Sync with HEAD
|
| 1.34.16.1 | 13-Dec-2007 |
yamt | sync with head.
|
| 1.34.14.2 | 26-Dec-2007 |
ad | Sync with head.
|
| 1.34.14.1 | 04-Dec-2007 |
ad | Pull the vmlocking changes into a new branch.
|
| 1.34.12.1 | 18-Feb-2008 |
mjf | Sync with HEAD.
|
| 1.34.6.1 | 09-Jan-2008 |
matt | sync with HEAD
|
| 1.36.10.2 | 19-Aug-2009 |
yamt | sync with head.
|
| 1.36.10.1 | 16-May-2008 |
yamt | sync with head.
|
| 1.36.8.1 | 18-May-2008 |
yamt | sync with head.
|
| 1.36.6.1 | 02-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.37.14.1 | 23-Jul-2009 |
jym | Sync with HEAD.
|
| 1.39.12.3 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.39.12.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.39.12.1 | 25-Feb-2013 |
tls | resync with head
|
| 1.39.2.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.39.2.1 | 23-Jan-2013 |
yamt | sync with head
|
| 1.43.10.1 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.54.24.1 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.54.22.1 | 17-Aug-2020 |
martin | Pull up following revision(s) (requested by riastradh in ticket #1050):
sys/ufs/lfs/lfs_subr.c: revision 1.101
sys/ufs/lfs/lfs_subr.c: revision 1.102
sys/ufs/lfs/lfs_inode.c: revision 1.158
sys/ufs/lfs/lfs_inode.h: revision 1.25
sys/ufs/lfs/lfs_balloc.c: revision 1.95
sys/ufs/lfs/lfs_pages.c: revision 1.21
sys/ufs/lfs/lfs_vnops.c: revision 1.330
sys/ufs/lfs/lfs_alloc.c: revision 1.140 (patch)
sys/ufs/lfs/lfs_alloc.c: revision 1.141 (patch)
lib/libp2k/p2k.c: revision 1.72
sys/ufs/lfs/lfs.h: revision 1.205
sys/ufs/lfs/lfs.h: revision 1.206
sys/ufs/lfs/lfs_segment.c: revision 1.284
sys/ufs/lfs/lfs.h: revision 1.207
sys/ufs/lfs/lfs_segment.c: revision 1.285
sys/ufs/lfs/lfs_debug.c: revision 1.55
sys/ufs/lfs/lfs_rename.c: revision 1.23
usr.sbin/dumplfs/dumplfs.c: revision 1.65
sys/ufs/lfs/lfs_vfsops.c: revision 1.371
sys/arch/i386/stand/efiboot/bootx64/Makefile: revision 1.3
sys/ufs/lfs/lfs_vfsops.c: revision 1.372
sys/ufs/lfs/lfs_vfsops.c: revision 1.373
sbin/fsck_lfs/pass1.c: revision 1.46
sys/ufs/lfs/lfs_vnops.c: revision 1.326
sys/ufs/lfs/lfs_vnops.c: revision 1.327
sys/ufs/lfs/lfs_vfsops.c: revision 1.375 (patch)
sys/ufs/lfs/lfs_vnops.c: revision 1.328
sys/ufs/lfs/lfs_subr.c: revision 1.98
sys/ufs/lfs/lfs_extern.h: revision 1.116
sys/ufs/lfs/lfs_vnops.c: revision 1.329
sys/ufs/lfs/lfs_subr.c: revision 1.99
sys/ufs/lfs/lfs_extern.h: revision 1.117
sys/ufs/lfs/lfs_accessors.h: revision 1.49
sys/ufs/lfs/lfs_extern.h: revision 1.118
sys/rump/fs/lib/liblfs/Makefile: revision 1.15
sys/ufs/lfs/lfs_bio.c: revision 1.146 (patch)
sys/ufs/lfs/lfs_bio.c: revision 1.147
sys/ufs/lfs/lfs_subr.c: revision 1.100
Fix kassert in lfs by initializing vp first.
Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
Teach lfs to transition ro<->rw.
Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
Assert lfs_writer where I think we can now prove it.
Serialize access to the splay tree with lfs_lock.
Change some cheap KDASSERT into KASSERT.
Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
Call lfs_orphan in lfs_rename while we're still in the dirop.
lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
Teach LFS_ORPHAN_NEXTFREE about lfs64.
Dust off the orphan detection code and try to make it work.
Fix !DIAGNOSTIC compile
Fix userland references to LFS_ORPHAN_NEXTFREE.
Forgot to grep for these or do a full distribution build, oops!
Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
OR into bp->b_cflags; don't overwrite.
CTASSERT lfs on-disk structure sizes.
Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
Fix clang build after packed lfs64 accessor change.
Suppress spurious address-of-packed error in rump lfs too.
|
| 1.54.18.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.123 | 20-Oct-2025 |
perseant | * Generalize the partial-segment parser introduced for roll-forward,
using it to facilitate an in-kernel segment rewriter (cleaner), and a
mechanism to check whether a segment is in fact empty (only used
with DEBUG).
* Add these new fcntl calls:
- LFCNFILESTATS: For each inode given, report its number of direct
blocks, how many gaps (discontinuities) there are between direct
blocks, and how large the total gap distance is. This will be
useful for a coalescing agent.
- LFCNREWRITEFILE: For each inode given, rewrite its direct blocks,
effectively coalescing it into as compact a form as possible.
- LFCNSCRAMBLE: As above, except that it only rewrites every other
block. This causes the file to have many gaps that can be
measured with LFCNFILESTATS and addressed with LFCNREWRITEFILE,
for testing purposes.
- LFCNREWRITESEGS: Rewrite any live data in the given segments.
This is intended to simplify the cleaner API and facilitate an
in-kernel cleaner.
- LFCNCLEANERINFO: Get the most current CLEANERINFO data from the
kernel.
- LFCNSEGUSE: Retrieve segment usage data from the kernel.
* Vnodes marked IN_CLEANING now take a reference. Add a new "cleaner
lock", which must be taken by the cleaner before the segment lock,
and before marking nodes IN_CLEANING. This allows us to flush
vnodes, if necessary, before the cleaning segment is written, and
never to flush vnodes being cleaned. When the cleaner lock is
released, the vnodes are cleared of IN_CLEANING and the reference
dropped.
* Track a potential infinite loop in lfs_gatherblock.
* Pull "needs to flush" and "needs to wait for flush" into functions
instead of inlining their definitions.
|
| 1.122 | 17-Sep-2025 |
perseant | Use a workqueue to handle the superblock callback.
|
| 1.121 | 17-Sep-2025 |
perseant | Add routines to check freelist consistency if compiled with DEBUG and
conditional on a kernel variable manipulated via sysctl.
Add checks before and after each routine that modifies the free list.
#if 0 a section of lfs_vfree() that was intended to keep the free list ordered
but instead corrupted it.
|
| 1.120 | 04-Sep-2025 |
perseant | Copy the flags from a full partial segment to its continuation, if
a continuation is necessary, so that partial-segment collections marked
with SS_DIROP|SS_CONT are properly completed wiht a partial-segment marked
SS_DIROP (without SS_CONT). Necessary for roll-forward.
|
| 1.119 | 02-Sep-2025 |
perseant | Use a workqueue to handle cluster iodone, rather than doing it in interrupt context.
|
| 1.118 | 23-Feb-2020 |
riastradh | Dust off the orphan detection code and try to make it work.
|
| 1.117 | 23-Feb-2020 |
riastradh | lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
|
| 1.116 | 23-Feb-2020 |
riastradh | Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
|
| 1.115 | 18-Feb-2020 |
chs | remove the aiodoned thread. I originally added this to provide a thread context
for doing page cache iodone work, but since then biodone() has changed to
hand off all iodone work to a softint thread, so we no longer need the
special-purpose aiodoned thread.
|
| 1.114 | 22-Aug-2018 |
msaitoh | branches: 1.114.4; 1.114.6;
- Cleanup for dynamic sysctl:
- Remove unused *_NAMES macros for sysctl.
- Remove unused *_MAXID for sysctls.
- Move CTL_MACHDEP sysctl definitions for m68k into m68k/include/cpu.h and
use them on all m68k machines.
|
| 1.113 | 26-Jul-2017 |
maya | branches: 1.113.2; 1.113.4;
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
|
| 1.112 | 08-Jun-2017 |
chs | move some buffer cache internals declarations from buf.h to vfs_bio.c.
this is needed to avoid name conflicts with ZFS and also
makes it clearer that other code shouldn't be messing with these.
remove the LFS debug code that poked around in bufqueues and
remove the BQ_EMPTY bufqueue since nothing uses it anymore.
provide a function to let LFS and wapbl read the value of nbuf for now.
|
| 1.111 | 20-Jun-2016 |
dholland | branches: 1.111.10;
u_int{8,16,32,64}_t -> uint{8,16,32,64}_t in remaining lfs headers.
|
| 1.110 | 12-Aug-2015 |
dholland | Hack up dinode usage to be 64 vs. 32 as needed. Part 1.
(This part changes the native lfs code; the ufs-derived code already
has 64 vs. 32 logic, but as aspects of it are unsafe, and don't
entirely interoperate cleanly with the lfs 64/32 stuff, pass 2 will be
rehashing that.)
|
| 1.109 | 12-Aug-2015 |
dholland | Move the security checks for lfs_bmapv/lfs_markv into those functions.
(instead of the system call entry points)
Avoids duplication.
While touching these, pass the lwp around instead of the proc -- the
latter was there for no other reason than because once upon a time
struct proc was the first argument of all syscalls.
(For that matter, why not just use curlwp instead of passing it around
all over the place? The cost of passing it to every syscall probably
exceeds the cost of loading it from curcpu, even on machines where
it's not just kept in a register all the time.)
|
| 1.108 | 12-Aug-2015 |
dholland | Fix assorted 64->32 truncations related to BLOCK_INFO.
Also make note of a cleaner limitation: it seems that when it goes to
coalesce discontiguous files, it mallocs an array with one BLOCK_INFO
for every block in the file. Therefore, with 64-bit LFS, on a 32-bit
platform it will be possible to have files large enough to overflow
the cleaner's address space. Currently these will be skipped and cause
warnings via syslog.
At some point someone should rewrite the logic to coalesce files to
use chunks of some reasonable size, as discontinuity between such
chunks is immaterial and mallocing this much space is silly and
fragile. Also, the kernel only accepts up to 65536 blocks at a time
for bmapv and markv, so processing more than this at once probably
isn't useful and may not even work currently. I don't want to change
this around just now as it's not entirely trivial.
|
| 1.107 | 02-Aug-2015 |
dholland | Add a (draft) 64-bit superblock. Make things build again.
Add pieces of support for using both superblock types where
convenient, and specifically to the superblock accessors, but don't
actually enable it anywhere.
First substantive step on PR 50000.
|
| 1.106 | 02-Aug-2015 |
dholland | Second batch of 64 -> 32 truncations in lfs, along with more minor
tidyups and corrections in passing.
|
| 1.105 | 31-May-2015 |
hannken | Change lfs from hash table to vcache.
- Change lfs_valloc() to return an inode number and version instead of
a vnode and move lfs_ialloc() and lfs_vcreate() to new lfs_init_vnode().
- Add lfs_valloc_fixed() to allocate a known inode, used by kernel
roll forward.
- Remove lfs_*ref(), these functions cannot coexist with vcache and
their commented behaviour is far away from their implementation.
- Add the cleaner lwp and blockinfo to struct ulfsmount so lfs_loadvnode()
may use hints from the cleaner.
- Remove vnode locks from ulfs_lookup() like we did with ufs_lookup().
|
| 1.104 | 31-May-2015 |
hannken | Make lfs_fastvget() private to lfs_syscalls.c, change it to take the
BLOCK_INFO and vnode lock type instead of the inode disk address and
return the vnode locked.
Change lfs_markv() and lfs_bmapv() to work on locked vnodes.
|
| 1.103 | 31-May-2015 |
hannken | Use VFS_PROTOS() for lfs.
Rename conflicting struct lfs field "lfs_start" to "lfs_s0addr".
No functional change.
|
| 1.102 | 27-Mar-2015 |
riastradh | Disentangle buffer-cached I/O from page-cached I/O in UFS.
Page-cached I/O is used for regular files, and is initiated by VFS
users such as userland and NFS.
Buffer-cached I/O is used for directories and symlinks, and is issued
only internally by UFS.
New UFS routine ufs_bufio replaces vn_rdwr for internal use.
ufs_bufio is implemented by new UFS operations uo_bufrd/uo_bufwr,
which sit in ufs_readwrite.c alongside the VOP_READ/VOP_WRITE
implementations.
I preserved the code as much as possible and will leave further
simplification for future commits. I kept the ulfs_readwrite.c
copypasta close to ufs_readwrite.c in case we ever want to merge them
back; likewise ext2fs_readwrite.c.
No externally visible semantic change. All atf fs tests still pass.
|
| 1.101 | 18-Mar-2014 |
riastradh | branches: 1.101.6;
Merge riastradh-drm2 to HEAD.
|
| 1.100 | 20-Jul-2013 |
dholland | Collect the pieces of lfs rename into lfs_rename.c, and sprinkle static.
|
| 1.99 | 06-Jun-2013 |
dholland | branches: 1.99.2; 1.99.4;
Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.98 | 23-Feb-2012 |
joerg | branches: 1.98.2;
Make sure that __BEGIN_DECLS and __END_DECLS are paired.
|
| 1.97 | 02-Jan-2012 |
perseant | * Remove PGO_RECLAIM during lfs_putpages()' call to genfs_putpages(),
to avoid a live lock in the latter when reclaiming a vnode with
dirty pages.
* Add a new segment flag, SEGM_RECLAIM, to note when a segment is
being written for vnode reclamation, and record which inode is being
reclaimed, to aid in forensic debugging.
* Add a new segment flag, SEGM_SINGLE, so that opportunistic writes
can write a single segment's worth of blocks and then stop, rather
than writing all the way up to the cleaner's reserved number of
segments.
* Add assert statements to check mutex ownership is the way it ought
to be, mostly in lfs_putpages; fix problems uncovered by this.
* Don't clear VU_DIROP until the inode actually makes its way to disk,
avoiding a problem where dirop inodes could become separated
(uncovered by a modified version of the "ckckp" forensic regression
test).
* Move the vfs_getopsbyname() call into lfs_writerd. Prepare code to
make lfs_writerd notice when there are no more LFSs, and exit losing
the reference, so that, in theory, the module can be unloaded. This
code is not enabled, since it causes a crash on exit.
* Set IN_MODIFIED on inodes flushed by lfs_flush_dirops. Really we
only need to set IN_MODIFIED if we are going to write them again
(e.g., to write pages); need to think about this more.
Finally, several changes to help avoid "no clean segments" panics:
* In lfs_bmapv, note when a vnode is loaded only to discover whether
its blocks are live, so it can immediately be recycled. Since the
cleaner will try to choose ~empty segments over full ones, this
prevents the cleaner from (1) filling the vnode cache with junk, and
(2) squeezing any unwritten writes to disk and running the fs out of
segments.
* Overestimate by half the amount of metadata that will be required
to fill the clean segments. This will make the disk appear smaller,
but should help avoid a "no clean segments" panic.
* Rearrange lfs_writerd. In particular, lfs_writerd now pays
attention to the number of clean segments available, and holds off
writing until there is room.
|
| 1.96 | 28-Jun-2008 |
rumble | branches: 1.96.30; 1.96.34;
Create sysctl entries during module initialisation and destroy them
appropriately.
Many of these file systems are now ready for modularisation.
|
| 1.95 | 28-Apr-2008 |
martin | branches: 1.95.2; 1.95.4;
Remove clause 3 and 4 from TNF licenses
|
| 1.94 | 02-Jan-2008 |
ad | branches: 1.94.6; 1.94.8; 1.94.10;
Merge vmlocking2 to head.
|
| 1.93 | 08-Dec-2007 |
pooka | branches: 1.93.4;
Remove cn_lwp from struct componentname. curlwp should be used
from on. The NDINIT() macro no longer takes the lwp parameter and
associates the credentials of the calling thread with the namei
structure.
|
| 1.92 | 26-Nov-2007 |
pooka | branches: 1.92.2;
Remove the "struct lwp *" argument from all VFS and VOP interfaces.
The general trend is to remove it from all kernel interfaces and
this is a start. In case the calling lwp is desired, curlwp should
be used.
quick consensus on tech-kern
|
| 1.91 | 31-Jul-2007 |
pooka | branches: 1.91.2; 1.91.4; 1.91.10; 1.91.12;
* nuke the nameidata parameter from VFS_MOUNT(). Nobody on tech-kern
knew what it was supposed to be used for and wrstuden gave a go-ahead
* while rototilling, convert file systems which went easily to
use VFS_PROTOS() instead of manually prototyping the methods
|
| 1.90 | 12-Jul-2007 |
dsl | branches: 1.90.2;
Change the VFS_MOUNT() interface so that the 'data' buffer passed to the
fs code is a kernel buffer, pass though the length of the buffer as well.
Since the length of the userspace buffer isn'it (yet) passed through the mount
system call, add a field to the vfsops structure containing the default length.
Split sys_mount() for calls from compat code.
Ride one of the recent kernel version changes - old fs LKMs will load, but
sys_mount() will reject any attempt to use them.
|
| 1.89 | 17-Apr-2007 |
perseant | Install a new sysctl, vfs.lfs.ignore_lazy_sync, which causes LFS to ignore
the "smooth" syncer, as if vfs.sync.*delay = 0, but only for LFS. The
default is "on", i.e., ignore lazy sync.
Reduce the amount of polling/busy-waiting done by lfs_putpages(). To
accomplish this, copied genfs_putpages() and modified it to indicate which
page it was that caused it to return with EDEADLK. fsync()/fdatasync()
should no longer ever fail with EAGAIN, and should not consume huge
quantities of cpu.
Also, try to make dirops less likely to be written as the result of a
VOP_PUTPAGES(), while ensuring that they are written regularly.
|
| 1.88 | 04-Mar-2007 |
christos | branches: 1.88.2; 1.88.4;
Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.
|
| 1.87 | 01-Sep-2006 |
perseant | branches: 1.87.6; 1.87.8; 1.87.12;
Changes to help the roll-forward agent, to wit:
* Mark being-deleted files in the Ifile so we can finish deleting them
at fs mount time.
* Flag the Ifile with "cleaner must clean" when writers are waiting for
the cleaner, rather than relying solely on the cleaner's estimation of
whether it should clean or not.
* Note partial segments written by a user agent (in particular,
fsck_lfs) so that repeated rolls forward don't interfere with one
another.
* Add a new fcntl, LFCNPASS, that allows the log to wrap exactly once,
for better testing of the validity of checkpoints.
* Keep track of the on-disk nlink count when cleaning, so that we don't
partially complete directory operations while cleaning.
* Ensure that every single Ifile inode write represents a consistent
view of the filesystem. In particular, the accounting for the segment
we are writing the inode into must be correct, and the accounting for
the segment that inode used to reside in must be correct. Rather than
just rewriting the inode if we wrote it wrong, rewrite the necessary
ifile blocks before writing the inode so we never write it wrong.
* Don't unmark any VDIROP vnodes if we haven't written them to disk,
avoiding yet another problem with the "wait for the cleaner" error
return from lfs_putpages().
Also, move the last callback to an aiodone call, so we no longer do any
memory management from interrupt context.
|
| 1.86 | 20-Jul-2006 |
perseant | Separate the (non-working) LFS kernel roll-forward code into its own file,
lfs_rfw.c.
|
| 1.85 | 13-Jul-2006 |
martin | Fix alignement problems for fhandle_t, exposed by gcc4.1.
While touching all vptofh/fhtovp functions, get rid of VFS_MAXFIDSIZ,
version the getfh(2) syscall and explicitly pass the size available in
the filehandle from userland.
Discussed on tech-kern, with lots of help from yamt (thanks!).
|
| 1.84 | 29-Jun-2006 |
perseant | Don't wake up the cleaner if the filesystem is unwrappable, and fix the
compatibility fcntls.
Also includes one-line fixes for an MP locking bug and a zero-length FINFO
problem that manifested during testing.
|
| 1.83 | 18-May-2006 |
perseant | branches: 1.83.4;
Break out the finfo array manipulation code into two new functions,
lfs_acquire_finfo() and lfs_release_finfo(). Add a debugging check
for zero-length finfo arrays in the segment summary to avoid future
regressions.
|
| 1.82 | 14-May-2006 |
elad | integrate kauth.
|
| 1.81 | 01-May-2006 |
perseant | Don't ever partially write dirops, even if we need the cleaner to run.
This increases the chances of the "no clean segments" panic slightly,
but allows us to run the ckckp regression test successfully to completion.
|
| 1.80 | 30-Apr-2006 |
perseant | Postpone the segment accounting changes coming from truncation until the
inode that makes those changes valid is either written to disk by
lfs_writeinode() or discarded by lfs_vfree().
A couple of locking fixes are also included as well.
|
| 1.79 | 23-Apr-2006 |
yamt | remove unused FFS_NAMES and LFS_NAMES.
|
| 1.78 | 08-Apr-2006 |
perseant | Implement a somewhat finer-grained mechanism for paging LFS-backed pages.
The writer daemon, if it does not need to flush the whole filesystem,
now only writes the vnodes for which the pagedaemon has requested pageouts
(although it does not pay attention to the page ranges the pagedaemon
supplies).
|
| 1.77 | 08-Apr-2006 |
perseant | Keep the free list ordered. This solves a problem first pointed out to me
by Michel Oey, in which an aged LFS writes up to an extra Ifile block for
every file created; and paves the way for the truncation of the Ifile when
many files are deleted.
|
| 1.76 | 24-Mar-2006 |
perseant | Improvements to LFS's paging mechanism, to wit:
* Acknowledge that sometimes there are more dirty pages to be written to
disk than clean segments. When we reach the danger line,
lfs_gop_write() now returns EAGAIN. The caller of VOP_PUTPAGES(), if
it holds the segment lock, drops it and waits for the cleaner to make
room before continuing.
* Note and avoid a three-way deadlock in lfs_putpages (a writer holding
a page busy blocks on the cleaner while the cleaner blocks on the
segment lock while lfs_putpages blocks on the page).
|
| 1.75 | 14-Jan-2006 |
yamt | branches: 1.75.2; 1.75.4; 1.75.6; 1.75.8; 1.75.10;
- unify ffs_blkatoff and lfs_blkatoff.
- remove ufs_ops::uo_blkatoff.
- add directory read-ahead code. (disabled for now.)
|
| 1.74 | 06-Jan-2006 |
yamt | remove an obsolete prototype.
|
| 1.73 | 11-Dec-2005 |
christos | branches: 1.73.2;
merge ktrace-lwp.
|
| 1.72 | 02-Nov-2005 |
yamt | merge yamt-vop branch. remove following VOPs.
VOP_BLKATOFF
VOP_VALLOC
VOP_BALLOC
VOP_REALLOCBLKS
VOP_VFREE
VOP_TRUNCATE
VOP_UPDATE
|
| 1.71 | 13-Sep-2005 |
christos | branches: 1.71.2;
split out lfs_itimes(). It is used in fsck_lfs.
|
| 1.70 | 12-Sep-2005 |
christos | Use nanotime() to update the time fields in filesystems. Convert the code
from macros to real functions. Original patch and review from chuq.
Note: ext2fs only keeps seconds in the on-disk inode, and msdosfs does not
have enough precision for all fields, so this is not very useful for those
two.
|
| 1.69 | 28-Jun-2005 |
yamt | branches: 1.69.2;
- constify genfs_ops.
- use member designators.
|
| 1.68 | 29-May-2005 |
christos | - sprinkle const
- avoid shadow variables.
|
| 1.67 | 23-Apr-2005 |
perseant | Provide a resize_lfs(8), including kernel and cleaner support. The current
implementation requires the fs to be mounted while resizing. Tested in both
directions, and everything appears to work happily, but ymmv.
|
| 1.66 | 16-Apr-2005 |
perseant | Use splay trees, rather than a hash table, to manage the accounting of
blocks allocated through VOP_BALLOC() for pages to be written to disk.
This accounting no longer takes a noticeable fraction of the system CPU.
|
| 1.65 | 14-Apr-2005 |
perseant | Consolidate the hash table we use to maintain the integrity of lfs_avail
into a single, system-wide table, rather than having a separate hash table
per inode. Significantly reduces the "system" cpu usage of your average
file write.
|
| 1.64 | 08-Mar-2005 |
perseant | branches: 1.64.2;
Straighten out the maze of ifdefs. Instead, consolidate all the debugging
stuff under '#ifdef DEBUG', and use sysctl knobs to turn on/off particular
parts of the debugging reporting (if DEBUG is enabled). Re-enable the LFS
statistics in sysctl, while I'm there. A bit of a rototill.
|
| 1.63 | 26-Feb-2005 |
perry | nuke trailing whitespace
|
| 1.62 | 26-Feb-2005 |
perseant | Various minor LFS improvements:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statvfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
|
| 1.61 | 20-May-2004 |
atatat | branches: 1.61.4; 1.61.6;
Tweak sysctl setup functions (the macros, actually) for use in lkms,
and tweak lkminit_*.c (where applicable) to call them, and to call
sysctl_teardown() when being unloaded.
This consists of (1) making setup functions not be static when being
compiled as lkms (change to sys/sysctl.h), (2) making prototypes
visible for the various setup functions in header files (changes to
various header files), and (3) making simple "load" and "unload"
functions in the actual lkminit stuff.
linux_sysctl.c also needs its root exposed (ie, made not static) for
this (when built as an lkm).
|
| 1.60 | 21-Apr-2004 |
christos | Replace the statfs() family of system calls with statvfs().
Retain binary compatibility.
|
| 1.59 | 09-Mar-2004 |
yamt | branches: 1.59.2;
calculate data checksum inline.
|
| 1.58 | 04-Dec-2003 |
atatat | Dynamic sysctl.
Gone are the old kern_sysctl(), cpu_sysctl(), hw_sysctl(),
vfs_sysctl(), etc, routines, along with sysctl_int() et al. Now all
nodes are registered with the tree, and nodes can be added (or
removed) easily, and I/O to and from the tree is handled generically.
Since the nodes are registered with the tree, the mapping from name to
number (and back again) can now be discovered, instead of having to be
hard coded. Adding new nodes to the tree is likewise much simpler --
the new infrastructure handles almost all the work for simple types,
and just about anything else can be done with a small helper function.
All existing nodes are where they were before (numerically speaking),
so all existing consumers of sysctl information should notice no
difference.
PS - I'm sorry, but there's a distinct lack of documentation at the
moment. I'm working on sysctl(3/8/9) right now, and I promise to
watch out for buses.
|
| 1.57 | 07-Nov-2003 |
yamt | - tweak lfs_update_single()'s prototype so that it can be used by
roll-forward code.
- reduce code duplication using the above in update_meta()
this also fixes fragment accounting.
|
| 1.56 | 07-Nov-2003 |
yamt | fix spec vnode aliasing.
|
| 1.55 | 29-Sep-2003 |
yamt | remove redundant prototypes.
|
| 1.54 | 23-Sep-2003 |
yamt | cleanup IN_ADIROP/VDIROP handling a little.
|
| 1.53 | 07-Aug-2003 |
agc | Move UCB-licensed code from 4-clause to 3-clause licence.
Patches provided by Joel Baker in PR 22364, verified by myself.
|
| 1.52 | 12-Jul-2003 |
yamt | - protect global resource counts with lfs_subsys_lock.
- clean up scattered externs a little.
|
| 1.51 | 02-Jul-2003 |
yamt | - add a new functions, lfs_writer_enter/leave, and use them instead of
duplicated code fragments.
- add an assertion.
|
| 1.50 | 29-Jun-2003 |
fvdl | branches: 1.50.2;
Back out the lwp/ktrace changes. They contained a lot of colateral damage,
and need to be examined and discussed more.
|
| 1.49 | 29-Jun-2003 |
thorpej | Undo part of the ktrace/lwp changes. In particular:
* Remove the "lwp *" argument that was added to vget(). Turns out
that nothing actually used it!
* Remove the "lwp *" arguments that were added to VFS_ROOT(), VFS_VGET(),
and VFS_FHTOVP(); all they did was pass it to vget() (which, as noted
above, didn't use it).
* Remove all of the "lwp *" arguments to internal functions that were added
just to appease the above.
|
| 1.48 | 28-Jun-2003 |
darrenr | Pass lwp pointers throughtout the kernel, as required, so that the lwpid can
be inserted into ktrace records. The general change has been to replace
"struct proc *" with "struct lwp *" in various function prototypes, pass
the lwp through and use l_proc to get the process pointer when needed.
Bump the kernel rev up to 1.6V
|
| 1.47 | 02-Apr-2003 |
fvdl | Add support for UFS2. UFS2 is an enhanced FFS, adding support for
64 bit block pointers, extended attribute storage, and a few
other things.
This commit does not yet include the code to manipulate the extended
storage (for e.g. ACLs), this will be done later.
Originally written by Kirk McKusick and Network Associates Laboratories for
FreeBSD.
|
| 1.46 | 20-Mar-2003 |
yamt | fix "more than one fragment" panics;
direct and indirect block pointers are not valid in the case of shortlinks.
while i'm here, move duplicated code in lfs_vget/fastvget into a new
function, lfs_vinit.
|
| 1.45 | 08-Mar-2003 |
perseant | Add an lfs_strategy() that checks to make sure we're not trying to read
where the cleaner is trying to write, instead of tying up the "live"
buffers (or pages).
Fix a bug in the LFS_UBC case where oversized buffers would not be
checksummed correctly, causing uncleanable segments.
Make sure that wakeup(fs->lfs_iocount) is done if fs->lfs_iocount is 1
as well as 0, since we wait in some places for it to drop to 1.
Activate all pages that make it into lfs_gop_write without the segment
lock held, since they must have been dirtied very recently, even if
PG_DELWRI is not set.
|
| 1.44 | 25-Feb-2003 |
perseant | Make fs-specific fcntl macros take three arguments (approved wrstuden).
Let LFS use fcntl for cleaner functions.
|
| 1.43 | 24-Feb-2003 |
perseant | Add lfs_ioctl vnode op, with ioctls to take over cleaner system call
functionality (not including segment clean, since that is now done
automatically as checkpoints happen).
|
| 1.42 | 23-Feb-2003 |
perseant | Fix a buffer overflow bug in the LFS_UBC case that manifested itself
either as a mysterious UVM error or as "panic: dirty bufs". Verify
maximum size in lfs_malloc.
Teach lfs_updatemeta and lfs_shellsort about oversized cluster blocks from
lfs_gop_write.
When unwiring pages in lfs_gop_write, deactivate them, under the theory
that the pagedaemon wanted to free them last we knew.
|
| 1.41 | 20-Feb-2003 |
perseant | Tabify, and fix some comment alignment problems.
|
| 1.40 | 18-Feb-2003 |
perseant | Make it compile again, grr....
|
| 1.39 | 17-Feb-2003 |
perseant | Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now
(there are still some details to work out) but expect that to go
away soon. To support these basic changes (creation of lfs_putpages,
lfs_gop_write, mods to lfs_balloc) several other changes were made, to
wit:
* Create a writer daemon kernel thread whose purpose is to handle page
writes for the pagedaemon, but which also takes over some of the
functions of lfs_check(). This thread is started the first time an
LFS is mounted.
* Add a "flags" parameter to GOP_SIZE. Current values are
GOP_SIZE_READ, meaning that the call should return the size of the
in-core version of the file, and GOP_SIZE_WRITE, meaning that it
should return the on-disk size. One of GOP_SIZE_READ or
GOP_SIZE_WRITE must be specified.
* Instead of using malloc(...M_WAITOK) for everything, reserve enough
resources to get by and use malloc(...M_NOWAIT), using the reserves if
necessary. Use the pool subsystem for structures small enough that
this is feasible. This also obsoletes LFS_THROTTLE.
And a few that are not strictly necessary:
* Moves the LFS inode extensions off onto a separately allocated
structure; getting closer to LFS as an LKM. "Welcome to 1.6O."
* Unified GOP_ALLOC between FFS and LFS.
* Update LFS copyright headers to correct values.
* Actually cast to unsigned in lfs_shellsort, like the comment says.
* Keep track of which segments were empty before the previous
checkpoint; any segments that pass two checkpoints both dirty and
empty can be summarily cleaned. Do this. Right now lfs_segclean
still works, but this should be turned into an effectless
compatibility syscall.
|
| 1.38 | 01-Feb-2003 |
tron | Only use MALLOC_DECLARE() in kernel namespace.
|
| 1.37 | 01-Feb-2003 |
thorpej | Add extensible malloc types, adapted from FreeBSD. This turns
malloc types into a structure, a pointer to which is passed around,
instead of an int constant. Allow the limit to be adjusted when the
malloc type is defined, or with a function call, as suggested by
Jonathan Stone.
|
| 1.36 | 30-Jan-2003 |
yamt | there's no need to treat VOP_WHITEOUT as dirop
because it modifies only one inode.
|
| 1.35 | 24-Jan-2003 |
fvdl | Bump daddr_t to 64 bits. Replace it with int32_t in all places where
it was used on-disk, so that on-disk formats remain the same.
Remove ufs_daddr_t and ufs_lbn_t for the time being.
|
| 1.34 | 28-Dec-2002 |
yamt | - in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't
try to reclaim them.
(workaround for deadlock noted in the comment in lfs_reserveavail)
- in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
|
| 1.33 | 17-Dec-2002 |
yamt | no need for cleaner to hold vnode locks.
cleaner and normal vnode operations are synchronized enough by
seglock/fraglock and buf's B_BUSY-ness.
|
| 1.32 | 01-Dec-2002 |
matt | Add multiple inclusion protection for headers. Fix mismatched
variable declarations (missing const's) as needed.
|
| 1.31 | 16-Jun-2002 |
perseant | For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.30 | 14-May-2002 |
perseant | branches: 1.30.2; 1.30.4;
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes
I found while making sure there weren't any new ones.
* Make the write clusters keep track of the buffers whose blocks they contain.
This should make it possible to (1) write clusters using a page mapping
instead of malloc, if desired, and (2) schedule blocks for rewriting
(somewhere else) if a write error occurs. Code is present to use
pagemove() to construct the clusters but that is untested and will go away
anyway in favor of page mapping.
* DEBUG now keeps a log of Ifile writes, so that any lingering instances of
the "dirty bufs" problem can be properly debugged.
* Keep track of whether the Ifile has been dirtied by various routines that
can be called by lfs_segwrite, and loop on that until it is clean, for
a checkpoint. Checkpoints need to be squeaky clean.
* Warn the user (once) if the Ifile grows larger than is reasonable for their
buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can
grow.
* If an inode is not found in a disk block, try rereading the block, under
the assumption that the block was copied to a cluster and then freed.
* Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
|
| 1.29 | 12-May-2002 |
matt | Eliminate commons.
|
| 1.28 | 11-Feb-2002 |
perseant | Include the space taken by inodes in the count made by lfs_check();
make VOP_SETATTR call lfs_check. This prevents large numbers of inode
changes (say, at the end of tar(1)) from filling the buffer cache.
|
| 1.27 | 18-Dec-2001 |
chs | use the new compatibility routines to allow mmap() to work
(in the same non-coherent fashion that it worked pre-UBC)
until someone has time to do it the right way.
|
| 1.26 | 15-Sep-2001 |
chs | add a new VFS op, vfs_reinit, which is called when desiredvnodes is
adjusted via sysctl. file systems that have hash tables which are
sized based on the value of this variable now resize those hash tables
using the new value. the max number of FFS softdeps is also recalculated.
convert various file systems to use the <sys/queue.h> macros for
their hash tables.
|
| 1.25 | 13-Jul-2001 |
perseant | branches: 1.25.2;
Merge the short-lived perseant-lfsv2 branch into the trunk.
Kernels and tools understand both v1 and v2 filesystems; newfs_lfs
generates v2 by default. Changes for the v2 layout include:
- Segments of non-PO2 size and arbitrary block offset, so these can be
matched to convenient physical characteristics of the partition (e.g.,
stripe or track size and offset).
- Address by fragment instead of by disk sector, paving the way for
non-512-byte-sector devices. In theory fragments can be as large
as you like, though in reality they must be smaller than MAXBSIZE in size.
- Use serial number and filesystem identifier to ensure that roll-forward
doesn't get old data and think it's new. Roll-forward is enabled for
v2 filesystems, though not for v1 filesystems by default.
- The inode free list is now a tailq, paving the way for undelete (undelete
is not yet implemented, but can be without further non-backwards-compatible
changes to disk structures).
- Inode atime information is kept in the Ifile, instead of on the inode;
that is, the inode is never written *just* because atime was changed.
Because of this the inodes remain near the file data on the disk, rather
than wandering all over as the disk is read repeatedly. This speeds up
repeated reads by a small but noticeable amount.
Other changes of note include:
- The ifile written by newfs_lfs can now be of arbitrary length, it is no
longer restricted to a single indirect block.
- Fixed an old bug where ctime was changed every time a vnode was created.
I need to look more closely to make sure that the times are only updated
during write(2) and friends, not after-the-fact during a segment write,
and certainly not by the cleaner.
|
| 1.24 | 03-Dec-2000 |
perseant | branches: 1.24.2; 1.24.4; 1.24.6;
Get rid of some old unnecessary code that cleared B_NEEDCOMMIT from buffers in
lfs_writeseg (possibly after they had been freed).
If MALLOCLOG is defined, make lfs_newbuf and lfs_freebuf pass along the
caller's file and line to _malloc and _free.
|
| 1.23 | 25-Nov-2000 |
perseant | Use u_int32_t instead of u_long to compute LFS checksums, since the
checksum is stored in a u_int32_t.
|
| 1.22 | 17-Nov-2000 |
perseant | Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes.
(PR #11468). In the case of fragment allocation, check to see if enough
space is available before extending a fragment already scheduled for writing.
The locked_queue_* variables indicate the number of buffer headers and bytes,
respectively, that are unavailable to getnewbuf() because they are locked up
waiting for LFS to flush them; make sure that that is actually what we're
counting, i.e., never count malloced buffers, and always use b_bufsize instead
of b_bcount.
If DEBUG is defined, the periodic calls to lfs_countlocked will now complain
if either counter is incorrect. (In the future lfs_countlocked will not need
to be called at all if DEBUG is not defined.)
|
| 1.21 | 09-Sep-2000 |
perseant | Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
|
| 1.20 | 05-Jul-2000 |
perseant | Clean up accounting of lfs_uinodes (dirty but unwritten inodes).
Make lfs_uinodes a signed quantity for debugging purposes, and set it to
zero as fs mount time.
Enclose setting/clearing of the dirty flags (IN_MODIFIED, IN_ACCESSED,
IN_CLEANING) in macros, and use those macros everywhere. Make
LFS_ITIMES use these macros; updated the ITIMES macro in inode.h to know
about this. Make ufs_getattr use ITIMES instead of FFS_ITIMES.
|
| 1.19 | 30-Jun-2000 |
fvdl | Rearrange code around getnewvnode as was already done for ffs, to avoid
locking against oneself because getnewvnode recycles a softdep-using vnode.
|
| 1.18 | 27-Jun-2000 |
perseant | Fixes associated with filling an LFS:
Change the space computation to appear to change the size of the *disk*
rather than the *bytes used* when more segment summaries and inode
blocks are written. Try to estimate the amount of space that these will
take up when more files are written, so the disk size doesn't change too
much.
Regularize error returns from lfs_valloc, lfs_balloc, lfs_truncate: they
now fail entirely, rather than succeeding half-way and leaving the fs in
an inconsistent state.
Rewrite lfs_truncate, mostly stealing from ffs_truncate. The old
lfs_truncate had difficulty truncating a large file to a non-zero size
(indirect blocks were not handled appropriately).
Unmark VDIROP on fvp after ufs_remove, ufs_rmdir, so these can be
reclaimed immediately: this vnode would not be written to disk again
anyway if the removal succeeded, and if it failed, no directory
operation occurred.
ufs_makeinode and ufs_mkdir now remove IN_ADIROP on error.
|
| 1.17 | 16-Mar-2000 |
jdolecek | branches: 1.17.4;
Add new VFS op routine - vfs_done and call it on filesystem detach
in vfs_detach(). vfs_done may free global filesystem's resources,
typically those allocated in respective filesystem's init function.
Needed so those filesystems which went in via LKM have a chance to
clean after themselves before unloading. This fixes random panics
when LKM for filesystem using pools was loaded and unloaded several
times.
For each leaf filesystem, add appropriate vfs_done routine.
|
| 1.16 | 19-Jan-2000 |
perseant | Changes to stabilize LFS. The first two of these should also apply to the
1.4 branch.
* Use a separate per-fs lock, instead of ufs_hashlock, to protect the Inode
free list. This seems to prevent the "lockmgr: %d, not exclusive lock holder
%d, unlocking" message I was mis-attributing last night to an unlocked vnode
being passed to vrele.
* Change calling semantics of lfs_ifind, to give better error reporting:
If fed a struct buf, it can report the block number of the offending inode
block as well as the inode number.
* Back out rev 1.10 of lfs_subr.c, since the replacement code was slightly
uglier while being functionally identical.
* Make lfs_vunref use the same free list convention as vrele/vput, so that
vget does not remove vnodes from a hash list they are not on.
|
| 1.15 | 15-Nov-1999 |
fvdl | Add Kirk McKusick's soft updates code to the trunk. Not enabled by
default, as the copyright on the main file (ffs_softdep.c) is such
that is has been put into gnusrc. options SOFTDEP will pull this
in. This code also contains the trickle syncer.
Bump version number to 1.4O
|
| 1.14 | 01-Jun-1999 |
perseant | branches: 1.14.2; 1.14.4; 1.14.8;
Fixed lfs_update (and related functions) so that calls from lfs_fsync
will DTRT with vnodes marked VDIROP. In particular, the message
"flushing VDIROP" will no longer appear, and the filesystem will remain
stable in the event of a crash.
This was particularly a problem with NFS-exported LFSes, since fsync
was called on every file close.
|
| 1.13 | 10-Mar-1999 |
perseant | branches: 1.13.2; 1.13.4;
New sources should leave the LFS in a more-or-less working state. Changes
include:
- DIROP segregation is enabled, and greater care is taken
to make sure that a checkpoint completes. Fsck is not
needed to remount the filesystem.
- Several checks to make sure that the LFS subsystem does not
overuse various resources (memory, in particular).
- The cleaner routines, lfs_markv in particular, are completely
rewritten. A buffer overflow is removed. Greater care is taken
to ensure that inodes come from where lfs_cleanerd say they come
from (so we know nothing has changed since lfs_bmapv was called).
- Fragment allocation is fixed, so that writes beyond end-of-file
do the right thing.
|
| 1.12 | 26-Feb-1999 |
wrstuden | Modify vfsops to seperate vfs_fhtovp() into two routines. vfs_fhtovp() now
only handles the file handle to vnode conversion, and a new call,
vfs_checkexp(), performs the export verification.
|
| 1.11 | 11-Sep-1998 |
pk | PR#6032: define fixed sized on-disk superblock structure.
|
| 1.10 | 01-Sep-1998 |
thorpej | Use the pool allocator and the "nointr" pool page allocator for LFS inodes.
|
| 1.9 | 24-Jun-1998 |
sommerfe | Always include fifos; "not an option any more".
|
| 1.8 | 22-Jun-1998 |
sommerfe | defopt for options FIFO
|
| 1.7 | 01-Mar-1998 |
fvdl | Merge with Lite2 + local changes
|
| 1.6 | 22-Dec-1996 |
cgd | Change the second and third args to struct vfsops' (*vfs_mount)() to
'const char *', and 'void *', respectively. The second arg is taken directly
from user arguments, and is const there, so must be const in the prototypes
and functions. The third arg is also taken directly from user arguments.
It doesn't have to be changed, but since it's cleaner to keep the type
the same as the user arg's type, and I'm already making the 'const char *'
change...
|
| 1.5 | 12-Feb-1996 |
christos | Add fwd declaration for struct ucred
|
| 1.4 | 09-Feb-1996 |
christos | lfs prototypes
|
| 1.3 | 14-Dec-1994 |
mycroft | Sync with CSRG.
|
| 1.2 | 29-Jun-1994 |
cgd | New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
|
| 1.1 | 08-Jun-1994 |
mycroft | branches: 1.1.1;
Update to 4.4-Lite fs code, with local changes.
|
| 1.1.1.2 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite2
|
| 1.1.1.1 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite for reference
|
| 1.13.4.3 | 31-Aug-1999 |
perseant | Rudimentary support for LFS under UBC:
- LFS-specific VOP_BALLOC and VOP_PUTPAGES vnode ops.
- getblk VREG panic #ifdef'd out (can be reinstated when Ifile is
internalized and Ifile can be made another type from VREG)
- interface to VOP_PUTPAGES changed to pass all pager flags, not
just sync. FS putpages routines must know about the pager flags.
- new LFS magic disk address, -2 ("unwritten"), meaning accounted for
but not assigned to a fixed disk location (since LFS does these two
things separately, and the previous accounting method using buffer
headers no longer will work). Changed references to (foo == (daddr_t)-1)
to (foo < 0). Since disk drivers reject all addresses < 0, this should
not present a problem for other FSs.
|
| 1.13.4.2 | 11-Jul-1999 |
chs | add placeholders for getpages/putpages.
|
| 1.13.4.1 | 21-Jun-1999 |
thorpej | Sync w/ -current.
|
| 1.13.2.2 | 20-Jan-2000 |
he | Pull up revision 1.16 (requested by perseant):
Files removed (through unlink, rmdir) are now really removed, though the
removal is postponed until the dirop is complete to ensure validity of
the filesystem through a crash. Use a separate per-fs lock, instead of
ufs_hashlock, to protect the inode free list. Change calling semantics
of lfs_ifind, to give better error reporting: If fed a struct buf, it
can report the block number of the offending inode block as well as the
inode number.
|
| 1.13.2.1 | 17-Dec-1999 |
he | Pull up revision 1.14 (requested by perseant):
Avoid flushing vnodes involved in a dirop, making lfs' promise
of "no fsck needed, even in the event of a crash" closer to
reality.
|
| 1.14.8.1 | 27-Dec-1999 |
wrstuden | Pull up to last week's -current.
|
| 1.14.4.1 | 19-Oct-1999 |
fvdl | Bring in Kirk McKusick's FFS softdep code on a branch.
|
| 1.14.2.3 | 08-Dec-2000 |
bouyer | Sync with HEAD.
|
| 1.14.2.2 | 22-Nov-2000 |
bouyer | Sync with HEAD.
|
| 1.14.2.1 | 20-Nov-2000 |
bouyer | Update thorpej_scsipi to -current as of a month ago
|
| 1.17.4.3 | 03-Feb-2001 |
he | Pull up revision 1.22 (requested by perseant):
o Close up accounting holes in LFS' accounting of immediately-
available-space, number of clean segments, and amount of dirty
space taken up by metadata (PR#11468, PR#11470, PR#11534).
|
| 1.17.4.2 | 14-Sep-2000 |
perseant | Pull up recent LFS kernel changes (approved by thorpej):
ufs/ufs/inode.h, 1.20--1.22 (add i_lfs_effnblks extension ;
make ITIMES aware of LFS_ITIMES;
_LKM protection so userland progs
compile)
ufs/ufs/ufs_vnops.c, 1.69, 1.71 (remove IN_ADIROP;
use ITIMES instead of FFS_ITIMES)
ufs/ufs/ufs_readwrite.c, 1.27 (use lfs_reserve in lfs_write)
ufs/lfs/lfs.h, 1.26--1.32 (define LFS_EST_* macros ;
change MIN_FREE_SEGS to lfs_minfreesegs ;
add avail and bfree to CLEANERINFO ;
change lfs_uinodes to signed ;
change lfs_dmeta to signed ;
add whitespace to line up structure
members ;
explicit cast to int32_t in LFS_EST_*
macros)
ufs/lfs/lfs_alloc.c, back out 1.34.2.3 (pullups of 1.39, 1.40);
then pull up 1.38 (clean up on error)
1.39--1.43 (restore fvdl's ufs_hashlock fix ;
restore fvdl's ufs_hashlock fix ;
set i_lfs_effnblks ;
use UINO macros ;
add comments and fix long lines)
ufs/lfs/lfs_balloc.c, 1.19 (don't succeed halfway)
1.21--1.25 (use i_lfs_effnblks ;
fix i_lfs_effnblks computation and
quieten ;
fix i_ffs_blocks in unwritten fragment ;
remove useless debugging check ;
add comments and (c) 2000)
ufs/lfs/lfs_bio.c, 1.24--1.30 (cleanup and make lfs_flush_fs take
"struct lfs *" instead of "struct
mount *" ;
use lfs_minfreeseg instead of
MIN_FREE_SEGS ;
use UINO macros, and copy bfree/avail
to CLEANERINFO ;
add lfs_reserve function ;
1.28--1.30 fix printf formatting)
ufs/lfs/lfs_cksum.c, 1.13 (add (c) 2000)
ufs/lfs/lfs_debug.c, 1.11 (use btodb instead of DEV_BSIZE)
ufs/lfs/lfs_extern.h, 1.18, 1.20--1.21 (function prototype changes)
ufs/lfs/lfs_inode.c, 1.38 (rewrite lfs_truncate from
ffs_truncate)
1.40--1.44 (count written and unwritten blocks
seperately ;
use disk block units instead of bytes ;
remove unnecessary "mod" variable ;
correct B_DELWRI to avoid bawrite panic ;
use lfs_reserve)
ufs/lfs/lfs_segment.c, 1.52-1.59 (use lfs_dmeta to note used summaries ;
check for UNWRITTEN in indirect blocks ;
more debugging stuff inside #ifdef
DEBUG_LFS ;
use LK_CANRECURSE ;
don't drop dirty indirect blocks ;
use UINO macros ;
don't hose the free list ;
use btodb() instead of DEV_BSIZE ;
make it compile again (oops))
ufs/lfs/lfs_subr.c, 1.16--1.17 (check for locked inodes before
changing ;
use btodb() instead of DEV_BSIZE, (c)
2000)
ufs/lfs/lfs_syscalls.c, back out 1.41.4.2 (fvdl's ufs_hashlock fix);
then pull up 1.43 (use lfs_dmeta)
1.44--1.45 (restore fvdl's ufs_hashlock fix)
1.46--1.47 (fix lfs_avail leakage from sblock
segments ;
use UINO macros)
1.49 (bounds-check inode numbers in
lfs_markv)
ufs/lfs/lfs_vfsops.c, 1.53 (use LFS_EST_* macros in lfs_statfs)
1.56--1.58 (initialize lfs_minfreeseg, lfs_effnblk ;
initialize lfs_uinodes ;
initialize lfs_ravail)
ufs/lfs/lfs_vnops.c, 1.40 (remove VDIROP from removed files)
1.42--1.44 (move SET_ENDOP below the removal of
VDIROP ;
use UINO macros and add lfs_itimes
function ;
use lfs_reserve in dirops)
|
| 1.17.4.1 | 03-Jul-2000 |
fvdl | pullup the fixes from the trunk to not hold ufs_hashlock across
getnewvnode()
|
| 1.24.6.4 | 23-Jun-2002 |
jdolecek | catch up with -current on kqueue branch
|
| 1.24.6.3 | 16-Mar-2002 |
jdolecek | Catch up with -current.
|
| 1.24.6.2 | 10-Jan-2002 |
thorpej | Sync kqueue branch with -current.
|
| 1.24.6.1 | 03-Aug-2001 |
lukem | update to -current
|
| 1.24.4.2 | 29-Jun-2001 |
perseant | Get rid of __P(), protoizing where it had not already been done
|
| 1.24.4.1 | 27-Jun-2001 |
perseant | Import of what I've been calling "LFSv2", that is, LFS with some features
added that require changes to the on-disk data structures. These include:
- 64-bit time in everything but inodes
- User-specified segment offset, and segment size no longer
restricted to PO2.
- Serial number on segment summaries in addition to timestamp, and
a new volume identifier, to make roll-forward feasible without
fear of finding old data and thinking it was new.
Although I think this version works at least as well as what's on the trunk,
we're not done yet; hence this commit is going in on a branch and not on
the trunk. Enhancements that are not here yet include fragment addressing,
like FFS does, instead of block addressing.
|
| 1.24.2.8 | 29-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.24.2.7 | 19-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.24.2.6 | 11-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.24.2.5 | 20-Jun-2002 |
nathanw | Catch up to -current.
|
| 1.24.2.4 | 28-Feb-2002 |
nathanw | Catch up to -current.
|
| 1.24.2.3 | 08-Jan-2002 |
nathanw | Catch up to -current.
|
| 1.24.2.2 | 21-Sep-2001 |
nathanw | Catch up to -current.
|
| 1.24.2.1 | 24-Aug-2001 |
nathanw | Catch up with -current.
|
| 1.25.2.1 | 01-Oct-2001 |
fvdl | Catch up with -current.
|
| 1.30.4.1 | 20-Jun-2002 |
lukem | Pull up revision 1.31 (requested by perseant in ticket #325):
For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.30.2.1 | 20-Jun-2002 |
gehenna | catch up with -current.
|
| 1.50.2.9 | 10-Nov-2005 |
skrll | Sync with HEAD. Here we go again...
|
| 1.50.2.8 | 08-Mar-2005 |
skrll | Sync with HEAD.
|
| 1.50.2.7 | 04-Mar-2005 |
skrll | Sync with HEAD.
Hi Perry!
|
| 1.50.2.6 | 30-Oct-2004 |
skrll | Oops, forgot this as part of the
"Reduced diff to HEAD by restoring the struct proc * argument to lfs_bmapv"
change
|
| 1.50.2.5 | 21-Sep-2004 |
skrll | Fix the sync with head I botched.
|
| 1.50.2.4 | 18-Sep-2004 |
skrll | Sync with HEAD.
|
| 1.50.2.3 | 24-Aug-2004 |
skrll | Undo part of the ktrace/lwp changes. In particular:
* Remove the "lwp *" argument that was added to vget(). Turns out
that nothing actually used it!
* Remove the "lwp *" arguments that were added to VFS_ROOT(), VFS_VGET(),
and VFS_FHTOVP(); all they did was pass it to vget() (which, as noted
above, didn't use it).
* Remove all of the "lwp *" arguments to internal functions that were added
just to appease the above.
|
| 1.50.2.2 | 03-Aug-2004 |
skrll | Sync with HEAD
|
| 1.50.2.1 | 02-Jul-2003 |
darrenr | Apply the aborted ktrace-lwp changes to a specific branch. This is just for
others to review, I'm concerned that patch fuziness may have resulted in some
errant code being generated but I'll look at that later by comparing the diff
from the base to the branch with the file I attempt to apply to it. This will,
at the very least, put the changes in a better context for others to review
them and attempt to tinker with removing passing of 'struct lwp' through
the kernel.
|
| 1.59.2.1 | 23-May-2004 |
tron | branches: 1.59.2.1.2;
Pull up revision 1.61 (requested by atatat in ticket #374):
Tweak sysctl setup functions (the macros, actually) for use in lkms,
and tweak lkminit_*.c (where applicable) to call them, and to call
sysctl_teardown() when being unloaded.
This consists of (1) making setup functions not be static when being
compiled as lkms (change to sys/sysctl.h), (2) making prototypes
visible for the various setup functions in header files (changes to
various header files), and (3) making simple "load" and "unload"
functions in the actual lkminit stuff.
linux_sysctl.c also needs its root exposed (ie, made not static) for
this (when built as an lkm).
|
| 1.59.2.1.2.1 | 10-May-2005 |
riz | Pull up the following revisions (requested by perseant in ticket #1281):
1.8 sys/ufs/lfs/TODO
1.75 sys/ufs/lfs/lfs.h (via patch)
1.74 sys/ufs/lfs/lfs_alloc.c (via patch)
1.49, 1.51 sys/ufs/lfs/lfs_balloc.c (1.51 via patch)
1.78 sys/ufs/lfs/lfs_bio.c
1.62 sys/ufs/lfs/lfs_extern.h (via patch)
1.156 sys/ufs/lfs/lfs_segment.c (via patch)
1.48 sys/ufs/lfs/lfs_subr.c
1.101 sys/ufs/lfs/lfs_syscalls.c
1.163 sys/ufs/lfs/lfs_vfsops.c (via patch)
1.134 sys/ufs/lfs/lfs_vnops.c (via patch)
1.61 sys/ufs/ufs/ufs_readwrite.c (via patch)
1.20 libexec/lfs_cleanerd/clean.h (via patch)
1.52 libexec/lfs_cleanerd/cleanerd.c (via patch)
1.41 libexec/lfs_cleanerd/library.c (via patch)
1.4 regress/sys/fs/lfs/newfs_fsck/Makefile
1.2 regress/sys/fs/lfs/newfs_fsck/mkfs_mount
1.2 regress/sys/fs/lfs/newfs_fsck/smallfiles
1.3 sbin/fsck_lfs/bufcache.c
1.3 sbin/fsck_lfs/bufcache.h
1.3 sbin/fsck_lfs/lfs.h
1.8 sbin/fsck_lfs/lfs.c (via patch)
1.8 sbin/fsck_lfs/pass3.c (via patch)
1.18 sbin/fsck_lfs/pass0.c (via patch)
1.18 sbin/fsck_lfs/utilities.c (via patch)
1.7 sbin/fsck_lfs/segwrite.c
1.19 sbin/fsck_lfs/setup.c (via patch)
1.3 sbin/newfs_lfs/Makefile
0 sbin/newfs_lfs/lfs.c (yes, remove it)
1.1 sbin/newfs_lfs/make_lfs.c
1.15 sbin/newfs_lfs/newfs.c (via patch)
Various minor LFS improvements.
Kernel:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this. Should fix PR #29045.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
Fixes PR #26680.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
cleaner:
* Adapt lfs_cleanerd to use the fcntl call to get the Ifile filehandle,
so it need not be in the namespace.
* Make lfs_cleanerd be more careful when there are very few available
segments.
* Make lfs_cleanerd less verbose when the filesystem is unmounted.
newfs_lfs, fsck_lfs, and regression:
* Extend the lfs library from fsck_lfs(8) so that it can be used with a
not-yet-existent LFS. Make newfs_lfs(8) use this library, so it can
create LFSs whose Ifile is larger than one segment. Addresses PR #11110.
* Make newfs_lfs(8) use strsuftoi64() for its arguments, a la newfs(8).
* Make fsck_lfs(8) respect the "file system is clean" flag.
* Don't let fsck_lfs(8) think it has dirty blocks when invoked with the
-n flag.
* Remove the Ifile from the filesystem namespace. The cleaner now uses
a fcntl call on the root inode to find the Ifile filehandle. (As a
side-effect, addresses PR #29144.)
|
| 1.61.6.1 | 19-Mar-2005 |
yamt | sync with head. xen and whitespace. xen part is not finished.
|
| 1.61.4.1 | 29-Apr-2005 |
kent | sync with -current
|
| 1.64.2.9 | 10-Aug-2006 |
tron | Apply patch (requested by fair in perseant #1457):
Bring LFS up to current, including a patch (1.95 lfs_alloc.c) that
should prevent the inode free list errors seen on the STABLE branch
subsequent to pullup ticket #1327.
|
| 1.64.2.8 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.171
sys/ufs/lfs/lfs_extern.h: revision 1.81
sys/ufs/lfs/lfs_segment.c: revision 1.177
Don't ever partially write dirops, even if we need the cleaner to run.
This increases the chances of the "no clean segments" panic slightly,
but allows us to run the ckckp regression test successfully to completion.
|
| 1.64.2.7 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.104
sys/ufs/lfs/lfs_vfsops.c: revision 1.206
sys/ufs/lfs/lfs_vnops.c: revision 1.170
sys/ufs/lfs/lfs_extern.h: revision 1.80
sys/ufs/lfs/lfs_segment.c: revision 1.176
sys/ufs/lfs/lfs_inode.c: revision 1.103 via patch
sys/ufs/lfs/lfs_alloc.c: revision 1.90
Postpone the segment accounting changes coming from truncation until the
inode that makes those changes valid is either written to disk by
lfs_writeinode() or discarded by lfs_vfree().
A couple of locking fixes are also included as well.
|
| 1.64.2.6 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.200
sys/ufs/lfs/lfs_vnops.c: revision 1.164
sys/ufs/lfs/lfs_inode.c: revision 1.101
sys/ufs/lfs/lfs_extern.h: revision 1.78
sys/ufs/lfs/lfs.h: revision 1.100
Implement a somewhat finer-grained mechanism for paging LFS-backed pages.
The writer daemon, if it does not need to flush the whole filesystem,
now only writes the vnodes for which the pagedaemon has requested pageouts
(although it does not pay attention to the page ranges the pagedaemon
supplies).
|
| 1.64.2.5 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.87
sys/ufs/lfs/lfs.h: revision 1.99
sys/ufs/lfs/lfs_vfsops.c: revision 1.199
sys/ufs/lfs/lfs_extern.h: revision 1.77 via patch
Keep the free list ordered. This solves a problem first pointed out to me
by Michel Oey, in which an aged LFS writes up to an extra Ifile block for
every file created; and paves the way for the truncation of the Ifile when
many files are deleted.
|
| 1.64.2.4 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.158
sys/ufs/lfs/lfs_subr.c: revision 1.57
sys/ufs/lfs/lfs_segment.c: revision 1.171
sys/ufs/lfs/lfs.h: revision 1.97
sys/ufs/lfs/lfs_vfsops.c: revision 1.195
sys/ufs/lfs/lfs_extern.h: revision 1.76
Improvements to LFS's paging mechanism, to wit:
* Acknowledge that sometimes there are more dirty pages to be written to
disk than clean segments. When we reach the danger line,
lfs_gop_write() now returns EAGAIN. The caller of VOP_PUTPAGES(), if
it holds the segment lock, drops it and waits for the cleaner to make
room before continuing.
* Note and avoid a three-way deadlock in lfs_putpages (a writer holding
a page busy blocks on the cleaner while the cleaner blocks on the
segment lock while lfs_putpages blocks on the page).
|
| 1.64.2.3 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.152
sys/ufs/lfs/lfs_debug.c: revision 1.31
sys/ufs/lfs/lfs_subr.c: revision 1.53
sys/ufs/lfs/lfs_extern.h: revision 1.68
sys/ufs/lfs/lfs_inode.c: revision 1.96
sys/ufs/lfs/lfs_bio.c: revision 1.86
sys/ufs/lfs/lfs_alloc.c: revision 1.83
sys/ufs/lfs/lfs_vfsops.c: revision 1.181
sys/ufs/lfs/lfs.h: revision 1.88
sys/ufs/lfs/lfs_segment.c: revision 1.164
- sprinkle const
- avoid shadow variables.
|
| 1.64.2.2 | 24-Aug-2005 |
riz | Pull up following revision(s) (requested by yamt in ticket #688):
sys/miscfs/genfs/genfs_vnops.c: revision 1.98 via patch
sys/ufs/ffs/ffs_vfsops.c: revision 1.165
sys/ufs/lfs/lfs_extern.h: revision 1.69
sys/fs/filecorefs/filecore_vfsops.c: revision 1.20
sys/nfs/nfs_node.c: revision 1.80
sys/fs/smbfs/smbfs_node.c: revision 1.24
sys/fs/cd9660/cd9660_vfsops.c: revision 1.24
sys/fs/msdosfs/msdosfs_denode.c: revision 1.8
sys/miscfs/genfs/genfs_node.h: revision 1.6
sys/ufs/lfs/lfs_vfsops.c: revision 1.183
sys/ufs/ext2fs/ext2fs_vfsops.c: revision 1.86
sys/fs/adosfs/advfsops.c: revision 1.23
sys/fs/ntfs/ntfs_vfsops.c: revision 1.31
- constify genfs_ops.
- use member designators.
sys/miscfs/genfs/genfs_vnops.c: revision 1.99 via patch
genfs_getpages: don't forget to put the vnode onto the syncer's work que
ue
even in the case of PGO_LOCKED.
sys/uvm/uvm_bio.c: revision 1.40
sys/uvm/uvm_pager.h: revision 1.29
sys/miscfs/genfs/genfs_vnops.c: revision 1.100 via patch
sys/ufs/ufs/ufs_inode.c: revision 1.50
- introduce PGO_NOBLOCKALLOC and use it for ubc mapping
to prevent unnecessary block allocations in the case that
page size > block size.
- ufs_balloc_range: use VM_PROT_WRITE+PGO_NOBLOCKALLOC rather than
VM_PROT_READ.
sys/uvm/uvm_fault.c: revision 1.96
sys/miscfs/genfs/genfs_vnops.c: revision 1.101 via patch
sys/uvm/uvm_object.h: revision 1.19
sys/miscfs/genfs/genfs_node.h: revision 1.7
ensure that vnodes with dirty pages are always on syncer's queue.
- genfs_putpages: wait for i/o completion of PG_RELEASED/PG_PAGEOUT pages by
setting "wasclean" false when encountering them.
suggested by Stephan Uphoff in PR/24596 (1).
- genfs_putpages: write protect pages when cleaning out, if
we're going to take the vnode off the syncer's queue.
uvm_fault: don't write-map pages unless its vnode is already on
the syncer's queue.
fix PR/24596 (3) but in the different way from the suggested fix.
(to keep our current behaviour, ie. not to require explicit msync.
discussed on tech-kern@.)
- genfs_putpages: don't mistakenly take a vnode off the queue
by introducing a generation number in genfs_node.
genfs_getpages: increment the generation number.
suggested by Stephan Uphoff in PR/24596 (2).
- add some assertions.
sys/miscfs/genfs/genfs_vnops.c: revision 1.102 via patch
genfs_putpages: don't bother to clean the vnode unless VONWORKLST.
sys/ufs/ffs/ffs_vnops.c: revision 1.71
ffs_full_fsync: because VBLK/VCHR can be mmap'ed,
do VOP_PUTPAGES for them as well.
sys/uvm/uvm_fault.c: revision 1.97
uvm_fault: check a correct object in the case of layered filesystems.
fix PR/30811 from Jukka Salmi.
sys/uvm/uvm_object.h: revision 1.20
sys/ufs/ffs/ffs_vfsops.c: revision 1.167
sys/uvm/uvm_bio.c: revision 1.41
sys/ufs/ufs/ufs_vnops.c: revision 1.129
sys/uvm/uvm_mmap.c: revision 1.92
sys/uvm/uvm_fault.c: revision 1.98
sys/kern/vfs_subr.c: revision 1.252
sys/fs/msdosfs/denode.h: revision 1.5
sys/miscfs/genfs/genfs_vnops.c: revision 1.103 via patch
sys/fs/msdosfs/msdosfs_denode.c: revision 1.9
sys/sys/vnode.h: revision 1.141
sys/ufs/ufs/ufs_inode.c: revision 1.51
sys/ufs/ufs/ufs_extern.h: revision 1.45 via patch
sys/miscfs/genfs/genfs_node.h: revision 1.8
sys/ufs/lfs/lfs_vfsops.c: revision 1.184
sys/uvm/uvm_pager.h: revision 1.30
sys/ufs/ext2fs/ext2fs_vfsops.c: revision 1.87
update file timestamps for nfsd loaned-read and mmap.
PR/25279. discussed on tech-kern@.
sys/miscfs/genfs/genfs_vnops.c: revision 1.104 via patch
don't write-protect wired pages. pointed by Chuck Silvers.
for now, leave a vnode on the syncer's queue, as suggested by him.
sys/ufs/ffs/ffs_vnops.c: revision 1.72
revert VCHR part of ffs_vnops.c 1.71.
as VCHR uses the device pager, no point to call VOP_PUTPAGES here.
pointed by Chuck Silvers.
|
| 1.64.2.1 | 07-May-2005 |
tron | Apply patch (requested by perseant in ticket #242):
* fsck_lfs buffer cache fixes, including PR #29151
* Change fsck_lfs phase 0 message to reflect reality
* fsck_lfs: check phase 5 (cleanerinfo accounting) even on
roll-forward
* Keep better track of the free list during roll-forward, avoiding
a core dump
* Improve hash table use for fsck_lfs buffer and vnode cache
* Document fsck_lfs flag -f, and implement -q
* Add resize_lfs, including kernel support
* Add LFS to mountd's list of exportable filesystem types
* Make the LFS lkm work again [christos@]
* Add MP locking to the LFS kernel subsystem
* Fix pager_map deadlock in lfs_putpages()
* Avoid incomplete file extension that looks like "partial
truncation" to fsck
* Use lfs_malloc for cleaner malloc, since the cleaner often runs
in low-memory conditions.
* Use splay trees, not hash table, to track page allocation for
write.
* Fix mkdir panic on full fs
* Fix page accounting leak by counting differently.
* Use rightly named structure for lfs_getattr [skrll@]
* Cosmetic changes for readability.
|
| 1.69.2.5 | 21-Jan-2008 |
yamt | sync with head
|
| 1.69.2.4 | 07-Dec-2007 |
yamt | sync with head
|
| 1.69.2.3 | 03-Sep-2007 |
yamt | sync with head.
|
| 1.69.2.2 | 30-Dec-2006 |
yamt | sync with head.
|
| 1.69.2.1 | 21-Jun-2006 |
yamt | sync with head.
|
| 1.71.2.1 | 20-Oct-2005 |
yamt | adapt ufs.
|
| 1.73.2.1 | 15-Jan-2006 |
yamt | sync with head.
|
| 1.75.10.2 | 24-May-2006 |
tron | Merge 2006-05-24 NetBSD-current into the "peter-altq" branch.
|
| 1.75.10.1 | 28-Mar-2006 |
tron | Merge 2006-03-28 NetBSD-current into the "peter-altq" branch.
|
| 1.75.8.5 | 11-May-2006 |
elad | sync with head
|
| 1.75.8.4 | 06-May-2006 |
christos | - Move kauth_cred_t declaration to <sys/types.h>
- Cleanup struct ucred; forward declarations that are unused.
- Don't include <sys/kauth.h> in any header, but include it in the c files
that need it.
Approved by core.
|
| 1.75.8.3 | 03-May-2006 |
yamt | wrap some decls with #ifdef _KERNEL. ok'ed by elad@.
|
| 1.75.8.2 | 19-Apr-2006 |
elad | sync with head.
|
| 1.75.8.1 | 08-Mar-2006 |
elad | Adapt to kernel authorization KPI.
|
| 1.75.6.5 | 03-Sep-2006 |
yamt | sync with head.
|
| 1.75.6.4 | 11-Aug-2006 |
yamt | sync with head
|
| 1.75.6.3 | 24-May-2006 |
yamt | sync with head.
|
| 1.75.6.2 | 11-Apr-2006 |
yamt | sync with head
|
| 1.75.6.1 | 01-Apr-2006 |
yamt | sync with head.
|
| 1.75.4.2 | 01-Jun-2006 |
kardel | Sync with head.
|
| 1.75.4.1 | 22-Apr-2006 |
simonb | Sync with head.
|
| 1.75.2.1 | 09-Sep-2006 |
rpaulo | sync with head
|
| 1.83.4.1 | 13-Jul-2006 |
gdamore | Merge from HEAD.
|
| 1.87.12.1 | 03-Sep-2007 |
wrstuden | Sync w/ NetBSD-4-RC_1
|
| 1.87.8.2 | 07-May-2007 |
yamt | sync with head.
|
| 1.87.8.1 | 12-Mar-2007 |
rmind | Sync with HEAD.
|
| 1.87.6.1 | 05-Jun-2007 |
bouyer | Pull up following revision(s) (requested by perseant in ticket #703):
sys/miscfs/genfs/genfs.h 1.21
sys/miscfs/genfs/genfs_vnops.c 1.151
sys/ufs/lfs/lfs.h 1.119, 1.120
sys/ufs/lfs/lfs_bio.c 1.99-101
sys/ufs/lfs/lfs_extern.h 1.89
sys/ufs/lfs/lfs_inode.c 1.108, 1.109
sys/ufs/lfs/lfs_segment.c 1.197, 1.199, 1.200
sys/ufs/lfs/lfs_subr.c 1.69, 1.70
sys/ufs/lfs/lfs_syscalls.c 1.119
sys/ufs/lfs/lfs_vfsops.c 1.234, 1.235
sys/ufs/lfs/lfs_vnops.c 1.195, 1.196, 1.200, 1.202-206
Reduce busy waiting in lfs_putpages(), and other LFS improvements.
|
| 1.88.4.1 | 11-Jul-2007 |
mjf | Sync with head.
|
| 1.88.2.4 | 20-Aug-2007 |
ad | Sync with HEAD.
|
| 1.88.2.3 | 15-Jul-2007 |
ad | Sync with head.
|
| 1.88.2.2 | 08-Jun-2007 |
ad | Sync with head.
|
| 1.88.2.1 | 13-Mar-2007 |
ad | Pull in the initial set of changes for the vmlocking branch.
|
| 1.90.2.1 | 15-Aug-2007 |
skrll | Sync with HEAD.
|
| 1.91.12.2 | 31-Jul-2007 |
pooka | * nuke the nameidata parameter from VFS_MOUNT(). Nobody on tech-kern
knew what it was supposed to be used for and wrstuden gave a go-ahead
* while rototilling, convert file systems which went easily to
use VFS_PROTOS() instead of manually prototyping the methods
|
| 1.91.12.1 | 31-Jul-2007 |
pooka | file lfs_extern.h was added on branch matt-mips64 on 2007-07-31 21:14:21 +0000
|
| 1.91.10.3 | 18-Feb-2008 |
mjf | Sync with HEAD.
|
| 1.91.10.2 | 27-Dec-2007 |
mjf | Sync with HEAD.
|
| 1.91.10.1 | 08-Dec-2007 |
mjf | Sync with HEAD.
|
| 1.91.4.1 | 09-Jan-2008 |
matt | sync with HEAD
|
| 1.91.2.2 | 09-Dec-2007 |
jmcneill | Sync with HEAD.
|
| 1.91.2.1 | 27-Nov-2007 |
joerg | Sync with HEAD. amd64 Xen support needs testing.
|
| 1.92.2.4 | 26-Dec-2007 |
ad | Sync with head.
|
| 1.92.2.3 | 19-Dec-2007 |
ad | Use a global lfs_lock.
|
| 1.92.2.2 | 19-Dec-2007 |
ad | Get lfs mostly working.
|
| 1.92.2.1 | 04-Dec-2007 |
ad | Pull the vmlocking changes into a new branch.
|
| 1.93.4.1 | 02-Jan-2008 |
bouyer | Sync with HEAD
|
| 1.94.10.2 | 04-May-2009 |
yamt | sync with head.
|
| 1.94.10.1 | 16-May-2008 |
yamt | sync with head.
|
| 1.94.8.1 | 18-May-2008 |
yamt | sync with head.
|
| 1.94.6.2 | 29-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.94.6.1 | 02-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.95.4.1 | 03-Jul-2008 |
simonb | Sync with head.
|
| 1.95.2.1 | 18-Sep-2008 |
wrstuden | Sync with wrstuden-revivesa-base-2.
|
| 1.96.34.2 | 24-Feb-2012 |
mrg | sync to -current.
|
| 1.96.34.1 | 18-Feb-2012 |
mrg | merge to -current.
|
| 1.96.30.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.96.30.1 | 17-Apr-2012 |
yamt | sync with head
|
| 1.98.2.3 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.98.2.2 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.98.2.1 | 23-Jun-2013 |
tls | resync from head
|
| 1.99.4.1 | 23-Jul-2013 |
riastradh | sync with HEAD
|
| 1.99.2.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.101.6.5 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.101.6.4 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.101.6.3 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.101.6.2 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.101.6.1 | 06-Apr-2015 |
skrll | Sync with HEAD
|
| 1.111.10.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.113.4.2 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.113.4.1 | 10-Jun-2019 |
christos | Sync with HEAD
|
| 1.113.2.1 | 06-Sep-2018 |
pgoyette | Sync with HEAD
Resolve a couple of conflicts (result of the uimin/uimax changes)
|
| 1.114.6.1 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.114.4.1 | 17-Aug-2020 |
martin | Pull up following revision(s) (requested by riastradh in ticket #1050):
sys/ufs/lfs/lfs_subr.c: revision 1.101
sys/ufs/lfs/lfs_subr.c: revision 1.102
sys/ufs/lfs/lfs_inode.c: revision 1.158
sys/ufs/lfs/lfs_inode.h: revision 1.25
sys/ufs/lfs/lfs_balloc.c: revision 1.95
sys/ufs/lfs/lfs_pages.c: revision 1.21
sys/ufs/lfs/lfs_vnops.c: revision 1.330
sys/ufs/lfs/lfs_alloc.c: revision 1.140 (patch)
sys/ufs/lfs/lfs_alloc.c: revision 1.141 (patch)
lib/libp2k/p2k.c: revision 1.72
sys/ufs/lfs/lfs.h: revision 1.205
sys/ufs/lfs/lfs.h: revision 1.206
sys/ufs/lfs/lfs_segment.c: revision 1.284
sys/ufs/lfs/lfs.h: revision 1.207
sys/ufs/lfs/lfs_segment.c: revision 1.285
sys/ufs/lfs/lfs_debug.c: revision 1.55
sys/ufs/lfs/lfs_rename.c: revision 1.23
usr.sbin/dumplfs/dumplfs.c: revision 1.65
sys/ufs/lfs/lfs_vfsops.c: revision 1.371
sys/arch/i386/stand/efiboot/bootx64/Makefile: revision 1.3
sys/ufs/lfs/lfs_vfsops.c: revision 1.372
sys/ufs/lfs/lfs_vfsops.c: revision 1.373
sbin/fsck_lfs/pass1.c: revision 1.46
sys/ufs/lfs/lfs_vnops.c: revision 1.326
sys/ufs/lfs/lfs_vnops.c: revision 1.327
sys/ufs/lfs/lfs_vfsops.c: revision 1.375 (patch)
sys/ufs/lfs/lfs_vnops.c: revision 1.328
sys/ufs/lfs/lfs_subr.c: revision 1.98
sys/ufs/lfs/lfs_extern.h: revision 1.116
sys/ufs/lfs/lfs_vnops.c: revision 1.329
sys/ufs/lfs/lfs_subr.c: revision 1.99
sys/ufs/lfs/lfs_extern.h: revision 1.117
sys/ufs/lfs/lfs_accessors.h: revision 1.49
sys/ufs/lfs/lfs_extern.h: revision 1.118
sys/rump/fs/lib/liblfs/Makefile: revision 1.15
sys/ufs/lfs/lfs_bio.c: revision 1.146 (patch)
sys/ufs/lfs/lfs_bio.c: revision 1.147
sys/ufs/lfs/lfs_subr.c: revision 1.100
Fix kassert in lfs by initializing vp first.
Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
Teach lfs to transition ro<->rw.
Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
Assert lfs_writer where I think we can now prove it.
Serialize access to the splay tree with lfs_lock.
Change some cheap KDASSERT into KASSERT.
Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
Call lfs_orphan in lfs_rename while we're still in the dirop.
lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
Teach LFS_ORPHAN_NEXTFREE about lfs64.
Dust off the orphan detection code and try to make it work.
Fix !DIAGNOSTIC compile
Fix userland references to LFS_ORPHAN_NEXTFREE.
Forgot to grep for these or do a full distribution build, oops!
Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
OR into bp->b_cflags; don't overwrite.
CTASSERT lfs on-disk structure sizes.
Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
Fix clang build after packed lfs64 accessor change.
Suppress spurious address-of-packed error in rump lfs too.
|
| 1.160 | 23-Apr-2020 |
ad | PR kern/54759 (vm.ubc_direct deadlock when read()/write() into mapping of itself)
- Add new flag UBC_ISMAPPED which tells ubc_uiomove() the object is mmap()ed
somewhere. Use it to decide whether to do direct-mapped copy, rather than
poking around directly in the vnode in ubc_uiomove(), which is ugly and
doesn't work for tmpfs. It would be nicer to contain all this in UVM but
the filesystem provides the needed locking here (VV_MAPPED) and to
reinvent that would suck more.
- Rename UBC_UNMAP_FLAG() to UBC_VNODE_FLAGS(). Pass in UBC_ISMAPPED where
appropriate.
|
| 1.159 | 23-Feb-2020 |
ad | branches: 1.159.4;
UVM locking changes, proposed on tech-kern:
- Change the lock on uvm_object, vm_amap and vm_anon to be a RW lock.
- Break v_interlock and vmobjlock apart. v_interlock remains a mutex.
- Do partial PV list locking in the x86 pmap. Others to follow later.
|
| 1.158 | 23-Feb-2020 |
riastradh | In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
|
| 1.157 | 10-Jun-2017 |
maya | branches: 1.157.6; 1.157.10; 1.157.12;
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.156 | 05-Jun-2017 |
maya | Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
|
| 1.155 | 01-Apr-2017 |
maya | branches: 1.155.6;
Simplify locking
|
| 1.154 | 31-Mar-2017 |
maya | stopgap fix- move lfs_lock to include calls to lfs_dino_{set,get}block
blocks new users that need seglock (need to take lfs_lock) that
setblock before the assert (truncate to 0 but 31 blks/31 effblks)
not proper, but lets me run firefox on lfs
|
| 1.153 | 21-Mar-2017 |
maya | Update mtime even if oip->i_size == length
PR kern/51762, LFS version.
|
| 1.152 | 19-Mar-2017 |
riastradh | Fix inadvertently reversed sense of comparisons.
|
| 1.151 | 18-Mar-2017 |
riastradh | #if DIAGNOSTIC panic ---> KASSERT
|
| 1.150 | 16-Mar-2017 |
maya | actually cast to unsigned long long and use %llu. certainly not use hex (oops)
suggested by dh
|
| 1.149 | 15-Mar-2017 |
maya | print inode number in an assert I keep hitting and the adjacent one.
use PRIx64 for printing inode number elsewhere.
|
| 1.148 | 13-Mar-2017 |
riastradh | #if DIAGNOSTIC panic ---> KASSERTMSG
|
| 1.147 | 01-Sep-2015 |
dholland | branches: 1.147.2; 1.147.4;
Fix up indirect block handling in truncate to be 32/64 clean.
|
| 1.146 | 01-Sep-2015 |
dholland | Use the lfs dinode accessors in place of the ufs-derived ones.
(Mostly.)
The ufs-derived ones are fake structure member macros, which are gross
and not very safe. Also, it seems that a lot of places in the lfs code
were using the ffsv1 branch of them unconditionally, and this way it's
guaranteed all those places have been updated.
Found while doing this: for non-devices, have getattr produce NODEV
in the rdev field instead of leaking the address of the first direct
block.
|
| 1.145 | 19-Aug-2015 |
dholland | Part two of dinodes; use the same union everywhere.
(previously the ufs-derived code had things set up slightly different)
Remove a bunch of associated mess.
|
| 1.144 | 12-Aug-2015 |
dholland | Hack up dinode usage to be 64 vs. 32 as needed. Part 1.
(This part changes the native lfs code; the ufs-derived code already
has 64 vs. 32 logic, but as aspects of it are unsafe, and don't
entirely interoperate cleanly with the lfs 64/32 stuff, pass 2 will be
rehashing that.)
|
| 1.143 | 02-Aug-2015 |
dholland | whoops, fix 32-bit build
|
| 1.142 | 02-Aug-2015 |
dholland | Make i_eff_nblks in the in-memory inode 64 bits wide.
|
| 1.141 | 02-Aug-2015 |
dholland | Fix assorted 64 -> 32 truncations in lfs. Also, some minor tidyups and
corrections in passing.
|
| 1.140 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.139 | 24-Jul-2015 |
dholland | More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)
The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.
|
| 1.138 | 24-Jul-2015 |
dholland | Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)
It also gets rid of cpp abuse in the form of fake structure member
macros.
Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)
XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.
|
| 1.137 | 16-Jul-2015 |
dholland | Don't cast the return value of malloc.
|
| 1.136 | 17-Oct-2013 |
christos | branches: 1.136.6;
- remove unused variables
- add debug ifdefs for debugging variables
- __USE() where appropriate.
|
| 1.135 | 28-Jul-2013 |
dholland | Add more of the bits for supporting quotas.
|
| 1.134 | 28-Jul-2013 |
dholland | Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.133 | 28-Jul-2013 |
dholland | Add lfs_kernel.h for declarations that don't need to be exposed to userland.
lfs currently has the following headers:
lfs.h - on-disk structures and stuff needed for userlevel tools
lfs_inode.h - additional restricted materials for userlevel tools
that operate the fs (newfs_lfs, fsck_lfs, lfs_cleanerd)
lfs_kernel.h - stuff needed only in the kernel
and the following legacy headers that are expected to be mopped up and
folded into one of the above:
lfs_extern.h - function prototypes
ulfs_bswap.h - endian-independent support
ulfs_dinode.h - now contains very little
ulfs_dirhash.h - dirhash support
ulfs_extattr.h - extattr support
ulfs_extern.h - more function prototypes
ulfs_inode.h - assorted kernel-only declarations
ulfs_quota.h - quota support
ulfs_quota1.h - more quota support
ulfs_quota2.h - more quota support
ulfs_quotacommon.h - more quota support
ulfsmount.h - legacy copy of ufsmount material
|
| 1.132 | 18-Jun-2013 |
christos | branches: 1.132.2;
Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.
|
| 1.131 | 06-Jun-2013 |
dholland | Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.130 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.129 | 06-Jun-2013 |
dholland | Split lfs from ufs step 3: rearrange config stuff.
Add new options:
LFS_EI
LFS_DIRHASH
LFS_EXTATTR
LFS_EXTATTR_AUTOSTART
LFS_QUOTA
LFS_QUOTA2
and update code referring to the corresponding FFS and UFS config
symbols to use the LFS versions. Disable the one extant reference
to APPLE_UFS in the ulfs files. Use opt_lfs.h only, not opt_ffs.h.
|
| 1.128 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.127 | 22-Jan-2013 |
dholland | Stuff UFS_ in front of a few of ufs's symbols to reduce namespace
pollution. Specifically:
ROOTINO -> UFS_ROOTINO
WINO -> UFS_WINO
NXADDR -> UFS_NXADDR
NDADDR -> UFS_NDADDR
NIADDR -> UFS_NIADDR
MAXSYMLINKLEN -> UFS_MAXSYMLINKLEN
MAXSYMLINKLEN_UFS[12] -> UFS[12]_MAXSYMLINKLEN (for consistency)
Sort out ext2fs's misuse of NDADDR and NIADDR; fortunately, these have
the same values in ext2fs and ffs.
No functional change intended.
|
| 1.126 | 23-Nov-2011 |
bouyer | branches: 1.126.8;
If ufs_balloc_range() fails, make sure to call ?fs_truncate() to
reset v_writesize to the right value.
If v_writesize is left larger than the allocated blocks, we may have
the same issue as the one described in
http://mail-index.netbsd.org/tech-kern/2010/02/02/msg007156.html
|
| 1.125 | 11-Jul-2011 |
hannken | branches: 1.125.2;
Change VOP_BWRITE() to take a vnode as its first argument like all other
VOPs do. Layered file systems no longer have to modify bp->b_vp and run
into trouble when an async VOP_BWRITE() uses the wrong vnode.
- change all occurences of VOP_BWRITE(bp) to VOP_BWRITE(bp->b_vp, bp).
- remove layer_bwrite().
- welcome to 5.99.55
Adresses PR kern/38762 panic: vwakeup: neg numoutput
No objections from tech-kern@.
|
| 1.124 | 16-Jun-2011 |
hannken | Rename uvm_vnp_zerorange(struct vnode *, off_t, size_t) to
ubc_zerorange(struct uvm_object *, off_t, size_t, int) changing
the first argument to an uvm_object and adding a flags argument.
Modify tmpfs_reg_resize() to zero the backing store (aobj) instead
of the vnode. Ubc_purge() no longer panics when unmounting tmpfs.
Keep uvm_vnp_zerorange() until the next kernel version bump.
|
| 1.123 | 12-Jun-2011 |
rmind | Welcome to 5.99.53! Merge rmind-uvmplock branch:
- Reorganize locking in UVM and provide extra serialisation for pmap(9).
New lock order: [vmpage-owner-lock] -> pmap-lock.
- Simplify locking in some pmap(9) modules by removing P->V locking.
- Use lock object on vmobjlock (and thus vnode_t::v_interlock) to share
the locks amongst UVM objects where necessary (tmpfs, layerfs, unionfs).
- Rewrite and optimise x86 TLB shootdown code, make it simpler and cleaner.
Add TLBSTATS option for x86 to collect statistics about TLB shootdowns.
- Unify /dev/mem et al in MI code and provide required locking (removes
kernel-lock on some ports). Also, avoid cache-aliasing issues.
Thanks to Andrew Doran and Joerg Sonnenberger, as their initial patches
formed the core changes of this branch.
|
| 1.122 | 16-Feb-2010 |
mlelstv | branches: 1.122.2; 1.122.8;
Three changes in a single commit.
- drop the notion of frags (LFS fragments) vs fsb (FFS fragments)
The code uses a complicated unity function that just makes the
code difficult to understand.
- support larger sector sizes. Fix disk address computations
to use DEV_BSIZE in the kernel as required by device drivers
and to use sector sizes in userland.
- Fix several locking bugs in lfs_bio.c and lfs_subr.c.
|
| 1.121 | 07-Feb-2010 |
bouyer | branches: 1.121.2;
- ufs_balloc_range(): on error, only PG_RELEASED the pages that were
allocated to extend the file to the new size. Releasing all pages
may release pages that contains previously-written data not yet flushed
to disk. Should fix PR kern/35704
- {ffs,lfs,ext2fs}_truncate(): Even if the inode's size is the same as
the new length, call uvm_vnp_setsize(). *_truncate() may have been
called by *_write() in the error path (e.g. block allocation failure
because of quota of file system full), and at this point v_writesize
has been set to the desired size of the file and not reverted to the
old size. Not adjusting v_writesize to the real size cause
genfs_do_io() to write to disk past the real end of the file.
|
| 1.120 | 28-Apr-2008 |
martin | branches: 1.120.10; 1.120.18;
Remove clause 3 and 4 from TNF licenses
|
| 1.119 | 27-Mar-2008 |
ad | branches: 1.119.2; 1.119.4;
Make rusage collection per-LWP and collate in the appropriate places.
cloned threads need a little bit more work but the locking needs to
be fixed first.
|
| 1.118 | 15-Feb-2008 |
ad | branches: 1.118.6;
Give bbusy() an interlock argument. If the we need to wait for the buffer,
the interlock is dropped and reacquired when awoken. This allows for
busying buffers attached to a list that is not locked by bufcache_lock.
|
| 1.117 | 15-Feb-2008 |
ad | The buffer LOCKED flag need not be under the protection of bufcache_lock,
BUSY is enough.
|
| 1.116 | 02-Jan-2008 |
ad | Merge vmlocking2 to head.
|
| 1.115 | 08-Dec-2007 |
pooka | branches: 1.115.4;
Remove cn_lwp from struct componentname. curlwp should be used
from on. The NDINIT() macro no longer takes the lwp parameter and
associates the credentials of the calling thread with the namei
structure.
|
| 1.114 | 10-Oct-2007 |
ad | branches: 1.114.4; 1.114.6;
Fix DEBUG builds.
|
| 1.113 | 10-Oct-2007 |
ad | Merge from vmlocking:
- Split vnode::v_flag into three fields, depending on field locking.
- simple_lock -> kmutex in a few places.
- Fix some simple locking problems.
|
| 1.112 | 08-Oct-2007 |
ad | Merge ffs locking & brelse changes from the vmlocking branch.
|
| 1.111 | 10-Jul-2007 |
hannken | branches: 1.111.6; 1.111.8; 1.111.10;
Move `struct dquot' and its supporting functions from quota.h to ufs_quota.c.
- Make quota-internal functions static.
- Clean up declarations in quota.h and ufs_extern.h. quota.h now has the
description of quota criterions, on-disk structure, user-kernel interface and
declaration of init/done functions. All ufs quota related function
prototypes go to ufs_extern.h.
- New functions ufsquota_init() and ufsquota_free() create or destroy the
quota fields of `struct inode'.
- chkdq() and chkiq() always update the quota fields of `struct inode' first.
- Only ufs_access() explicitely calls getinoquota().
No objections on tech-kern@
|
| 1.110 | 05-Jun-2007 |
yamt | improve post-ubc file overwrite performance in common cases.
ie. when it's safe, actually overwrite blocks rather than doing
read-modify-write.
also fixes PR/33152 and PR/36303.
|
| 1.109 | 16-May-2007 |
perseant | Change references to SEGM_W_DIROPS to SEGM_CKP, and replace the logic that
formerly used SEGM_W_DIROPS in lfs_segwrite() appropriately. This prevents
a problem in which processes could get stuck in "buffers" sleep forever.
|
| 1.108 | 18-Apr-2007 |
perseant | Remember to write dirops when the vnode we are trying to flush is a dirop.
|
| 1.107 | 04-Mar-2007 |
christos | branches: 1.107.2; 1.107.4;
Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.
|
| 1.106 | 14-Oct-2006 |
yamt | branches: 1.106.4;
don't use g_glock directly.
|
| 1.105 | 14-May-2006 |
elad | branches: 1.105.8; 1.105.10;
integrate kauth.
|
| 1.104 | 14-May-2006 |
christos | Correct a bogus expression gcc4 found.
|
| 1.103 | 30-Apr-2006 |
perseant | Postpone the segment accounting changes coming from truncation until the
inode that makes those changes valid is either written to disk by
lfs_writeinode() or discarded by lfs_vfree().
A couple of locking fixes are also included as well.
|
| 1.102 | 19-Apr-2006 |
perseant | Avoid a possible sign overflow condition in lfs_truncate, which would result
in a buffer overflow (underflow). Coverity CID 1521.
|
| 1.101 | 08-Apr-2006 |
perseant | Implement a somewhat finer-grained mechanism for paging LFS-backed pages.
The writer daemon, if it does not need to flush the whole filesystem,
now only writes the vnodes for which the pagedaemon has requested pageouts
(although it does not pay attention to the page ranges the pagedaemon
supplies).
|
| 1.100 | 11-Dec-2005 |
christos | branches: 1.100.4; 1.100.6; 1.100.8; 1.100.10; 1.100.12;
merge ktrace-lwp.
|
| 1.99 | 11-Nov-2005 |
yamt | - ignore truncation for VCHR/VBLK/VFIFO as it used to be
before yamt-vop merge. PR/32049 from Atsushi Onoe.
- reject setattr which attempts to change size of VLNK/VSOCK.
|
| 1.98 | 02-Nov-2005 |
yamt | merge yamt-vop branch. remove following VOPs.
VOP_BLKATOFF
VOP_VALLOC
VOP_BALLOC
VOP_REALLOCBLKS
VOP_VFREE
VOP_TRUNCATE
VOP_UPDATE
|
| 1.97 | 12-Sep-2005 |
christos | branches: 1.97.2;
Use nanotime() to update the time fields in filesystems. Convert the code
from macros to real functions. Original patch and review from chuq.
Note: ext2fs only keeps seconds in the on-disk inode, and msdosfs does not
have enough precision for all fields, so this is not very useful for those
two.
|
| 1.96 | 29-May-2005 |
christos | branches: 1.96.2;
- sprinkle const
- avoid shadow variables.
|
| 1.95 | 23-Apr-2005 |
perseant | Provide a resize_lfs(8), including kernel and cleaner support. The current
implementation requires the fs to be mounted while resizing. Tested in both
directions, and everything appears to work happily, but ymmv.
|
| 1.94 | 16-Apr-2005 |
perseant | Use splay trees, rather than a hash table, to manage the accounting of
blocks allocated through VOP_BALLOC() for pages to be written to disk.
This accounting no longer takes a noticeable fraction of the system CPU.
|
| 1.93 | 16-Apr-2005 |
perseant | Use lfs_malloc() to manage the blkiov arrays that the cleaner functions use,
since the cleaner is likely to operate in a low-memory condition.
|
| 1.92 | 14-Apr-2005 |
perseant | Keep track of the highest block held by an LFS inode, so that we can
be assured that the last byte of a file is always allocated. Previously
a file extension could cause the filesystem to be flushed, writing an
inconsistent inode to disk. Although this condition would be corrected
the next time blocks were written to disk, an intervening crash would leave
the filesystem in an inconsistent state, leaving fsck_lfs to complain
of an inode "partially truncated".
|
| 1.91 | 01-Apr-2005 |
perseant | Protect various per-fs structures with fs->lfs_interlock simple_lock, to
improve behavior in the multiprocessor case. Add debugging segment-lock
assertion statements.
|
| 1.90 | 08-Mar-2005 |
perseant | branches: 1.90.2;
Straighten out the maze of ifdefs. Instead, consolidate all the debugging
stuff under '#ifdef DEBUG', and use sysctl knobs to turn on/off particular
parts of the debugging reporting (if DEBUG is enabled). Re-enable the LFS
statistics in sysctl, while I'm there. A bit of a rototill.
|
| 1.89 | 26-Feb-2005 |
perry | nuke trailing whitespace
|
| 1.88 | 15-Aug-2004 |
mycroft | branches: 1.88.4; 1.88.6;
Don't write out the extra zero pages with PGO_SYNCIO. We start an asynchronous
write anyway, and they will not be freed until that write is finished.
|
| 1.87 | 15-Aug-2004 |
mycroft | Copy the current partial-truncate logic from FFS. In the process, fix a
potential overrun when truncating a fragment.
|
| 1.86 | 15-Aug-2004 |
mycroft | Minor simplification to some arithmetic.
|
| 1.85 | 15-Aug-2004 |
mycroft | Fixing age old cruft:
* Rather than using mnt_maxsymlinklen to indicate that a file systems returns
d_type fields(!), add a new internal flag, IMNT_DTYPE.
Add 3 new elements to ufsmount:
* um_maxsymlinklen, replaces mnt_maxsymlinklen (which never should have existed
in the first place).
* um_dirblksiz, which tracks the current directory block size, eliminating the
FS-specific checks littered throughout the code. This may be used later to
make the block size variable.
* um_maxfilesize, which is the maximum file size, possibly adjusted lower due
to implementation issues.
Sync some bug fixes from FFS into ext2fs, particularly:
* ffs_lookup.c 1.21, 1.28, 1.33, 1.48
* ffs_inode.c 1.43, 1.44, 1.45, 1.66, 1.67
* ffs_vnops.c 1.84, 1.85, 1.86
Clean up some crappy pointer frobnication.
|
| 1.84 | 14-Aug-2004 |
mycroft | Add a new flag, IN_MODIFY. This is like IN_UPDATE|IN_CHANGE, but unlike
setting those flags, it does not cause the inode to be written in the periodic
sync. This is used for writes to special files (devices and named pipes) and
FIFOs.
Do not preemptively sync updates to access times and modification times. They
are now updated in the inode only opportunistically, or when the file or device
is closed. (Really, it should be delayed beyond close, but this is enough to
help substantially with device nodes.)
And the most amusing part:
Trickle sync was broken on both FFS and ext2fs, in different ways. In FFS, the
periodic call to VFS_SYNC(MNT_LAZY) was still causing all file data to be
synced. In ext2fs, it was causing the metadata to *not* be synced. We now
only call VOP_UPDATE() on the node if we're doing MNT_LAZY. I've confirmed
that we do in fact trickle correctly now.
|
| 1.83 | 30-Mar-2004 |
oster | If we bail out due to an error, we need 'unreserve' the space that
we'd reserved earlier.
Approved by: yamt
|
| 1.82 | 25-Jan-2004 |
hannken | branches: 1.82.2;
Make VOP_STRATEGY(bp) a real VOP as discussed on tech-kern.
VOP_STRATEGY(bp) is replaced by one of two new functions:
- VOP_STRATEGY(vp, bp) Call the strategy routine of vp for bp.
- DEV_STRATEGY(bp) Call the d_strategy routine of bp->b_dev for bp.
DEV_STRATEGY(bp) is used only for block-to-block device situations.
|
| 1.81 | 30-Dec-2003 |
pk | Replace the traditional buffer memory management -- based on fixed per buffer
virtual memory reservation and a private pool of memory pages -- by a scheme
based on memory pools.
This allows better utilization of memory because buffers can now be allocated
with a granularity finer than the system's native page size (useful for
filesystems with e.g. 1k or 2k fragment sizes). It also avoids fragmentation
of virtual to physical memory mappings (due to the former fixed virtual
address reservation) resulting in better utilization of MMU resources on some
platforms. Finally, the scheme is more flexible by allowing run-time decisions
on the amount of memory to be used for buffers.
On the other hand, the effectiveness of the LRU queue for buffer recycling
may be somewhat reduced compared to the traditional method since, due to the
nature of the pool based memory allocation, the actual least recently used
buffer may release its memory to a pool different from the one needed by a
newly allocated buffer. However, this effect will kick in only if the
system is under memory pressure.
|
| 1.80 | 07-Nov-2003 |
yamt | more assertion about file truncation to zero.
|
| 1.79 | 07-Aug-2003 |
agc | Move UCB-licensed code from 4-clause to 3-clause licence.
Patches provided by Joel Baker in PR 22364, verified by myself.
|
| 1.78 | 12-Jul-2003 |
yamt | - protect global resource counts with lfs_subsys_lock.
- clean up scattered externs a little.
|
| 1.77 | 29-Jun-2003 |
fvdl | branches: 1.77.2;
Back out the lwp/ktrace changes. They contained a lot of colateral damage,
and need to be examined and discussed more.
|
| 1.76 | 28-Jun-2003 |
darrenr | Pass lwp pointers throughtout the kernel, as required, so that the lwpid can
be inserted into ktrace records. The general change has been to replace
"struct proc *" with "struct lwp *" in various function prototypes, pass
the lwp through and use l_proc to get the process pointer when needed.
Bump the kernel rev up to 1.6V
|
| 1.75 | 27-Apr-2003 |
yamt | fix b_interlock lock/unlock mismatches.
|
| 1.74 | 23-Apr-2003 |
perseant | Make LFS work better (though still not "well") as an NFS-exported
filesystem (and other things that needed to be fixed before the tests
would complete), to wit:
* Include the fs ident in the filehandle; improve stale filehandle checks.
* Change definition of blksize() to use the on-dinode size instead of
the inode's i_size, so that fsck_lfs will work properly again.
* Use b_interlock in lfs_vtruncbuf.
* Postpone dirop reclamation until after the seglock has been released,
so that lfs_truncate is not called with the segment lock held.
* Don't loop in lfs_fsync(), just write everything and wait.
* Be more careful about the interlock/uobjlock in lfs_putpages: when we
lose this lock, we have to resynchronize dirtiness of pages in each
block.
* Be sure to always write indirect blocks and update metadata in
lfs_putpages; fixes a bug that caused blocks to be accounted to the
wrong segment.
|
| 1.73 | 10-Apr-2003 |
simonb | '#if 0' out a variable that is currently only used in other '#if 0'd out
code.
|
| 1.72 | 02-Apr-2003 |
fvdl | Add support for UFS2. UFS2 is an enhanced FFS, adding support for
64 bit block pointers, extended attribute storage, and a few
other things.
This commit does not yet include the code to manipulate the extended
storage (for e.g. ACLs), this will be done later.
Originally written by Kirk McKusick and Network Associates Laboratories for
FreeBSD.
|
| 1.71 | 20-Mar-2003 |
perseant | Hold the segment lock during truncation to prevent indirect blocks from
being written by lfs_updatemeta while lfs_truncate is also writing them,
a bug pointed out by YAMAMOTO Takashi <yamt@netbsd.org>.
|
| 1.70 | 08-Mar-2003 |
perseant | Take away "#ifdef LFS_UBC".
|
| 1.69 | 04-Mar-2003 |
perseant | Don't force all truncations to be synchronous
|
| 1.68 | 01-Mar-2003 |
perseant | Be careful to always zero pages on truncation/fragment extension,
in the case where the filesystem block size is larger than PAGE_SIZE.
|
| 1.67 | 28-Feb-2003 |
perseant | Make lfs_truncate handle file extension correctly, in the LFS_UBC case.
|
| 1.66 | 28-Feb-2003 |
perseant | Quell a hasty panic in lfs_truncate: on-inode disk addresses can be
different between the beginning and end of the call.
|
| 1.65 | 20-Feb-2003 |
perseant | Tabify, and fix some comment alignment problems.
|
| 1.64 | 17-Feb-2003 |
perseant | Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now
(there are still some details to work out) but expect that to go
away soon. To support these basic changes (creation of lfs_putpages,
lfs_gop_write, mods to lfs_balloc) several other changes were made, to
wit:
* Create a writer daemon kernel thread whose purpose is to handle page
writes for the pagedaemon, but which also takes over some of the
functions of lfs_check(). This thread is started the first time an
LFS is mounted.
* Add a "flags" parameter to GOP_SIZE. Current values are
GOP_SIZE_READ, meaning that the call should return the size of the
in-core version of the file, and GOP_SIZE_WRITE, meaning that it
should return the on-disk size. One of GOP_SIZE_READ or
GOP_SIZE_WRITE must be specified.
* Instead of using malloc(...M_WAITOK) for everything, reserve enough
resources to get by and use malloc(...M_NOWAIT), using the reserves if
necessary. Use the pool subsystem for structures small enough that
this is feasible. This also obsoletes LFS_THROTTLE.
And a few that are not strictly necessary:
* Moves the LFS inode extensions off onto a separately allocated
structure; getting closer to LFS as an LKM. "Welcome to 1.6O."
* Unified GOP_ALLOC between FFS and LFS.
* Update LFS copyright headers to correct values.
* Actually cast to unsigned in lfs_shellsort, like the comment says.
* Keep track of which segments were empty before the previous
checkpoint; any segments that pass two checkpoints both dirty and
empty can be summarily cleaned. Do this. Right now lfs_segclean
still works, but this should be turned into an effectless
compatibility syscall.
|
| 1.63 | 25-Jan-2003 |
fvdl | The oldblks and newblks arrays are used to store direct copies of
on-disk block pointers, so they should be int32_t. Error found
by Izumi Tsutsui.
|
| 1.62 | 24-Jan-2003 |
fvdl | Bump daddr_t to 64 bits. Replace it with int32_t in all places where
it was used on-disk, so that on-disk formats remain the same.
Remove ufs_daddr_t and ufs_lbn_t for the time being.
|
| 1.61 | 28-Dec-2002 |
yamt | - in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't
try to reclaim them.
(workaround for deadlock noted in the comment in lfs_reserveavail)
- in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
|
| 1.60 | 27-Sep-2002 |
provos | remove trailing \n in panic(). approved perry.
|
| 1.59 | 06-Jul-2002 |
perseant | Deal with fragment size changes better. For each fragment that can
exist on an on-disk inode, we keep a record of its size in struct inode,
which is updated when we write the block to disk. The cleaner routines
thus have ready access to what size is the correct size for this block,
on disk.
Fixed a related bug: if a file with fragments is being cleaned
(fragments being cleaned) at the same time it is being extended beyond
NDADDR blocks, we could write a bogus FINFO record that has a frag in the
middle; when it was cleaned this would give back bogus file data. Don't
write the indirect blocks in this case, since there is no need.
lfs_fragextend and lfs_truncate no longer require the seglock, but instead
take a shared lock, which the seglock locks exclusively.
|
| 1.58 | 02-Jul-2002 |
yamt | fix printf format for DEBUG_LFS.
|
| 1.57 | 14-May-2002 |
perseant | branches: 1.57.2;
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes
I found while making sure there weren't any new ones.
* Make the write clusters keep track of the buffers whose blocks they contain.
This should make it possible to (1) write clusters using a page mapping
instead of malloc, if desired, and (2) schedule blocks for rewriting
(somewhere else) if a write error occurs. Code is present to use
pagemove() to construct the clusters but that is untested and will go away
anyway in favor of page mapping.
* DEBUG now keeps a log of Ifile writes, so that any lingering instances of
the "dirty bufs" problem can be properly debugged.
* Keep track of whether the Ifile has been dirtied by various routines that
can be called by lfs_segwrite, and loop on that until it is clean, for
a checkpoint. Checkpoints need to be squeaky clean.
* Warn the user (once) if the Ifile grows larger than is reasonable for their
buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can
grow.
* If an inode is not found in a disk block, try rereading the block, under
the assumption that the block was copied to a cluster and then freed.
* Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
|
| 1.56 | 23-Nov-2001 |
chs | add spaces for KNF. confirmed to produce identical objects.
|
| 1.55 | 08-Nov-2001 |
lukem | add RCSID
|
| 1.54 | 06-Nov-2001 |
simonb | Remove some variables that are set but never used.
|
| 1.53 | 15-Sep-2001 |
chs | branches: 1.53.2;
a whole bunch of changes to improve performance and robustness under load:
- remove special treatment of pager_map mappings in pmaps. this is
required now, since I've removed the globals that expose the address range.
pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's
no longer any need to special-case it.
- eliminate struct uvm_vnode by moving its fields into struct vnode.
- rewrite the pageout path. the pager is now responsible for handling the
high-level requests instead of only getting control after a bunch of work
has already been done on its behalf. this will allow us to UBCify LFS,
which needs tighter control over its pages than other filesystems do.
writing a page to disk no longer requires making it read-only, which
allows us to write wired pages without causing all kinds of havoc.
- use a new PG_PAGEOUT flag to indicate that a page should be freed
on behalf of the pagedaemon when it's unlocked. this flag is very similar
to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the
pageout fails due to eg. an indirect-block buffer being locked.
this allows us to remove the "version" field from struct vm_page,
and together with shrinking "loan_count" from 32 bits to 16,
struct vm_page is now 4 bytes smaller.
- no longer use PG_RELEASED for swap-backed pages. if the page is busy
because it's being paged out, we can't release the swap slot to be
reallocated until that write is complete, but unlike with vnodes we
don't keep a count of in-progress writes so there's no good way to
know when the write is done. instead, when we need to free a busy
swap-backed page, just sleep until we can get it busy ourselves.
- implement a fast-path for extending writes which allows us to avoid
zeroing new pages. this substantially reduces cpu usage.
- encapsulate the data used by the genfs code in a struct genfs_node,
which must be the first element of the filesystem-specific vnode data
for filesystems which use genfs_{get,put}pages().
- eliminate many of the UVM pagerops, since they aren't needed anymore
now that the pager "put" operation is a higher-level operation.
- enhance the genfs code to allow NFS to use the genfs_{get,put}pages
instead of a modified copy.
- clean up struct vnode by removing all the fields that used to be used by
the vfs_cluster.c code (which we don't use anymore with UBC).
- remove kmem_object and mb_object since they were useless.
instead of allocating pages to these objects, we now just allocate
pages with no object. such pages are mapped in the kernel until they
are freed, so we can use the mapping to find the page to free it.
this allows us to remove splvm() protection in several places.
The sum of all these changes improves write throughput on my
decstation 5000/200 to within 1% of the rate of NetBSD 1.5
and reduces the elapsed time for "make release" of a NetBSD 1.5
source tree on my 128MB pc to 10% less than a 1.5 kernel took.
|
| 1.52 | 13-Jul-2001 |
perseant | branches: 1.52.2;
Merge the short-lived perseant-lfsv2 branch into the trunk.
Kernels and tools understand both v1 and v2 filesystems; newfs_lfs
generates v2 by default. Changes for the v2 layout include:
- Segments of non-PO2 size and arbitrary block offset, so these can be
matched to convenient physical characteristics of the partition (e.g.,
stripe or track size and offset).
- Address by fragment instead of by disk sector, paving the way for
non-512-byte-sector devices. In theory fragments can be as large
as you like, though in reality they must be smaller than MAXBSIZE in size.
- Use serial number and filesystem identifier to ensure that roll-forward
doesn't get old data and think it's new. Roll-forward is enabled for
v2 filesystems, though not for v1 filesystems by default.
- The inode free list is now a tailq, paving the way for undelete (undelete
is not yet implemented, but can be without further non-backwards-compatible
changes to disk structures).
- Inode atime information is kept in the Ifile, instead of on the inode;
that is, the inode is never written *just* because atime was changed.
Because of this the inodes remain near the file data on the disk, rather
than wandering all over as the disk is read repeatedly. This speeds up
repeated reads by a small but noticeable amount.
Other changes of note include:
- The ifile written by newfs_lfs can now be of arbitrary length, it is no
longer restricted to a single indirect block.
- Fixed an old bug where ctime was changed every time a vnode was created.
I need to look more closely to make sure that the times are only updated
during write(2) and friends, not after-the-fact during a segment write,
and certainly not by the cleaner.
|
| 1.51 | 30-May-2001 |
mrg | branches: 1.51.2; 1.51.4;
use _KERNEL_OPT
|
| 1.50 | 03-Dec-2000 |
perseant | branches: 1.50.2;
Get rid of some old unnecessary code that cleared B_NEEDCOMMIT from buffers in
lfs_writeseg (possibly after they had been freed).
If MALLOCLOG is defined, make lfs_newbuf and lfs_freebuf pass along the
caller's file and line to _malloc and _free.
|
| 1.49 | 27-Nov-2000 |
chs | Initial integration of the Unified Buffer Cache project.
|
| 1.48 | 27-Nov-2000 |
perseant | If LFS_DO_ROLLFORWARD is defined, roll forward from the older checkpoint
on mount, through the newer checkpoint and on through any newer
partial-segments that may have been written but not checkpointed because
of an intervening crash.
LFS_DO_ROLLFORWARD is not defined by default.
|
| 1.47 | 21-Nov-2000 |
perseant | More locked_queue_* and lfs_avail accounting fixes from Jesse Off
<joff@gci-net.com>. Remove a specious btodb() in lfs_fragextend, and
count blocks shrunk or removed by VOP_TRUNCATE in lfs_avail.
|
| 1.46 | 17-Nov-2000 |
perseant | Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes.
(PR #11468). In the case of fragment allocation, check to see if enough
space is available before extending a fragment already scheduled for writing.
The locked_queue_* variables indicate the number of buffer headers and bytes,
respectively, that are unavailable to getnewbuf() because they are locked up
waiting for LFS to flush them; make sure that that is actually what we're
counting, i.e., never count malloced buffers, and always use b_bufsize instead
of b_bcount.
If DEBUG is defined, the periodic calls to lfs_countlocked will now complain
if either counter is incorrect. (In the future lfs_countlocked will not need
to be called at all if DEBUG is not defined.)
|
| 1.45 | 14-Oct-2000 |
perseant | In lfs_truncate, don't overcount the real blocks removed from the inode,
when deallocating a fragment that has not made it to disk yet.
Also, during dirops, give the directory vnode an extra reference in
SET_DIROP, to ensure its continued existence during SET_ENDOP, preventing
a possible NULL-dereference there.
These two changes should close PR #11064.
|
| 1.44 | 09-Sep-2000 |
perseant | Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
|
| 1.43 | 09-Sep-2000 |
perseant | Make sure to unmark B_DELWRI on blocks freed due to truncation to a non-zero
file length. Should fix PR #s 10551 and 10831.
|
| 1.42 | 05-Jul-2000 |
perseant | Clean up accounting of lfs_uinodes (dirty but unwritten inodes).
Make lfs_uinodes a signed quantity for debugging purposes, and set it to
zero as fs mount time.
Enclose setting/clearing of the dirty flags (IN_MODIFIED, IN_ACCESSED,
IN_CLEANING) in macros, and use those macros everywhere. Make
LFS_ITIMES use these macros; updated the ITIMES macro in inode.h to know
about this. Make ufs_getattr use ITIMES instead of FFS_ITIMES.
|
| 1.41 | 04-Jul-2000 |
perseant | Fix errors observed while trying to fill the filesystem with yesterday's
fixes:
- Write copies of bfree and avail in the CLEANERINFO block, so the
cleaner doesn't have to guess which superblock has the current
information (if indeed any do).
- Tighten up accounting of lfs_avail (more needs to be done).
- When cleansing indirect blocks of UNWRITTEN, make sure not to mark
them clean, since they'll need to be rewritten later.
|
| 1.40 | 03-Jul-2000 |
perseant | Allow the number of free segments reserved for the cleaner to be
parametrized in the filesystem, defaulting to MIN_FREE_SEGS = 2 but set
to something more reasonable at newfs_lfs time.
Note the number of blocks that have been scheduled for writing but which
are not yet on disk in an inode extension, i_lfs_effnblks. Move
i_ffs_effnlink out of the ffs extension and onto the main inode, since
it's used all over the shared code and the lfs extension would clobber
it.
At inode write time, indirect blocks and inode-held blocks of inodes
that have i_lfs_effnblks != i_ffs_blocks are cleansed of UNWRITTEN disk
addresses, so that these never make it to disk.
|
| 1.39 | 28-Jun-2000 |
mrg | remove include of <vm/vm.h> and <uvm/uvm_extern.h>
|
| 1.38 | 27-Jun-2000 |
perseant | Fixes associated with filling an LFS:
Change the space computation to appear to change the size of the *disk*
rather than the *bytes used* when more segment summaries and inode
blocks are written. Try to estimate the amount of space that these will
take up when more files are written, so the disk size doesn't change too
much.
Regularize error returns from lfs_valloc, lfs_balloc, lfs_truncate: they
now fail entirely, rather than succeeding half-way and leaving the fs in
an inconsistent state.
Rewrite lfs_truncate, mostly stealing from ffs_truncate. The old
lfs_truncate had difficulty truncating a large file to a non-zero size
(indirect blocks were not handled appropriately).
Unmark VDIROP on fvp after ufs_remove, ufs_rmdir, so these can be
reclaimed immediately: this vnode would not be written to disk again
anyway if the removal succeeded, and if it failed, no directory
operation occurred.
ufs_makeinode and ufs_mkdir now remove IN_ADIROP on error.
|
| 1.37 | 31-May-2000 |
perseant | branches: 1.37.2;
update for IN_ACCESSED changes
|
| 1.36 | 13-May-2000 |
perseant | branches: 1.36.2;
Change the sementics of the last parameter from a boolean ("waitfor") to
a set of flags ("flags"). Two flags are defined, UPDATE_WAIT and
UPDATE_DIROP.
Under the old semantics, VOP_UPDATE would block if waitfor were set,
under the assumption that directory operations should be done
synchronously. At least LFS and FFS+softdep do not make this
assumption; FFS+softdep got around the problem by enclosing all relevant
calls to VOP_UPDATE in a "if(!DOINGSOFTDEP(vp))", while LFS simply
ignored waitfor, one of the reasons why NFS-serving an LFS filesystem
did not work properly.
Under the new semantics, the UPDATE_DIROP flag is a hint to the
fs-specific update routine that the call comes from a dirop routine, and
should be wait for, or not, accordingly.
Closes PR#8996.
|
| 1.35 | 05-May-2000 |
perseant | Change the way LFS does block accounting, from trying to infer from the
buffer cache flags, to marking the inode and/or indirect blocks with a
special disk address UNWRITTEN==-2 when a block is accounted for. (This
address is never written to disk, but only used in-core. This is essentially
the same method of block accounting as on the UBC branch, where the buffer
headers don't exist.) Make sure that truncation is handled properly,
especially in the case of holey files.
Fixes PR#9994.
|
| 1.34 | 24-Apr-2000 |
perseant | get rid of unused variable aflags
|
| 1.33 | 23-Apr-2000 |
perseant | Fix problems outlined in PR#9926:
- lfs_truncate extends the file if called with length > i_ffs_size;
- lfs_truncate errors out if called with length < 0;
- lfs_balloc block accounting corrected for the case of blocks read
into the cache before they exist on disk;
- mp->mnt_stat.f_iosize is initialized in lfs_mountfs.
|
| 1.32 | 30-Mar-2000 |
augustss | Remove register declarations.
|
| 1.31 | 12-Mar-2000 |
bouyer | lfs_truncate: handle synlinks with length > maxsymlink_len as regular files.
For symlinks > 60 chars we were bzero'ing part of (struct inode) past the
actual inode struct, corrupting memory following the current (struct inode)
resuling in a 'panic: pool_get(lfsinopl): free list modified' later.
This could also be the cause of random panics. With this fix LFS seems to be
useable for me now.
|
| 1.30 | 19-Jan-2000 |
perseant | Changes to stabilize LFS. The first two of these should also apply to the
1.4 branch.
* Use a separate per-fs lock, instead of ufs_hashlock, to protect the Inode
free list. This seems to prevent the "lockmgr: %d, not exclusive lock holder
%d, unlocking" message I was mis-attributing last night to an unlocked vnode
being passed to vrele.
* Change calling semantics of lfs_ifind, to give better error reporting:
If fed a struct buf, it can report the block number of the offending inode
block as well as the inode number.
* Back out rev 1.10 of lfs_subr.c, since the replacement code was slightly
uglier while being functionally identical.
* Make lfs_vunref use the same free list convention as vrele/vput, so that
vget does not remove vnodes from a hash list they are not on.
|
| 1.29 | 16-Jan-2000 |
perseant | Fix a problem in my changes of Dec 14th, that prevents removed vnodes
from being inactivated under some conditions. Removed vnodes are now
inactivated when the VDIROP flag is cleared, and to prevent block
accounting problems this clearing has been postponed until
lfs_segunlock.
|
| 1.28 | 23-Nov-1999 |
fvdl | Be more careful to block bio interrupts for some data structures. There
were at least a few missed cases where vp->v_{clean,dirty}blkhd were
unprotected since the softdep/trickle sync merge.
|
| 1.27 | 03-Sep-1999 |
perseant | branches: 1.27.2; 1.27.8;
Make changes that will allow an LFS filesystem to be used as the root
filesystem. In particular,
- Fix mknod deadlock, described in PR 8172.
- Enable lfs_mountroot.
- Make lfs_writevnodes treat filesystems mounted on lfs device nodes properly,
by flushing that device rather than trying to add blocks to the device inode.
This, in combination with lfs boot blocks, will allow operation of an all-lfs
system.
|
| 1.26 | 15-Jun-1999 |
perseant | Minor changes to the segment live bytes calculation. In particular, fixed
a bug in fragment extension that could run the count negative. Also, don't
overcount for inodes, and don't count segment summaries. Thus, for empty
segments the live bytes count should now be exactly zero.
|
| 1.25 | 01-Jun-1999 |
perseant | Fixed lfs_update (and related functions) so that calls from lfs_fsync
will DTRT with vnodes marked VDIROP. In particular, the message
"flushing VDIROP" will no longer appear, and the filesystem will remain
stable in the event of a crash.
This was particularly a problem with NFS-exported LFSes, since fsync
was called on every file close.
|
| 1.24 | 12-Apr-1999 |
perseant | Fix block counting during file truncation, if not truncating to zero.
|
| 1.23 | 12-Apr-1999 |
perseant | Make sure that the wakeup occurs for vnodes that lfs_update might be sleeping
on (nodes which are not marked IN_MODIFIED/IN_CLEANING, but which have dirty
buffers), by marking them with the appropriate flag if dirtybuffers were added
while the write was in progress.
|
| 1.22 | 01-Apr-1999 |
perseant | branches: 1.22.2;
Fix buffer handling problems in lfs_vinvalbuf
|
| 1.21 | 29-Mar-1999 |
perseant | lfs_truncate calls vinvalbuf to invalidate all currently-hald buffers, which
in turn forces a flush of the vnode, whether or not it is involved in a dirop.
(This can happen during a remove or rmdir, when the directory is shrunk.)
Because of the nature of dirops, however, flushing a vnode involved in a dirop
is disallowed (and was marked with a panic). This patch has lfs_truncate
call a specialized vinvalbuf that only invalidates buffers following the new
end-of-file, and thus does not require a flush. Also the panic is demoted,
in case I missed any other path to lfs_vflush.
|
| 1.20 | 25-Mar-1999 |
perseant | clean up unused/required #ifdefs
|
| 1.19 | 24-Mar-1999 |
mrg | completely remove Mach VM support. all that is left is the all the
header files as UVM still uses (most of) these.
|
| 1.18 | 10-Mar-1999 |
perseant | New sources should leave the LFS in a more-or-less working state. Changes
include:
- DIROP segregation is enabled, and greater care is taken
to make sure that a checkpoint completes. Fsck is not
needed to remount the filesystem.
- Several checks to make sure that the LFS subsystem does not
overuse various resources (memory, in particular).
- The cleaner routines, lfs_markv in particular, are completely
rewritten. A buffer overflow is removed. Greater care is taken
to ensure that inodes come from where lfs_cleanerd say they come
from (so we know nothing has changed since lfs_bmapv was called).
- Fragment allocation is fixed, so that writes beyond end-of-file
do the right thing.
|
| 1.17 | 05-Mar-1999 |
mycroft | Pass null pointers to VOP_UPDATE rather than having all the callers fetch the
current time themselves.
|
| 1.16 | 05-Mar-1999 |
mycroft | Permit the access and modify time pointers passed to VOP_UPDATE to be null,
meaning the current time.
|
| 1.15 | 10-Feb-1999 |
bouyer | Make sure a buffer optained from bread() is always bresle()'d in case of
error. Closes PR kern/1448 from Wolfgang Solfrank.
|
| 1.14 | 09-Jun-1998 |
scottr | Protect various config(8)-generated files from inclusion while
building LKMs. Fixes PR 5557.
|
| 1.13 | 08-Jun-1998 |
scottr | Use the newly-defined opt_quota.h.
|
| 1.12 | 01-Mar-1998 |
fvdl | Merge with Lite2 + local changes
|
| 1.11 | 07-Feb-1998 |
chs | add UVM stuff.
|
| 1.10 | 04-Jul-1997 |
drochner | Don't cast 64bit (off_t) file sizes to vm_offset_t (32bit on many
architectures), truncate them intelligently instead.
The truncation is done centralized in vnode_pager.c.
This prevents from wrap-over effects when parts of large (>2^32 byte) files
are mmapped.
Don't allow to mmap above the numerical range of vm_offset_t.
This is considered a temporary solution until the vm system handles the
object sizes/offsets more cleanly.
|
| 1.9 | 11-Jun-1997 |
bouyer | Add support for ext2fs, this needed a few modifications to ufs/ufs/inode.h:
- added an "union inode_ext" to struct inode, for the per-fs extentions.
For now only ext2fs uses it.
- i_din is now an union:
union {
struct dinode ffs_din; /* 128 bytes of the on-disk dinode. */
struct ext2fs_dinode e2fs_din; /* 128 bytes of the on-disk dinode. */
} i_din
Added a lot of #define i_ffs_* and i_e2fs_* to access the fields.
- Added two macros: FFS_ITIMES and EXT2FS_ITIMES. ITIMES calls the rigth
macro, depending on the time of the inode. ITIMES is used where necessary,
FFS_ITIMES and EXT2FS_ITIMES in other places.
|
| 1.8 | 12-Oct-1996 |
christos | revert previous kprintf changes
|
| 1.7 | 10-Oct-1996 |
christos | printf -> kprintf, sprintf -> ksprintf
|
| 1.6 | 01-Sep-1996 |
mycroft | Add a set of generic file system operations that most file systems use.
Also, fix some time stamp bogosities.
|
| 1.5 | 11-May-1996 |
mycroft | Change VOP_UPDATE() semantics:
* Make 2nd and 3rd args timespecs, not timevals.
* Consistently pass a Boolean as the 4th arg (except in LFS).
Also, fix ffs_update() and lfs_update() to actually change the nsec fields.
|
| 1.4 | 09-Feb-1996 |
christos | lfs prototypes
|
| 1.3 | 15-Jun-1995 |
cgd | compensate for timeval/timespec/stat structure changes.
|
| 1.2 | 29-Jun-1994 |
cgd | New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
|
| 1.1 | 08-Jun-1994 |
mycroft | branches: 1.1.1;
Update to 4.4-Lite fs code, with local changes.
|
| 1.1.1.2 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite2
|
| 1.1.1.1 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite for reference
|
| 1.22.2.5 | 20-Jan-2000 |
he | Pull up revisions 1.29-1.30 (requested by perseant):
Files removed (through unlink, rmdir) are now really removed, though the
removal is postponed until the dirop is complete to ensure validity of
the filesystem through a crash. Use a separate per-fs lock, instead of
ufs_hashlock, to protect the inode free list. Change calling semantics
of lfs_ifind, to give better error reporting: If fed a struct buf, it
can report the block number of the offending inode block as well as the
inode number.
|
| 1.22.2.4 | 15-Jan-2000 |
he | Pull up revision 1.27 (requested by perseant):
Address problems related to using an LFS filesystem as the root
filesystem, including mknod hangs. Fixes PR#8172 and PR#9072.
|
| 1.22.2.3 | 17-Dec-1999 |
he | Pull up revision 1.25 (requested by perseant):
Avoid flushing vnodes involved in a dirop, making lfs' promise
of "no fsck needed, even in the event of a crash" closer to
reality.
|
| 1.22.2.2 | 25-Jun-1999 |
perry | pullup 1.25->1.26 (perseant)
|
| 1.22.2.1 | 13-Apr-1999 |
perseant | branches: 1.22.2.1.2; 1.22.2.1.4;
Pull-up of changes made to the trunk on Sunday [1.22->1.24], to wit:
Take out the `#ifdef USE_UFSHASH'; use ufs_hashlock to lock the inode free
list instead of free_lock.
Fix inode reporting in lfs_statfs (the meaning of f_files and f_ffree was
reversed).
Fix "lfs_ifind: dinode xxx not found" panic. When inodes were freed, then
immediately reloaded, their dinodes were located in an inode block which
was not on disk at the advertized location, nor in the cache (although it
would be flushed to disk next segment write). Fix this by using getblk()
instead of lfs_newbuf() for inode blocks.
Better checking for held inode locks in lfs_fastvget, for a number of
error conditions. Also change the default setting of lfs_clean_vnhead to
0, which seems to make the locking problems go away (although this is
difficult to test as I can't reliably reproduce them).
Make sure that the wakeup occurs for vnodes that lfs_update might be
sleeping on (nodes which are not marked IN_MODIFIED/IN_CLEANING, but which
have dirty buffers), by marking them with the appropriate flag if
dirtybuffers were added while the write was in progress.
Fix block counting during file truncation, if not truncating to zero.
Disallow threshold-initiated cache flush when dirops are active. Also,
make SET_ENDOP use lfs_check instead of inlining most of it.
Improve the debugging printfs in the cleaner syscalls (in particular, make
it obvious that they're coming from lfs).
Check the superblock version field, and refuse to mount the filesystem if
the version number is higher than we know about. This allows, e.g.,
changes in the format of the ifile, segment size restrictions and
boundaries, etc., which would not affect existing fields in the
superblock, but which would drastically affect the filesystem, to be
smoothly integrated at a later date.
|
| 1.22.2.1.4.1 | 30-Nov-1999 |
itojun | bring in latest KAME (as of 19991130, KAME/NetBSD141) into kame branch
just for reference purposes.
This commit includes 1.4 -> 1.4.1 sync for kame branch.
The branch does not compile at all (due to the lack of ALTQ and some other
source code). Please do not try to modify the branch, this is just for
referenre purposes.
synchronization to latest KAME will take place on HEAD branch soon.
|
| 1.22.2.1.2.3 | 31-Aug-1999 |
perseant | Rudimentary support for LFS under UBC:
- LFS-specific VOP_BALLOC and VOP_PUTPAGES vnode ops.
- getblk VREG panic #ifdef'd out (can be reinstated when Ifile is
internalized and Ifile can be made another type from VREG)
- interface to VOP_PUTPAGES changed to pass all pager flags, not
just sync. FS putpages routines must know about the pager flags.
- new LFS magic disk address, -2 ("unwritten"), meaning accounted for
but not assigned to a fixed disk location (since LFS does these two
things separately, and the previous accounting method using buffer
headers no longer will work). Changed references to (foo == (daddr_t)-1)
to (foo < 0). Since disk drivers reject all addresses < 0, this should
not present a problem for other FSs.
|
| 1.22.2.1.2.2 | 11-Jul-1999 |
chs | remove uvm_vnp_uncache(), it's no longer needed.
|
| 1.22.2.1.2.1 | 21-Jun-1999 |
thorpej | Sync w/ -current.
|
| 1.27.8.2 | 27-Dec-1999 |
wrstuden | Pull up to last week's -current.
|
| 1.27.8.1 | 21-Dec-1999 |
wrstuden | Initial commit of recent changes to make DEV_BSIZE go away.
Runs on i386, needs work on other arch's. Main kernel routines should be
fine, but a number of the stand programs need help.
cd, fd, ccd, wd, and sd have been updated. sd has been tested with non-512
byte block devices. vnd, raidframe, and lfs need work.
Non 2**n block support is automatic for LKM's and conditional for kernels
on "options NON_PO2_BLOCKS".
|
| 1.27.2.3 | 08-Dec-2000 |
bouyer | Sync with HEAD.
|
| 1.27.2.2 | 22-Nov-2000 |
bouyer | Sync with HEAD.
|
| 1.27.2.1 | 20-Nov-2000 |
bouyer | Update thorpej_scsipi to -current as of a month ago
|
| 1.36.2.1 | 22-Jun-2000 |
minoura | Sync w/ netbsd-1-5-base.
|
| 1.37.2.3 | 23-Mar-2001 |
he | Pull up revisions 1.46-1.47 (via patch, requested by perseant):
o Close up accounting holes in LFS' accounting of immediately-
available-space, number of clean segments, and amount of dirty
space taken up by metadata (PR#11468, PR#11470, PR#11534).
This one got left out when the rest was pulled up. Sorry.
|
| 1.37.2.2 | 01-Nov-2000 |
tv | Pullup 1.45 [perseant, toshii]:
In lfs_truncate, don't overcount the real blocks removed from the inode,
when deallocating a fragment that has not made it to disk yet.
Also, during dirops, give the directory vnode an extra reference in
SET_DIROP, to ensure its continued existence during SET_ENDOP, preventing
a possible NULL-dereference there.
These two changes should close PR #11064.
|
| 1.37.2.1 | 14-Sep-2000 |
perseant | Pull up recent LFS kernel changes (approved by thorpej):
ufs/ufs/inode.h, 1.20--1.22 (add i_lfs_effnblks extension ;
make ITIMES aware of LFS_ITIMES;
_LKM protection so userland progs
compile)
ufs/ufs/ufs_vnops.c, 1.69, 1.71 (remove IN_ADIROP;
use ITIMES instead of FFS_ITIMES)
ufs/ufs/ufs_readwrite.c, 1.27 (use lfs_reserve in lfs_write)
ufs/lfs/lfs.h, 1.26--1.32 (define LFS_EST_* macros ;
change MIN_FREE_SEGS to lfs_minfreesegs ;
add avail and bfree to CLEANERINFO ;
change lfs_uinodes to signed ;
change lfs_dmeta to signed ;
add whitespace to line up structure
members ;
explicit cast to int32_t in LFS_EST_*
macros)
ufs/lfs/lfs_alloc.c, back out 1.34.2.3 (pullups of 1.39, 1.40);
then pull up 1.38 (clean up on error)
1.39--1.43 (restore fvdl's ufs_hashlock fix ;
restore fvdl's ufs_hashlock fix ;
set i_lfs_effnblks ;
use UINO macros ;
add comments and fix long lines)
ufs/lfs/lfs_balloc.c, 1.19 (don't succeed halfway)
1.21--1.25 (use i_lfs_effnblks ;
fix i_lfs_effnblks computation and
quieten ;
fix i_ffs_blocks in unwritten fragment ;
remove useless debugging check ;
add comments and (c) 2000)
ufs/lfs/lfs_bio.c, 1.24--1.30 (cleanup and make lfs_flush_fs take
"struct lfs *" instead of "struct
mount *" ;
use lfs_minfreeseg instead of
MIN_FREE_SEGS ;
use UINO macros, and copy bfree/avail
to CLEANERINFO ;
add lfs_reserve function ;
1.28--1.30 fix printf formatting)
ufs/lfs/lfs_cksum.c, 1.13 (add (c) 2000)
ufs/lfs/lfs_debug.c, 1.11 (use btodb instead of DEV_BSIZE)
ufs/lfs/lfs_extern.h, 1.18, 1.20--1.21 (function prototype changes)
ufs/lfs/lfs_inode.c, 1.38 (rewrite lfs_truncate from
ffs_truncate)
1.40--1.44 (count written and unwritten blocks
seperately ;
use disk block units instead of bytes ;
remove unnecessary "mod" variable ;
correct B_DELWRI to avoid bawrite panic ;
use lfs_reserve)
ufs/lfs/lfs_segment.c, 1.52-1.59 (use lfs_dmeta to note used summaries ;
check for UNWRITTEN in indirect blocks ;
more debugging stuff inside #ifdef
DEBUG_LFS ;
use LK_CANRECURSE ;
don't drop dirty indirect blocks ;
use UINO macros ;
don't hose the free list ;
use btodb() instead of DEV_BSIZE ;
make it compile again (oops))
ufs/lfs/lfs_subr.c, 1.16--1.17 (check for locked inodes before
changing ;
use btodb() instead of DEV_BSIZE, (c)
2000)
ufs/lfs/lfs_syscalls.c, back out 1.41.4.2 (fvdl's ufs_hashlock fix);
then pull up 1.43 (use lfs_dmeta)
1.44--1.45 (restore fvdl's ufs_hashlock fix)
1.46--1.47 (fix lfs_avail leakage from sblock
segments ;
use UINO macros)
1.49 (bounds-check inode numbers in
lfs_markv)
ufs/lfs/lfs_vfsops.c, 1.53 (use LFS_EST_* macros in lfs_statfs)
1.56--1.58 (initialize lfs_minfreeseg, lfs_effnblk ;
initialize lfs_uinodes ;
initialize lfs_ravail)
ufs/lfs/lfs_vnops.c, 1.40 (remove VDIROP from removed files)
1.42--1.44 (move SET_ENDOP below the removal of
VDIROP ;
use UINO macros and add lfs_itimes
function ;
use lfs_reserve in dirops)
|
| 1.50.2.9 | 29-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.50.2.8 | 18-Oct-2002 |
nathanw | Catch up to -current.
|
| 1.50.2.7 | 01-Aug-2002 |
nathanw | Catch up to -current.
|
| 1.50.2.6 | 20-Jun-2002 |
nathanw | Catch up to -current.
|
| 1.50.2.5 | 08-Jan-2002 |
nathanw | Catch up to -current.
|
| 1.50.2.4 | 14-Nov-2001 |
nathanw | Catch up to -current.
|
| 1.50.2.3 | 21-Sep-2001 |
nathanw | Catch up to -current.
|
| 1.50.2.2 | 24-Aug-2001 |
nathanw | Catch up with -current.
|
| 1.50.2.1 | 21-Jun-2001 |
nathanw | Catch up to -current.
|
| 1.51.4.5 | 10-Oct-2002 |
jdolecek | sync kqueue with -current; this includes merge of gehenna-devsw branch,
merge of i386 MP branch, and part of autoconf rototil work
|
| 1.51.4.4 | 06-Sep-2002 |
jdolecek | sync kqueue branch with HEAD
|
| 1.51.4.3 | 23-Jun-2002 |
jdolecek | catch up with -current on kqueue branch
|
| 1.51.4.2 | 10-Jan-2002 |
thorpej | Sync kqueue branch with -current.
|
| 1.51.4.1 | 03-Aug-2001 |
lukem | update to -current
|
| 1.51.2.3 | 02-Jul-2001 |
perseant | Change disk addressing unit to be the fragment, instead of the disk sector.
All quantities in the superblock, inodes, indirect blocks, etc. refer now
to this abstract unit (called "fsb" as it is in FFS) instead of disk sectors;
as a consequence segment summary blocks have to be multiples of a fragment in
size. In v1 filesystems, compatibility code ensures that 1 fsb == 1 sector,
regardless of fragment size.
Fragments can now range in size between 512 and 32k; in the event that
LFS_LABELPAD (8k) is smaller than the disk address unit size, an extra
proto-superblock is kept at 8k from the beginning of the disk, to be used
*only* to locate the real superblocks. (Not all of the userland knows about
this yet.)
Almost all of this was done not by me, but by joff.
|
| 1.51.2.2 | 29-Jun-2001 |
perseant | Get rid of __P(), protoizing where it had not already been done
|
| 1.51.2.1 | 27-Jun-2001 |
perseant | Import of what I've been calling "LFSv2", that is, LFS with some features
added that require changes to the on-disk data structures. These include:
- 64-bit time in everything but inodes
- User-specified segment offset, and segment size no longer
restricted to PO2.
- Serial number on segment summaries in addition to timestamp, and
a new volume identifier, to make roll-forward feasible without
fear of finding old data and thinking it was new.
Although I think this version works at least as well as what's on the trunk,
we're not done yet; hence this commit is going in on a branch and not on
the trunk. Enhancements that are not here yet include fragment addressing,
like FFS does, instead of block addressing.
|
| 1.52.2.1 | 01-Oct-2001 |
fvdl | Catch up with -current.
|
| 1.53.2.1 | 12-Nov-2001 |
thorpej | Sync the thorpej-mips-cache branch with -current.
|
| 1.57.2.1 | 15-Jul-2002 |
gehenna | catch up with -current.
|
| 1.77.2.9 | 11-Dec-2005 |
christos | Sync with head.
|
| 1.77.2.8 | 10-Nov-2005 |
skrll | Sync with HEAD. Here we go again...
|
| 1.77.2.7 | 08-Mar-2005 |
skrll | Sync with HEAD.
|
| 1.77.2.6 | 04-Mar-2005 |
skrll | Sync with HEAD.
Hi Perry!
|
| 1.77.2.5 | 21-Sep-2004 |
skrll | Fix the sync with head I botched.
|
| 1.77.2.4 | 18-Sep-2004 |
skrll | Sync with HEAD.
|
| 1.77.2.3 | 25-Aug-2004 |
skrll | Sync with HEAD.
|
| 1.77.2.2 | 03-Aug-2004 |
skrll | Sync with HEAD
|
| 1.77.2.1 | 02-Jul-2003 |
darrenr | Apply the aborted ktrace-lwp changes to a specific branch. This is just for
others to review, I'm concerned that patch fuziness may have resulted in some
errant code being generated but I'll look at that later by comparing the diff
from the base to the branch with the file I attempt to apply to it. This will,
at the very least, put the changes in a better context for others to review
them and attempt to tinker with removing passing of 'struct lwp' through
the kernel.
|
| 1.82.2.1 | 09-Apr-2004 |
jmc | Pullup rev 1.83 (requested by oster in ticket #112)
If we bail out due to an error, we need 'unreserve' the space that
we'd reserved earlier.
|
| 1.88.6.1 | 19-Mar-2005 |
yamt | sync with head. xen and whitespace. xen part is not finished.
|
| 1.88.4.1 | 29-Apr-2005 |
kent | sync with -current
|
| 1.90.2.5 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.104
sys/ufs/lfs/lfs_vfsops.c: revision 1.206
sys/ufs/lfs/lfs_vnops.c: revision 1.170
sys/ufs/lfs/lfs_extern.h: revision 1.80
sys/ufs/lfs/lfs_segment.c: revision 1.176
sys/ufs/lfs/lfs_inode.c: revision 1.103 via patch
sys/ufs/lfs/lfs_alloc.c: revision 1.90
Postpone the segment accounting changes coming from truncation until the
inode that makes those changes valid is either written to disk by
lfs_writeinode() or discarded by lfs_vfree().
A couple of locking fixes are also included as well.
|
| 1.90.2.4 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_inode.c: revision 1.102
Avoid a possible sign overflow condition in lfs_truncate, which would result
in a buffer overflow (underflow). Coverity CID 1521.
|
| 1.90.2.3 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.200
sys/ufs/lfs/lfs_vnops.c: revision 1.164
sys/ufs/lfs/lfs_inode.c: revision 1.101
sys/ufs/lfs/lfs_extern.h: revision 1.78
sys/ufs/lfs/lfs.h: revision 1.100
Implement a somewhat finer-grained mechanism for paging LFS-backed pages.
The writer daemon, if it does not need to flush the whole filesystem,
now only writes the vnodes for which the pagedaemon has requested pageouts
(although it does not pay attention to the page ranges the pagedaemon
supplies).
|
| 1.90.2.2 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.152
sys/ufs/lfs/lfs_debug.c: revision 1.31
sys/ufs/lfs/lfs_subr.c: revision 1.53
sys/ufs/lfs/lfs_extern.h: revision 1.68
sys/ufs/lfs/lfs_inode.c: revision 1.96
sys/ufs/lfs/lfs_bio.c: revision 1.86
sys/ufs/lfs/lfs_alloc.c: revision 1.83
sys/ufs/lfs/lfs_vfsops.c: revision 1.181
sys/ufs/lfs/lfs.h: revision 1.88
sys/ufs/lfs/lfs_segment.c: revision 1.164
- sprinkle const
- avoid shadow variables.
|
| 1.90.2.1 | 07-May-2005 |
tron | Apply patch (requested by perseant in ticket #242):
* fsck_lfs buffer cache fixes, including PR #29151
* Change fsck_lfs phase 0 message to reflect reality
* fsck_lfs: check phase 5 (cleanerinfo accounting) even on
roll-forward
* Keep better track of the free list during roll-forward, avoiding
a core dump
* Improve hash table use for fsck_lfs buffer and vnode cache
* Document fsck_lfs flag -f, and implement -q
* Add resize_lfs, including kernel support
* Add LFS to mountd's list of exportable filesystem types
* Make the LFS lkm work again [christos@]
* Add MP locking to the LFS kernel subsystem
* Fix pager_map deadlock in lfs_putpages()
* Avoid incomplete file extension that looks like "partial
truncation" to fsck
* Use lfs_malloc for cleaner malloc, since the cleaner often runs
in low-memory conditions.
* Use splay trees, not hash table, to track page allocation for
write.
* Fix mkdir panic on full fs
* Fix page accounting leak by counting differently.
* Use rightly named structure for lfs_getattr [skrll@]
* Cosmetic changes for readability.
|
| 1.96.2.6 | 27-Feb-2008 |
yamt | sync with head.
|
| 1.96.2.5 | 21-Jan-2008 |
yamt | sync with head
|
| 1.96.2.4 | 27-Oct-2007 |
yamt | sync with head.
|
| 1.96.2.3 | 03-Sep-2007 |
yamt | sync with head.
|
| 1.96.2.2 | 30-Dec-2006 |
yamt | sync with head.
|
| 1.96.2.1 | 21-Jun-2006 |
yamt | sync with head.
|
| 1.97.2.2 | 29-Oct-2005 |
yamt | use lfs_* directly rather than via ufs_ops.
suggested by Chuck Silvers.
|
| 1.97.2.1 | 20-Oct-2005 |
yamt | adapt ufs.
|
| 1.100.12.1 | 24-May-2006 |
tron | Merge 2006-05-24 NetBSD-current into the "peter-altq" branch.
|
| 1.100.10.4 | 11-May-2006 |
elad | sync with head
|
| 1.100.10.3 | 06-May-2006 |
christos | - Move kauth_cred_t declaration to <sys/types.h>
- Cleanup struct ucred; forward declarations that are unused.
- Don't include <sys/kauth.h> in any header, but include it in the c files
that need it.
Approved by core.
|
| 1.100.10.2 | 19-Apr-2006 |
elad | sync with head.
|
| 1.100.10.1 | 08-Mar-2006 |
elad | Adapt to kernel authorization KPI.
|
| 1.100.8.2 | 24-May-2006 |
yamt | sync with head.
|
| 1.100.8.1 | 11-Apr-2006 |
yamt | sync with head
|
| 1.100.6.2 | 01-Jun-2006 |
kardel | Sync with head.
|
| 1.100.6.1 | 22-Apr-2006 |
simonb | Sync with head.
|
| 1.100.4.1 | 09-Sep-2006 |
rpaulo | sync with head
|
| 1.105.10.1 | 22-Oct-2006 |
yamt | sync with head
|
| 1.105.8.1 | 18-Nov-2006 |
ad | Sync with head.
|
| 1.106.4.3 | 17-May-2007 |
yamt | sync with head.
|
| 1.106.4.2 | 07-May-2007 |
yamt | sync with head.
|
| 1.106.4.1 | 12-Mar-2007 |
rmind | Sync with HEAD.
|
| 1.107.4.1 | 11-Jul-2007 |
mjf | Sync with head.
|
| 1.107.2.7 | 28-Aug-2007 |
yamt | make this compilable with DEBUG.
|
| 1.107.2.6 | 24-Aug-2007 |
ad | Sync with buffer cache locking changes. See buf.h/vfs_bio.c for details.
Some minor portions are incomplete and needs to be verified as a whole.
|
| 1.107.2.5 | 15-Jul-2007 |
ad | Sync with head.
|
| 1.107.2.4 | 17-Jun-2007 |
ad | - Increase the number of thread priorities from 128 to 256. How the space
is set up is to be revisited.
- Implement soft interrupts as kernel threads. A generic implementation
is provided, with hooks for fast-path MD code that can run the interrupt
threads over the top of other threads executing in the kernel.
- Split vnode::v_flag into three fields, depending on how the flag is
locked (by the interlock, by the vnode lock, by the file system).
- Miscellaneous locking fixes and improvements.
|
| 1.107.2.3 | 09-Jun-2007 |
ad | Sync with head.
|
| 1.107.2.2 | 13-May-2007 |
ad | - Pass the error number and residual count to biodone(), and let it handle
setting error indicators. Prepare to eliminate B_ERROR.
- Add a flag argument to brelse() to be set into the buf's flags, instead
of doing it directly. Typically used to set B_INVAL.
- Add a "struct cpu_info *" argument to kthread_create(), to be used to
create bound threads. Change "bool mpsafe" to "int flags".
- Allow exit of LWPs in the IDL state when (l != curlwp).
- More locking fixes & conversion to the new API.
|
| 1.107.2.1 | 13-Mar-2007 |
ad | Pull in the initial set of changes for the vmlocking branch.
|
| 1.111.10.1 | 14-Oct-2007 |
yamt | sync with head.
|
| 1.111.8.3 | 23-Mar-2008 |
matt | sync with HEAD
|
| 1.111.8.2 | 09-Jan-2008 |
matt | sync with HEAD
|
| 1.111.8.1 | 06-Nov-2007 |
matt | sync with HEAD
|
| 1.111.6.2 | 09-Dec-2007 |
jmcneill | Sync with HEAD.
|
| 1.111.6.1 | 26-Oct-2007 |
joerg | Sync with HEAD.
Follow the merge of pmap.c on i386 and amd64 and move
pmap_init_tmp_pgtbl into arch/x86/x86/pmap.c. Modify the ACPI wakeup
code to restore CR4 before jumping back into kernel space as the large
page option might cover that.
|
| 1.114.6.5 | 26-Dec-2007 |
ad | Sync with head.
|
| 1.114.6.4 | 19-Dec-2007 |
ad | Use a global lfs_lock.
|
| 1.114.6.3 | 19-Dec-2007 |
ad | Fix some more problems w/lfs on this branch.
|
| 1.114.6.2 | 19-Dec-2007 |
ad | Get lfs mostly working.
|
| 1.114.6.1 | 04-Dec-2007 |
ad | Pull the vmlocking changes into a new branch.
|
| 1.114.4.1 | 18-Feb-2008 |
mjf | Sync with HEAD.
|
| 1.115.4.1 | 02-Jan-2008 |
bouyer | Sync with HEAD
|
| 1.118.6.2 | 02-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.118.6.1 | 03-Apr-2008 |
mjf | Sync with HEAD.
|
| 1.119.4.2 | 11-Mar-2010 |
yamt | sync with head
|
| 1.119.4.1 | 16-May-2008 |
yamt | sync with head.
|
| 1.119.2.1 | 18-May-2008 |
yamt | sync with head.
|
| 1.120.18.1 | 21-Apr-2010 |
matt | sync to netbsd-5
|
| 1.120.10.2 | 25-Jan-2012 |
riz | Pull up following revision(s) (requested by bouyer in ticket #1702):
sys/ufs/lfs/lfs_inode.c: revision 1.126
sys/ufs/ffs/ffs_inode.c: revision 1.108
If ufs_balloc_range() fails, make sure to call ?fs_truncate() to
reset v_writesize to the right value.
If v_writesize is left larger than the allocated blocks, we may have
the same issue as the one described in
http://mail-index.netbsd.org/tech-kern/2010/02/02/msg007156.html
|
| 1.120.10.1 | 22-Feb-2010 |
snj | Pull up following revision(s) (requested by bouyer in ticket #1302):
sys/ufs/ext2fs/ext2fs_inode.c: revision 1.71
sys/ufs/ffs/ffs_inode.c: revision 1.104
sys/ufs/lfs/lfs_inode.c: revision 1.121
sys/ufs/ufs/ufs_inode.c: revision 1.79
- ufs_balloc_range(): on error, only PG_RELEASED the pages that were
allocated to extend the file to the new size. Releasing all pages
may release pages that contains previously-written data not yet flushed
to disk. Should fix PR kern/35704
- {ffs,lfs,ext2fs}_truncate(): Even if the inode's size is the same as
the new length, call uvm_vnp_setsize(). *_truncate() may have been
called by *_write() in the error path (e.g. block allocation failure
because of quota of file system full), and at this point v_writesize
has been set to the desired size of the file and not reverted to the
old size. Not adjusting v_writesize to the real size cause
genfs_do_io() to write to disk past the real end of the file.
|
| 1.121.2.1 | 30-Apr-2010 |
uebayasi | Sync with HEAD.
|
| 1.122.8.1 | 23-Jun-2011 |
cherry | Catchup with rmind-uvmplock merge.
|
| 1.122.2.1 | 16-Mar-2010 |
rmind | Change struct uvm_object::vmobjlock to be dynamically allocated with
mutex_obj_alloc(). It allows us to share the locks among UVM objects.
|
| 1.125.2.3 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.125.2.2 | 23-Jan-2013 |
yamt | sync with head
|
| 1.125.2.1 | 17-Apr-2012 |
yamt | sync with head
|
| 1.126.8.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.126.8.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.126.8.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.126.8.1 | 25-Feb-2013 |
tls | resync with head
|
| 1.132.2.2 | 18-May-2014 |
rmind | sync with head
|
| 1.132.2.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.136.6.2 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.136.6.1 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.147.4.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.147.2.2 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.147.2.1 | 20-Mar-2017 |
pgoyette | Sync with HEAD
|
| 1.155.6.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.157.12.1 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.157.10.1 | 17-Aug-2020 |
martin | Pull up following revision(s) (requested by riastradh in ticket #1050):
sys/ufs/lfs/lfs_subr.c: revision 1.101
sys/ufs/lfs/lfs_subr.c: revision 1.102
sys/ufs/lfs/lfs_inode.c: revision 1.158
sys/ufs/lfs/lfs_inode.h: revision 1.25
sys/ufs/lfs/lfs_balloc.c: revision 1.95
sys/ufs/lfs/lfs_pages.c: revision 1.21
sys/ufs/lfs/lfs_vnops.c: revision 1.330
sys/ufs/lfs/lfs_alloc.c: revision 1.140 (patch)
sys/ufs/lfs/lfs_alloc.c: revision 1.141 (patch)
lib/libp2k/p2k.c: revision 1.72
sys/ufs/lfs/lfs.h: revision 1.205
sys/ufs/lfs/lfs.h: revision 1.206
sys/ufs/lfs/lfs_segment.c: revision 1.284
sys/ufs/lfs/lfs.h: revision 1.207
sys/ufs/lfs/lfs_segment.c: revision 1.285
sys/ufs/lfs/lfs_debug.c: revision 1.55
sys/ufs/lfs/lfs_rename.c: revision 1.23
usr.sbin/dumplfs/dumplfs.c: revision 1.65
sys/ufs/lfs/lfs_vfsops.c: revision 1.371
sys/arch/i386/stand/efiboot/bootx64/Makefile: revision 1.3
sys/ufs/lfs/lfs_vfsops.c: revision 1.372
sys/ufs/lfs/lfs_vfsops.c: revision 1.373
sbin/fsck_lfs/pass1.c: revision 1.46
sys/ufs/lfs/lfs_vnops.c: revision 1.326
sys/ufs/lfs/lfs_vnops.c: revision 1.327
sys/ufs/lfs/lfs_vfsops.c: revision 1.375 (patch)
sys/ufs/lfs/lfs_vnops.c: revision 1.328
sys/ufs/lfs/lfs_subr.c: revision 1.98
sys/ufs/lfs/lfs_extern.h: revision 1.116
sys/ufs/lfs/lfs_vnops.c: revision 1.329
sys/ufs/lfs/lfs_subr.c: revision 1.99
sys/ufs/lfs/lfs_extern.h: revision 1.117
sys/ufs/lfs/lfs_accessors.h: revision 1.49
sys/ufs/lfs/lfs_extern.h: revision 1.118
sys/rump/fs/lib/liblfs/Makefile: revision 1.15
sys/ufs/lfs/lfs_bio.c: revision 1.146 (patch)
sys/ufs/lfs/lfs_bio.c: revision 1.147
sys/ufs/lfs/lfs_subr.c: revision 1.100
Fix kassert in lfs by initializing vp first.
Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
Teach lfs to transition ro<->rw.
Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
Assert lfs_writer where I think we can now prove it.
Serialize access to the splay tree with lfs_lock.
Change some cheap KDASSERT into KASSERT.
Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
Call lfs_orphan in lfs_rename while we're still in the dirop.
lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
Teach LFS_ORPHAN_NEXTFREE about lfs64.
Dust off the orphan detection code and try to make it work.
Fix !DIAGNOSTIC compile
Fix userland references to LFS_ORPHAN_NEXTFREE.
Forgot to grep for these or do a full distribution build, oops!
Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
OR into bp->b_cflags; don't overwrite.
CTASSERT lfs on-disk structure sizes.
Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
Fix clang build after packed lfs64 accessor change.
Suppress spurious address-of-packed error in rump lfs too.
|
| 1.157.6.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.159.4.1 | 25-Apr-2020 |
bouyer | Sync with bouyer-xenpvh-base2 (HEAD)
|
| 1.27 | 20-Oct-2025 |
perseant | * Generalize the partial-segment parser introduced for roll-forward,
using it to facilitate an in-kernel segment rewriter (cleaner), and a
mechanism to check whether a segment is in fact empty (only used
with DEBUG).
* Add these new fcntl calls:
- LFCNFILESTATS: For each inode given, report its number of direct
blocks, how many gaps (discontinuities) there are between direct
blocks, and how large the total gap distance is. This will be
useful for a coalescing agent.
- LFCNREWRITEFILE: For each inode given, rewrite its direct blocks,
effectively coalescing it into as compact a form as possible.
- LFCNSCRAMBLE: As above, except that it only rewrites every other
block. This causes the file to have many gaps that can be
measured with LFCNFILESTATS and addressed with LFCNREWRITEFILE,
for testing purposes.
- LFCNREWRITESEGS: Rewrite any live data in the given segments.
This is intended to simplify the cleaner API and facilitate an
in-kernel cleaner.
- LFCNCLEANERINFO: Get the most current CLEANERINFO data from the
kernel.
- LFCNSEGUSE: Retrieve segment usage data from the kernel.
* Vnodes marked IN_CLEANING now take a reference. Add a new "cleaner
lock", which must be taken by the cleaner before the segment lock,
and before marking nodes IN_CLEANING. This allows us to flush
vnodes, if necessary, before the cleaning segment is written, and
never to flush vnodes being cleaned. When the cleaner lock is
released, the vnodes are cleared of IN_CLEANING and the reference
dropped.
* Track a potential infinite loop in lfs_gatherblock.
* Pull "needs to flush" and "needs to wait for flush" into functions
instead of inlining their definitions.
|
| 1.26 | 23-Mar-2022 |
andvar | fix few typos for word "previous(ly)" in comments.
|
| 1.25 | 23-Feb-2020 |
riastradh | Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
|
| 1.24 | 18-Feb-2020 |
chs | remove the aiodoned thread. I originally added this to provide a thread context
for doing page cache iodone work, but since then biodone() has changed to
hand off all iodone work to a softint thread, so we no longer need the
special-purpose aiodoned thread.
|
| 1.23 | 10-Jun-2017 |
maya | branches: 1.23.6; 1.23.10; 1.23.12;
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.22 | 08-Jun-2017 |
chs | move some buffer cache internals declarations from buf.h to vfs_bio.c.
this is needed to avoid name conflicts with ZFS and also
makes it clearer that other code shouldn't be messing with these.
remove the LFS debug code that poked around in bufqueues and
remove the BQ_EMPTY bufqueue since nothing uses it anymore.
provide a function to let LFS and wapbl read the value of nbuf for now.
|
| 1.21 | 05-Jun-2017 |
maya | Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
|
| 1.20 | 05-Jun-2017 |
maya | Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
|
| 1.19 | 06-Apr-2017 |
maya | branches: 1.19.6;
don't guard lfs_sbactive or lfs_log with splbio, lfs_lock is plenty.
|
| 1.18 | 06-Apr-2017 |
maya | Provide a LFS_ENTER_LOG (__nothing) in the !DEBUG case.
so I can drop lots of #ifdef DEBUG around this macro. NFCI
|
| 1.17 | 06-Apr-2017 |
maya | Drop single use macro LFS_BCLEAN_LOG with an inlined implementation.
LFS_ENTER_LOG currently macro grabs lfs_lock, so I'd like to have just one
name for it.
|
| 1.16 | 20-Jun-2016 |
dholland | branches: 1.16.2; 1.16.4;
u_int{8,16,32,64}_t -> uint{8,16,32,64}_t in remaining lfs headers.
|
| 1.15 | 20-Jun-2016 |
dholland | Note more already-merged versions:
inode.h 1.68 is subsumed by ulfs_inode.h 1.19
inode.h 1.69-1.72 do not apply to lfs
ufs_extern.h 1.74 was covered when lfs was moved to the new vnode cache
ufs_extern.h 1.75 is equivalent to ulfs_extern.h 1.13
ufs_extern.h 1.76-1.77 do not apply to lfs
ufsmount.h 1.42 does not apply to lfs
ufs_inode.c 1.90 is subsumed by ulfs_inode.c 1.10
ufs_inode.c 1.91-1.92 do not apply to lfs
ufs_lookup.c 1.130 is subsumed by ulfs_lookup.c 1.24
ufs_lookup.c 1.131 is equivalent to ulfs_lookup.c 1.20
ufs_lookup.c 1.132 is equivalent to ulfs_lookup.c 1.21
ufs_lookup.c 1.133 is equivalent to ulfs_lookup.c 1.22
ufs_lookup.c 1.134 is equivalent to ulfs_lookup.c 1.23
ufs_lookup.c 1.135 is equivalent to ulfs_lookup.c 1.25
ufs_quota2.c 1.38 is equivalent to ulfs_quota2.c 1.17
ufs_quota2.c 1.39 is equivalent to ulfs_quota2.c 1.16
ufs_quota2.c 1.40 is equivalent to ulfs_quota2.c 1.18
ufs_vfsops.c 1.53 is subsumed by lfs_vfsops.c 1.324
ufs_vfsops.c 1.54 is subsumed by lfs_vfsops.c 1.324
ufs_vnops.c 1.223-1.224 do not apply to lfs
|
| 1.14 | 20-Jun-2016 |
dholland | ufs/inode.h -r1.67 is effectively merged into here too.
|
| 1.13 | 20-Jun-2016 |
dholland | Merge ufs/inode.h 1.66: remove i_hash from struct inode. This is the
hash table entry link from the old per-fs vnode cache and we don't
need it any more.
|
| 1.12 | 19-Jun-2016 |
dholland | Mark ufs file versions we're already synced with.
|
| 1.11 | 01-Sep-2015 |
dholland | Use the lfs dinode accessors in place of the ufs-derived ones.
(Mostly.)
The ufs-derived ones are fake structure member macros, which are gross
and not very safe. Also, it seems that a lot of places in the lfs code
were using the ffsv1 branch of them unconditionally, and this way it's
guaranteed all those places have been updated.
Found while doing this: for non-devices, have getattr produce NODEV
in the rdev field instead of leaking the address of the first direct
block.
|
| 1.10 | 19-Aug-2015 |
dholland | Part two of dinodes; use the same union everywhere.
(previously the ufs-derived code had things set up slightly different)
Remove a bunch of associated mess.
|
| 1.9 | 12-Aug-2015 |
dholland | Hack up dinode usage to be 64 vs. 32 as needed. Part 1.
(This part changes the native lfs code; the ufs-derived code already
has 64 vs. 32 logic, but as aspects of it are unsafe, and don't
entirely interoperate cleanly with the lfs 64/32 stuff, pass 2 will be
rehashing that.)
|
| 1.8 | 02-Aug-2015 |
dholland | Make i_eff_nblks in the in-memory inode 64 bits wide.
|
| 1.7 | 26-May-2014 |
ryoon | branches: 1.7.4;
Close comments
|
| 1.6 | 26-May-2014 |
dholland | remove ffs-only IN_SPACECOUNTED
|
| 1.5 | 18-Jun-2013 |
dholland | branches: 1.5.2; 1.5.8; 1.5.10;
Tuck away a bunch of symbols that don't need to be public.
|
| 1.4 | 09-Jun-2013 |
dholland | Move struct lfs_inode_ext to lfs_inode.h; it doesn't need to be public.
|
| 1.3 | 08-Jun-2013 |
dholland | G/C another unneeded union
|
| 1.2 | 08-Jun-2013 |
dholland | Remove stale union and accessor macros.
|
| 1.1 | 08-Jun-2013 |
dholland | Split the definitions suitable for userland out of ulfs_inode.h into
lfs_inode.h. Since fsck_lfs, newfs_lfs, and lfs_cleanerd want to reuse
the inode structure for their own internal use, and some of them share
parts of the kernel code as well, the best way forward is to provide a
relatively sanitized header that doesn't bring in stray material.
Shuffle a few other definitions around so that lfs_inode.h depends
only on lfs.h.
Install lfs_inode.h into /usr/include.
|
| 1.5.10.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.5.10.1 | 18-Jun-2013 |
yamt | file lfs_inode.h was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.5.8.1 | 10-Aug-2014 |
tls | Rebase.
|
| 1.5.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.5.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.5.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.5.2.1 | 18-Jun-2013 |
tls | file lfs_inode.h was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.7.4.3 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.7.4.2 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.7.4.1 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.16.4.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.16.2.1 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.19.6.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.23.12.1 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.23.10.1 | 17-Aug-2020 |
martin | Pull up following revision(s) (requested by riastradh in ticket #1050):
sys/ufs/lfs/lfs_subr.c: revision 1.101
sys/ufs/lfs/lfs_subr.c: revision 1.102
sys/ufs/lfs/lfs_inode.c: revision 1.158
sys/ufs/lfs/lfs_inode.h: revision 1.25
sys/ufs/lfs/lfs_balloc.c: revision 1.95
sys/ufs/lfs/lfs_pages.c: revision 1.21
sys/ufs/lfs/lfs_vnops.c: revision 1.330
sys/ufs/lfs/lfs_alloc.c: revision 1.140 (patch)
sys/ufs/lfs/lfs_alloc.c: revision 1.141 (patch)
lib/libp2k/p2k.c: revision 1.72
sys/ufs/lfs/lfs.h: revision 1.205
sys/ufs/lfs/lfs.h: revision 1.206
sys/ufs/lfs/lfs_segment.c: revision 1.284
sys/ufs/lfs/lfs.h: revision 1.207
sys/ufs/lfs/lfs_segment.c: revision 1.285
sys/ufs/lfs/lfs_debug.c: revision 1.55
sys/ufs/lfs/lfs_rename.c: revision 1.23
usr.sbin/dumplfs/dumplfs.c: revision 1.65
sys/ufs/lfs/lfs_vfsops.c: revision 1.371
sys/arch/i386/stand/efiboot/bootx64/Makefile: revision 1.3
sys/ufs/lfs/lfs_vfsops.c: revision 1.372
sys/ufs/lfs/lfs_vfsops.c: revision 1.373
sbin/fsck_lfs/pass1.c: revision 1.46
sys/ufs/lfs/lfs_vnops.c: revision 1.326
sys/ufs/lfs/lfs_vnops.c: revision 1.327
sys/ufs/lfs/lfs_vfsops.c: revision 1.375 (patch)
sys/ufs/lfs/lfs_vnops.c: revision 1.328
sys/ufs/lfs/lfs_subr.c: revision 1.98
sys/ufs/lfs/lfs_extern.h: revision 1.116
sys/ufs/lfs/lfs_vnops.c: revision 1.329
sys/ufs/lfs/lfs_subr.c: revision 1.99
sys/ufs/lfs/lfs_extern.h: revision 1.117
sys/ufs/lfs/lfs_accessors.h: revision 1.49
sys/ufs/lfs/lfs_extern.h: revision 1.118
sys/rump/fs/lib/liblfs/Makefile: revision 1.15
sys/ufs/lfs/lfs_bio.c: revision 1.146 (patch)
sys/ufs/lfs/lfs_bio.c: revision 1.147
sys/ufs/lfs/lfs_subr.c: revision 1.100
Fix kassert in lfs by initializing vp first.
Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
Teach lfs to transition ro<->rw.
Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
Assert lfs_writer where I think we can now prove it.
Serialize access to the splay tree with lfs_lock.
Change some cheap KDASSERT into KASSERT.
Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
Call lfs_orphan in lfs_rename while we're still in the dirop.
lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
Teach LFS_ORPHAN_NEXTFREE about lfs64.
Dust off the orphan detection code and try to make it work.
Fix !DIAGNOSTIC compile
Fix userland references to LFS_ORPHAN_NEXTFREE.
Forgot to grep for these or do a full distribution build, oops!
Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
OR into bp->b_cflags; don't overwrite.
CTASSERT lfs on-disk structure sizes.
Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
Fix clang build after packed lfs64 accessor change.
Suppress spurious address-of-packed error in rump lfs too.
|
| 1.23.6.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.20 | 10-Jun-2017 |
maya | Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.19 | 01-Sep-2015 |
dholland | branches: 1.19.10;
Use the lfs dinode accessors in place of the ufs-derived ones.
(Mostly.)
The ufs-derived ones are fake structure member macros, which are gross
and not very safe. Also, it seems that a lot of places in the lfs code
were using the ffsv1 branch of them unconditionally, and this way it's
guaranteed all those places have been updated.
Found while doing this: for non-devices, have getattr produce NODEV
in the rdev field instead of leaking the address of the first direct
block.
|
| 1.18 | 12-Aug-2015 |
dholland | Add IFILE32 and IFILE64 structures for the on-disk ifile entries.
Add and use accessors. There are also a bunch of places that cast and
I hope I've found them all...
|
| 1.17 | 02-Aug-2015 |
dholland | Use accessor functions for the version field of the lfs superblock.
I thought at first maybe the cases that test the version should be
rolled into the accessors, but on the whole I think the conclusion on
that is no.
|
| 1.16 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.15 | 08-Jun-2013 |
dholland | branches: 1.15.10;
Tidy up the LFS userland build hacks.
Don't use -I${NETBSDSRCDIR}/sys; don't include files other than the
exported LFS headers, which are lfs.h, lfs_inode.h, and (for now)
lfs_extern.h.
|
| 1.14 | 06-Jun-2013 |
dholland | Cleanups and hacks to make lfs userland stuff build:
- lfs_cksum.c doesn't actually need ulfs_inode.h any more.
- neither does lfs_itimes.c.
- add hacks to fsck_lfs to make it compile.
- add hacks to newfs_lfs to make it compile.
- fix warning in ulfs_quota.c when quotas are fully disabled
(as I guess is happening with the rumpity version)
XXX: This commit adds -I${NETBSDSRCDIR}/sys to the Makefiles for
XXX: fsck_lfs, newfs_lfs, and lfs_cleanerd. This needs to be cleaned
XXX: up ASAP; but I consider this less problematic in the short term
XXX: than spewing ulfs_*.h into /usr/include.
|
| 1.13 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.12 | 28-Apr-2008 |
martin | branches: 1.12.34; 1.12.44;
Remove clause 3 and 4 from TNF licenses
|
| 1.11 | 02-Jan-2008 |
ad | branches: 1.11.6; 1.11.8; 1.11.10;
Merge vmlocking2 to head.
|
| 1.10 | 23-Jun-2006 |
yamt | branches: 1.10.14; 1.10.30; 1.10.36; 1.10.40; 1.10.44;
fix a simonb-timecounters regression.
the precision of getnanotime() is not suitable for file timestamps.
esp. when it's nfs-exported.
- introduce vfs_timestamp().
(the name is from freebsd. currently merely a wrapper of nanotime())
- for ufs-like filesystems, use it rather than getnanotime().
XXX check other filesystems.
|
| 1.9 | 07-Jun-2006 |
kardel | branches: 1.9.2; 1.9.4;
merge FreeBSD timecounters from branch simonb-timecounters
- struct timeval time is gone
time.tv_sec -> time_second
- struct timeval mono_time is gone
mono_time.tv_sec -> time_uptime
- access to time via
{get,}{micro,nano,bin}time()
get* versions are fast but less precise
- support NTP nanokernel implementation (NTP API 4)
- further reading:
Timecounter Paper: http://phk.freebsd.dk/pubs/timecounter.pdf
NTP Nanokernel: http://www.eecis.udel.edu/~mills/ntp/html/kern.html
|
| 1.8 | 15-May-2006 |
christos | branches: 1.8.2;
we need <sys/kauth.h> for the kernel.
|
| 1.7 | 15-May-2006 |
christos | Don't include <sys/kauth.h>; breaks userland (newfs_lfs)
|
| 1.6 | 14-May-2006 |
elad | integrate kauth.
|
| 1.5 | 19-Mar-2006 |
rtr | don't bother checking of ts == NULL before assigning since we know that
it is.
solves coverity 2725 / run 6
|
| 1.4 | 11-Dec-2005 |
christos | branches: 1.4.4; 1.4.6; 1.4.8; 1.4.10; 1.4.12;
merge ktrace-lwp.
|
| 1.3 | 30-Oct-2005 |
simonb | branches: 1.3.2;
We don't need <sys/systm.h> here.
|
| 1.2 | 13-Sep-2005 |
christos | branches: 1.2.2;
redefine panic if we are a user program.
|
| 1.1 | 13-Sep-2005 |
christos | split out lfs_itimes(). It is used in fsck_lfs.
|
| 1.2.2.1 | 02-Nov-2005 |
yamt | sync with head.
|
| 1.3.2.2 | 10-Nov-2005 |
skrll | Sync with HEAD. Here we go again...
|
| 1.3.2.1 | 30-Oct-2005 |
skrll | file lfs_itimes.c was added on branch ktrace-lwp on 2005-11-10 14:12:32 +0000
|
| 1.4.12.2 | 24-May-2006 |
tron | Merge 2006-05-24 NetBSD-current into the "peter-altq" branch.
|
| 1.4.12.1 | 28-Mar-2006 |
tron | Merge 2006-03-28 NetBSD-current into the "peter-altq" branch.
|
| 1.4.10.2 | 06-May-2006 |
christos | - Move kauth_cred_t declaration to <sys/types.h>
- Cleanup struct ucred; forward declarations that are unused.
- Don't include <sys/kauth.h> in any header, but include it in the c files
that need it.
Approved by core.
|
| 1.4.10.1 | 19-Apr-2006 |
elad | sync with head.
|
| 1.4.8.3 | 26-Jun-2006 |
yamt | sync with head.
|
| 1.4.8.2 | 24-May-2006 |
yamt | sync with head.
|
| 1.4.8.1 | 01-Apr-2006 |
yamt | sync with head.
|
| 1.4.6.4 | 01-Jun-2006 |
kardel | Sync with head.
|
| 1.4.6.3 | 22-Apr-2006 |
simonb | Sync with head.
|
| 1.4.6.2 | 05-Feb-2006 |
simonb | In the *itimes functions, just call getnanotime() at the start of
the function and use the result if needed, rather than the previous
conditional calls/assignments method. The code is clearer this way,
and benchmarks at about the same speed.
|
| 1.4.6.1 | 04-Feb-2006 |
simonb | Adapt for timecounters: mostly use get*time() and use "time_second"
instead of "time.tv_sec".
|
| 1.4.4.1 | 09-Sep-2006 |
rpaulo | sync with head
|
| 1.8.2.1 | 19-Jun-2006 |
chap | Sync with head.
|
| 1.9.4.4 | 21-Jan-2008 |
yamt | sync with head
|
| 1.9.4.3 | 30-Dec-2006 |
yamt | sync with head.
|
| 1.9.4.2 | 21-Jun-2006 |
yamt | sync with head.
|
| 1.9.4.1 | 07-Jun-2006 |
yamt | file lfs_itimes.c was added on branch yamt-lazymbuf on 2006-06-21 15:12:39 +0000
|
| 1.9.2.1 | 13-Jul-2006 |
gdamore | Merge from HEAD.
|
| 1.10.44.1 | 02-Jan-2008 |
bouyer | Sync with HEAD
|
| 1.10.40.2 | 19-Dec-2007 |
ad | Use a global lfs_lock.
|
| 1.10.40.1 | 04-Dec-2007 |
ad | Pull the vmlocking changes into a new branch.
|
| 1.10.36.1 | 18-Feb-2008 |
mjf | Sync with HEAD.
|
| 1.10.30.1 | 09-Jan-2008 |
matt | sync with HEAD
|
| 1.10.14.1 | 13-Mar-2007 |
ad | Pull in the initial set of changes for the vmlocking branch.
|
| 1.11.10.1 | 16-May-2008 |
yamt | sync with head.
|
| 1.11.8.1 | 18-May-2008 |
yamt | sync with head.
|
| 1.11.6.1 | 02-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.12.44.2 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.12.44.1 | 23-Jun-2013 |
tls | resync from head
|
| 1.12.34.1 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.15.10.2 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.15.10.1 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.19.10.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.4 | 20-Oct-2025 |
perseant | * Generalize the partial-segment parser introduced for roll-forward,
using it to facilitate an in-kernel segment rewriter (cleaner), and a
mechanism to check whether a segment is in fact empty (only used
with DEBUG).
* Add these new fcntl calls:
- LFCNFILESTATS: For each inode given, report its number of direct
blocks, how many gaps (discontinuities) there are between direct
blocks, and how large the total gap distance is. This will be
useful for a coalescing agent.
- LFCNREWRITEFILE: For each inode given, rewrite its direct blocks,
effectively coalescing it into as compact a form as possible.
- LFCNSCRAMBLE: As above, except that it only rewrites every other
block. This causes the file to have many gaps that can be
measured with LFCNFILESTATS and addressed with LFCNREWRITEFILE,
for testing purposes.
- LFCNREWRITESEGS: Rewrite any live data in the given segments.
This is intended to simplify the cleaner API and facilitate an
in-kernel cleaner.
- LFCNCLEANERINFO: Get the most current CLEANERINFO data from the
kernel.
- LFCNSEGUSE: Retrieve segment usage data from the kernel.
* Vnodes marked IN_CLEANING now take a reference. Add a new "cleaner
lock", which must be taken by the cleaner before the segment lock,
and before marking nodes IN_CLEANING. This allows us to flush
vnodes, if necessary, before the cleaning segment is written, and
never to flush vnodes being cleaned. When the cleaner lock is
released, the vnodes are cleared of IN_CLEANING and the reference
dropped.
* Track a potential infinite loop in lfs_gatherblock.
* Pull "needs to flush" and "needs to wait for flush" into functions
instead of inlining their definitions.
|
| 1.3 | 20-Jun-2016 |
dholland | u_int{8,16,32,64}_t -> uint{8,16,32,64}_t in remaining lfs headers.
|
| 1.2 | 12-Aug-2015 |
dholland | Widen several of the fields of BLOCK_INFO to 64 bits.
Keep the old BLOCK_INFO as BLOCK_INFO_70, and version the fcntls that
use it.
Note that BLOCK_INFO_70 has 64-bit padding issues so that it's
different on 32-bit and 64-bit machines. This has been fixed. However,
BLOCK_INFO also contains a pointer, so compat32 stuff for 32-on-64 is
still needed and doesn't currently exist.
|
| 1.1 | 28-Jul-2013 |
dholland | branches: 1.1.2; 1.1.6; 1.1.10; 1.1.12;
Add lfs_kernel.h for declarations that don't need to be exposed to userland.
lfs currently has the following headers:
lfs.h - on-disk structures and stuff needed for userlevel tools
lfs_inode.h - additional restricted materials for userlevel tools
that operate the fs (newfs_lfs, fsck_lfs, lfs_cleanerd)
lfs_kernel.h - stuff needed only in the kernel
and the following legacy headers that are expected to be mopped up and
folded into one of the above:
lfs_extern.h - function prototypes
ulfs_bswap.h - endian-independent support
ulfs_dinode.h - now contains very little
ulfs_dirhash.h - dirhash support
ulfs_extattr.h - extattr support
ulfs_extern.h - more function prototypes
ulfs_inode.h - assorted kernel-only declarations
ulfs_quota.h - quota support
ulfs_quota1.h - more quota support
ulfs_quota2.h - more quota support
ulfs_quotacommon.h - more quota support
ulfsmount.h - legacy copy of ufsmount material
|
| 1.1.12.2 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.1.12.1 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.1.10.3 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.1.10.2 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.1.10.1 | 28-Jul-2013 |
tls | file lfs_kernel.h was added on branch tls-maxphys on 2014-08-20 00:04:44 +0000
|
| 1.1.6.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.1.6.1 | 28-Jul-2013 |
yamt | file lfs_kernel.h was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.1.2.2 | 28-Aug-2013 |
rmind | sync with head
|
| 1.1.2.1 | 28-Jul-2013 |
rmind | file lfs_kernel.h was added on branch rmind-smpnet on 2013-08-28 23:59:38 +0000
|
| 1.27 | 11-Apr-2023 |
riastradh | lfs: Assert page identity doesn't change.
Forgot what I was debugging when I inserted a relookup in my local
tree months or years ago, but whatever it was, if that solved a
problem, this KDASSERT will make the problem more obvious.
|
| 1.26 | 05-Sep-2020 |
riastradh | Round of uvm.h cleanup.
The poorly named uvm.h is generally supposed to be for uvm-internal
users only.
- Narrow it to files that actually need it -- mostly files that need
to query whether curlwp is the pagedaemon, which should maybe be
exposed by an external header.
- Use uvm_extern.h where feasible and uvm_*.h for things not exposed
by it. We should split up uvm_extern.h but this will serve for now
to reduce the uvm.h dependencies.
- Use uvm_stat.h and #ifdef UVMHIST uvm.h for files that use
UVMHIST(ubchist), since ubchist is declared in uvm.h but the
reference evaporates if UVMHIST is not defined, so we reduce header
file dependencies.
- Make uvm_device.h and uvm_swap.h independently includable while
here.
ok chs@
|
| 1.25 | 17-Mar-2020 |
ad | Tweak the March 14th change to make page waits interlocked by pg->interlock.
Remove unneeded changes and only deal with the PQ_WANTED flag, to exclude
possible bugs.
|
| 1.24 | 14-Mar-2020 |
ad | Make uvm_pagemarkdirty() responsible for putting vnodes onto the syncer
work list. Proposed on tech-kern@.
|
| 1.23 | 14-Mar-2020 |
ad | Make page waits (WANTED vs BUSY) interlocked by pg->interlock. Gets RW
locks out of the equation for sleep/wakeup, and allows observing+waiting
for busy pages when holding only a read lock. Proposed on tech-kern.
|
| 1.22 | 23-Feb-2020 |
ad | UVM locking changes, proposed on tech-kern:
- Change the lock on uvm_object, vm_amap and vm_anon to be a RW lock.
- Break v_interlock and vmobjlock apart. v_interlock remains a mutex.
- Do partial PV list locking in the x86 pmap. Others to follow later.
|
| 1.21 | 23-Feb-2020 |
riastradh | Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
|
| 1.20 | 15-Jan-2020 |
ad | Merge from yamt-pagecache (after much testing):
- Reduce unnecessary page scan in putpages esp. when an object has a ton of
pages cached but only a few of them are dirty.
- Reduce the number of pmap operations by tracking page dirtiness more
precisely in uvm layer.
|
| 1.19 | 31-Dec-2019 |
ad | branches: 1.19.2;
- Add and use wrapper functions that take and acquire page interlocks, and pairs
of page interlocks. Require that the page interlock be held over calls to
uvm_pageactivate(), uvm_pagewire() and similar.
- Solve the concurrency problem with page replacement state. Rather than
updating the global state synchronously, set an intended state on
individual pages (active, inactive, enqueued, dequeued) while holding the
page interlock. After the interlock is released put the pages on a 128
entry per-CPU queue for their state changes to be made real in batch.
This results in in a ~400 fold decrease in contention on my test system.
Proposed on tech-kern but modified to use the page interlock rather than
atomics to synchronise as it's much easier to maintain that way, and
cheaper.
|
| 1.18 | 20-Dec-2019 |
ad | Fix lfs_putpages() for bsize < nbpg.
|
| 1.17 | 15-Dec-2019 |
ad | Merge from yamt-pagecache:
- do gang lookup of pages using radixtree.
- remove now unused uvm_object::uo_memq and vm_page::listq.queue.
|
| 1.16 | 13-Dec-2019 |
ad | Break the global uvm_pageqlock into a per-page identity lock and a private
lock for use of the pagedaemon policy code. Discussed on tech-kern.
PR kern/54209: NetBSD 8 large memory performance extremely low
PR kern/54210: NetBSD-8 processes presumably not exiting
PR kern/54727: writing a large file causes unreasonable system behaviour
|
| 1.15 | 19-Aug-2017 |
maya | branches: 1.15.4; 1.15.8;
Ask some question about the code in a XXX comment
|
| 1.14 | 10-Jun-2017 |
maya | Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.13 | 05-Jun-2017 |
maya | Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
|
| 1.12 | 04-Jun-2017 |
hannken | Operations fstrans_start() and fstrans_start_nowait() now always
use FSTRANS_SHARED as lock type so remove the lock type argument.
File system state FSTRANS_SUSPENDING is now unused so remove it.
Regen vnode_if files.
Ride 8.99.1 less than a hour ago.
|
| 1.11 | 01-Apr-2017 |
maya | branches: 1.11.6;
Switch lfs_writer_daemon to use condvar instead of mtsleep.
track thread existence with struct lwp instead of pid + lid,
it's more useful from ddb.
|
| 1.10 | 30-Mar-2017 |
hannken | Remove now redundant calls to fstrans_start()/fstrans_done().
Add fstrans_start()/fstrans_done() to lfs_putpages().
|
| 1.9 | 04-Oct-2016 |
christos | branches: 1.9.2;
Grr, the optimizer on mips64 can't handle this... Use MIN_PAGE_SIZE.
|
| 1.8 | 21-Jul-2016 |
christos | Don't do variable stack allocations for systems with non-const PAGE_SIZE;
instead assume that the smallest pagesize is 1024.
|
| 1.7 | 12-Aug-2015 |
dholland | branches: 1.7.2;
Make 32-bit and 64-bit versions of SEGSUM.
Also fix some of the FINFO handling as it's closely entangled.
|
| 1.6 | 02-Aug-2015 |
dholland | Second batch of 64 -> 32 truncations in lfs, along with more minor
tidyups and corrections in passing.
|
| 1.5 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.4 | 25-Jul-2015 |
martin | Use accessors in DEBUG and DIAGNOSTIC code as well
|
| 1.3 | 24-Jul-2015 |
dholland | More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)
The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.
|
| 1.2 | 24-Jul-2015 |
dholland | Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)
It also gets rid of cpp abuse in the form of fake structure member
macros.
Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)
XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.
|
| 1.1 | 16-May-2014 |
dholland | branches: 1.1.2; 1.1.4; 1.1.8; 1.1.10;
Move lfs_getpages and lfs_putpages to their own file.
|
| 1.1.10.4 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.1.10.3 | 05-Dec-2016 |
skrll | Sync with HEAD
|
| 1.1.10.2 | 05-Oct-2016 |
skrll | Sync with HEAD
|
| 1.1.10.1 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.1.8.3 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.1.8.2 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.1.8.1 | 16-May-2014 |
tls | file lfs_pages.c was added on branch tls-maxphys on 2014-08-20 00:04:44 +0000
|
| 1.1.4.2 | 10-Aug-2014 |
tls | Rebase.
|
| 1.1.4.1 | 16-May-2014 |
tls | file lfs_pages.c was added on branch tls-earlyentropy on 2014-08-10 06:56:58 +0000
|
| 1.1.2.2 | 18-May-2014 |
rmind | sync with head
|
| 1.1.2.1 | 16-May-2014 |
rmind | file lfs_pages.c was added on branch rmind-smpnet on 2014-05-18 17:46:21 +0000
|
| 1.7.2.3 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.7.2.2 | 04-Nov-2016 |
pgoyette | Sync with HEAD
|
| 1.7.2.1 | 26-Jul-2016 |
pgoyette | Sync with HEAD
|
| 1.9.2.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.11.6.2 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.11.6.1 | 04-Jun-2017 |
bouyer | pullup the following revisions, requested by hannken in ticket #2:
src/share/man/man9/fstrans.9 1.25
src/sys/kern/vfs_mount.c 1.66
src/sys/kern/vfs_subr.c 1.468
src/sys/kern/vfs_trans.c 1.46
src/sys/kern/vfs_vnode.c 1.94, 1.95, 1.96
src/sys/kern/vnode_if.c 1.105, 1.106
src/sys/kern/vnode_if.sh 1.65, 1.66
src/sys/kern/vnode_if.src 1.76
src/sys/miscfs/genfs/genfs_io.c 1.69
src/sys/miscfs/genfs/genfs_vnops.c 1.196, 1.197
src/sys/miscfs/genfs/layer_extern.h 1.40
src/sys/miscfs/genfs/layer_vfsops.c 1.51
src/sys/miscfs/genfs/layer_vnops.c 1.67
src/sys/miscfs/nullfs/null_vnops.c 1.42
src/sys/miscfs/overlay/overlay_vnops.c 1.24
src/sys/miscfs/umapfs/umap_vnops.c 1.60
src/sys/rump/include/rump/rumpvnode_if.h 1.29, 1.30
src/sys/rump/librump/rumpkern/emul.c 1.182
src/sys/rump/librump/rumpvfs/rumpvnode_if.c 1.29, 1.30
src/sys/sys/fstrans.h 1.11
src/sys/sys/vnode.h 1.278
src/sys/sys/vnode_if.h 1.100, 1.101
src/sys/sys/vnode_impl.h 1.14, 1.15
src/sys/ufs/lfs/lfs_pages.c 1.12
Vnode state, lock and fstrans cleanup:
- Rename vnode state "VS_ACTIVE" to "VS_LOADED" and add synthetic
state "VS_ACTIVE" to assert a loaded vnode with usecount > 0.
- Redo FSTRANS in vnode_if.c and use it for VOP_LOCK and VOP_UNLOCK.
- Cleanup the genfs lock operations.
- Make "struct vnode_impl" member "vi_lock" a krwlock_t again.
- Remove the lock type argument from fstrans_start and
fstrans_start_nowait,
remove now unused FSTRANS state "FSTRANS_SUSPENDING".
|
| 1.15.8.1 | 17-Aug-2020 |
martin | Pull up following revision(s) (requested by riastradh in ticket #1050):
sys/ufs/lfs/lfs_subr.c: revision 1.101
sys/ufs/lfs/lfs_subr.c: revision 1.102
sys/ufs/lfs/lfs_inode.c: revision 1.158
sys/ufs/lfs/lfs_inode.h: revision 1.25
sys/ufs/lfs/lfs_balloc.c: revision 1.95
sys/ufs/lfs/lfs_pages.c: revision 1.21
sys/ufs/lfs/lfs_vnops.c: revision 1.330
sys/ufs/lfs/lfs_alloc.c: revision 1.140 (patch)
sys/ufs/lfs/lfs_alloc.c: revision 1.141 (patch)
lib/libp2k/p2k.c: revision 1.72
sys/ufs/lfs/lfs.h: revision 1.205
sys/ufs/lfs/lfs.h: revision 1.206
sys/ufs/lfs/lfs_segment.c: revision 1.284
sys/ufs/lfs/lfs.h: revision 1.207
sys/ufs/lfs/lfs_segment.c: revision 1.285
sys/ufs/lfs/lfs_debug.c: revision 1.55
sys/ufs/lfs/lfs_rename.c: revision 1.23
usr.sbin/dumplfs/dumplfs.c: revision 1.65
sys/ufs/lfs/lfs_vfsops.c: revision 1.371
sys/arch/i386/stand/efiboot/bootx64/Makefile: revision 1.3
sys/ufs/lfs/lfs_vfsops.c: revision 1.372
sys/ufs/lfs/lfs_vfsops.c: revision 1.373
sbin/fsck_lfs/pass1.c: revision 1.46
sys/ufs/lfs/lfs_vnops.c: revision 1.326
sys/ufs/lfs/lfs_vnops.c: revision 1.327
sys/ufs/lfs/lfs_vfsops.c: revision 1.375 (patch)
sys/ufs/lfs/lfs_vnops.c: revision 1.328
sys/ufs/lfs/lfs_subr.c: revision 1.98
sys/ufs/lfs/lfs_extern.h: revision 1.116
sys/ufs/lfs/lfs_vnops.c: revision 1.329
sys/ufs/lfs/lfs_subr.c: revision 1.99
sys/ufs/lfs/lfs_extern.h: revision 1.117
sys/ufs/lfs/lfs_accessors.h: revision 1.49
sys/ufs/lfs/lfs_extern.h: revision 1.118
sys/rump/fs/lib/liblfs/Makefile: revision 1.15
sys/ufs/lfs/lfs_bio.c: revision 1.146 (patch)
sys/ufs/lfs/lfs_bio.c: revision 1.147
sys/ufs/lfs/lfs_subr.c: revision 1.100
Fix kassert in lfs by initializing vp first.
Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
Teach lfs to transition ro<->rw.
Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
Assert lfs_writer where I think we can now prove it.
Serialize access to the splay tree with lfs_lock.
Change some cheap KDASSERT into KASSERT.
Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
Call lfs_orphan in lfs_rename while we're still in the dirop.
lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
Teach LFS_ORPHAN_NEXTFREE about lfs64.
Dust off the orphan detection code and try to make it work.
Fix !DIAGNOSTIC compile
Fix userland references to LFS_ORPHAN_NEXTFREE.
Forgot to grep for these or do a full distribution build, oops!
Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
OR into bp->b_cflags; don't overwrite.
CTASSERT lfs on-disk structure sizes.
Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
Fix clang build after packed lfs64 accessor change.
Suppress spurious address-of-packed error in rump lfs too.
|
| 1.15.4.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.19.2.2 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.19.2.1 | 17-Jan-2020 |
ad | Sync with head.
|
| 1.25 | 20-Oct-2021 |
thorpej | Overhaul of the EVFILT_VNODE kevent(2) filter:
- Centralize vnode kevent handling in the VOP_*() wrappers, rather than
forcing each individual file system to deal with it (except VOP_RENAME(),
because VOP_RENAME() is a mess and we currently have 2 different ways
of handling it; at least it's reasonably well-centralized in the "new"
way).
- Add support for NOTE_OPEN, NOTE_CLOSE, NOTE_CLOSE_WRITE, and NOTE_READ,
compatible with the same events in FreeBSD.
- Track which kevent notifications clients are interested in receiving
to avoid doing work for events no one cares about (avoiding, e.g.
taking locks and traversing the klist to send a NOTE_WRITE when
someone is merely watching for a file to be deleted, for example).
In support of the above:
- Add support in vnode_if.sh for specifying PRE- and POST-op handlers,
to be invoked before and after vop_pre() and vop_post(), respectively.
Basic idea from FreeBSD, but implemented differently.
- Add support in vnode_if.sh for specifying CONTEXT fields in the
vop_*_args structures. These context fields are used to convey information
between the file system VOP function and the VOP wrapper, but do not
occupy an argument slot in the VOP_*() call itself. These context fields
are initialized and subsequently interpreted by PRE- and POST-op handlers.
- Version VOP_REMOVE(), uses the a context field for the file system to report
back the resulting link count of the target vnode. Return this in tmpfs,
udf, nfs, chfs, ext2fs, lfs, and ufs.
NetBSD 9.99.92.
|
| 1.24 | 05-Sep-2020 |
riastradh | Round of uvm.h cleanup.
The poorly named uvm.h is generally supposed to be for uvm-internal
users only.
- Narrow it to files that actually need it -- mostly files that need
to query whether curlwp is the pagedaemon, which should maybe be
exposed by an external header.
- Use uvm_extern.h where feasible and uvm_*.h for things not exposed
by it. We should split up uvm_extern.h but this will serve for now
to reduce the uvm.h dependencies.
- Use uvm_stat.h and #ifdef UVMHIST uvm.h for files that use
UVMHIST(ubchist), since ubchist is declared in uvm.h but the
reference evaporates if UVMHIST is not defined, so we reduce header
file dependencies.
- Make uvm_device.h and uvm_swap.h independently includable while
here.
ok chs@
|
| 1.23 | 23-Feb-2020 |
riastradh | Call lfs_orphan in lfs_rename while we're still in the dirop.
|
| 1.22 | 10-Jun-2017 |
maya | branches: 1.22.6; 1.22.10; 1.22.12;
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.21 | 20-Jun-2016 |
dholland | branches: 1.21.10;
One more batch of already-synced ufs changes:
ufs_extern.h 1.79 is equivalent to ulfs_extern.h 1.14
ufsmount.h 1.43 is (roughly) equivalent to lfs_extern.h 1.102
ufs_inode.c 1.94 does not apply to lfs
ufs_inode.c 1.95 does not apply to lfs either
ufs_readwrite.c 1.108 is equivalent to ulfs_readwrite.c 1.8
ufs_readwrite.c 1.109 is equivalent to ulfs_readwrite.c 1.9
ufs_readwrite.c 1.110 is equivalent to ulfs_readwrite.c 1.10
ufs_readwrite.c 1.111 does not apply to lfs
ufs_readwrite.c 1.112 is equivalent to ulfs_readwrite.c 1.11
ufs_readwrite.c 1.113 is equivalent to ulfs_readwrite.c 1.13
ufs_readwrite.c 1.114 is equivalent to ulfs_readwrite.c 1.14
ufs_readwrite.c 1.115 is equivalent to ulfs_readwrite.c 1.15
ufs_readwrite.c 1.116-1.118 does not apply to lfs
ufs_readwrite.c 1.119-1.120 are equivalent to ulfs_readwrite.c 1.16
ufs_rename.c 1.12 is equivalent to lfs_rename.c 1.8
ufs_vnops.c 1.226 is equivalent to ulfs_vnops.c 1.22 and lfs_vnops.c 1.270
ufs_vnops.c 1.227 is equivalent to ulfs_vnops.c 1.23
ufs_vnops.c 1.228-1.229 are equivalent to ulfs_vnops.c 1.24
ufs_vnops.c 1.230 is equivalent to ulfs_vnops.c 1.25 and lfs_vnops.c 1.271
ufs_vnops.c 1.231 originated in lfs
ufs_vnops.c 1.232 does not apply to lfs
|
| 1.20 | 20-Jun-2016 |
dholland | fix typo in previous
|
| 1.19 | 20-Jun-2016 |
dholland | Merge ufs_rename.c 1.11: ufs_gro_genealogy: use vcache_get() to lookup DOTDOT.
|
| 1.18 | 20-Jun-2016 |
dholland | More already-merged or equivalent changes:
ufs_dirhash.c 1.36 corresponds to ulfs_dirhash.c 1.8
ufs_extattr.c 1.43 corresponds to ulfs_extattr.c 1.7
ufs_lookup.c 1.126 does not apply to lfs
ufs_lookup.c 1.127 we already have
ufs_lookup.c 1.128 does not apply to lfs
ufs_lookup.c 1.129 corresponds to ulfs_lookup.c 1.19
ufs_quota1.c 1.19 corresponds to ulfs_quota1.c 1.7
ufs_quota1.c 1.20 corresponds to ulfs_quota1.c 1.8
ufs_quota2.c 1.36 we have equivalent changes for
ufs_rename.c 1.9 corresponds to lfs_rename.c 1.5
ufs_rename.c 1.10 corresponds to lfs_rename.c 1.6
ufs_vnops.c 1.219 corresponds to lfs_vnops.c 1.260 and ulfs_vnops.c 1.19
ufs_vnops.c 1.220 corresponds to lfs_vnops.c 1.261 and ulfs_vnops.c 1.20
ufs_vnops.c 1.221 was superseded by later changes
ufs_vnops.c 1.222 got fixed independently in lfs
|
| 1.17 | 19-Jun-2016 |
dholland | Mark ufs file versions we're already synced with.
|
| 1.16 | 21-Sep-2015 |
dholland | Add 64-bit directory entry structures, and adjust accessors accordingly.
The LFS64 directory entry has a 64-bit inode number. This is stored as
two 32-bit values to avoid inducing 64-bit alignment requirements.
The exposed type for manipulating directory entries is now
LFS_DIRHEADER, following the same convention as e.g. IFILE and SEGUSE.
(But with LFS_ on it, because.)
|
| 1.15 | 21-Sep-2015 |
dholland | Oops; LFS_DIRECTSIZ() is going to need the fs as an argument.
Also, it turns out that dirhash needs a compile-time-constant version
of LFS_DIRECTSIZ(LFS_MAXNAMLEN+1), independent of 64-vs-32, so create
LFS_MAXDIRENTRYSIZE for this. Sigh.
|
| 1.14 | 20-Sep-2015 |
dholland | Clean up struct lfs_dirtemplate.
|
| 1.13 | 15-Sep-2015 |
dholland | Pass around struct lfs_dirheader instead of struct lfs_direct.
|
| 1.12 | 15-Sep-2015 |
dholland | Add an accessor function for directory names.
|
| 1.11 | 15-Sep-2015 |
dholland | Kill off ulfs_makedirentry; just pass the data to ulfs_direnter instead.
For now, move one copy of the code that allocates and fills in a
temporary struct lfs_direct to the top of ulfs_direnter; but it should
go away shortly.
|
| 1.10 | 15-Sep-2015 |
dholland | Add and use accessor functions for more of the directory entry fields.
|
| 1.9 | 01-Sep-2015 |
dholland | Add new accessors for the d_type and d_namlen fields of struct lfs_direct.
Napalm the old byteswap access logic for these.
|
| 1.8 | 27-Mar-2015 |
riastradh | Disentangle buffer-cached I/O from page-cached I/O in UFS.
Page-cached I/O is used for regular files, and is initiated by VFS
users such as userland and NFS.
Buffer-cached I/O is used for directories and symlinks, and is issued
only internally by UFS.
New UFS routine ufs_bufio replaces vn_rdwr for internal use.
ufs_bufio is implemented by new UFS operations uo_bufrd/uo_bufwr,
which sit in ufs_readwrite.c alongside the VOP_READ/VOP_WRITE
implementations.
I preserved the code as much as possible and will leave further
simplification for future commits. I kept the ulfs_readwrite.c
copypasta close to ufs_readwrite.c in case we ever want to merge them
back; likewise ext2fs_readwrite.c.
No externally visible semantic change. All atf fs tests still pass.
|
| 1.7 | 17-May-2014 |
dholland | branches: 1.7.2; 1.7.6; 1.7.8;
Remove the DIROP macros. They are evil, especially the CREATE ones.
This results in some duplicate logic in the creation vnops (symlink,
mknod, create, mkdir) but we will probably be able to factor it out in
a more sensible way later.
Now the creation vnops call getnewvnode explicitly instead of under
multiple layers of obscure gunk. Then we explicitly do lfs_set_dirop,
and afterwards lfs_unset_dirop.
|
| 1.6 | 06-Feb-2014 |
hannken | branches: 1.6.2;
Move fstrans_start()/fstrans_done() into genfs_insane_rename() to protect
the complete rename operation like we do for all other vnode operations.
|
| 1.5 | 28-Jan-2014 |
martin | Quell a gcc 4.8 maybe-unitialized false positive
|
| 1.4 | 28-Jul-2013 |
dholland | branches: 1.4.2;
Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.3 | 28-Jul-2013 |
dholland | Remove the now-pointless ulfs ops macros.
|
| 1.2 | 20-Jul-2013 |
dholland | branches: 1.2.2;
G/C unused pieces.
|
| 1.1 | 20-Jul-2013 |
dholland | Collect the pieces of lfs rename into lfs_rename.c, and sprinkle static.
|
| 1.2.2.2 | 23-Jul-2013 |
riastradh | sync with HEAD
|
| 1.2.2.1 | 20-Jul-2013 |
riastradh | file lfs_rename.c was added on branch riastradh-drm2 on 2013-07-23 21:07:38 +0000
|
| 1.4.2.3 | 18-May-2014 |
rmind | sync with head
|
| 1.4.2.2 | 28-Aug-2013 |
rmind | sync with head
|
| 1.4.2.1 | 28-Jul-2013 |
rmind | file lfs_rename.c was added on branch rmind-smpnet on 2013-08-28 23:59:38 +0000
|
| 1.6.2.1 | 10-Aug-2014 |
tls | Rebase.
|
| 1.7.8.4 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.7.8.3 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.7.8.2 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.7.8.1 | 06-Apr-2015 |
skrll | Sync with HEAD
|
| 1.7.6.3 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.7.6.2 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.7.6.1 | 17-May-2014 |
tls | file lfs_rename.c was added on branch tls-maxphys on 2014-08-20 00:04:44 +0000
|
| 1.7.2.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.7.2.1 | 17-May-2014 |
yamt | file lfs_rename.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.21.10.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.22.12.1 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.22.10.1 | 17-Aug-2020 |
martin | Pull up following revision(s) (requested by riastradh in ticket #1050):
sys/ufs/lfs/lfs_subr.c: revision 1.101
sys/ufs/lfs/lfs_subr.c: revision 1.102
sys/ufs/lfs/lfs_inode.c: revision 1.158
sys/ufs/lfs/lfs_inode.h: revision 1.25
sys/ufs/lfs/lfs_balloc.c: revision 1.95
sys/ufs/lfs/lfs_pages.c: revision 1.21
sys/ufs/lfs/lfs_vnops.c: revision 1.330
sys/ufs/lfs/lfs_alloc.c: revision 1.140 (patch)
sys/ufs/lfs/lfs_alloc.c: revision 1.141 (patch)
lib/libp2k/p2k.c: revision 1.72
sys/ufs/lfs/lfs.h: revision 1.205
sys/ufs/lfs/lfs.h: revision 1.206
sys/ufs/lfs/lfs_segment.c: revision 1.284
sys/ufs/lfs/lfs.h: revision 1.207
sys/ufs/lfs/lfs_segment.c: revision 1.285
sys/ufs/lfs/lfs_debug.c: revision 1.55
sys/ufs/lfs/lfs_rename.c: revision 1.23
usr.sbin/dumplfs/dumplfs.c: revision 1.65
sys/ufs/lfs/lfs_vfsops.c: revision 1.371
sys/arch/i386/stand/efiboot/bootx64/Makefile: revision 1.3
sys/ufs/lfs/lfs_vfsops.c: revision 1.372
sys/ufs/lfs/lfs_vfsops.c: revision 1.373
sbin/fsck_lfs/pass1.c: revision 1.46
sys/ufs/lfs/lfs_vnops.c: revision 1.326
sys/ufs/lfs/lfs_vnops.c: revision 1.327
sys/ufs/lfs/lfs_vfsops.c: revision 1.375 (patch)
sys/ufs/lfs/lfs_vnops.c: revision 1.328
sys/ufs/lfs/lfs_subr.c: revision 1.98
sys/ufs/lfs/lfs_extern.h: revision 1.116
sys/ufs/lfs/lfs_vnops.c: revision 1.329
sys/ufs/lfs/lfs_subr.c: revision 1.99
sys/ufs/lfs/lfs_extern.h: revision 1.117
sys/ufs/lfs/lfs_accessors.h: revision 1.49
sys/ufs/lfs/lfs_extern.h: revision 1.118
sys/rump/fs/lib/liblfs/Makefile: revision 1.15
sys/ufs/lfs/lfs_bio.c: revision 1.146 (patch)
sys/ufs/lfs/lfs_bio.c: revision 1.147
sys/ufs/lfs/lfs_subr.c: revision 1.100
Fix kassert in lfs by initializing vp first.
Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
Teach lfs to transition ro<->rw.
Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
Assert lfs_writer where I think we can now prove it.
Serialize access to the splay tree with lfs_lock.
Change some cheap KDASSERT into KASSERT.
Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
Call lfs_orphan in lfs_rename while we're still in the dirop.
lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
Teach LFS_ORPHAN_NEXTFREE about lfs64.
Dust off the orphan detection code and try to make it work.
Fix !DIAGNOSTIC compile
Fix userland references to LFS_ORPHAN_NEXTFREE.
Forgot to grep for these or do a full distribution build, oops!
Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
OR into bp->b_cflags; don't overwrite.
CTASSERT lfs on-disk structure sizes.
Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
Fix clang build after packed lfs64 accessor change.
Suppress spurious address-of-packed error in rump lfs too.
|
| 1.22.6.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.40 | 20-Oct-2025 |
perseant | * Generalize the partial-segment parser introduced for roll-forward,
using it to facilitate an in-kernel segment rewriter (cleaner), and a
mechanism to check whether a segment is in fact empty (only used
with DEBUG).
* Add these new fcntl calls:
- LFCNFILESTATS: For each inode given, report its number of direct
blocks, how many gaps (discontinuities) there are between direct
blocks, and how large the total gap distance is. This will be
useful for a coalescing agent.
- LFCNREWRITEFILE: For each inode given, rewrite its direct blocks,
effectively coalescing it into as compact a form as possible.
- LFCNSCRAMBLE: As above, except that it only rewrites every other
block. This causes the file to have many gaps that can be
measured with LFCNFILESTATS and addressed with LFCNREWRITEFILE,
for testing purposes.
- LFCNREWRITESEGS: Rewrite any live data in the given segments.
This is intended to simplify the cleaner API and facilitate an
in-kernel cleaner.
- LFCNCLEANERINFO: Get the most current CLEANERINFO data from the
kernel.
- LFCNSEGUSE: Retrieve segment usage data from the kernel.
* Vnodes marked IN_CLEANING now take a reference. Add a new "cleaner
lock", which must be taken by the cleaner before the segment lock,
and before marking nodes IN_CLEANING. This allows us to flush
vnodes, if necessary, before the cleaning segment is written, and
never to flush vnodes being cleaned. When the cleaner lock is
released, the vnodes are cleared of IN_CLEANING and the reference
dropped.
* Track a potential infinite loop in lfs_gatherblock.
* Pull "needs to flush" and "needs to wait for flush" into functions
instead of inlining their definitions.
|
| 1.39 | 14-Oct-2025 |
perseant | Check the existing inode address against LFS_UNUSED_DADDR before checking
whether it is in the same segment, to prevent a byte undercount in segment 0
during roll forward. This was most often expressed in the fs/lfs/t_rfw rfw64
test case, though it affected both 32- and 64-bit LFSs equally.
|
| 1.38 | 06-Oct-2025 |
perseant | Don't stop recovery when we find a partial-segment with neither inodes nor
finfos. Under normal conditions, we should never be producing such a partial
segment. However, these do sometimes appear and they need not prevent us
from continuing.
|
| 1.37 | 17-Sep-2025 |
perseant | Add working in-kernel roll forward.
|
| 1.36 | 05-Sep-2020 |
riastradh | Round of uvm.h cleanup.
The poorly named uvm.h is generally supposed to be for uvm-internal
users only.
- Narrow it to files that actually need it -- mostly files that need
to query whether curlwp is the pagedaemon, which should maybe be
exposed by an external header.
- Use uvm_extern.h where feasible and uvm_*.h for things not exposed
by it. We should split up uvm_extern.h but this will serve for now
to reduce the uvm.h dependencies.
- Use uvm_stat.h and #ifdef UVMHIST uvm.h for files that use
UVMHIST(ubchist), since ubchist is declared in uvm.h but the
reference evaporates if UVMHIST is not defined, so we reduce header
file dependencies.
- Make uvm_device.h and uvm_swap.h independently includable while
here.
ok chs@
|
| 1.35 | 17-Jan-2020 |
ad | VFS_VGET(), VFS_ROOT(), VFS_FHTOVP(): give them a "int lktype" argument, to
allow us to get shared locks (or no lock) on the returned vnode. Matches
FreeBSD.
|
| 1.34 | 01-Jan-2019 |
hannken | branches: 1.34.6;
Add "void *extra" argument to vcache_new() so a file system may
pass more information about the file to create.
Welcome to 8.99.30
|
| 1.33 | 10-Dec-2018 |
maxv | Remove unused mbuf.h includes.
|
| 1.32 | 03-Oct-2015 |
dholland | branches: 1.32.16; 1.32.18;
Use the new IINFO in the rfw code, eliminating hardwired 32-bit values.
|
| 1.31 | 01-Sep-2015 |
dholland | Use the lfs dinode accessors in place of the ufs-derived ones.
(Mostly.)
The ufs-derived ones are fake structure member macros, which are gross
and not very safe. Also, it seems that a lot of places in the lfs code
were using the ffsv1 branch of them unconditionally, and this way it's
guaranteed all those places have been updated.
Found while doing this: for non-devices, have getattr produce NODEV
in the rdev field instead of leaking the address of the first direct
block.
|
| 1.30 | 19-Aug-2015 |
dholland | Part two of dinodes; use the same union everywhere.
(previously the ufs-derived code had things set up slightly different)
Remove a bunch of associated mess.
|
| 1.29 | 12-Aug-2015 |
dholland | Hack up dinode usage to be 64 vs. 32 as needed. Part 1.
(This part changes the native lfs code; the ufs-derived code already
has 64 vs. 32 logic, but as aspects of it are unsafe, and don't
entirely interoperate cleanly with the lfs 64/32 stuff, pass 2 will be
rehashing that.)
|
| 1.28 | 12-Aug-2015 |
dholland | Provide 32-bit and 64-bit versions of FINFO.
This also entailed sorting out part of struct segment, as that
contains a pointer into the current FINFO data.
|
| 1.27 | 12-Aug-2015 |
dholland | Make 32-bit and 64-bit versions of SEGSUM.
Also fix some of the FINFO handling as it's closely entangled.
|
| 1.26 | 12-Aug-2015 |
dholland | Add IFILE32 and IFILE64 structures for the on-disk ifile entries.
Add and use accessors. There are also a bunch of places that cast and
I hope I've found them all...
|
| 1.25 | 02-Aug-2015 |
dholland | Use accessor functions for the version field of the lfs superblock.
I thought at first maybe the cases that test the version should be
rolled into the accessors, but on the whole I think the conclusion on
that is no.
|
| 1.24 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.23 | 24-Jul-2015 |
dholland | More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)
The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.
|
| 1.22 | 24-Jul-2015 |
dholland | Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)
It also gets rid of cpp abuse in the form of fake structure member
macros.
Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)
XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.
|
| 1.21 | 16-Jul-2015 |
dholland | Don't cast the return value of malloc.
|
| 1.20 | 31-May-2015 |
hannken | Change lfs from hash table to vcache.
- Change lfs_valloc() to return an inode number and version instead of
a vnode and move lfs_ialloc() and lfs_vcreate() to new lfs_init_vnode().
- Add lfs_valloc_fixed() to allocate a known inode, used by kernel
roll forward.
- Remove lfs_*ref(), these functions cannot coexist with vcache and
their commented behaviour is far away from their implementation.
- Add the cleaner lwp and blockinfo to struct ulfsmount so lfs_loadvnode()
may use hints from the cleaner.
- Remove vnode locks from ulfs_lookup() like we did with ufs_lookup().
|
| 1.19 | 28-Mar-2015 |
maxv | Remove the 'cred' argument from bread(). Remove a now unused var in
ffs_snapshot.c. Update the man page accordingly.
ok hannken@
|
| 1.18 | 28-Jul-2013 |
dholland | branches: 1.18.6;
Add lfs_kernel.h for declarations that don't need to be exposed to userland.
lfs currently has the following headers:
lfs.h - on-disk structures and stuff needed for userlevel tools
lfs_inode.h - additional restricted materials for userlevel tools
that operate the fs (newfs_lfs, fsck_lfs, lfs_cleanerd)
lfs_kernel.h - stuff needed only in the kernel
and the following legacy headers that are expected to be mopped up and
folded into one of the above:
lfs_extern.h - function prototypes
ulfs_bswap.h - endian-independent support
ulfs_dinode.h - now contains very little
ulfs_dirhash.h - dirhash support
ulfs_extattr.h - extattr support
ulfs_extern.h - more function prototypes
ulfs_inode.h - assorted kernel-only declarations
ulfs_quota.h - quota support
ulfs_quota1.h - more quota support
ulfs_quota2.h - more quota support
ulfs_quotacommon.h - more quota support
ulfsmount.h - legacy copy of ufsmount material
|
| 1.17 | 18-Jun-2013 |
christos | branches: 1.17.2;
Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.
|
| 1.16 | 08-Jun-2013 |
dholland | Stick LFS_ in front of IFMT, IFIFO, IFREG, etc. so as not to conflict
with the UFS copies of these symbols. (Which themselves ought to have
UFS_ stuck on.)
|
| 1.15 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.14 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.13 | 22-Jan-2013 |
dholland | Stuff UFS_ in front of a few of ufs's symbols to reduce namespace
pollution. Specifically:
ROOTINO -> UFS_ROOTINO
WINO -> UFS_WINO
NXADDR -> UFS_NXADDR
NDADDR -> UFS_NDADDR
NIADDR -> UFS_NIADDR
MAXSYMLINKLEN -> UFS_MAXSYMLINKLEN
MAXSYMLINKLEN_UFS[12] -> UFS[12]_MAXSYMLINKLEN (for consistency)
Sort out ext2fs's misuse of NDADDR and NIADDR; fortunately, these have
the same values in ext2fs and ffs.
No functional change intended.
|
| 1.12 | 22-Feb-2009 |
ad | branches: 1.12.12; 1.12.22;
PR kern/26878 FFSv2 + softdep = livelock (no free ram)
PR kern/16942 panic with softdep and quotas
PR kern/19565 panic: softdep_write_inodeblock: indirect pointer #1 mismatch
PR kern/26274 softdep panic: allocdirect_merge: ...
PR kern/26374 Long delay before non-root users can write to softdep partitions
PR kern/28621 1.6.x "vp != NULL" panic in ffs_softdep.c:4653 while unmounting a softdep (+quota) filesystem
PR kern/29513 FFS+Softdep panic with unfsck-able file-corruption
PR kern/31544 The ffs softdep code appears to fail to write dirty bits to disk
PR kern/31981 stopping scsi disk can cause panic (softdep)
PR kern/32116 kernel panic in softdep (assertion failure)
PR kern/32532 softdep_trackbufs deadlock
PR kern/37191 softdep: locking against myself
PR kern/40474 Kernel panic after remounting raid root with softdep
Retire softdep, pass 2. As discussed and later formally announced on the
mailing lists.
|
| 1.11 | 16-May-2008 |
hannken | branches: 1.11.6; 1.11.12;
Make sure all cached buffers with valid, not yet written data have been
run through copy-on-write. Call fscow_run() with valid data where possible.
The LP_UFSCOW hack is no longer needed to protect ffs_copyonwrite() against
endless recursion.
- Add a flag B_MODIFY to bread(), breada() and breadn(). If set the caller
intends to modify the buffer returned.
- Always run copy-on-write on buffers returned from ffs_balloc().
- Add new function ffs_getblk() that gets a buffer, assigns a new blkno,
may clear the buffer and runs copy-on-write. Process possible errors
from getblk() or fscow_run(). Part of PR kern/38664.
Welcome to 4.99.63
Reviewed by: YAMAMOTO Takashi <yamt@netbsd.org>
|
| 1.10 | 28-Apr-2008 |
martin | branches: 1.10.2;
Remove clause 3 and 4 from TNF licenses
|
| 1.9 | 02-Jan-2008 |
ad | branches: 1.9.6; 1.9.8; 1.9.10;
Merge vmlocking2 to head.
|
| 1.8 | 12-Dec-2007 |
he | Make this build again, as part of sys/lkm/dev/vnd/:
- lfs_truncate() has lost its lwp argument.
- Cast from void* to char* before doing pointer arithmetic.
|
| 1.7 | 12-Dec-2007 |
ad | Fix a stray brelse() that got missed.
|
| 1.6 | 12-Dec-2007 |
lukem | defflag LFS_KERNEL_RFW (in opt_lfs.h).
Note: lfs_rfw.c doesn't compile if you define the option; locking API fallout?
|
| 1.5 | 10-Oct-2007 |
ad | branches: 1.5.4; 1.5.6; 1.5.8; 1.5.10;
Merge from vmlocking:
- Split vnode::v_flag into three fields, depending on field locking.
- simple_lock -> kmutex in a few places.
- Fix some simple locking problems.
|
| 1.4 | 08-Oct-2007 |
ad | Merge ffs locking & brelse changes from the vmlocking branch.
|
| 1.3 | 01-Sep-2006 |
perseant | branches: 1.3.4; 1.3.10; 1.3.16; 1.3.30; 1.3.32; 1.3.34;
Changes to help the roll-forward agent, to wit:
* Mark being-deleted files in the Ifile so we can finish deleting them
at fs mount time.
* Flag the Ifile with "cleaner must clean" when writers are waiting for
the cleaner, rather than relying solely on the cleaner's estimation of
whether it should clean or not.
* Note partial segments written by a user agent (in particular,
fsck_lfs) so that repeated rolls forward don't interfere with one
another.
* Add a new fcntl, LFCNPASS, that allows the log to wrap exactly once,
for better testing of the validity of checkpoints.
* Keep track of the on-disk nlink count when cleaning, so that we don't
partially complete directory operations while cleaning.
* Ensure that every single Ifile inode write represents a consistent
view of the filesystem. In particular, the accounting for the segment
we are writing the inode into must be correct, and the accounting for
the segment that inode used to reside in must be correct. Rather than
just rewriting the inode if we wrote it wrong, rewrite the necessary
ifile blocks before writing the inode so we never write it wrong.
* Don't unmark any VDIROP vnodes if we haven't written them to disk,
avoiding yet another problem with the "wait for the cleaner" error
return from lfs_putpages().
Also, move the last callback to an aiodone call, so we no longer do any
memory management from interrupt context.
|
| 1.2 | 20-Jul-2006 |
perseant | branches: 1.2.4;
Oops, commit the correct version of lfs_rfw.c. The roll-forward functionality
is known not to work in this version (as it did not previously) but it should
at least compile.
|
| 1.1 | 20-Jul-2006 |
perseant | Separate the (non-working) LFS kernel roll-forward code into its own file,
lfs_rfw.c.
|
| 1.2.4.3 | 03-Sep-2006 |
yamt | sync with head.
|
| 1.2.4.2 | 11-Aug-2006 |
yamt | sync with head
|
| 1.2.4.1 | 20-Jul-2006 |
yamt | file lfs_rfw.c was added on branch yamt-pdpolicy on 2006-08-11 15:47:37 +0000
|
| 1.3.34.1 | 14-Oct-2007 |
yamt | sync with head.
|
| 1.3.32.2 | 09-Jan-2008 |
matt | sync with HEAD
|
| 1.3.32.1 | 06-Nov-2007 |
matt | sync with HEAD
|
| 1.3.30.1 | 26-Oct-2007 |
joerg | Sync with HEAD.
Follow the merge of pmap.c on i386 and amd64 and move
pmap_init_tmp_pgtbl into arch/x86/x86/pmap.c. Modify the ACPI wakeup
code to restore CR4 before jumping back into kernel space as the large
page option might cover that.
|
| 1.3.16.3 | 24-Aug-2007 |
ad | Sync with buffer cache locking changes. See buf.h/vfs_bio.c for details.
Some minor portions are incomplete and needs to be verified as a whole.
|
| 1.3.16.2 | 13-May-2007 |
ad | - Pass the error number and residual count to biodone(), and let it handle
setting error indicators. Prepare to eliminate B_ERROR.
- Add a flag argument to brelse() to be set into the buf's flags, instead
of doing it directly. Typically used to set B_INVAL.
- Add a "struct cpu_info *" argument to kthread_create(), to be used to
create bound threads. Change "bool mpsafe" to "int flags".
- Allow exit of LWPs in the IDL state when (l != curlwp).
- More locking fixes & conversion to the new API.
|
| 1.3.16.1 | 13-Mar-2007 |
ad | Pull in the initial set of changes for the vmlocking branch.
|
| 1.3.10.4 | 21-Jan-2008 |
yamt | sync with head
|
| 1.3.10.3 | 27-Oct-2007 |
yamt | sync with head.
|
| 1.3.10.2 | 30-Dec-2006 |
yamt | sync with head.
|
| 1.3.10.1 | 01-Sep-2006 |
yamt | file lfs_rfw.c was added on branch yamt-lazymbuf on 2006-12-30 20:51:01 +0000
|
| 1.3.4.2 | 09-Sep-2006 |
rpaulo | sync with head
|
| 1.3.4.1 | 01-Sep-2006 |
rpaulo | file lfs_rfw.c was added on branch rpaulo-netinet-merge-pcb on 2006-09-09 03:00:00 +0000
|
| 1.5.10.2 | 02-Jan-2008 |
bouyer | Sync with HEAD
|
| 1.5.10.1 | 13-Dec-2007 |
bouyer | Sync with HEAD
|
| 1.5.8.1 | 13-Dec-2007 |
yamt | sync with head.
|
| 1.5.6.5 | 28-Dec-2007 |
ad | Make it compile.
|
| 1.5.6.4 | 26-Dec-2007 |
ad | Sync with head.
|
| 1.5.6.3 | 19-Dec-2007 |
ad | Use a global lfs_lock.
|
| 1.5.6.2 | 19-Dec-2007 |
ad | Fix some more problems w/lfs on this branch.
|
| 1.5.6.1 | 04-Dec-2007 |
ad | Pull the vmlocking changes into a new branch.
|
| 1.5.4.1 | 18-Feb-2008 |
mjf | Sync with HEAD.
|
| 1.9.10.2 | 04-May-2009 |
yamt | sync with head.
|
| 1.9.10.1 | 16-May-2008 |
yamt | sync with head.
|
| 1.9.8.1 | 18-May-2008 |
yamt | sync with head.
|
| 1.9.6.1 | 02-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.10.2.1 | 23-Jun-2008 |
wrstuden | Sync w/ -current. 34 merge conflicts to follow.
|
| 1.11.12.1 | 13-May-2009 |
jym | Sync with HEAD.
Commit is split, to avoid a "too many arguments" protocol error.
|
| 1.11.6.1 | 03-Mar-2009 |
skrll | Sync with HEAD.
|
| 1.12.22.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.12.22.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.12.22.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.12.22.1 | 25-Feb-2013 |
tls | resync with head
|
| 1.12.12.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.12.12.1 | 23-Jan-2013 |
yamt | sync with head
|
| 1.17.2.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.18.6.4 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.18.6.3 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.18.6.2 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.18.6.1 | 06-Apr-2015 |
skrll | Sync with HEAD
|
| 1.32.18.2 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.32.18.1 | 10-Jun-2019 |
christos | Sync with HEAD
|
| 1.32.16.2 | 18-Jan-2019 |
pgoyette | Synch with HEAD
|
| 1.32.16.1 | 26-Dec-2018 |
pgoyette | Sync with HEAD, resolve a few conflicts
|
| 1.34.6.1 | 17-Jan-2020 |
ad | Sync with head.
|
| 1.295 | 29-Oct-2025 |
perseant | Use IINFOSIZE and LFS_BLKPTRSIZE, rather than sizeof(int32_t), to represent
the size of inode numbers and logical block numbers, respectively, in the
segment summary header. Prevents an overrun in LFS64.
|
| 1.294 | 20-Oct-2025 |
perseant | * Generalize the partial-segment parser introduced for roll-forward,
using it to facilitate an in-kernel segment rewriter (cleaner), and a
mechanism to check whether a segment is in fact empty (only used
with DEBUG).
* Add these new fcntl calls:
- LFCNFILESTATS: For each inode given, report its number of direct
blocks, how many gaps (discontinuities) there are between direct
blocks, and how large the total gap distance is. This will be
useful for a coalescing agent.
- LFCNREWRITEFILE: For each inode given, rewrite its direct blocks,
effectively coalescing it into as compact a form as possible.
- LFCNSCRAMBLE: As above, except that it only rewrites every other
block. This causes the file to have many gaps that can be
measured with LFCNFILESTATS and addressed with LFCNREWRITEFILE,
for testing purposes.
- LFCNREWRITESEGS: Rewrite any live data in the given segments.
This is intended to simplify the cleaner API and facilitate an
in-kernel cleaner.
- LFCNCLEANERINFO: Get the most current CLEANERINFO data from the
kernel.
- LFCNSEGUSE: Retrieve segment usage data from the kernel.
* Vnodes marked IN_CLEANING now take a reference. Add a new "cleaner
lock", which must be taken by the cleaner before the segment lock,
and before marking nodes IN_CLEANING. This allows us to flush
vnodes, if necessary, before the cleaning segment is written, and
never to flush vnodes being cleaned. When the cleaner lock is
released, the vnodes are cleared of IN_CLEANING and the reference
dropped.
* Track a potential infinite loop in lfs_gatherblock.
* Pull "needs to flush" and "needs to wait for flush" into functions
instead of inlining their definitions.
|
| 1.293 | 17-Sep-2025 |
perseant | Add working in-kernel roll forward.
|
| 1.292 | 17-Sep-2025 |
perseant | Use a workqueue to handle the superblock callback.
|
| 1.291 | 17-Sep-2025 |
perseant | Add routines to check freelist consistency if compiled with DEBUG and
conditional on a kernel variable manipulated via sysctl.
Add checks before and after each routine that modifies the free list.
#if 0 a section of lfs_vfree() that was intended to keep the free list ordered
but instead corrupted it.
|
| 1.290 | 04-Sep-2025 |
perseant | Copy the flags from a full partial segment to its continuation, if
a continuation is necessary, so that partial-segment collections marked
with SS_DIROP|SS_CONT are properly completed wiht a partial-segment marked
SS_DIROP (without SS_CONT). Necessary for roll-forward.
|
| 1.289 | 02-Sep-2025 |
perseant | Use a workqueue to handle cluster iodone, rather than doing it in interrupt context.
|
| 1.288 | 05-Sep-2020 |
riastradh | Round of uvm.h cleanup.
The poorly named uvm.h is generally supposed to be for uvm-internal
users only.
- Narrow it to files that actually need it -- mostly files that need
to query whether curlwp is the pagedaemon, which should maybe be
exposed by an external header.
- Use uvm_extern.h where feasible and uvm_*.h for things not exposed
by it. We should split up uvm_extern.h but this will serve for now
to reduce the uvm.h dependencies.
- Use uvm_stat.h and #ifdef UVMHIST uvm.h for files that use
UVMHIST(ubchist), since ubchist is declared in uvm.h but the
reference evaporates if UVMHIST is not defined, so we reduce header
file dependencies.
- Make uvm_device.h and uvm_swap.h independently includable while
here.
ok chs@
|
| 1.287 | 13-Aug-2020 |
riastradh | Skip unlinked inodes.
They no longer matter on disk so we don't need to write anything out
for them.
|
| 1.286 | 23-Feb-2020 |
ad | UVM locking changes, proposed on tech-kern:
- Change the lock on uvm_object, vm_amap and vm_anon to be a RW lock.
- Break v_interlock and vmobjlock apart. v_interlock remains a mutex.
- Do partial PV list locking in the x86 pmap. Others to follow later.
|
| 1.285 | 23-Feb-2020 |
riastradh | Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
|
| 1.284 | 23-Feb-2020 |
riastradh | Change some cheap KDASSERT into KASSERT.
|
| 1.283 | 22-Feb-2020 |
ad | Make LFS/rump play nice with aiodoned removal.
PR kern/55004 (Hundreds of file system tests now fail on real hardware)
|
| 1.282 | 18-Feb-2020 |
chs | remove the aiodoned thread. I originally added this to provide a thread context
for doing page cache iodone work, but since then biodone() has changed to
hand off all iodone work to a softint thread, so we no longer need the
special-purpose aiodoned thread.
|
| 1.281 | 15-Jan-2020 |
ad | Merge from yamt-pagecache (after much testing):
- Reduce unnecessary page scan in putpages esp. when an object has a ton of
pages cached but only a few of them are dirty.
- Reduce the number of pmap operations by tracking page dirtiness more
precisely in uvm layer.
|
| 1.280 | 08-Dec-2019 |
ad | branches: 1.280.2;
Revert previous. No performance gain worth the potential headaches
with buffers in these contexts.
|
| 1.279 | 08-Dec-2019 |
ad | Avoid thundering herd: cv_broadcast(&bp->b_busy) -> cv_signal(&bp->b_busy)
|
| 1.278 | 03-Sep-2018 |
riastradh | branches: 1.278.4;
Rename min/max -> uimin/uimax for better honesty.
These functions are defined on unsigned int. The generic name
min/max should not silently truncate to 32 bits on 64-bit systems.
This is purely a name change -- no functional change intended.
HOWEVER! Some subsystems have
#define min(a, b) ((a) < (b) ? (a) : (b))
#define max(a, b) ((a) > (b) ? (a) : (b))
even though our standard name for that is MIN/MAX. Although these
may invite multiple evaluation bugs, these do _not_ cause integer
truncation.
To avoid `fixing' these cases, I first changed the name in libkern,
and then compile-tested every file where min/max occurred in order to
confirm that it failed -- and thus confirm that nothing shadowed
min/max -- before changing it.
I have left a handful of bootloaders that are too annoying to
compile-test, and some dead code:
cobalt ews4800mips hp300 hppa ia64 luna68k vax
acorn32/if_ie.c (not included in any kernels)
macppc/if_gm.c (superseded by gem(4))
It should be easy to fix the fallout once identified -- this way of
doing things fails safe, and the goal here, after all, is to _avoid_
silent integer truncations, not introduce them.
Maybe one day we can reintroduce min/max as type-generic things that
never silently truncate. But we should avoid doing that for a while,
so that existing code has a chance to be detected by the compiler for
conversion to uimin/uimax without changing the semantics until we can
properly audit it all. (Who knows, maybe in some cases integer
truncation is actually intended!)
|
| 1.277 | 09-Jun-2018 |
zafer | branches: 1.277.2;
Add missing b_cflags and b_oflags.
Ok dholland@
Addresses PR kern/42342 by Yoshihiro Nakajima
|
| 1.276 | 06-Jun-2018 |
maya | Remove duplicate ;
|
| 1.275 | 20-Aug-2017 |
maya | branches: 1.275.2;
XXX question our double-flushing of dirops
|
| 1.274 | 26-Jul-2017 |
maya | change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
|
| 1.273 | 26-Jul-2017 |
maya | Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
|
| 1.272 | 15-Jun-2017 |
maya | It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
|
| 1.271 | 12-Jun-2017 |
maya | Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
|
| 1.270 | 10-Jun-2017 |
maya | Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.269 | 06-Apr-2017 |
maya | branches: 1.269.6;
don't guard lfs_sbactive or lfs_log with splbio, lfs_lock is plenty.
|
| 1.268 | 06-Apr-2017 |
maya | remove deprecated comment (and move it below assert)
there's no spl dance for I/O here.
|
| 1.267 | 06-Apr-2017 |
maya | Provide a LFS_ENTER_LOG (__nothing) in the !DEBUG case.
so I can drop lots of #ifdef DEBUG around this macro. NFCI
|
| 1.266 | 06-Apr-2017 |
maya | Drop single use macro LFS_BCLEAN_LOG with an inlined implementation.
LFS_ENTER_LOG currently macro grabs lfs_lock, so I'd like to have just one
name for it.
|
| 1.265 | 01-Apr-2017 |
riastradh | KASSERT(mutex_owned(vp->v_interlock)) in vnode iterator selector.
|
| 1.264 | 13-Mar-2017 |
riastradh | #if DIAGNOSTIC panic ---> KASSERT
Replace some #if DEBUG by this too. DEBUG is only for expensive
assertions; these are not.
|
| 1.263 | 19-Oct-2015 |
dholland | branches: 1.263.2; 1.263.4;
improve some panic messages
|
| 1.262 | 10-Oct-2015 |
dholland | Fix minor bitrot in #if 0 or otherwise disabled code.
|
| 1.261 | 10-Oct-2015 |
dholland | Use accessors for some more indirect block manipulations.
|
| 1.260 | 03-Oct-2015 |
dholland | Use IINFO in lfs_writeinode().
(both the kernel and the userland copies)
|
| 1.259 | 01-Sep-2015 |
dholland | Use the lfs dinode accessors in place of the ufs-derived ones.
(Mostly.)
The ufs-derived ones are fake structure member macros, which are gross
and not very safe. Also, it seems that a lot of places in the lfs code
were using the ffsv1 branch of them unconditionally, and this way it's
guaranteed all those places have been updated.
Found while doing this: for non-devices, have getattr produce NODEV
in the rdev field instead of leaking the address of the first direct
block.
|
| 1.258 | 21-Aug-2015 |
hannken | lfs_writevnodes: replace mnt_vnodelist traversal with vfs_vnode_iterator.
|
| 1.257 | 19-Aug-2015 |
dholland | Part two of dinodes; use the same union everywhere.
(previously the ufs-derived code had things set up slightly different)
Remove a bunch of associated mess.
|
| 1.256 | 12-Aug-2015 |
dholland | Hack up dinode usage to be 64 vs. 32 as needed. Part 1.
(This part changes the native lfs code; the ufs-derived code already
has 64 vs. 32 logic, but as aspects of it are unsafe, and don't
entirely interoperate cleanly with the lfs 64/32 stuff, pass 2 will be
rehashing that.)
|
| 1.255 | 12-Aug-2015 |
dholland | Provide 32-bit and 64-bit versions of FINFO.
This also entailed sorting out part of struct segment, as that
contains a pointer into the current FINFO data.
|
| 1.254 | 12-Aug-2015 |
dholland | Make 32-bit and 64-bit versions of SEGSUM.
Also fix some of the FINFO handling as it's closely entangled.
|
| 1.253 | 12-Aug-2015 |
dholland | Add IFILE32 and IFILE64 structures for the on-disk ifile entries.
Add and use accessors. There are also a bunch of places that cast and
I hope I've found them all...
|
| 1.252 | 12-Aug-2015 |
dholland | Make 32-bit and 64-bit versions of CLEANERINFO.
XXX: while this is written to disk, it seems like much of it would
XXX: be better set up as a commpage shared with the cleaner.
|
| 1.251 | 02-Aug-2015 |
dholland | Pass the fs object to LFS_MAX_DADDR so it can check lfs_is64.
Remove some hackish intentional 64->32 truncations next to the checks
using LFS_MAX_DADDR, and tackle the problem they handled in bmap
instead.
The problem: the magic block pointer value UNWRITTEN has magic value
-2, and if it's not handled specifically, uint32 -> uint64 promotion
turns it into 4294967294, which then causes consternation and
monkeyhouse downstream.
What's here is still kind of a hack, but it's a step forward.
|
| 1.250 | 02-Aug-2015 |
dholland | Add a (draft) 64-bit superblock. Make things build again.
Add pieces of support for using both superblock types where
convenient, and specifically to the superblock accessors, but don't
actually enable it anywhere.
First substantive step on PR 50000.
|
| 1.249 | 02-Aug-2015 |
dholland | Use accessor functions for the version field of the lfs superblock.
I thought at first maybe the cases that test the version should be
rolled into the accessors, but on the whole I think the conclusion on
that is no.
|
| 1.248 | 02-Aug-2015 |
dholland | Make i_eff_nblks in the in-memory inode 64 bits wide.
|
| 1.247 | 02-Aug-2015 |
dholland | Fix catastrophic bug in lfs_rewind() that changed segment numbers
(lfs_curseg/lfs_nextseg in the superblock) using the wrong units.
These fields are for whatever reason the start addresses of segments
(measured in frags) rather than the segment numbers 0..n.
This only apparently affects dumping from a mounted fs; however, it
trashes the fs.
I would really, really like to have a static analysis tool that can
keep track of the units things are measured in, since fs code is full
of conversion macros and the macros are named inscrutable things like
"sntod" whose letters don't necessarily even correspond to the units
they convert. It is surprising that more of these are not wrong.
|
| 1.246 | 02-Aug-2015 |
dholland | Second batch of 64 -> 32 truncations in lfs, along with more minor
tidyups and corrections in passing.
|
| 1.245 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.244 | 25-Jul-2015 |
martin | Use accessors in DEBUG and DIAGNOSTIC code as well
|
| 1.243 | 24-Jul-2015 |
dholland | More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)
The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.
|
| 1.242 | 24-Jul-2015 |
dholland | Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)
It also gets rid of cpp abuse in the form of fake structure member
macros.
Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)
XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.
|
| 1.241 | 07-Jun-2015 |
hannken | Fix copy and paste errors from last commits.
- Kernel i386/ALL and amd64/ALL compile again.
- Resolves CID 1304138 (DEADCODE) and 1304139 (IDENTICAL_BRANCHES).
|
| 1.240 | 31-May-2015 |
hannken | Change lfs from hash table to vcache.
- Change lfs_valloc() to return an inode number and version instead of
a vnode and move lfs_ialloc() and lfs_vcreate() to new lfs_init_vnode().
- Add lfs_valloc_fixed() to allocate a known inode, used by kernel
roll forward.
- Remove lfs_*ref(), these functions cannot coexist with vcache and
their commented behaviour is far away from their implementation.
- Add the cleaner lwp and blockinfo to struct ulfsmount so lfs_loadvnode()
may use hints from the cleaner.
- Remove vnode locks from ulfs_lookup() like we did with ufs_lookup().
|
| 1.239 | 31-May-2015 |
hannken | Use VFS_PROTOS() for lfs.
Rename conflicting struct lfs field "lfs_start" to "lfs_s0addr".
No functional change.
|
| 1.238 | 20-Apr-2015 |
riastradh | Make vget always return vnode unlocked.
Convert callers who want locks to use vn_lock afterward.
Add extra argument so the compiler will report stragglers.
|
| 1.237 | 28-Mar-2015 |
maxv | Remove the 'cred' argument from bread(). Remove a now unused var in
ffs_snapshot.c. Update the man page accordingly.
ok hannken@
|
| 1.236 | 24-Mar-2014 |
hannken | branches: 1.236.4; 1.236.6;
- Make VI_XLOCK, VI_CLEAN and VI_LOCKSHARE private to kern/vfs_*.c.
- Make vwait() static.
- Add vdead_check() to check a vnode for being or becoming dead.
Discussed on tech-kern.
Welcome to 6.99.38
|
| 1.235 | 18-Mar-2014 |
hannken | Operations vmark(), vunmark() and vismarker() have been replaced by
vfs_vnode_iterator_*(), remove them.
Document vfs_vnode_iterator_*().
Make VI_MARKER private to vfs_vnode.c, vfs_mount.c and unfortunately
to ufs/lfs/lfs_segment.c.
Welcome to 6.99.37
|
| 1.234 | 17-Mar-2014 |
hannken | Change vismarker() to VI_MARKER for lfs_writevnodes().
This operation has to be changed to vfs_vnode_iterator.
|
| 1.233 | 29-Oct-2013 |
hannken | Vnode API cleanup pass 1.
- Make these defines and functions private to vfs_vnode.c:
VC_MASK, VC_LOCK, DOCLOSE, VI_IANCTREDO and VI_INACTNOW
vclean() and vrelel()
- Remove the long time unused lwp argument from vrecycle().
- Remove vtryget(), it is responsible for ugly hacks and doesn't
look that effective.
Presented on tech-kern.
Welcome to 6.99.25
|
| 1.232 | 17-Oct-2013 |
christos | - remove unused variables
- add debug ifdefs for debugging variables
- __USE() where appropriate.
|
| 1.231 | 28-Jul-2013 |
dholland | Add lfs_kernel.h for declarations that don't need to be exposed to userland.
lfs currently has the following headers:
lfs.h - on-disk structures and stuff needed for userlevel tools
lfs_inode.h - additional restricted materials for userlevel tools
that operate the fs (newfs_lfs, fsck_lfs, lfs_cleanerd)
lfs_kernel.h - stuff needed only in the kernel
and the following legacy headers that are expected to be mopped up and
folded into one of the above:
lfs_extern.h - function prototypes
ulfs_bswap.h - endian-independent support
ulfs_dinode.h - now contains very little
ulfs_dirhash.h - dirhash support
ulfs_extattr.h - extattr support
ulfs_extern.h - more function prototypes
ulfs_inode.h - assorted kernel-only declarations
ulfs_quota.h - quota support
ulfs_quota1.h - more quota support
ulfs_quota2.h - more quota support
ulfs_quotacommon.h - more quota support
ulfsmount.h - legacy copy of ufsmount material
|
| 1.230 | 18-Jun-2013 |
christos | branches: 1.230.2;
Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.
|
| 1.229 | 08-Jun-2013 |
dholland | ulfs_dir.h has been emptied; remove it.
|
| 1.228 | 08-Jun-2013 |
dholland | Stick LFS_ in front of IFMT, IFIFO, IFREG, etc. so as not to conflict
with the UFS copies of these symbols. (Which themselves ought to have
UFS_ stuck on.)
|
| 1.227 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.226 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.225 | 22-Jan-2013 |
dholland | Stuff UFS_ in front of a few of ufs's symbols to reduce namespace
pollution. Specifically:
ROOTINO -> UFS_ROOTINO
WINO -> UFS_WINO
NXADDR -> UFS_NXADDR
NDADDR -> UFS_NDADDR
NIADDR -> UFS_NIADDR
MAXSYMLINKLEN -> UFS_MAXSYMLINKLEN
MAXSYMLINKLEN_UFS[12] -> UFS[12]_MAXSYMLINKLEN (for consistency)
Sort out ext2fs's misuse of NDADDR and NIADDR; fortunately, these have
the same values in ext2fs and ffs.
No functional change intended.
|
| 1.224 | 16-Feb-2012 |
perseant | branches: 1.224.2;
Pass t_renamerace and t_rmdirrace tests.
Adapt dholland@'s fix to ufs_rename to fix PR kern/43582. Address several
other MP locking issues discovered during the course of investigating the
same problem.
Removed extraneous vn_lock() calls on the Ifile, since the Ifile writes
are controlled by the segment lock.
Fix PR kern/45982 by deemphasizing the estimate of how much metadata
will fill the empty space on disk when the disk is nearly empty
(t_renamerace crates a lot of inode blocks on a tiny empty disk).
|
| 1.223 | 02-Jan-2012 |
perseant | branches: 1.223.2;
* Remove PGO_RECLAIM during lfs_putpages()' call to genfs_putpages(),
to avoid a live lock in the latter when reclaiming a vnode with
dirty pages.
* Add a new segment flag, SEGM_RECLAIM, to note when a segment is
being written for vnode reclamation, and record which inode is being
reclaimed, to aid in forensic debugging.
* Add a new segment flag, SEGM_SINGLE, so that opportunistic writes
can write a single segment's worth of blocks and then stop, rather
than writing all the way up to the cleaner's reserved number of
segments.
* Add assert statements to check mutex ownership is the way it ought
to be, mostly in lfs_putpages; fix problems uncovered by this.
* Don't clear VU_DIROP until the inode actually makes its way to disk,
avoiding a problem where dirop inodes could become separated
(uncovered by a modified version of the "ckckp" forensic regression
test).
* Move the vfs_getopsbyname() call into lfs_writerd. Prepare code to
make lfs_writerd notice when there are no more LFSs, and exit losing
the reference, so that, in theory, the module can be unloaded. This
code is not enabled, since it causes a crash on exit.
* Set IN_MODIFIED on inodes flushed by lfs_flush_dirops. Really we
only need to set IN_MODIFIED if we are going to write them again
(e.g., to write pages); need to think about this more.
Finally, several changes to help avoid "no clean segments" panics:
* In lfs_bmapv, note when a vnode is loaded only to discover whether
its blocks are live, so it can immediately be recycled. Since the
cleaner will try to choose ~empty segments over full ones, this
prevents the cleaner from (1) filling the vnode cache with junk, and
(2) squeezing any unwritten writes to disk and running the fs out of
segments.
* Overestimate by half the amount of metadata that will be required
to fill the clean segments. This will make the disk appear smaller,
but should help avoid a "no clean segments" panic.
* Rearrange lfs_writerd. In particular, lfs_writerd now pays
attention to the number of clean segments available, and holds off
writing until there is room.
|
| 1.222 | 11-Jul-2011 |
hannken | branches: 1.222.2; 1.222.6;
Change VOP_BWRITE() to take a vnode as its first argument like all other
VOPs do. Layered file systems no longer have to modify bp->b_vp and run
into trouble when an async VOP_BWRITE() uses the wrong vnode.
- change all occurences of VOP_BWRITE(bp) to VOP_BWRITE(bp->b_vp, bp).
- remove layer_bwrite().
- welcome to 5.99.55
Adresses PR kern/38762 panic: vwakeup: neg numoutput
No objections from tech-kern@.
|
| 1.221 | 12-Jun-2011 |
rmind | Welcome to 5.99.53! Merge rmind-uvmplock branch:
- Reorganize locking in UVM and provide extra serialisation for pmap(9).
New lock order: [vmpage-owner-lock] -> pmap-lock.
- Simplify locking in some pmap(9) modules by removing P->V locking.
- Use lock object on vmobjlock (and thus vnode_t::v_interlock) to share
the locks amongst UVM objects where necessary (tmpfs, layerfs, unionfs).
- Rewrite and optimise x86 TLB shootdown code, make it simpler and cleaner.
Add TLBSTATS option for x86 to collect statistics about TLB shootdowns.
- Unify /dev/mem et al in MI code and provide required locking (removes
kernel-lock on some ports). Also, avoid cache-aliasing issues.
Thanks to Andrew Doran and Joerg Sonnenberger, as their initial patches
formed the core changes of this branch.
|
| 1.220 | 03-Apr-2011 |
rmind | branches: 1.220.2;
- Use offsetof() in VOPARG_OFFSETOF() instead of re-implementing it.
- Remove VDESC_NOMAP_VPP and VDESC_VPP_WILLRELE.
- Remove VRELEL_NOINACTIVE and VRELEL_ONHEAD.
|
| 1.219 | 02-Apr-2011 |
rmind | Split off parts of vfs_subr.c into vfs_vnode.c and vfs_mount.c modules.
No functional change. Discussed on tech-kern@.
|
| 1.218 | 23-Mar-2011 |
rmind | G/C count_lock_queue (unused for 12 years)
|
| 1.217 | 21-Jul-2010 |
hannken | branches: 1.217.2;
Make holding v_interlock mandatory for callers of vget().
Announced some time ago on tech-kern.
|
| 1.216 | 24-Jun-2010 |
hannken | Clean up vnode lock operations pass 2:
VOP_UNLOCK(vp, flags) -> VOP_UNLOCK(vp): Remove the unneeded flags argument.
Welcome to 5.99.32.
Discussed on tech-kern.
|
| 1.215 | 16-Feb-2010 |
mlelstv | branches: 1.215.2;
Three changes in a single commit.
- drop the notion of frags (LFS fragments) vs fsb (FFS fragments)
The code uses a complicated unity function that just makes the
code difficult to understand.
- support larger sector sizes. Fix disk address computations
to use DEV_BSIZE in the kernel as required by device drivers
and to use sector sizes in userland.
- Fix several locking bugs in lfs_bio.c and lfs_subr.c.
|
| 1.214 | 07-Aug-2009 |
wiz | branches: 1.214.2;
Add missing parenthesis in #ifdef LFS_USE_B_INVAL.
From Henning Petersen in PR 41841.
|
| 1.213 | 02-Jun-2008 |
ad | branches: 1.213.8; 1.213.18; 1.213.22;
Use atomics to maintain v_usecount.
|
| 1.212 | 16-May-2008 |
hannken | Make sure all cached buffers with valid, not yet written data have been
run through copy-on-write. Call fscow_run() with valid data where possible.
The LP_UFSCOW hack is no longer needed to protect ffs_copyonwrite() against
endless recursion.
- Add a flag B_MODIFY to bread(), breada() and breadn(). If set the caller
intends to modify the buffer returned.
- Always run copy-on-write on buffers returned from ffs_balloc().
- Add new function ffs_getblk() that gets a buffer, assigns a new blkno,
may clear the buffer and runs copy-on-write. Process possible errors
from getblk() or fscow_run(). Part of PR kern/38664.
Welcome to 4.99.63
Reviewed by: YAMAMOTO Takashi <yamt@netbsd.org>
|
| 1.211 | 28-Apr-2008 |
martin | branches: 1.211.2;
Remove clause 3 and 4 from TNF licenses
|
| 1.210 | 27-Mar-2008 |
ad | branches: 1.210.2; 1.210.4;
Make rusage collection per-LWP and collate in the appropriate places.
cloned threads need a little bit more work but the locking needs to
be fixed first.
|
| 1.209 | 15-Feb-2008 |
ad | branches: 1.209.6;
The buffer LOCKED flag need not be under the protection of bufcache_lock,
BUSY is enough.
|
| 1.208 | 27-Jan-2008 |
pooka | Replace vrelel() 010101-mania with a flags parameter. However,
leave flags unimplemented for a while (no change in functionality).
|
| 1.207 | 02-Jan-2008 |
ad | Merge vmlocking2 to head.
|
| 1.206 | 10-Oct-2007 |
ad | branches: 1.206.4; 1.206.6; 1.206.10;
Merge from vmlocking:
- Split vnode::v_flag into three fields, depending on field locking.
- simple_lock -> kmutex in a few places.
- Fix some simple locking problems.
|
| 1.205 | 08-Oct-2007 |
ad | Merge ffs locking & brelse changes from the vmlocking branch.
|
| 1.204 | 09-Aug-2007 |
pooka | branches: 1.204.2; 1.204.4;
Instead of having lfs muck directly about with vnode free lists,
introduce vrele2(), which allows to release vnodes the way lfs
sometimes wants it:
+ without calling inactive
+ inserting the vnode at the head of the freelist (this is a very
questionable optimization that isn't even enabled by default,
but I went along with the same semantics for now)
|
| 1.203 | 29-Jul-2007 |
ad | branches: 1.203.4; 1.203.6;
It's not a good idea for device drivers to modify b_flags, as they don't
need to understand the locking around that field. Instead of setting
B_ERROR, set b_error instead. b_error is 'owned' by whoever completes
the I/O request.
|
| 1.202 | 12-Jul-2007 |
rmind | branches: 1.202.2;
Implementation of per-CPU work-queues support for workqueue(9) interface.
WQ_PERCPU flag for workqueue and additional argument for workqueue_enqueue()
to assign a CPU might be used. Notes:
- For now, the list is used for workqueue_queue, which is non-optimal,
and will be changed with array, where index would be CPU ID.
- The data structures should be changed to be cache-friendly.
Reviewed by: <yamt>, <tech-kern>
|
| 1.201 | 30-Jun-2007 |
pooka | Using POOL_INIT here makes no sense, since file systems always have
an init method. So get rid of it and #ifdef _LKM and just always
init in the init method. Give malloc types the same treatment.
Makes file systems nicer to work with in linksetless environments
and fixes a few LKM discrepancies.
|
| 1.200 | 16-May-2007 |
perseant | Change references to SEGM_W_DIROPS to SEGM_CKP, and replace the logic that
formerly used SEGM_W_DIROPS in lfs_segwrite() appropriately. This prevents
a problem in which processes could get stuck in "buffers" sleep forever.
|
| 1.199 | 17-Apr-2007 |
perseant | Install a new sysctl, vfs.lfs.ignore_lazy_sync, which causes LFS to ignore
the "smooth" syncer, as if vfs.sync.*delay = 0, but only for LFS. The
default is "on", i.e., ignore lazy sync.
Reduce the amount of polling/busy-waiting done by lfs_putpages(). To
accomplish this, copied genfs_putpages() and modified it to indicate which
page it was that caused it to return with EDEADLK. fsync()/fdatasync()
should no longer ever fail with EAGAIN, and should not consume huge
quantities of cpu.
Also, try to make dirops less likely to be written as the result of a
VOP_PUTPAGES(), while ensuring that they are written regularly.
|
| 1.198 | 04-Mar-2007 |
christos | branches: 1.198.2; 1.198.4;
Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.
|
| 1.197 | 23-Feb-2007 |
perseant | Reverse the order of searching the vnode list in lfs_writevnodes(). This
should speed up e.g. "chown -R" on LFS filesystems; e.g. it shows a 100%
increase in the 'seq_stat' column of bonnie++.
|
| 1.196 | 21-Dec-2006 |
yamt | branches: 1.196.2;
merge yamt-splraiseipl branch.
- finish implementing splraiseipl (and makeiplcookie).
http://mail-index.NetBSD.org/tech-kern/2006/07/01/0000.html
- complete workqueue(9) and fix its ipl problem, which is reported
to cause audio skipping.
- fix netbt (at least compilation problems) for some ports.
- fix PR/33218.
|
| 1.195 | 16-Nov-2006 |
christos | branches: 1.195.2; 1.195.4;
__unused removal on arguments; approved by core.
|
| 1.194 | 20-Oct-2006 |
reinoud | Replace the LIST structure mp->mnt_vnodelist to a TAILQ structure since all
vnodes were synced and processed backwards. This meant that the last
accessed node was processed first and the earlierst last.
An extra benefit is the removal of the ugly hack from the Berkly days on
LFS.
In the proces, i've also replaced the various variations hand written loops
by the TAILQ_FOREACH() macro's.
|
| 1.193 | 12-Oct-2006 |
christos | - sprinkle __unused on function decls.
- fix a couple of unused bugs
- no more -Wno-unused for i386
|
| 1.192 | 04-Oct-2006 |
christos | fix empty if
|
| 1.191 | 28-Sep-2006 |
perseant | Use lockstatus instead of a homebrewed locking system to control
LFCNWRAPSTOP and LFCNWRAPGO.
Be less verbose about the various looping checks: use log() rather than
printf(), and only log anything if we are really looping ("count = 2" is
not an error condition).
Allow dirops sleeping on available space to be interruptible.
|
| 1.190 | 02-Sep-2006 |
christos | branches: 1.190.2; 1.190.4;
remove impossible test
|
| 1.189 | 01-Sep-2006 |
perseant | Changes to help the roll-forward agent, to wit:
* Mark being-deleted files in the Ifile so we can finish deleting them
at fs mount time.
* Flag the Ifile with "cleaner must clean" when writers are waiting for
the cleaner, rather than relying solely on the cleaner's estimation of
whether it should clean or not.
* Note partial segments written by a user agent (in particular,
fsck_lfs) so that repeated rolls forward don't interfere with one
another.
* Add a new fcntl, LFCNPASS, that allows the log to wrap exactly once,
for better testing of the validity of checkpoints.
* Keep track of the on-disk nlink count when cleaning, so that we don't
partially complete directory operations while cleaning.
* Ensure that every single Ifile inode write represents a consistent
view of the filesystem. In particular, the accounting for the segment
we are writing the inode into must be correct, and the accounting for
the segment that inode used to reside in must be correct. Rather than
just rewriting the inode if we wrote it wrong, rewrite the necessary
ifile blocks before writing the inode so we never write it wrong.
* Don't unmark any VDIROP vnodes if we haven't written them to disk,
avoiding yet another problem with the "wait for the cleaner" error
return from lfs_putpages().
Also, move the last callback to an aiodone call, so we no longer do any
memory management from interrupt context.
|
| 1.188 | 20-Jul-2006 |
perseant | Note partial segments that are written by the cleaner, to help out the
roll-forward agent.
|
| 1.187 | 20-Jul-2006 |
perseant | Loop on the check for lfs_nowrap, so we don't allow a process to squeeze by.
|
| 1.186 | 20-Jul-2006 |
perseant | Don't try to write all the vnodes, when the cleaner needs a vnode to be
recycled.
|
| 1.185 | 29-Jun-2006 |
perseant | Don't wake up the cleaner if the filesystem is unwrappable, and fix the
compatibility fcntls.
Also includes one-line fixes for an MP locking bug and a zero-length FINFO
problem that manifested during testing.
|
| 1.184 | 24-Jun-2006 |
perseant | Change LFCNWRAP{STOP,GO} to make them more suitable for snapshotting; in
particular, the caller can now choose whether to wait for the condition
to be met, and if the caller of LFCNWRAPSTOP dies or otherwise closes
the descriptor, the filesystem is started again. Updated the ckckp
regression test to use the new semantics.
dump_lfs(8) now uses the fcntls to implement LFS-style snapshotting through
the -X flag, addressing PR#33457 albeit not using fss(4). Fixed a couple
other problems with dump_lfs that manifested themselves during testing.
|
| 1.183 | 23-Jun-2006 |
yamt | fix a simonb-timecounters regression.
the precision of getnanotime() is not suitable for file timestamps.
esp. when it's nfs-exported.
- introduce vfs_timestamp().
(the name is from freebsd. currently merely a wrapper of nanotime())
- for ufs-like filesystems, use it rather than getnanotime().
XXX check other filesystems.
|
| 1.182 | 07-Jun-2006 |
kardel | branches: 1.182.2;
merge FreeBSD timecounters from branch simonb-timecounters
- struct timeval time is gone
time.tv_sec -> time_second
- struct timeval mono_time is gone
mono_time.tv_sec -> time_uptime
- access to time via
{get,}{micro,nano,bin}time()
get* versions are fast but less precise
- support NTP nanokernel implementation (NTP API 4)
- further reading:
Timecounter Paper: http://phk.freebsd.dk/pubs/timecounter.pdf
NTP Nanokernel: http://www.eecis.udel.edu/~mills/ntp/html/kern.html
|
| 1.181 | 20-May-2006 |
perseant | Fix a bug in which FINFOs were written with a version number of zero.
Add assertions and add this to the DEBUG fip test in lfs_writeseg.
|
| 1.180 | 18-May-2006 |
perseant | branches: 1.180.2;
Break out the finfo array manipulation code into two new functions,
lfs_acquire_finfo() and lfs_release_finfo(). Add a debugging check
for zero-length finfo arrays in the segment summary to avoid future
regressions.
|
| 1.179 | 14-May-2006 |
elad | integrate kauth.
|
| 1.178 | 12-May-2006 |
perseant | Fixes to address the "vinvalbuf: dirty blocks" panic that can occur when
many inodes are cleaned at once. Make sure that we write all the pages
on vnodes that are being flushed, even if we don't think there's room;
drain v_numoutput before lfs_vflush() completes.
Also, don't allow a vnode that is in the process of being cleaned to be
chosen by getnewvnode(); this avoids a segment accounting panic in the case
that a large number of inodes are fed to lfs_markv() all at once.
|
| 1.177 | 01-May-2006 |
perseant | Don't ever partially write dirops, even if we need the cleaner to run.
This increases the chances of the "no clean segments" panic slightly,
but allows us to run the ckckp regression test successfully to completion.
|
| 1.176 | 30-Apr-2006 |
perseant | Postpone the segment accounting changes coming from truncation until the
inode that makes those changes valid is either written to disk by
lfs_writeinode() or discarded by lfs_vfree().
A couple of locking fixes are also included as well.
|
| 1.175 | 22-Apr-2006 |
perseant | Regression test improvements:
Move the stop for LFCNWRAPSTOP to the point at which writing at segment 0
is really about to commence, since this is what the test expects (and
incidentally what a snapshotting utility wants as well).
More correctly reconstruct the on-disk state at every checkpoint, rather
than relying on the entire state at the point of wrapping to be accurate
(that is only true the first time we wrap). Add a "make abort" target to
make rerunning the test more convenient when it has failed and we're done
analyzing the failure.
|
| 1.174 | 17-Apr-2006 |
perseant | Introduce two fcntl calls that freeze the filesystem right at the point
where segment 0 is being considered for writing. This allows for automated
checkpoint vailidity scanning, and could be used (in conjunction with the
existing LFCNREWIND) for e.g. snapshot dumps as well.
Include a regression test that does such scanning.
When writing the Ifile, loop through the dirty block list three times to
make sure that the checkpoint is always consistent (the first and second
times the Ifile blocks can cross a segment boundary; not so the third time
unless the segments are very small). Discovered by using the aforementioned
regression test.
|
| 1.173 | 13-Apr-2006 |
perseant | Make lfs_vref/lfs_vunref not need to know about VXLOCK and VFREEING
explicitly (especially since we didn't know about VFREEING at all before),
but notice the EBUSY return from vget() instead.
Fix some more MP locking protocol issues, most of which were pointed out by
Christian Ehrhardt this morning on tech-kern.
|
| 1.172 | 07-Apr-2006 |
perseant | Several minor bug fixes:
* Correct (weak) segment lock assertions in lfs_fragextend and lfs_putpages.
* Keep IN_MODIFIED set if we run out of avail in lfs_putpages.
* Don't try to (re)write buffers on a VBLK vnode; fixes a panic I found
while running with an LFS root.
* Raise priority of LFCNSEGWAIT to PVFS; PUSER is way too low for
something the pagedaemon is relying on.
|
| 1.171 | 24-Mar-2006 |
perseant | Improvements to LFS's paging mechanism, to wit:
* Acknowledge that sometimes there are more dirty pages to be written to
disk than clean segments. When we reach the danger line,
lfs_gop_write() now returns EAGAIN. The caller of VOP_PUTPAGES(), if
it holds the segment lock, drops it and waits for the cleaner to make
room before continuing.
* Note and avoid a three-way deadlock in lfs_putpages (a writer holding
a page busy blocks on the cleaner while the cleaner blocks on the
segment lock while lfs_putpages blocks on the page).
|
| 1.170 | 17-Mar-2006 |
tls | From Konrad Schroeder, in response to strange df output on anoncvs.netbsd.org:
We were returning the wrong value for free space. Now we're not.
|
| 1.169 | 04-Jan-2006 |
yamt | branches: 1.169.2; 1.169.4; 1.169.6; 1.169.8; 1.169.10;
- add simple functions to allocate/free a buffer for i/o.
- make bufpool static.
|
| 1.168 | 11-Dec-2005 |
christos | branches: 1.168.2;
merge ktrace-lwp.
|
| 1.167 | 26-Sep-2005 |
yamt | always use nanotime rather than time.
it's bad to mix nanotime and time because it sometimes
make timestamps go backwards.
|
| 1.166 | 12-Sep-2005 |
christos | Use nanotime() to update the time fields in filesystems. Convert the code
from macros to real functions. Original patch and review from chuq.
Note: ext2fs only keeps seconds in the on-disk inode, and msdosfs does not
have enough precision for all fields, so this is not very useful for those
two.
|
| 1.165 | 19-Aug-2005 |
christos | 64 bit inode changes.
|
| 1.164 | 29-May-2005 |
christos | branches: 1.164.2;
- sprinkle const
- avoid shadow variables.
|
| 1.163 | 23-Apr-2005 |
perseant | Provide a resize_lfs(8), including kernel and cleaner support. The current
implementation requires the fs to be mounted while resizing. Tested in both
directions, and everything appears to work happily, but ymmv.
|
| 1.162 | 19-Apr-2005 |
perseant | Keep per-inode, per-fs, and subsystem-wide counts of blocks allocated through
lfs_balloc(), and use that to estimate the number of dirty pages belonging
to LFS (subsystem or filesystem). This is almost certainly wrong for
the case of a large mmap()ed region, but the accounting is tighter than
what we had before, and performs much better in the typical case of pages
dirtied through write().
|
| 1.161 | 18-Apr-2005 |
perseant | Check the to-be-on-disk consistency of directories as well (correct a typo
in an earlier commit).
|
| 1.160 | 14-Apr-2005 |
perseant | Keep track of the highest block held by an LFS inode, so that we can
be assured that the last byte of a file is always allocated. Previously
a file extension could cause the filesystem to be flushed, writing an
inconsistent inode to disk. Although this condition would be corrected
the next time blocks were written to disk, an intervening crash would leave
the filesystem in an inconsistent state, leaving fsck_lfs to complain
of an inode "partially truncated".
|
| 1.159 | 01-Apr-2005 |
perseant | Protect various per-fs structures with fs->lfs_interlock simple_lock, to
improve behavior in the multiprocessor case. Add debugging segment-lock
assertion statements.
|
| 1.158 | 08-Mar-2005 |
perseant | branches: 1.158.2;
Straighten out the maze of ifdefs. Instead, consolidate all the debugging
stuff under '#ifdef DEBUG', and use sysctl knobs to turn on/off particular
parts of the debugging reporting (if DEBUG is enabled). Re-enable the LFS
statistics in sysctl, while I'm there. A bit of a rototill.
|
| 1.157 | 26-Feb-2005 |
perry | nuke trailing whitespace
|
| 1.156 | 26-Feb-2005 |
perseant | Various minor LFS improvements:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statvfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
|
| 1.155 | 18-Sep-2004 |
yamt | branches: 1.155.4; 1.155.6;
change some members of struct buf from long to int.
ride on 2.0H.
|
| 1.154 | 14-Aug-2004 |
mycroft | Add a new flag, IN_MODIFY. This is like IN_UPDATE|IN_CHANGE, but unlike
setting those flags, it does not cause the inode to be written in the periodic
sync. This is used for writes to special files (devices and named pipes) and
FIFOs.
Do not preemptively sync updates to access times and modification times. They
are now updated in the inode only opportunistically, or when the file or device
is closed. (Really, it should be delayed beyond close, but this is enough to
help substantially with device nodes.)
And the most amusing part:
Trickle sync was broken on both FFS and ext2fs, in different ways. In FFS, the
periodic call to VFS_SYNC(MNT_LAZY) was still causing all file data to be
synced. In ext2fs, it was causing the metadata to *not* be synced. We now
only call VOP_UPDATE() on the node if we're doing MNT_LAZY. I've confirmed
that we do in fact trickle correctly now.
|
| 1.153 | 19-May-2004 |
yamt | lfs_cluster_aiodone: turn an invariant condition into an assertion.
|
| 1.152 | 09-Mar-2004 |
yamt | branches: 1.152.4;
calculate data checksum inline.
|
| 1.151 | 09-Mar-2004 |
yamt | use correct segment size. this fixes memory corruption when using lfsv1.
|
| 1.150 | 29-Jan-2004 |
yamt | lfs_update_single: add an assertion.
|
| 1.149 | 28-Jan-2004 |
yamt | eliminate tricky usages of VOP_STRATEGY which are (no longer?) necessary.
|
| 1.148 | 25-Jan-2004 |
hannken | Make VOP_STRATEGY(bp) a real VOP as discussed on tech-kern.
VOP_STRATEGY(bp) is replaced by one of two new functions:
- VOP_STRATEGY(vp, bp) Call the strategy routine of vp for bp.
- DEV_STRATEGY(bp) Call the d_strategy routine of bp->b_dev for bp.
DEV_STRATEGY(bp) is used only for block-to-block device situations.
|
| 1.147 | 10-Jan-2004 |
yamt | store a i/o priority hint in struct buf for buffer queue discipline.
|
| 1.146 | 17-Dec-2003 |
yamt | set VBWAIT when waiting v_numoutput to be drained.
|
| 1.145 | 17-Dec-2003 |
yamt | remove a redundant substitution.
|
| 1.144 | 04-Dec-2003 |
yamt | use b_private rather than b_saveaddr.
XXX LFS_USE_B_INVAL
|
| 1.143 | 07-Nov-2003 |
yamt | - tweak lfs_update_single()'s prototype so that it can be used by
roll-forward code.
- reduce code duplication using the above in update_meta()
this also fixes fragment accounting.
|
| 1.142 | 25-Oct-2003 |
christos | Fix uninitialized variable warnings.
|
| 1.141 | 18-Oct-2003 |
yamt | be more strict about sa->vp.
(make sure the last lfs_updatemata in lfs_putpages takes effect.)
|
| 1.140 | 18-Oct-2003 |
simonb | Remove assigned-to but otherwise unused variable.
|
| 1.139 | 17-Oct-2003 |
yamt | add comments and tweak code a little for readability.
(no behaviour changes)
|
| 1.138 | 14-Oct-2003 |
yamt | remove a redundant definition of LFS_MAX_ACTIVE.
|
| 1.137 | 08-Oct-2003 |
yamt | - a comment.
- bcopy -> memcpy
- increase 'p' only when needed.
|
| 1.136 | 03-Oct-2003 |
yamt | assertions.
|
| 1.135 | 03-Oct-2003 |
yamt | reassignbuf() when lfs_writeseg() takes away B_DELWRI.
|
| 1.134 | 03-Oct-2003 |
yamt | when inactivating segments, compare segment numbers correctly.
|
| 1.133 | 29-Sep-2003 |
yamt | remove redundant prototypes.
|
| 1.132 | 07-Sep-2003 |
yamt | - buffer cache MP locks.
- avoid changing buffer state on the free queue.
|
| 1.131 | 07-Aug-2003 |
agc | Move UCB-licensed code from 4-clause to 3-clause licence.
Patches provided by Joel Baker in PR 22364, verified by myself.
|
| 1.130 | 30-Jul-2003 |
yamt | using normal bufcache buffer for cluster buffer head.
|
| 1.129 | 23-Jul-2003 |
yamt | KNF.
|
| 1.128 | 12-Jul-2003 |
yamt | - wrap long lines.
- remove a mysterious blank line.
|
| 1.127 | 12-Jul-2003 |
yamt | - protect global resource counts with lfs_subsys_lock.
- clean up scattered externs a little.
|
| 1.126 | 02-Jul-2003 |
yamt | use queue.h macros.
|
| 1.125 | 02-Jul-2003 |
yamt | - add a new functions, lfs_writer_enter/leave, and use them instead of
duplicated code fragments.
- add an assertion.
|
| 1.124 | 29-Jun-2003 |
fvdl | branches: 1.124.2;
Back out the lwp/ktrace changes. They contained a lot of colateral damage,
and need to be examined and discussed more.
|
| 1.123 | 29-Jun-2003 |
thorpej | Undo part of the ktrace/lwp changes. In particular:
* Remove the "lwp *" argument that was added to vget(). Turns out
that nothing actually used it!
* Remove the "lwp *" arguments that were added to VFS_ROOT(), VFS_VGET(),
and VFS_FHTOVP(); all they did was pass it to vget() (which, as noted
above, didn't use it).
* Remove all of the "lwp *" arguments to internal functions that were added
just to appease the above.
|
| 1.122 | 28-Jun-2003 |
darrenr | Pass lwp pointers throughtout the kernel, as required, so that the lwpid can
be inserted into ktrace records. The general change has been to replace
"struct proc *" with "struct lwp *" in various function prototypes, pass
the lwp through and use l_proc to get the process pointer when needed.
Bump the kernel rev up to 1.6V
|
| 1.121 | 18-May-2003 |
yamt | make is_sequential a callback in order to achieve better lfs write clustering.
since lfs always rewrite blocks into the new segment,
current on-disk place of the block doesn't affect to write clustering.
ok'ed by Konrad Schroder.
|
| 1.120 | 23-Apr-2003 |
perseant | Make LFS work better (though still not "well") as an NFS-exported
filesystem (and other things that needed to be fixed before the tests
would complete), to wit:
* Include the fs ident in the filehandle; improve stale filehandle checks.
* Change definition of blksize() to use the on-dinode size instead of
the inode's i_size, so that fsck_lfs will work properly again.
* Use b_interlock in lfs_vtruncbuf.
* Postpone dirop reclamation until after the seglock has been released,
so that lfs_truncate is not called with the segment lock held.
* Don't loop in lfs_fsync(), just write everything and wait.
* Be more careful about the interlock/uobjlock in lfs_putpages: when we
lose this lock, we have to resynchronize dirtiness of pages in each
block.
* Be sure to always write indirect blocks and update metadata in
lfs_putpages; fixes a bug that caused blocks to be accounted to the
wrong segment.
|
| 1.119 | 02-Apr-2003 |
fvdl | Add support for UFS2. UFS2 is an enhanced FFS, adding support for
64 bit block pointers, extended attribute storage, and a few
other things.
This commit does not yet include the code to manipulate the extended
storage (for e.g. ACLs), this will be done later.
Originally written by Kirk McKusick and Network Associates Laboratories for
FreeBSD.
|
| 1.118 | 01-Apr-2003 |
yamt | add assertions and a debug check.
|
| 1.117 | 28-Mar-2003 |
fvdl | The checkpoint loop always used (multiples of) lfs_sepb as the number
of segments to mark. However, this may be much more than lfs_nseg.
Originally this wasn't a big problem, since only the structures in the
diskblock were changed, but nowadays there's a mirror of the segflags
in the in-core superblock. This problem caused the code to walk
way past the end of that allocated area, causing memory corruption
in other kernel structures. So, use lfs_nseg as the maximum, as it should be.
While here, simplify the loop; it had become an obfuscated piece of
code overtime.
|
| 1.116 | 28-Mar-2003 |
perseant | Add a sleeper count, to prevent the cleaner from panicing the kernel
when the filesystem is unmounted, relocking the Ifile when its lock is
draining. (We can't use vfs_busy() since the process is sleeping for a
good long time.) Clean up / organize lfs.h, while I'm here.
In lfs_update_single, assert that disk addresses are either negative, or
are still positive when converted to int32_t, to prevent recurrence of a
negative/positive block problem.
|
| 1.115 | 21-Mar-2003 |
perseant | KNF (space after keywords).
|
| 1.114 | 21-Mar-2003 |
perseant | Use VONWORKLST as a heuristic for vnode emptiness, rather than exhaustively
checking the memq.
Take greater care not to dirty the Ifile vnode when unmounting the filesystem.
This should fix a "(vp->v_flag & VONWORKLST) == 0" assertion panic in vgonel
that could occur when unmounting.
Do not allow the Ifile to be mapped for writing.
|
| 1.113 | 20-Mar-2003 |
yamt | lfs_writevnodes:
in the case of "starting over", kick lfs_writeseg
in order to avoid deadlock in check_dirty.
|
| 1.112 | 20-Mar-2003 |
perseant | Don't break out of Ifile-writing loop in lfs_segwrite until nothing is left.
Note however that blocks can be added to the Ifile even when the segment
block is held because of inodes' atime. Do not panic with "dirty blocks"
if these blocks are present.
|
| 1.111 | 15-Mar-2003 |
perseant | Add simple_lock protection for lfs_seglock and lfs_subsys_pages; these will
be expanded to cover other per-fs and subsystem-wide data as well.
Fix a case of IN_MODIFIED being set without updating lfs_uinodes, resulting
in a "lfs_uinodes < 0" panic.
Fix a deadlock in lfs_putpages arising from the need to busy all pages in a
block; unbusy any that had already been busied before starting over.
|
| 1.110 | 15-Mar-2003 |
kristerw | SO C requires a statement after a label.
|
| 1.109 | 11-Mar-2003 |
perseant | - Get rid of unused #ifdefs LFS_NO_PAGEMOVE and LFS_MALLOC_SUMMARY (both
always true) and accompanying dead code.
- When constructing write clusters in lfs_writeseg, if the block we are
about to add is itself a cluster from GOP_WRITE, don't put a cluster
in a cluster, just write the GOP_WRITE cluster on its own. This seems
to represent a slight performance gain on my test machine.
- Charge someone's rusage for writes on LFSes. It's difficult to tell
who the "right" process to charge is; just charge whoever triggered
the write.
|
| 1.108 | 08-Mar-2003 |
perseant | Take away "#ifdef LFS_UBC".
|
| 1.107 | 08-Mar-2003 |
perseant | Add an lfs_strategy() that checks to make sure we're not trying to read
where the cleaner is trying to write, instead of tying up the "live"
buffers (or pages).
Fix a bug in the LFS_UBC case where oversized buffers would not be
checksummed correctly, causing uncleanable segments.
Make sure that wakeup(fs->lfs_iocount) is done if fs->lfs_iocount is 1
as well as 0, since we wait in some places for it to drop to 1.
Activate all pages that make it into lfs_gop_write without the segment
lock held, since they must have been dirtied very recently, even if
PG_DELWRI is not set.
|
| 1.106 | 04-Mar-2003 |
perseant | Make sure we hold the uobjlock when checking for dirty pages, in lfs_vflush.
Note that pages can become dirty without our knowing it, anyway; don't
panic if that happens.
|
| 1.105 | 02-Mar-2003 |
perseant | Account SEGUSE_ACTIVE correctly so that the automatic segment cleaning
actually happens.
Add a new fcntl call that will write the minimum necessary to checkpoint
(i.e., for on-disk directory structure to be consistent, not including
updates to file data) so that the cleaner can clean segments more quickly
without sacrificing three-way commit for cleaning.
|
| 1.104 | 23-Feb-2003 |
perseant | Fix a buffer overflow bug in the LFS_UBC case that manifested itself
either as a mysterious UVM error or as "panic: dirty bufs". Verify
maximum size in lfs_malloc.
Teach lfs_updatemeta and lfs_shellsort about oversized cluster blocks from
lfs_gop_write.
When unwiring pages in lfs_gop_write, deactivate them, under the theory
that the pagedaemon wanted to free them last we knew.
|
| 1.103 | 20-Feb-2003 |
perseant | Tabify, and fix some comment alignment problems.
|
| 1.102 | 19-Feb-2003 |
yamt | acquire v_interlock before calling VOP_PUTPAGES.
|
| 1.101 | 17-Feb-2003 |
perseant | Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now
(there are still some details to work out) but expect that to go
away soon. To support these basic changes (creation of lfs_putpages,
lfs_gop_write, mods to lfs_balloc) several other changes were made, to
wit:
* Create a writer daemon kernel thread whose purpose is to handle page
writes for the pagedaemon, but which also takes over some of the
functions of lfs_check(). This thread is started the first time an
LFS is mounted.
* Add a "flags" parameter to GOP_SIZE. Current values are
GOP_SIZE_READ, meaning that the call should return the size of the
in-core version of the file, and GOP_SIZE_WRITE, meaning that it
should return the on-disk size. One of GOP_SIZE_READ or
GOP_SIZE_WRITE must be specified.
* Instead of using malloc(...M_WAITOK) for everything, reserve enough
resources to get by and use malloc(...M_NOWAIT), using the reserves if
necessary. Use the pool subsystem for structures small enough that
this is feasible. This also obsoletes LFS_THROTTLE.
And a few that are not strictly necessary:
* Moves the LFS inode extensions off onto a separately allocated
structure; getting closer to LFS as an LKM. "Welcome to 1.6O."
* Unified GOP_ALLOC between FFS and LFS.
* Update LFS copyright headers to correct values.
* Actually cast to unsigned in lfs_shellsort, like the comment says.
* Keep track of which segments were empty before the previous
checkpoint; any segments that pass two checkpoints both dirty and
empty can be summarily cleaned. Do this. Right now lfs_segclean
still works, but this should be turned into an effectless
compatibility syscall.
|
| 1.100 | 05-Feb-2003 |
pk | Make the buffer cache code MP-safe.
|
| 1.99 | 01-Feb-2003 |
thorpej | Add extensible malloc types, adapted from FreeBSD. This turns
malloc types into a structure, a pointer to which is passed around,
instead of an int constant. Allow the limit to be adjusted when the
malloc type is defined, or with a function call, as suggested by
Jonathan Stone.
|
| 1.98 | 29-Jan-2003 |
yamt | don't use daddr_t for segment summary since it's an on-disk structure.
|
| 1.97 | 29-Jan-2003 |
simonb | Remove variable that is only assigned to but not referenced.
|
| 1.96 | 27-Jan-2003 |
yamt | make these compilable with lfs debug options.
(follow daddr_t change)
XXX maybe segment number should be 64bit.
|
| 1.95 | 27-Jan-2003 |
kleink | Further printf format fixes in the wake of daddr_t.
Note that PRI?64 and long long int arguments aren't made for each other,
nor are %lld and int64_t arguments.
|
| 1.94 | 25-Jan-2003 |
kleink | Fix further printf format warnings for DEBUG, in the wake of daddr_t
having changed.
|
| 1.93 | 25-Jan-2003 |
tron | Use PRId64 instead of hard coding "%lld" to fix build problems under
LP64 ports.
|
| 1.92 | 25-Jan-2003 |
tron | Fix printf() format strings problems caused by "daddr_t" change.
|
| 1.91 | 24-Jan-2003 |
fvdl | Bump daddr_t to 64 bits. Replace it with int32_t in all places where
it was used on-disk, so that on-disk formats remain the same.
Remove ufs_daddr_t and ufs_lbn_t for the time being.
|
| 1.90 | 08-Jan-2003 |
yamt | backout wrong assertions that i added.
|
| 1.89 | 08-Jan-2003 |
yamt | add assertions.
|
| 1.88 | 31-Dec-2002 |
yamt | write ifile only when it has dirty buffers.
|
| 1.87 | 17-Dec-2002 |
yamt | no need for cleaner to hold vnode locks.
cleaner and normal vnode operations are synchronized enough by
seglock/fraglock and buf's B_BUSY-ness.
|
| 1.86 | 17-Dec-2002 |
yamt | use ufs_daddr_t instead of int where appropriate.
|
| 1.85 | 14-Dec-2002 |
yamt | in lfs_writefile, check v_type==VNON earlier.
to avoid null dereference with DEBUG_LFS_VERBOSE.
|
| 1.84 | 13-Dec-2002 |
yamt | save a segment write when doing checkpoint.
|
| 1.83 | 12-Dec-2002 |
yamt | correct DIAGNOSTIC code for duplicated inodes in a segment and su_nbytes.
|
| 1.82 | 27-Sep-2002 |
provos | remove trailing \n in panic(). approved perry.
|
| 1.81 | 22-Sep-2002 |
jdolecek | don't need <sys/conf.h> here
|
| 1.80 | 06-Jul-2002 |
perseant | Deal with fragment size changes better. For each fragment that can
exist on an on-disk inode, we keep a record of its size in struct inode,
which is updated when we write the block to disk. The cleaner routines
thus have ready access to what size is the correct size for this block,
on disk.
Fixed a related bug: if a file with fragments is being cleaned
(fragments being cleaned) at the same time it is being extended beyond
NDADDR blocks, we could write a bogus FINFO record that has a frag in the
middle; when it was cleaned this would give back bogus file data. Don't
write the indirect blocks in this case, since there is no need.
lfs_fragextend and lfs_truncate no longer require the seglock, but instead
take a shared lock, which the seglock locks exclusively.
|
| 1.79 | 16-Jun-2002 |
perseant | For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.78 | 24-May-2002 |
perseant | Fix a couple of instances where reassignbuf() was not done at splbio.
Tested on i386.
|
| 1.77 | 23-May-2002 |
perseant | Back out rev 1.174 of vfs_subr.c, because the splbio() wasn't protecting
enough to be useful, and broadening it so that it did would have meant
that operations possibly requiring synchronous disk activity would have
to be done in splbio(). This clearly was not going to work.
Worked around this in the LFS case by having lfs_cluster_callback put an
extra hold on the vnode before calling biodone(), and taking the hold
off without HOLDRELE's problematic list swapping. lfs_vunref() will take
care of that---in thread context---on the next write if need be.
Also, ensure that the list walking in lfs_{writevnodes,segunlock,gather}
takes into account the possibility that the list may change
underneath it (possibly because it itself deleted an element).
Tested on i386, test-compiled on alpha.
|
| 1.76 | 20-May-2002 |
perseant | branches: 1.76.2;
Protect v_freelist with splbio(), since HOLDRELE can be called in
interrupt context (through brelvp). (LFS may be the only subsystem
affected by this problem.)
Tested on i386.
|
| 1.75 | 17-May-2002 |
perseant | use macros from <sys/queue.h>
|
| 1.74 | 14-May-2002 |
perseant | branches: 1.74.2;
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes
I found while making sure there weren't any new ones.
* Make the write clusters keep track of the buffers whose blocks they contain.
This should make it possible to (1) write clusters using a page mapping
instead of malloc, if desired, and (2) schedule blocks for rewriting
(somewhere else) if a write error occurs. Code is present to use
pagemove() to construct the clusters but that is untested and will go away
anyway in favor of page mapping.
* DEBUG now keeps a log of Ifile writes, so that any lingering instances of
the "dirty bufs" problem can be properly debugged.
* Keep track of whether the Ifile has been dirtied by various routines that
can be called by lfs_segwrite, and loop on that until it is clean, for
a checkpoint. Checkpoints need to be squeaky clean.
* Warn the user (once) if the Ifile grows larger than is reasonable for their
buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can
grow.
* If an inode is not found in a disk block, try rereading the block, under
the assumption that the block was copied to a cluster and then freed.
* Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
|
| 1.73 | 23-Nov-2001 |
chs | add spaces for KNF. confirmed to produce identical objects.
|
| 1.72 | 08-Nov-2001 |
lukem | add RCSID
|
| 1.71 | 26-Oct-2001 |
lukem | remove #include <ufs/ufs/quota.h> where it was just to appease
<ufs/ufs/inode.h>, since the latter now includes the former. leave the former
in source that obviously uses specific bits of it (for completeness.)
|
| 1.70 | 26-Jul-2001 |
jdolecek | branches: 1.70.2; 1.70.4;
lfs_writeseg(): make el_size a size_t (cosmetic only, no functional change)
|
| 1.69 | 13-Jul-2001 |
perseant | Merge the short-lived perseant-lfsv2 branch into the trunk.
Kernels and tools understand both v1 and v2 filesystems; newfs_lfs
generates v2 by default. Changes for the v2 layout include:
- Segments of non-PO2 size and arbitrary block offset, so these can be
matched to convenient physical characteristics of the partition (e.g.,
stripe or track size and offset).
- Address by fragment instead of by disk sector, paving the way for
non-512-byte-sector devices. In theory fragments can be as large
as you like, though in reality they must be smaller than MAXBSIZE in size.
- Use serial number and filesystem identifier to ensure that roll-forward
doesn't get old data and think it's new. Roll-forward is enabled for
v2 filesystems, though not for v1 filesystems by default.
- The inode free list is now a tailq, paving the way for undelete (undelete
is not yet implemented, but can be without further non-backwards-compatible
changes to disk structures).
- Inode atime information is kept in the Ifile, instead of on the inode;
that is, the inode is never written *just* because atime was changed.
Because of this the inodes remain near the file data on the disk, rather
than wandering all over as the disk is read repeatedly. This speeds up
repeated reads by a small but noticeable amount.
Other changes of note include:
- The ifile written by newfs_lfs can now be of arbitrary length, it is no
longer restricted to a single indirect block.
- Fixed an old bug where ctime was changed every time a vnode was created.
I need to look more closely to make sure that the times are only updated
during write(2) and friends, not after-the-fact during a segment write,
and certainly not by the cleaner.
|
| 1.68 | 30-May-2001 |
mrg | branches: 1.68.2; 1.68.4;
use _KERNEL_OPT
|
| 1.67 | 09-Jan-2001 |
joff | branches: 1.67.2;
If DIAGNOSTIC and the segment writer gets a badly sized buffer, panic()
instead of silently corrupting the filesystem.
|
| 1.66 | 03-Dec-2000 |
perseant | Get rid of some old unnecessary code that cleared B_NEEDCOMMIT from buffers in
lfs_writeseg (possibly after they had been freed).
If MALLOCLOG is defined, make lfs_newbuf and lfs_freebuf pass along the
caller's file and line to _malloc and _free.
|
| 1.65 | 30-Nov-2000 |
jdolecek | only include opt_ddb.h for !LKM
|
| 1.64 | 27-Nov-2000 |
chs | Initial integration of the Unified Buffer Cache project.
|
| 1.63 | 27-Nov-2000 |
perseant | If LFS_DO_ROLLFORWARD is defined, roll forward from the older checkpoint
on mount, through the newer checkpoint and on through any newer
partial-segments that may have been written but not checkpointed because
of an intervening crash.
LFS_DO_ROLLFORWARD is not defined by default.
|
| 1.62 | 17-Nov-2000 |
perseant | Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes.
(PR #11468). In the case of fragment allocation, check to see if enough
space is available before extending a fragment already scheduled for writing.
The locked_queue_* variables indicate the number of buffer headers and bytes,
respectively, that are unavailable to getnewbuf() because they are locked up
waiting for LFS to flush them; make sure that that is actually what we're
counting, i.e., never count malloced buffers, and always use b_bufsize instead
of b_bcount.
If DEBUG is defined, the periodic calls to lfs_countlocked will now complain
if either counter is incorrect. (In the future lfs_countlocked will not need
to be called at all if DEBUG is not defined.)
|
| 1.61 | 12-Nov-2000 |
perseant | Do not needlessly dirty segment table blocks during lfs_segwrite,
preventing needless disk activity when the filesystem is idle. (PR #10979.)
|
| 1.60 | 12-Nov-2000 |
toshii | Fix obsolete comments in lfs_writeinode since rev. 1.27.
New comments are mostly from perseant, with my additions.
|
| 1.59 | 09-Sep-2000 |
perseant | oops
|
| 1.58 | 09-Sep-2000 |
perseant | Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
|
| 1.57 | 09-Sep-2000 |
perseant | Fix a buffer-cache corrupting bug in lfs_writeseg, where brelse could
be improperly used on an already-queued buffer.
|
| 1.56 | 05-Jul-2000 |
perseant | Clean up accounting of lfs_uinodes (dirty but unwritten inodes).
Make lfs_uinodes a signed quantity for debugging purposes, and set it to
zero as fs mount time.
Enclose setting/clearing of the dirty flags (IN_MODIFIED, IN_ACCESSED,
IN_CLEANING) in macros, and use those macros everywhere. Make
LFS_ITIMES use these macros; updated the ITIMES macro in inode.h to know
about this. Make ufs_getattr use ITIMES instead of FFS_ITIMES.
|
| 1.55 | 04-Jul-2000 |
perseant | Fix errors observed while trying to fill the filesystem with yesterday's
fixes:
- Write copies of bfree and avail in the CLEANERINFO block, so the
cleaner doesn't have to guess which superblock has the current
information (if indeed any do).
- Tighten up accounting of lfs_avail (more needs to be done).
- When cleansing indirect blocks of UNWRITTEN, make sure not to mark
them clean, since they'll need to be rewritten later.
|
| 1.54 | 03-Jul-2000 |
perseant | i_lfs_effnblks fixes. Put debugging printfs under #ifdef DEBUG_LFS.
|
| 1.53 | 03-Jul-2000 |
perseant | Allow the number of free segments reserved for the cleaner to be
parametrized in the filesystem, defaulting to MIN_FREE_SEGS = 2 but set
to something more reasonable at newfs_lfs time.
Note the number of blocks that have been scheduled for writing but which
are not yet on disk in an inode extension, i_lfs_effnblks. Move
i_ffs_effnlink out of the ffs extension and onto the main inode, since
it's used all over the shared code and the lfs extension would clobber
it.
At inode write time, indirect blocks and inode-held blocks of inodes
that have i_lfs_effnblks != i_ffs_blocks are cleansed of UNWRITTEN disk
addresses, so that these never make it to disk.
|
| 1.52 | 27-Jun-2000 |
perseant | Fixes associated with filling an LFS:
Change the space computation to appear to change the size of the *disk*
rather than the *bytes used* when more segment summaries and inode
blocks are written. Try to estimate the amount of space that these will
take up when more files are written, so the disk size doesn't change too
much.
Regularize error returns from lfs_valloc, lfs_balloc, lfs_truncate: they
now fail entirely, rather than succeeding half-way and leaving the fs in
an inconsistent state.
Rewrite lfs_truncate, mostly stealing from ffs_truncate. The old
lfs_truncate had difficulty truncating a large file to a non-zero size
(indirect blocks were not handled appropriately).
Unmark VDIROP on fvp after ufs_remove, ufs_rmdir, so these can be
reclaimed immediately: this vnode would not be written to disk again
anyway if the removal succeeded, and if it failed, no directory
operation occurred.
ufs_makeinode and ufs_mkdir now remove IN_ADIROP on error.
|
| 1.51 | 27-Jun-2000 |
perseant | From John Evans <jevans@cray.com>: use datosn() to convert to segment
number, when remarking the current segment ACTIVE. See PR #10463.
|
| 1.50 | 22-Jun-2000 |
perseant | Update lfs_vunref for the fact that now a vnode can be locked with no
references (locked for VOP_INACTIVE at the end of vrele) and it's okay.
Check the return value of lfs_vref where appropriate.
Fixes PR #s 10285 and 10352.
|
| 1.49 | 06-Jun-2000 |
perseant | branches: 1.49.2;
Protect inode free list with seglock, instead of separate lock, so that
the head of the inode free list (on the superblock) always matches the
rest of the free list (in the ifile).
Protect lfs_fragextend with seglock, to prevent the segment byte count
fudging from making its way to disk.
Don't try to inactivate dirop vnodes that are still in the middle of
their dirop (may address PR#10285).
|
| 1.48 | 31-May-2000 |
fredb | Make this build. (Balance parenthesis.
|
| 1.47 | 31-May-2000 |
perseant | update for IN_ACCESSED changes
|
| 1.46 | 27-May-2000 |
perseant | branches: 1.46.2;
Prevent dirops from getting around lfs_check and wedging the buffer cache.
All the dirop vnops now mark the inodes with a new flag, IN_ADIROP, which
is removed as soon as the dirop is done (as opposed to VDIROP which stays
until the file is written). To address one issue raised in PR#9357.
|
| 1.45 | 19-May-2000 |
thorpej | NULL != 0
|
| 1.44 | 10-May-2000 |
perseant | stop vnode reference leak introduced in patch to PR#9994
|
| 1.43 | 05-May-2000 |
perseant | Change the way LFS does block accounting, from trying to infer from the
buffer cache flags, to marking the inode and/or indirect blocks with a
special disk address UNWRITTEN==-2 when a block is accounted for. (This
address is never written to disk, but only used in-core. This is essentially
the same method of block accounting as on the UBC branch, where the buffer
headers don't exist.) Make sure that truncation is handled properly,
especially in the case of holey files.
Fixes PR#9994.
|
| 1.42 | 30-Mar-2000 |
augustss | Remove register declarations.
|
| 1.41 | 13-Mar-2000 |
soren | Fix doubled 'the's in comments.
|
| 1.40 | 19-Jan-2000 |
perseant | Changes to stabilize LFS. The first two of these should also apply to the
1.4 branch.
* Use a separate per-fs lock, instead of ufs_hashlock, to protect the Inode
free list. This seems to prevent the "lockmgr: %d, not exclusive lock holder
%d, unlocking" message I was mis-attributing last night to an unlocked vnode
being passed to vrele.
* Change calling semantics of lfs_ifind, to give better error reporting:
If fed a struct buf, it can report the block number of the offending inode
block as well as the inode number.
* Back out rev 1.10 of lfs_subr.c, since the replacement code was slightly
uglier while being functionally identical.
* Make lfs_vunref use the same free list convention as vrele/vput, so that
vget does not remove vnodes from a hash list they are not on.
|
| 1.39 | 16-Jan-2000 |
perseant | Fix a problem in my changes of Dec 14th, that prevents removed vnodes
from being inactivated under some conditions. Removed vnodes are now
inactivated when the VDIROP flag is cleared, and to prevent block
accounting problems this clearing has been postponed until
lfs_segunlock.
|
| 1.38 | 14-Jan-2000 |
perseant | Better handling of various combinations of cleaning, vnode flushing, and
dirop writing. In particular, lfs_writevnodes now writes all buffers from
a flushed vnode whether cleaning or not, and the same with the Ifile; and
lfs_segwrite does not attempt to write data from other non-cleaning vnodes,
even if a vnode is being flushed.
|
| 1.37 | 03-Dec-1999 |
perseant | Handle the case of a vnode flush while dirops are active correctly in
lfs_segwrite. Also, make sure a flush is called in SET_DIROP before sleeping
on its results. Addresses PR #8863.
|
| 1.36 | 17-Nov-1999 |
perseant | Fix spllevel problem with superblock exclusion and with segment write throttle.
May address PR#8383.
|
| 1.35 | 15-Nov-1999 |
fvdl | Add Kirk McKusick's soft updates code to the trunk. Not enabled by
default, as the copyright on the main file (ffs_softdep.c) is such
that is has been put into gnusrc. options SOFTDEP will pull this
in. This code also contains the trickle syncer.
Bump version number to 1.4O
|
| 1.34 | 12-Nov-1999 |
perseant | Back out my patch of the 8th (to address unreferenced inode problem).
Apparently this needs more thought.
|
| 1.33 | 09-Nov-1999 |
perseant | If ifile blocks were written before dirops were complete, and then the
system crashed, inodes could be allocated that were not referenced. (Though
not a serious problem, it evidences itself in phase 4 of fsck_lfs.) Fix
this by marking if_daddr with UNASSIGNED before the inodes are actually
written; at mount time the ifile is checked for UNASSIGNED entries and
any that are found are linked back into the free list. (The latter
functionality should move into the roll-forward agent when it materializes.)
|
| 1.32 | 06-Nov-1999 |
perseant | branches: 1.32.2;
Address ufs_hashlock/ufs_ihashins protocol bug, discovered while doing a
post-mortem of a production machine. Also, take the active dirop
count off of the fs and make it global (since it is measuring a global
resource) and tie the threshold value LFS_MAXDIROP to desiredvnodes.
|
| 1.31 | 01-Oct-1999 |
mycroft | branches: 1.31.2; 1.31.4; 1.31.6;
Fix printf() formats.
|
| 1.30 | 03-Sep-1999 |
perseant | Make changes that will allow an LFS filesystem to be used as the root
filesystem. In particular,
- Fix mknod deadlock, described in PR 8172.
- Enable lfs_mountroot.
- Make lfs_writevnodes treat filesystems mounted on lfs device nodes properly,
by flushing that device rather than trying to add blocks to the device inode.
This, in combination with lfs boot blocks, will allow operation of an all-lfs
system.
|
| 1.29 | 08-Jul-1999 |
wrstuden | Modify file systems to deal with struct lock in struct vnode. All leaf
fs's other than nfs use genfs_lock() for locking.
Modify lookup routines to set PDIRUNLOCK when they unlock the parrent.
|
| 1.28 | 17-Jun-1999 |
tls | squash some compiler warnings on debug printfs by casting to int
|
| 1.27 | 15-Jun-1999 |
perseant | Minor changes to the segment live bytes calculation. In particular, fixed
a bug in fragment extension that could run the count negative. Also, don't
overcount for inodes, and don't count segment summaries. Thus, for empty
segments the live bytes count should now be exactly zero.
|
| 1.26 | 12-Apr-1999 |
perseant | Make sure that the wakeup occurs for vnodes that lfs_update might be sleeping
on (nodes which are not marked IN_MODIFIED/IN_CLEANING, but which have dirty
buffers), by marking them with the appropriate flag if dirtybuffers were added
while the write was in progress.
|
| 1.25 | 12-Apr-1999 |
perseant | Better checking for held inode locks in lfs_fastvget, for a number of error
conditions. Also change the default setting of lfs_clean_vnhead to 0, which
seems to make the locking problems go away (although this is difficult to
test as I can't reliably reproduce them).
|
| 1.24 | 12-Apr-1999 |
perseant | Fix "lfs_ifind: dinode xxx not found" panic. When inodes were freed,
then immediately reloaded, their dinodes were located in an inode block
which was not on disk at the advertized location, nor in the cache (although
it would be flushed to disk next segment write). Fix this by using getblk()
instead of lfs_newbuf() for inode blocks.
|
| 1.23 | 30-Mar-1999 |
perseant | branches: 1.23.2;
Add initialization to quell compiler warning (only on some platforms?)
|
| 1.22 | 30-Mar-1999 |
perseant | Move variable initialization to the top of lfs_vflush
|
| 1.21 | 29-Mar-1999 |
perseant | lfs_truncate calls vinvalbuf to invalidate all currently-hald buffers, which
in turn forces a flush of the vnode, whether or not it is involved in a dirop.
(This can happen during a remove or rmdir, when the directory is shrunk.)
Because of the nature of dirops, however, flushing a vnode involved in a dirop
is disallowed (and was marked with a panic). This patch has lfs_truncate
call a specialized vinvalbuf that only invalidates buffers following the new
end-of-file, and thus does not require a flush. Also the panic is demoted,
in case I missed any other path to lfs_vflush.
|
| 1.20 | 25-Mar-1999 |
perseant | Make sysctl variable lfs_clean_vnhead do what it was supposed to do,
namely, toggle whether vnodes loaded only for cleaning (as opposed to
normal filesystem use) are freed to the *head* of the vnode free list,
rather than the tail. This should avoid a possible cache flushing
effect, if the cleaner cleans a segment containing a large number of
live inodes.
|
| 1.19 | 25-Mar-1999 |
perseant | Fixes to make dirops and lfs_vflush play together well. In particular,
if we are short on vnodes, lfs_vflush from another process can grab a
vnode that lfs_markv has already processed but not yet written; but
lfs_markv holds the seglock. When lfs_vflush gets around to writing it,
the context for copyin is gone. So, now lfs_markv calls copyin itself,
rather than having lfs_writeseg do it.
|
| 1.18 | 25-Mar-1999 |
perseant | Lock buffers with B_BUSY between data checksum calculation and write, so
some other process doesn't change the data after it was checksummed.
|
| 1.17 | 25-Mar-1999 |
perseant | Change lfs_sb_cksum to use offsetof() instead of an inlined version.
Fix lfs_vref/lfs_vunredf to ignore VXLOCKed vnodes that are also being
flushed.
Improve the debugging messages somewhat.
|
| 1.16 | 25-Mar-1999 |
perseant | clean up unused/required #ifdefs
|
| 1.15 | 10-Mar-1999 |
perseant | New sources should leave the LFS in a more-or-less working state. Changes
include:
- DIROP segregation is enabled, and greater care is taken
to make sure that a checkpoint completes. Fsck is not
needed to remount the filesystem.
- Several checks to make sure that the LFS subsystem does not
overuse various resources (memory, in particular).
- The cleaner routines, lfs_markv in particular, are completely
rewritten. A buffer overflow is removed. Greater care is taken
to ensure that inodes come from where lfs_cleanerd say they come
from (so we know nothing has changed since lfs_bmapv was called).
- Fragment allocation is fixed, so that writes beyond end-of-file
do the right thing.
|
| 1.14 | 09-Nov-1998 |
mycroft | GC the B_CACHE bit.
|
| 1.13 | 23-Oct-1998 |
thorpej | Use DINODE_SIZE rather than sizeof(struct dinode).
|
| 1.12 | 11-Sep-1998 |
pk | PR#6032: define fixed sized on-disk superblock structure.
|
| 1.11 | 08-May-1998 |
kleink | Fix some arithmetics lossage on typeless pointers.
|
| 1.10 | 01-Mar-1998 |
fvdl | Merge with Lite2 + local changes
|
| 1.9 | 13-Jun-1997 |
pk | TIMESPEC_TO_TIMEVAL => TIMEVAL_TO_TIMESPEC
|
| 1.8 | 11-Jun-1997 |
bouyer | Add support for ext2fs, this needed a few modifications to ufs/ufs/inode.h:
- added an "union inode_ext" to struct inode, for the per-fs extentions.
For now only ext2fs uses it.
- i_din is now an union:
union {
struct dinode ffs_din; /* 128 bytes of the on-disk dinode. */
struct ext2fs_dinode e2fs_din; /* 128 bytes of the on-disk dinode. */
} i_din
Added a lot of #define i_ffs_* and i_e2fs_* to access the fields.
- Added two macros: FFS_ITIMES and EXT2FS_ITIMES. ITIMES calls the rigth
macro, depending on the time of the inode. ITIMES is used where necessary,
FFS_ITIMES and EXT2FS_ITIMES in other places.
|
| 1.7 | 12-Oct-1996 |
christos | revert previous kprintf changes
|
| 1.6 | 10-Oct-1996 |
christos | printf -> kprintf, sprintf -> ksprintf
|
| 1.5 | 01-Sep-1996 |
mycroft | Add a set of generic file system operations that most file systems use.
Also, fix some time stamp bogosities.
|
| 1.4 | 09-Feb-1996 |
christos | lfs prototypes
|
| 1.3 | 21-Aug-1994 |
cgd | C syntax fix, and syscall args style (For later.)
|
| 1.2 | 29-Jun-1994 |
cgd | New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
|
| 1.1 | 08-Jun-1994 |
mycroft | branches: 1.1.1;
Update to 4.4-Lite fs code, with local changes.
|
| 1.1.1.2 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite2
|
| 1.1.1.1 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite for reference
|
| 1.23.2.10 | 20-Jan-2000 |
he | Pull up revision 1.39 (requested by perseant):
Files removed (through unlink, rmdir) are now really removed, though the
removal is postponed until the dirop is complete to ensure validity of
the filesystem through a crash. Use a separate per-fs lock, instead of
ufs_hashlock, to protect the inode free list. Change calling semantics
of lfs_ifind, to give better error reporting: If fed a struct buf, it
can report the block number of the offending inode block as well as the
inode number.
|
| 1.23.2.9 | 15-Jan-2000 |
he | Pull up revision 1.38 (requested by perseant):
Handle flushing a vnode during cleaning, and cleaning the Ifile,
more correctly, avoiding possible disk corruption in some cases.
|
| 1.23.2.8 | 15-Jan-2000 |
he | Pull up revision 1.30 (requested by perseant):
Address problems related to using an LFS filesystem as the root
filesystem, including mknod hangs. Fixes PR#8172 and PR#9072.
|
| 1.23.2.7 | 18-Dec-1999 |
he | Pull up revision 1.37 (requested by perseant):
Handle the case of a vnode flush while dirops are active correctly
in lfs_segwrite. Also, make sure a flush is called in SET_DIROP
before sleeping on its results. Addresses PR#8863.
|
| 1.23.2.6 | 17-Dec-1999 |
he | Pull up revision 1.32 (requested by perseant):
Address locking protocol error for inode hash, and make the
maximum number of active dirops a global quantity.
|
| 1.23.2.5 | 16-Dec-1999 |
he | Pull up revision 1.36 (requested by perseant):
Fix spllevel problem with superblock exclusion and with write
throttle. Addressess PR#8383.
|
| 1.23.2.4 | 10-Oct-1999 |
cgd | pull up rev 1.31 from trunk (requested by mycroft):
Fix potential overflow of v_usecount and v_writecount (and panics
resulting from this) by widening them to `long'. Mostly affects
systems where maxvnodes>=32768.
|
| 1.23.2.3 | 03-Sep-1999 |
he | Pull up revision 1.28:
Fix a printf format bug that gives compiler warnings/errors on
64-bit platforms, fixing PR#8241. (perseant)
|
| 1.23.2.2 | 25-Jun-1999 |
perry | pullup 1.26->1.27 (perseant)
|
| 1.23.2.1 | 13-Apr-1999 |
perseant | branches: 1.23.2.1.2; 1.23.2.1.4;
Pull-up of changes made to the trunk on Sunday [1.23->1.26], to wit:
Take out the `#ifdef USE_UFSHASH'; use ufs_hashlock to lock the inode free
list instead of free_lock.
Fix inode reporting in lfs_statfs (the meaning of f_files and f_ffree was
reversed).
Fix "lfs_ifind: dinode xxx not found" panic. When inodes were freed, then
immediately reloaded, their dinodes were located in an inode block which
was not on disk at the advertized location, nor in the cache (although it
would be flushed to disk next segment write). Fix this by using getblk()
instead of lfs_newbuf() for inode blocks.
Better checking for held inode locks in lfs_fastvget, for a number of
error conditions. Also change the default setting of lfs_clean_vnhead to
0, which seems to make the locking problems go away (although this is
difficult to test as I can't reliably reproduce them).
Make sure that the wakeup occurs for vnodes that lfs_update might be
sleeping on (nodes which are not marked IN_MODIFIED/IN_CLEANING, but which
have dirty buffers), by marking them with the appropriate flag if
dirtybuffers were added while the write was in progress.
Fix block counting during file truncation, if not truncating to zero.
Disallow threshold-initiated cache flush when dirops are active. Also,
make SET_ENDOP use lfs_check instead of inlining most of it.
Improve the debugging printfs in the cleaner syscalls (in particular, make
it obvious that they're coming from lfs).
Check the superblock version field, and refuse to mount the filesystem if
the version number is higher than we know about. This allows, e.g.,
changes in the format of the ifile, segment size restrictions and
boundaries, etc., which would not affect existing fields in the
superblock, but which would drastically affect the filesystem, to be
smoothly integrated at a later date.
|
| 1.23.2.1.4.1 | 30-Nov-1999 |
itojun | bring in latest KAME (as of 19991130, KAME/NetBSD141) into kame branch
just for reference purposes.
This commit includes 1.4 -> 1.4.1 sync for kame branch.
The branch does not compile at all (due to the lack of ALTQ and some other
source code). Please do not try to modify the branch, this is just for
referenre purposes.
synchronization to latest KAME will take place on HEAD branch soon.
|
| 1.23.2.1.2.4 | 31-Aug-1999 |
perseant | Rudimentary support for LFS under UBC:
- LFS-specific VOP_BALLOC and VOP_PUTPAGES vnode ops.
- getblk VREG panic #ifdef'd out (can be reinstated when Ifile is
internalized and Ifile can be made another type from VREG)
- interface to VOP_PUTPAGES changed to pass all pager flags, not
just sync. FS putpages routines must know about the pager flags.
- new LFS magic disk address, -2 ("unwritten"), meaning accounted for
but not assigned to a fixed disk location (since LFS does these two
things separately, and the previous accounting method using buffer
headers no longer will work). Changed references to (foo == (daddr_t)-1)
to (foo < 0). Since disk drivers reject all addresses < 0, this should
not present a problem for other FSs.
|
| 1.23.2.1.2.3 | 02-Aug-1999 |
thorpej | Update from trunk.
|
| 1.23.2.1.2.2 | 21-Jun-1999 |
thorpej | Correct a printf format now that vnode flags are an int (in the uvm_vnode
structure).
|
| 1.23.2.1.2.1 | 21-Jun-1999 |
thorpej | Sync w/ -current.
|
| 1.31.6.2 | 27-Dec-1999 |
wrstuden | Pull up to last week's -current.
|
| 1.31.6.1 | 21-Dec-1999 |
wrstuden | Initial commit of recent changes to make DEV_BSIZE go away.
Runs on i386, needs work on other arch's. Main kernel routines should be
fine, but a number of the stand programs need help.
cd, fd, ccd, wd, and sd have been updated. sd has been tested with non-512
byte block devices. vnd, raidframe, and lfs need work.
Non 2**n block support is automatic for LKM's and conditional for kernels
on "options NON_PO2_BLOCKS".
|
| 1.31.4.2 | 15-Nov-1999 |
fvdl | Sync with -current
|
| 1.31.4.1 | 19-Oct-1999 |
fvdl | Bring in Kirk McKusick's FFS softdep code on a branch.
|
| 1.31.2.4 | 18-Jan-2001 |
bouyer | Sync with head (for UBC+NFS fixes, mostly).
|
| 1.31.2.3 | 08-Dec-2000 |
bouyer | Sync with HEAD.
|
| 1.31.2.2 | 22-Nov-2000 |
bouyer | Sync with HEAD.
|
| 1.31.2.1 | 20-Nov-2000 |
bouyer | Update thorpej_scsipi to -current as of a month ago
|
| 1.32.2.2 | 06-Nov-1999 |
perseant | Address ufs_hashlock/ufs_ihashins protocol bug, discovered while doing a
post-mortem of a production machine. Also, take the active dirop
count off of the fs and make it global (since it is measuring a global
resource) and tie the threshold value LFS_MAXDIROP to desiredvnodes.
|
| 1.32.2.1 | 06-Nov-1999 |
perseant | file lfs_segment.c was added on branch comdex-fall-1999 on 1999-11-06 20:33:06 +0000
|
| 1.46.2.1 | 22-Jun-2000 |
minoura | Sync w/ netbsd-1-5-base.
|
| 1.49.2.4 | 03-Feb-2001 |
he | Pull up revisions 1.60-1.62 (requested by perseant):
o Don't write anything if the filesystem is idle (PR#10979).
o Close up accounting holes in LFS' accounting of immediately-
available-space, number of clean segments, and amount of dirty
space taken up by metadata (PR#11468, PR#11470, PR#11534).
|
| 1.49.2.3 | 14-Sep-2000 |
perseant | Pull up recent LFS kernel changes (approved by thorpej):
ufs/ufs/inode.h, 1.20--1.22 (add i_lfs_effnblks extension ;
make ITIMES aware of LFS_ITIMES;
_LKM protection so userland progs
compile)
ufs/ufs/ufs_vnops.c, 1.69, 1.71 (remove IN_ADIROP;
use ITIMES instead of FFS_ITIMES)
ufs/ufs/ufs_readwrite.c, 1.27 (use lfs_reserve in lfs_write)
ufs/lfs/lfs.h, 1.26--1.32 (define LFS_EST_* macros ;
change MIN_FREE_SEGS to lfs_minfreesegs ;
add avail and bfree to CLEANERINFO ;
change lfs_uinodes to signed ;
change lfs_dmeta to signed ;
add whitespace to line up structure
members ;
explicit cast to int32_t in LFS_EST_*
macros)
ufs/lfs/lfs_alloc.c, back out 1.34.2.3 (pullups of 1.39, 1.40);
then pull up 1.38 (clean up on error)
1.39--1.43 (restore fvdl's ufs_hashlock fix ;
restore fvdl's ufs_hashlock fix ;
set i_lfs_effnblks ;
use UINO macros ;
add comments and fix long lines)
ufs/lfs/lfs_balloc.c, 1.19 (don't succeed halfway)
1.21--1.25 (use i_lfs_effnblks ;
fix i_lfs_effnblks computation and
quieten ;
fix i_ffs_blocks in unwritten fragment ;
remove useless debugging check ;
add comments and (c) 2000)
ufs/lfs/lfs_bio.c, 1.24--1.30 (cleanup and make lfs_flush_fs take
"struct lfs *" instead of "struct
mount *" ;
use lfs_minfreeseg instead of
MIN_FREE_SEGS ;
use UINO macros, and copy bfree/avail
to CLEANERINFO ;
add lfs_reserve function ;
1.28--1.30 fix printf formatting)
ufs/lfs/lfs_cksum.c, 1.13 (add (c) 2000)
ufs/lfs/lfs_debug.c, 1.11 (use btodb instead of DEV_BSIZE)
ufs/lfs/lfs_extern.h, 1.18, 1.20--1.21 (function prototype changes)
ufs/lfs/lfs_inode.c, 1.38 (rewrite lfs_truncate from
ffs_truncate)
1.40--1.44 (count written and unwritten blocks
seperately ;
use disk block units instead of bytes ;
remove unnecessary "mod" variable ;
correct B_DELWRI to avoid bawrite panic ;
use lfs_reserve)
ufs/lfs/lfs_segment.c, 1.52-1.59 (use lfs_dmeta to note used summaries ;
check for UNWRITTEN in indirect blocks ;
more debugging stuff inside #ifdef
DEBUG_LFS ;
use LK_CANRECURSE ;
don't drop dirty indirect blocks ;
use UINO macros ;
don't hose the free list ;
use btodb() instead of DEV_BSIZE ;
make it compile again (oops))
ufs/lfs/lfs_subr.c, 1.16--1.17 (check for locked inodes before
changing ;
use btodb() instead of DEV_BSIZE, (c)
2000)
ufs/lfs/lfs_syscalls.c, back out 1.41.4.2 (fvdl's ufs_hashlock fix);
then pull up 1.43 (use lfs_dmeta)
1.44--1.45 (restore fvdl's ufs_hashlock fix)
1.46--1.47 (fix lfs_avail leakage from sblock
segments ;
use UINO macros)
1.49 (bounds-check inode numbers in
lfs_markv)
ufs/lfs/lfs_vfsops.c, 1.53 (use LFS_EST_* macros in lfs_statfs)
1.56--1.58 (initialize lfs_minfreeseg, lfs_effnblk ;
initialize lfs_uinodes ;
initialize lfs_ravail)
ufs/lfs/lfs_vnops.c, 1.40 (remove VDIROP from removed files)
1.42--1.44 (move SET_ENDOP below the removal of
VDIROP ;
use UINO macros and add lfs_itimes
function ;
use lfs_reserve in dirops)
|
| 1.49.2.2 | 28-Jun-2000 |
perseant | pull up active current segment patch from trunk
|
| 1.49.2.1 | 22-Jun-2000 |
perseant | Pull up lfs_vunref fix from the trunk.
|
| 1.67.2.13 | 08-Jan-2003 |
thorpej | Oh my aching HEAD.
|
| 1.67.2.12 | 08-Jan-2003 |
thorpej | Sync with HEAD.
|
| 1.67.2.11 | 03-Jan-2003 |
thorpej | Sync with HEAD.
|
| 1.67.2.10 | 19-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.67.2.9 | 18-Oct-2002 |
nathanw | Catch up to -current.
|
| 1.67.2.8 | 01-Aug-2002 |
nathanw | Catch up to -current.
|
| 1.67.2.7 | 24-Jun-2002 |
nathanw | Curproc->curlwp renaming.
Change uses of "curproc->l_proc" back to "curproc", which is more like the
original use. Bare uses of "curproc" are now "curlwp".
"curproc" is now #defined in proc.h as ((curlwp) ? (curlwp)->l_proc) : NULL)
so that it is always safe to reference curproc (*de*referencing curproc
is another story, but that's always been true).
|
| 1.67.2.6 | 20-Jun-2002 |
nathanw | Catch up to -current.
|
| 1.67.2.5 | 08-Jan-2002 |
nathanw | Catch up to -current.
|
| 1.67.2.4 | 14-Nov-2001 |
nathanw | Catch up to -current.
|
| 1.67.2.3 | 24-Aug-2001 |
nathanw | Catch up with -current.
|
| 1.67.2.2 | 21-Jun-2001 |
nathanw | Catch up to -current.
|
| 1.67.2.1 | 05-Mar-2001 |
nathanw | Initial commit of scheduler activations and lightweight process support.
|
| 1.68.4.5 | 10-Oct-2002 |
jdolecek | sync kqueue with -current; this includes merge of gehenna-devsw branch,
merge of i386 MP branch, and part of autoconf rototil work
|
| 1.68.4.4 | 06-Sep-2002 |
jdolecek | sync kqueue branch with HEAD
|
| 1.68.4.3 | 23-Jun-2002 |
jdolecek | catch up with -current on kqueue branch
|
| 1.68.4.2 | 10-Jan-2002 |
thorpej | Sync kqueue branch with -current.
|
| 1.68.4.1 | 03-Aug-2001 |
lukem | update to -current
|
| 1.68.2.3 | 02-Jul-2001 |
perseant | Change disk addressing unit to be the fragment, instead of the disk sector.
All quantities in the superblock, inodes, indirect blocks, etc. refer now
to this abstract unit (called "fsb" as it is in FFS) instead of disk sectors;
as a consequence segment summary blocks have to be multiples of a fragment in
size. In v1 filesystems, compatibility code ensures that 1 fsb == 1 sector,
regardless of fragment size.
Fragments can now range in size between 512 and 32k; in the event that
LFS_LABELPAD (8k) is smaller than the disk address unit size, an extra
proto-superblock is kept at 8k from the beginning of the disk, to be used
*only* to locate the real superblocks. (Not all of the userland knows about
this yet.)
Almost all of this was done not by me, but by joff.
|
| 1.68.2.2 | 29-Jun-2001 |
perseant | Get rid of __P(), protoizing where it had not already been done
|
| 1.68.2.1 | 27-Jun-2001 |
perseant | Import of what I've been calling "LFSv2", that is, LFS with some features
added that require changes to the on-disk data structures. These include:
- 64-bit time in everything but inodes
- User-specified segment offset, and segment size no longer
restricted to PO2.
- Serial number on segment summaries in addition to timestamp, and
a new volume identifier, to make roll-forward feasible without
fear of finding old data and thinking it was new.
Although I think this version works at least as well as what's on the trunk,
we're not done yet; hence this commit is going in on a branch and not on
the trunk. Enhancements that are not here yet include fragment addressing,
like FFS does, instead of block addressing.
|
| 1.70.4.1 | 12-Nov-2001 |
thorpej | Sync the thorpej-mips-cache branch with -current.
|
| 1.70.2.1 | 07-Sep-2001 |
thorpej | Commit my "devvp" changes to the thorpej-devvp branch. This
replaces the use of dev_t in most places with a struct vnode *.
This will form the basic infrastructure for real cloning device
support (besides being architecurally cleaner -- it'll be good
to get away from using numbers to represent objects).
|
| 1.74.2.3 | 15-Jul-2002 |
gehenna | catch up with -current.
|
| 1.74.2.2 | 20-Jun-2002 |
gehenna | catch up with -current.
|
| 1.74.2.1 | 30-May-2002 |
gehenna | Catch up with -current.
|
| 1.76.2.3 | 20-Jun-2002 |
lukem | Pull up revision 1.79 (requested by perseant in ticket #325):
For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.76.2.2 | 02-Jun-2002 |
tv | Pull up revision 1.78 (requested by perseant in ticket #135):
Fix a couple of instances where reassignbuf() was not done at splbio.
Tested on i386.
|
| 1.76.2.1 | 02-Jun-2002 |
tv | Pull up revision 1.77 (requested by perseant in ticket #132):
Back out rev 1.174 of vfs_subr.c, because the splbio() wasn't protecting
enough to be useful, and broadening it so that it did would have meant
that operations possibly requiring synchronous disk activity would have
to be done in splbio(). This clearly was not going to work.
Worked around this in the LFS case by having lfs_cluster_callback put an
extra hold on the vnode before calling biodone(), and taking the hold
off without HOLDRELE's problematic list swapping. lfs_vunref() will take
care of that---in thread context---on the next write if need be.
Also, ensure that the list walking in lfs_{writevnodes,segunlock,gather}
takes into account the possibility that the list may change
underneath it (possibly because it itself deleted an element).
Tested on i386, test-compiled on alpha.
|
| 1.124.2.10 | 10-Nov-2005 |
skrll | Sync with HEAD. Here we go again...
|
| 1.124.2.9 | 08-Mar-2005 |
skrll | Sync with HEAD.
|
| 1.124.2.8 | 04-Mar-2005 |
skrll | Sync with HEAD.
Hi Perry!
|
| 1.124.2.7 | 24-Sep-2004 |
skrll | Sync with HEAD.
|
| 1.124.2.6 | 21-Sep-2004 |
skrll | Fix the sync with head I botched.
|
| 1.124.2.5 | 18-Sep-2004 |
skrll | Sync with HEAD.
|
| 1.124.2.4 | 25-Aug-2004 |
skrll | Sync with HEAD.
|
| 1.124.2.3 | 24-Aug-2004 |
skrll | Undo part of the ktrace/lwp changes. In particular:
* Remove the "lwp *" argument that was added to vget(). Turns out
that nothing actually used it!
* Remove the "lwp *" arguments that were added to VFS_ROOT(), VFS_VGET(),
and VFS_FHTOVP(); all they did was pass it to vget() (which, as noted
above, didn't use it).
* Remove all of the "lwp *" arguments to internal functions that were added
just to appease the above.
|
| 1.124.2.2 | 03-Aug-2004 |
skrll | Sync with HEAD
|
| 1.124.2.1 | 02-Jul-2003 |
darrenr | Apply the aborted ktrace-lwp changes to a specific branch. This is just for
others to review, I'm concerned that patch fuziness may have resulted in some
errant code being generated but I'll look at that later by comparing the diff
from the base to the branch with the file I attempt to apply to it. This will,
at the very least, put the changes in a better context for others to review
them and attempt to tinker with removing passing of 'struct lwp' through
the kernel.
|
| 1.152.4.1 | 10-May-2005 |
riz | Pull up the following revisions (requested by perseant in ticket #1281):
1.8 sys/ufs/lfs/TODO
1.75 sys/ufs/lfs/lfs.h (via patch)
1.74 sys/ufs/lfs/lfs_alloc.c (via patch)
1.49, 1.51 sys/ufs/lfs/lfs_balloc.c (1.51 via patch)
1.78 sys/ufs/lfs/lfs_bio.c
1.62 sys/ufs/lfs/lfs_extern.h (via patch)
1.156 sys/ufs/lfs/lfs_segment.c (via patch)
1.48 sys/ufs/lfs/lfs_subr.c
1.101 sys/ufs/lfs/lfs_syscalls.c
1.163 sys/ufs/lfs/lfs_vfsops.c (via patch)
1.134 sys/ufs/lfs/lfs_vnops.c (via patch)
1.61 sys/ufs/ufs/ufs_readwrite.c (via patch)
1.20 libexec/lfs_cleanerd/clean.h (via patch)
1.52 libexec/lfs_cleanerd/cleanerd.c (via patch)
1.41 libexec/lfs_cleanerd/library.c (via patch)
1.4 regress/sys/fs/lfs/newfs_fsck/Makefile
1.2 regress/sys/fs/lfs/newfs_fsck/mkfs_mount
1.2 regress/sys/fs/lfs/newfs_fsck/smallfiles
1.3 sbin/fsck_lfs/bufcache.c
1.3 sbin/fsck_lfs/bufcache.h
1.3 sbin/fsck_lfs/lfs.h
1.8 sbin/fsck_lfs/lfs.c (via patch)
1.8 sbin/fsck_lfs/pass3.c (via patch)
1.18 sbin/fsck_lfs/pass0.c (via patch)
1.18 sbin/fsck_lfs/utilities.c (via patch)
1.7 sbin/fsck_lfs/segwrite.c
1.19 sbin/fsck_lfs/setup.c (via patch)
1.3 sbin/newfs_lfs/Makefile
0 sbin/newfs_lfs/lfs.c (yes, remove it)
1.1 sbin/newfs_lfs/make_lfs.c
1.15 sbin/newfs_lfs/newfs.c (via patch)
Various minor LFS improvements.
Kernel:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this. Should fix PR #29045.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
Fixes PR #26680.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
cleaner:
* Adapt lfs_cleanerd to use the fcntl call to get the Ifile filehandle,
so it need not be in the namespace.
* Make lfs_cleanerd be more careful when there are very few available
segments.
* Make lfs_cleanerd less verbose when the filesystem is unmounted.
newfs_lfs, fsck_lfs, and regression:
* Extend the lfs library from fsck_lfs(8) so that it can be used with a
not-yet-existent LFS. Make newfs_lfs(8) use this library, so it can
create LFSs whose Ifile is larger than one segment. Addresses PR #11110.
* Make newfs_lfs(8) use strsuftoi64() for its arguments, a la newfs(8).
* Make fsck_lfs(8) respect the "file system is clean" flag.
* Don't let fsck_lfs(8) think it has dirty blocks when invoked with the
-n flag.
* Remove the Ifile from the filesystem namespace. The cleaner now uses
a fcntl call on the root inode to find the Ifile filehandle. (As a
side-effect, addresses PR #29144.)
|
| 1.155.6.1 | 19-Mar-2005 |
yamt | sync with head. xen and whitespace. xen part is not finished.
|
| 1.155.4.1 | 29-Apr-2005 |
kent | sync with -current
|
| 1.158.2.13 | 10-Aug-2006 |
tron | Apply patch (requested by fair in perseant #1457):
Bring LFS up to current, including a patch (1.95 lfs_alloc.c) that
should prevent the inode free list errors seen on the STABLE branch
subsequent to pullup ticket #1327.
|
| 1.158.2.12 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.93
sys/ufs/lfs/lfs.h: revision 1.106
sys/ufs/lfs/lfs_vfsops.c: revision 1.209
sys/ufs/lfs/lfs_vnops.c: revision 1.175
sys/ufs/lfs/lfs_segment.c: revision 1.178
Fixes to address the "vinvalbuf: dirty blocks" panic that can occur when
many inodes are cleaned at once. Make sure that we write all the pages
on vnodes that are being flushed, even if we don't think there's room;
drain v_numoutput before lfs_vflush() completes.
Also, don't allow a vnode that is in the process of being cleaned to be
chosen by getnewvnode(); this avoids a segment accounting panic in the case
that a large number of inodes are fed to lfs_markv() all at once.
|
| 1.158.2.11 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.171
sys/ufs/lfs/lfs_extern.h: revision 1.81
sys/ufs/lfs/lfs_segment.c: revision 1.177
Don't ever partially write dirops, even if we need the cleaner to run.
This increases the chances of the "no clean segments" panic slightly,
but allows us to run the ckckp regression test successfully to completion.
|
| 1.158.2.10 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.104
sys/ufs/lfs/lfs_vfsops.c: revision 1.206
sys/ufs/lfs/lfs_vnops.c: revision 1.170
sys/ufs/lfs/lfs_extern.h: revision 1.80
sys/ufs/lfs/lfs_segment.c: revision 1.176
sys/ufs/lfs/lfs_inode.c: revision 1.103 via patch
sys/ufs/lfs/lfs_alloc.c: revision 1.90
Postpone the segment accounting changes coming from truncation until the
inode that makes those changes valid is either written to disk by
lfs_writeinode() or discarded by lfs_vfree().
A couple of locking fixes are also included as well.
|
| 1.158.2.9 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_segment.c: revision 1.175
Regression test improvements:
Move the stop for LFCNWRAPSTOP to the point at which writing at segment 0
is really about to commence, since this is what the test expects (and
incidentally what a snapshotting utility wants as well).
More correctly reconstruct the on-disk state at every checkpoint, rather
than relying on the entire state at the point of wrapping to be accurate
(that is only true the first time we wrap). Add a "make abort" target to
make rerunning the test more convenient when it has failed and we're done
analyzing the failure.
|
| 1.158.2.8 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.103
sys/ufs/lfs/lfs_segment.c: revision 1.174
sys/ufs/lfs/lfs_vnops.c: revision 1.168
Introduce two fcntl calls that freeze the filesystem right at the point
where segment 0 is being considered for writing. This allows for automated
checkpoint vailidity scanning, and could be used (in conjunction with the
existing LFCNREWIND) for e.g. snapshot dumps as well.
Include a regression test that does such scanning.
When writing the Ifile, loop through the dirty block list three times to
make sure that the checkpoint is always consistent (the first and second
times the Ifile blocks can cross a segment boundary; not so the third time
unless the segments are very small). Discovered by using the aforementioned
regression test.
|
| 1.158.2.7 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.102
sys/ufs/lfs/lfs_segment.c: revision 1.173
sys/ufs/lfs/lfs_vnops.c: revision 1.167 via patch
sys/ufs/lfs/lfs_bio.c: revision 1.91
Make lfs_vref/lfs_vunref not need to know about VXLOCK and VFREEING
explicitly (especially since we didn't know about VFREEING at all before),
but notice the EBUSY return from vget() instead.
Fix some more MP locking protocol issues, most of which were pointed out by
Christian Ehrhardt this morning on tech-kern.
|
| 1.158.2.6 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_balloc.c: revision 1.60
sys/ufs/lfs/lfs_syscalls.c: revision 1.111
sys/ufs/lfs/lfs_segment.c: revision 1.172
sys/ufs/lfs/lfs_vnops.c: revision 1.163
Several minor bug fixes:
* Correct (weak) segment lock assertions in lfs_fragextend and lfs_putpages.
* Keep IN_MODIFIED set if we run out of avail in lfs_putpages.
* Don't try to (re)write buffers on a VBLK vnode; fixes a panic I found
while running with an LFS root.
* Raise priority of LFCNSEGWAIT to PVFS; PUSER is way too low for
something the pagedaemon is relying on.
|
| 1.158.2.5 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.158
sys/ufs/lfs/lfs_subr.c: revision 1.57
sys/ufs/lfs/lfs_segment.c: revision 1.171
sys/ufs/lfs/lfs.h: revision 1.97
sys/ufs/lfs/lfs_vfsops.c: revision 1.195
sys/ufs/lfs/lfs_extern.h: revision 1.76
Improvements to LFS's paging mechanism, to wit:
* Acknowledge that sometimes there are more dirty pages to be written to
disk than clean segments. When we reach the danger line,
lfs_gop_write() now returns EAGAIN. The caller of VOP_PUTPAGES(), if
it holds the segment lock, drops it and waits for the cleaner to make
room before continuing.
* Note and avoid a three-way deadlock in lfs_putpages (a writer holding
a page busy blocks on the cleaner while the cleaner blocks on the
segment lock while lfs_putpages blocks on the page).
|
| 1.158.2.4 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_segment.c: revision 1.170
sys/ufs/lfs/lfs.h: revision 1.96
sys/ufs/lfs/lfs_vfsops.c: revision 1.194
sys/ufs/lfs/lfs_syscalls.c: revision 1.109
From Konrad Schroeder, in response to strange df output on anoncvs.netbsd.org:
We were returning the wrong value for free space. Now we're not.
|
| 1.158.2.3 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.153
sys/ufs/lfs/lfs_debug.c: revision 1.32
sys/ufs/lfs/lfs_alloc.c: revision 1.84
sys/ufs/lfs/lfs_vfsops.c: revision 1.185
sys/ufs/lfs/lfs_segment.c: revision 1.165
64 bit inode changes.
|
| 1.158.2.2 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.152
sys/ufs/lfs/lfs_debug.c: revision 1.31
sys/ufs/lfs/lfs_subr.c: revision 1.53
sys/ufs/lfs/lfs_extern.h: revision 1.68
sys/ufs/lfs/lfs_inode.c: revision 1.96
sys/ufs/lfs/lfs_bio.c: revision 1.86
sys/ufs/lfs/lfs_alloc.c: revision 1.83
sys/ufs/lfs/lfs_vfsops.c: revision 1.181
sys/ufs/lfs/lfs.h: revision 1.88
sys/ufs/lfs/lfs_segment.c: revision 1.164
- sprinkle const
- avoid shadow variables.
|
| 1.158.2.1 | 07-May-2005 |
tron | Apply patch (requested by perseant in ticket #242):
* fsck_lfs buffer cache fixes, including PR #29151
* Change fsck_lfs phase 0 message to reflect reality
* fsck_lfs: check phase 5 (cleanerinfo accounting) even on
roll-forward
* Keep better track of the free list during roll-forward, avoiding
a core dump
* Improve hash table use for fsck_lfs buffer and vnode cache
* Document fsck_lfs flag -f, and implement -q
* Add resize_lfs, including kernel support
* Add LFS to mountd's list of exportable filesystem types
* Make the LFS lkm work again [christos@]
* Add MP locking to the LFS kernel subsystem
* Fix pager_map deadlock in lfs_putpages()
* Avoid incomplete file extension that looks like "partial
truncation" to fsck
* Use lfs_malloc for cleaner malloc, since the cleaner often runs
in low-memory conditions.
* Use splay trees, not hash table, to track page allocation for
write.
* Fix mkdir panic on full fs
* Fix page accounting leak by counting differently.
* Use rightly named structure for lfs_getattr [skrll@]
* Cosmetic changes for readability.
|
| 1.164.2.8 | 27-Feb-2008 |
yamt | sync with head.
|
| 1.164.2.7 | 04-Feb-2008 |
yamt | sync with head.
|
| 1.164.2.6 | 21-Jan-2008 |
yamt | sync with head
|
| 1.164.2.5 | 27-Oct-2007 |
yamt | sync with head.
|
| 1.164.2.4 | 03-Sep-2007 |
yamt | sync with head.
|
| 1.164.2.3 | 26-Feb-2007 |
yamt | sync with head.
|
| 1.164.2.2 | 30-Dec-2006 |
yamt | sync with head.
|
| 1.164.2.1 | 21-Jun-2006 |
yamt | sync with head.
|
| 1.168.2.1 | 15-Jan-2006 |
yamt | sync with head.
|
| 1.169.10.2 | 24-May-2006 |
tron | Merge 2006-05-24 NetBSD-current into the "peter-altq" branch.
|
| 1.169.10.1 | 28-Mar-2006 |
tron | Merge 2006-03-28 NetBSD-current into the "peter-altq" branch.
|
| 1.169.8.3 | 11-May-2006 |
elad | sync with head
|
| 1.169.8.2 | 06-May-2006 |
christos | - Move kauth_cred_t declaration to <sys/types.h>
- Cleanup struct ucred; forward declarations that are unused.
- Don't include <sys/kauth.h> in any header, but include it in the c files
that need it.
Approved by core.
|
| 1.169.8.1 | 19-Apr-2006 |
elad | sync with head.
|
| 1.169.6.6 | 03-Sep-2006 |
yamt | sync with head.
|
| 1.169.6.5 | 11-Aug-2006 |
yamt | sync with head
|
| 1.169.6.4 | 26-Jun-2006 |
yamt | sync with head.
|
| 1.169.6.3 | 24-May-2006 |
yamt | sync with head.
|
| 1.169.6.2 | 11-Apr-2006 |
yamt | sync with head
|
| 1.169.6.1 | 01-Apr-2006 |
yamt | sync with head.
|
| 1.169.4.3 | 01-Jun-2006 |
kardel | Sync with head.
|
| 1.169.4.2 | 22-Apr-2006 |
simonb | Sync with head.
|
| 1.169.4.1 | 04-Feb-2006 |
simonb | Adapt for timecounters: mostly use get*time() and use "time_second"
instead of "time.tv_sec".
|
| 1.169.2.1 | 09-Sep-2006 |
rpaulo | sync with head
|
| 1.180.2.1 | 19-Jun-2006 |
chap | Sync with head.
|
| 1.182.2.1 | 13-Jul-2006 |
gdamore | Merge from HEAD.
|
| 1.190.4.3 | 10-Dec-2006 |
yamt | sync with head.
|
| 1.190.4.2 | 22-Oct-2006 |
yamt | use workqueue for aiodoned.
|
| 1.190.4.1 | 22-Oct-2006 |
yamt | sync with head
|
| 1.190.2.2 | 12-Jan-2007 |
ad | Sync with head.
|
| 1.190.2.1 | 18-Nov-2006 |
ad | Sync with head.
|
| 1.195.4.1 | 03-Sep-2007 |
wrstuden | Sync w/ NetBSD-4-RC_1
|
| 1.195.2.1 | 05-Jun-2007 |
bouyer | Pull up following revision(s) (requested by perseant in ticket #703):
sys/miscfs/genfs/genfs.h 1.21
sys/miscfs/genfs/genfs_vnops.c 1.151
sys/ufs/lfs/lfs.h 1.119, 1.120
sys/ufs/lfs/lfs_bio.c 1.99-101
sys/ufs/lfs/lfs_extern.h 1.89
sys/ufs/lfs/lfs_inode.c 1.108, 1.109
sys/ufs/lfs/lfs_segment.c 1.197, 1.199, 1.200
sys/ufs/lfs/lfs_subr.c 1.69, 1.70
sys/ufs/lfs/lfs_syscalls.c 1.119
sys/ufs/lfs/lfs_vfsops.c 1.234, 1.235
sys/ufs/lfs/lfs_vnops.c 1.195, 1.196, 1.200, 1.202-206
Reduce busy waiting in lfs_putpages(), and other LFS improvements.
|
| 1.196.2.4 | 17-May-2007 |
yamt | sync with head.
|
| 1.196.2.3 | 07-May-2007 |
yamt | sync with head.
|
| 1.196.2.2 | 12-Mar-2007 |
rmind | Sync with HEAD.
|
| 1.196.2.1 | 27-Feb-2007 |
yamt | - sync with head.
- move sched_changepri back to kern_synch.c as it doesn't know PPQ anymore.
|
| 1.198.4.1 | 11-Jul-2007 |
mjf | Sync with head.
|
| 1.198.2.13 | 01-Oct-2007 |
ad | Make it compile (XXX not correct).
|
| 1.198.2.12 | 28-Aug-2007 |
yamt | - mark aiodone workqueue MPSAFE.
- make lfs callbacks acquire kernel_lock by themselves.
ok'ed by Andrew Doran.
|
| 1.198.2.11 | 28-Aug-2007 |
yamt | make this compilable with DEBUG.
|
| 1.198.2.10 | 24-Aug-2007 |
ad | Sync with buffer cache locking changes. See buf.h/vfs_bio.c for details.
Some minor portions are incomplete and needs to be verified as a whole.
|
| 1.198.2.9 | 20-Aug-2007 |
ad | Sync with HEAD.
|
| 1.198.2.8 | 19-Aug-2007 |
ad | - Back out the biodone() changes.
- Eliminate B_ERROR (from HEAD).
|
| 1.198.2.7 | 15-Jul-2007 |
ad | Sync with head.
|
| 1.198.2.6 | 23-Jun-2007 |
ad | - Lock v_cleanblkhd, v_dirtyblkhd, v_numoutput with the vnode's interlock.
Get rid of global_v_numoutput_lock. Partially incomplete as the buffer
cache locking doesn't work very well and needs an overhaul.
- Some changes to try and make softdep MP safe. Untested.
|
| 1.198.2.5 | 17-Jun-2007 |
ad | - Increase the number of thread priorities from 128 to 256. How the space
is set up is to be revisited.
- Implement soft interrupts as kernel threads. A generic implementation
is provided, with hooks for fast-path MD code that can run the interrupt
threads over the top of other threads executing in the kernel.
- Split vnode::v_flag into three fields, depending on how the flag is
locked (by the interlock, by the vnode lock, by the file system).
- Miscellaneous locking fixes and improvements.
|
| 1.198.2.4 | 08-Jun-2007 |
ad | Sync with head.
|
| 1.198.2.3 | 13-May-2007 |
ad | - Pass the error number and residual count to biodone(), and let it handle
setting error indicators. Prepare to eliminate B_ERROR.
- Add a flag argument to brelse() to be set into the buf's flags, instead
of doing it directly. Typically used to set B_INVAL.
- Add a "struct cpu_info *" argument to kthread_create(), to be used to
create bound threads. Change "bool mpsafe" to "int flags".
- Allow exit of LWPs in the IDL state when (l != curlwp).
- More locking fixes & conversion to the new API.
|
| 1.198.2.2 | 21-Mar-2007 |
ad | - Replace more simple_locks, and fix up in a few places.
- Use condition variables.
- LOCK_ASSERT -> KASSERT.
|
| 1.198.2.1 | 13-Mar-2007 |
ad | Pull in the initial set of changes for the vmlocking branch.
|
| 1.202.2.1 | 15-Aug-2007 |
skrll | Sync with HEAD.
|
| 1.203.6.2 | 29-Jul-2007 |
ad | It's not a good idea for device drivers to modify b_flags, as they don't
need to understand the locking around that field. Instead of setting
B_ERROR, set b_error instead. b_error is 'owned' by whoever completes
the I/O request.
|
| 1.203.6.1 | 29-Jul-2007 |
ad | file lfs_segment.c was added on branch matt-mips64 on 2007-07-29 13:31:15 +0000
|
| 1.203.4.2 | 26-Oct-2007 |
joerg | Sync with HEAD.
Follow the merge of pmap.c on i386 and amd64 and move
pmap_init_tmp_pgtbl into arch/x86/x86/pmap.c. Modify the ACPI wakeup
code to restore CR4 before jumping back into kernel space as the large
page option might cover that.
|
| 1.203.4.1 | 16-Aug-2007 |
jmcneill | Sync with HEAD.
|
| 1.204.4.1 | 14-Oct-2007 |
yamt | sync with head.
|
| 1.204.2.3 | 23-Mar-2008 |
matt | sync with HEAD
|
| 1.204.2.2 | 09-Jan-2008 |
matt | sync with HEAD
|
| 1.204.2.1 | 06-Nov-2007 |
matt | sync with HEAD
|
| 1.206.10.1 | 02-Jan-2008 |
bouyer | Sync with HEAD
|
| 1.206.6.4 | 19-Dec-2007 |
ad | Use a global lfs_lock.
|
| 1.206.6.3 | 19-Dec-2007 |
ad | Fix some more problems w/lfs on this branch.
|
| 1.206.6.2 | 19-Dec-2007 |
ad | Get lfs mostly working.
|
| 1.206.6.1 | 04-Dec-2007 |
ad | Pull the vmlocking changes into a new branch.
|
| 1.206.4.1 | 18-Feb-2008 |
mjf | Sync with HEAD.
|
| 1.209.6.3 | 05-Jun-2008 |
mjf | Sync with HEAD.
Also fix build.
|
| 1.209.6.2 | 02-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.209.6.1 | 03-Apr-2008 |
mjf | Sync with HEAD.
|
| 1.210.4.5 | 11-Aug-2010 |
yamt | sync with head.
|
| 1.210.4.4 | 11-Mar-2010 |
yamt | sync with head
|
| 1.210.4.3 | 19-Aug-2009 |
yamt | sync with head.
|
| 1.210.4.2 | 04-May-2009 |
yamt | sync with head.
|
| 1.210.4.1 | 16-May-2008 |
yamt | sync with head.
|
| 1.210.2.2 | 04-Jun-2008 |
yamt | sync with head
|
| 1.210.2.1 | 18-May-2008 |
yamt | sync with head.
|
| 1.211.2.1 | 23-Jun-2008 |
wrstuden | Sync w/ -current. 34 merge conflicts to follow.
|
| 1.213.22.2 | 09-Nov-2015 |
snj | Fix ticket #1974 fallout.
|
| 1.213.22.1 | 07-Nov-2015 |
snj | Pull up following revision(s) (requested by dholland in ticket #1974):
sys/ufs/lfs/lfs_segment.c: revision 1.247 via patch
Fix catastrophic bug in lfs_rewind() that changed segment numbers
(lfs_curseg/lfs_nextseg in the superblock) using the wrong units.
These fields are for whatever reason the start addresses of segments
(measured in frags) rather than the segment numbers 0..n.
This only apparently affects dumping from a mounted fs; however, it
trashes the fs.
I would really, really like to have a static analysis tool that can
keep track of the units things are measured in, since fs code is full
of conversion macros and the macros are named inscrutable things like
"sntod" whose letters don't necessarily even correspond to the units
they convert. It is surprising that more of these are not wrong.
|
| 1.213.18.2 | 09-Nov-2015 |
snj | Fix ticket #1974 fallout.
|
| 1.213.18.1 | 07-Nov-2015 |
snj | Pull up following revision(s) (requested by dholland in ticket #1974):
sys/ufs/lfs/lfs_segment.c: revision 1.247 via patch
Fix catastrophic bug in lfs_rewind() that changed segment numbers
(lfs_curseg/lfs_nextseg in the superblock) using the wrong units.
These fields are for whatever reason the start addresses of segments
(measured in frags) rather than the segment numbers 0..n.
This only apparently affects dumping from a mounted fs; however, it
trashes the fs.
I would really, really like to have a static analysis tool that can
keep track of the units things are measured in, since fs code is full
of conversion macros and the macros are named inscrutable things like
"sntod" whose letters don't necessarily even correspond to the units
they convert. It is surprising that more of these are not wrong.
|
| 1.213.8.2 | 09-Nov-2015 |
sborrill | Fix breakage from ticket #1974
|
| 1.213.8.1 | 07-Nov-2015 |
snj | Pull up following revision(s) (requested by dholland in ticket #1974):
sys/ufs/lfs/lfs_segment.c: revision 1.247 via patch
Fix catastrophic bug in lfs_rewind() that changed segment numbers
(lfs_curseg/lfs_nextseg in the superblock) using the wrong units.
These fields are for whatever reason the start addresses of segments
(measured in frags) rather than the segment numbers 0..n.
This only apparently affects dumping from a mounted fs; however, it
trashes the fs.
I would really, really like to have a static analysis tool that can
keep track of the units things are measured in, since fs code is full
of conversion macros and the macros are named inscrutable things like
"sntod" whose letters don't necessarily even correspond to the units
they convert. It is surprising that more of these are not wrong.
|
| 1.214.2.2 | 17-Aug-2010 |
uebayasi | Sync with HEAD.
|
| 1.214.2.1 | 30-Apr-2010 |
uebayasi | Sync with HEAD.
|
| 1.215.2.4 | 21-Apr-2011 |
rmind | sync with head
|
| 1.215.2.3 | 05-Mar-2011 |
rmind | sync with head
|
| 1.215.2.2 | 03-Jul-2010 |
rmind | sync with head
|
| 1.215.2.1 | 16-Mar-2010 |
rmind | Change struct uvm_object::vmobjlock to be dynamically allocated with
mutex_obj_alloc(). It allows us to share the locks among UVM objects.
|
| 1.217.2.1 | 06-Jun-2011 |
jruoho | Sync with HEAD.
|
| 1.220.2.1 | 23-Jun-2011 |
cherry | Catchup with rmind-uvmplock merge.
|
| 1.222.6.1 | 18-Feb-2012 |
mrg | merge to -current.
|
| 1.222.2.4 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.222.2.3 | 23-Jan-2013 |
yamt | sync with head
|
| 1.222.2.2 | 17-Apr-2012 |
yamt | sync with head
|
| 1.222.2.1 | 02-Nov-2011 |
yamt | page cache related changes
- maintain object pages in radix tree rather than rb tree.
- reduce unnecessary page scan in putpages. esp. when an object has a ton of
pages cached but only a few of them are dirty.
- reduce the number of pmap operations by tracking page dirtiness more
precisely in uvm layer.
- fix nfs commit range tracking.
- fix nfs write clustering. XXX hack
|
| 1.223.2.2 | 15-Nov-2015 |
bouyer | Pull up following revision(s) (requested by dholland in ticket #1319):
sys/ufs/lfs/lfs_segment.c: revision 1.247 via patch
Fix catastrophic bug in lfs_rewind() that changed segment numbers
(lfs_curseg/lfs_nextseg in the superblock) using the wrong units.
These fields are for whatever reason the start addresses of segments
(measured in frags) rather than the segment numbers 0..n.
This only apparently affects dumping from a mounted fs; however, it
trashes the fs.
I would really, really like to have a static analysis tool that can
keep track of the units things are measured in, since fs code is full
of conversion macros and the macros are named inscrutable things like
"sntod" whose letters don't necessarily even correspond to the units
they convert. It is surprising that more of these are not wrong.
|
| 1.223.2.1 | 17-Mar-2012 |
bouyer | Pull up following revision(s) (requested by perseant in ticket #116):
sys/ufs/lfs/lfs_alloc.c: revision 1.112
tests/fs/vfs/t_rmdirrace.c: revision 1.9
tests/fs/vfs/t_renamerace.c: revision 1.25
sys/ufs/lfs/lfs_vnops.c: revision 1.240
sys/ufs/lfs/lfs_segment.c: revision 1.224
sys/ufs/lfs/lfs_bio.c: revision 1.122
sys/ufs/lfs/lfs_vfsops.c: revision 1.294
sbin/newfs_lfs/make_lfs.c: revision 1.19
sys/ufs/lfs/lfs.h: revision 1.136
Pass t_renamerace and t_rmdirrace tests.
Adapt dholland@'s fix to ufs_rename to fix PR kern/43582. Address several
other MP locking issues discovered during the course of investigating the
same problem.
Removed extraneous vn_lock() calls on the Ifile, since the Ifile writes
are controlled by the segment lock.
Fix PR kern/45982 by deemphasizing the estimate of how much metadata
will fill the empty space on disk when the disk is nearly empty
(t_renamerace crates a lot of inode blocks on a tiny empty disk).
|
| 1.224.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.224.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.224.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.224.2.1 | 25-Feb-2013 |
tls | resync with head
|
| 1.230.2.2 | 18-May-2014 |
rmind | sync with head
|
| 1.230.2.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.236.6.5 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.236.6.4 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.236.6.3 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.236.6.2 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.236.6.1 | 06-Apr-2015 |
skrll | Sync with HEAD
|
| 1.236.4.1 | 04-Aug-2015 |
snj | Pull up following revision(s) (requested by dholland in ticket #932):
sys/ufs/lfs/lfs_segment.c: revision 1.247 via patch
Fix catastrophic bug in lfs_rewind() that changed segment numbers
(lfs_curseg/lfs_nextseg in the superblock) using the wrong units.
These fields are for whatever reason the start addresses of segments
(measured in frags) rather than the segment numbers 0..n.
This only apparently affects dumping from a mounted fs; however, it
trashes the fs.
I would really, really like to have a static analysis tool that can
keep track of the units things are measured in, since fs code is full
of conversion macros and the macros are named inscrutable things like
"sntod" whose letters don't necessarily even correspond to the units
they convert. It is surprising that more of these are not wrong.
|
| 1.263.4.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.263.2.2 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.263.2.1 | 20-Mar-2017 |
pgoyette | Sync with HEAD
|
| 1.269.6.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.275.2.2 | 06-Sep-2018 |
pgoyette | Sync with HEAD
Resolve a couple of conflicts (result of the uimin/uimax changes)
|
| 1.275.2.1 | 25-Jun-2018 |
pgoyette | Sync with HEAD
|
| 1.277.2.2 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.277.2.1 | 10-Jun-2019 |
christos | Sync with HEAD
|
| 1.278.4.1 | 17-Aug-2020 |
martin | Pull up following revision(s) (requested by riastradh in ticket #1050):
sys/ufs/lfs/lfs_subr.c: revision 1.101
sys/ufs/lfs/lfs_subr.c: revision 1.102
sys/ufs/lfs/lfs_inode.c: revision 1.158
sys/ufs/lfs/lfs_inode.h: revision 1.25
sys/ufs/lfs/lfs_balloc.c: revision 1.95
sys/ufs/lfs/lfs_pages.c: revision 1.21
sys/ufs/lfs/lfs_vnops.c: revision 1.330
sys/ufs/lfs/lfs_alloc.c: revision 1.140 (patch)
sys/ufs/lfs/lfs_alloc.c: revision 1.141 (patch)
lib/libp2k/p2k.c: revision 1.72
sys/ufs/lfs/lfs.h: revision 1.205
sys/ufs/lfs/lfs.h: revision 1.206
sys/ufs/lfs/lfs_segment.c: revision 1.284
sys/ufs/lfs/lfs.h: revision 1.207
sys/ufs/lfs/lfs_segment.c: revision 1.285
sys/ufs/lfs/lfs_debug.c: revision 1.55
sys/ufs/lfs/lfs_rename.c: revision 1.23
usr.sbin/dumplfs/dumplfs.c: revision 1.65
sys/ufs/lfs/lfs_vfsops.c: revision 1.371
sys/arch/i386/stand/efiboot/bootx64/Makefile: revision 1.3
sys/ufs/lfs/lfs_vfsops.c: revision 1.372
sys/ufs/lfs/lfs_vfsops.c: revision 1.373
sbin/fsck_lfs/pass1.c: revision 1.46
sys/ufs/lfs/lfs_vnops.c: revision 1.326
sys/ufs/lfs/lfs_vnops.c: revision 1.327
sys/ufs/lfs/lfs_vfsops.c: revision 1.375 (patch)
sys/ufs/lfs/lfs_vnops.c: revision 1.328
sys/ufs/lfs/lfs_subr.c: revision 1.98
sys/ufs/lfs/lfs_extern.h: revision 1.116
sys/ufs/lfs/lfs_vnops.c: revision 1.329
sys/ufs/lfs/lfs_subr.c: revision 1.99
sys/ufs/lfs/lfs_extern.h: revision 1.117
sys/ufs/lfs/lfs_accessors.h: revision 1.49
sys/ufs/lfs/lfs_extern.h: revision 1.118
sys/rump/fs/lib/liblfs/Makefile: revision 1.15
sys/ufs/lfs/lfs_bio.c: revision 1.146 (patch)
sys/ufs/lfs/lfs_bio.c: revision 1.147
sys/ufs/lfs/lfs_subr.c: revision 1.100
Fix kassert in lfs by initializing vp first.
Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
Teach lfs to transition ro<->rw.
Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
Assert lfs_writer where I think we can now prove it.
Serialize access to the splay tree with lfs_lock.
Change some cheap KDASSERT into KASSERT.
Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
Call lfs_orphan in lfs_rename while we're still in the dirop.
lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
Teach LFS_ORPHAN_NEXTFREE about lfs64.
Dust off the orphan detection code and try to make it work.
Fix !DIAGNOSTIC compile
Fix userland references to LFS_ORPHAN_NEXTFREE.
Forgot to grep for these or do a full distribution build, oops!
Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
OR into bp->b_cflags; don't overwrite.
CTASSERT lfs on-disk structure sizes.
Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
Fix clang build after packed lfs64 accessor change.
Suppress spurious address-of-packed error in rump lfs too.
|
| 1.280.2.2 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.280.2.1 | 17-Jan-2020 |
ad | Sync with head.
|
| 1.105 | 20-Oct-2025 |
perseant | * Generalize the partial-segment parser introduced for roll-forward,
using it to facilitate an in-kernel segment rewriter (cleaner), and a
mechanism to check whether a segment is in fact empty (only used
with DEBUG).
* Add these new fcntl calls:
- LFCNFILESTATS: For each inode given, report its number of direct
blocks, how many gaps (discontinuities) there are between direct
blocks, and how large the total gap distance is. This will be
useful for a coalescing agent.
- LFCNREWRITEFILE: For each inode given, rewrite its direct blocks,
effectively coalescing it into as compact a form as possible.
- LFCNSCRAMBLE: As above, except that it only rewrites every other
block. This causes the file to have many gaps that can be
measured with LFCNFILESTATS and addressed with LFCNREWRITEFILE,
for testing purposes.
- LFCNREWRITESEGS: Rewrite any live data in the given segments.
This is intended to simplify the cleaner API and facilitate an
in-kernel cleaner.
- LFCNCLEANERINFO: Get the most current CLEANERINFO data from the
kernel.
- LFCNSEGUSE: Retrieve segment usage data from the kernel.
* Vnodes marked IN_CLEANING now take a reference. Add a new "cleaner
lock", which must be taken by the cleaner before the segment lock,
and before marking nodes IN_CLEANING. This allows us to flush
vnodes, if necessary, before the cleaning segment is written, and
never to flush vnodes being cleaned. When the cleaner lock is
released, the vnodes are cleared of IN_CLEANING and the reference
dropped.
* Track a potential infinite loop in lfs_gatherblock.
* Pull "needs to flush" and "needs to wait for flush" into functions
instead of inlining their definitions.
|
| 1.104 | 04-Sep-2025 |
perseant | Copy the flags from a full partial segment to its continuation, if
a continuation is necessary, so that partial-segment collections marked
with SS_DIROP|SS_CONT are properly completed wiht a partial-segment marked
SS_DIROP (without SS_CONT). Necessary for roll-forward.
|
| 1.103 | 05-Sep-2020 |
riastradh | Round of uvm.h cleanup.
The poorly named uvm.h is generally supposed to be for uvm-internal
users only.
- Narrow it to files that actually need it -- mostly files that need
to query whether curlwp is the pagedaemon, which should maybe be
exposed by an external header.
- Use uvm_extern.h where feasible and uvm_*.h for things not exposed
by it. We should split up uvm_extern.h but this will serve for now
to reduce the uvm.h dependencies.
- Use uvm_stat.h and #ifdef UVMHIST uvm.h for files that use
UVMHIST(ubchist), since ubchist is declared in uvm.h but the
reference evaporates if UVMHIST is not defined, so we reduce header
file dependencies.
- Make uvm_device.h and uvm_swap.h independently includable while
here.
ok chs@
|
| 1.102 | 23-Feb-2020 |
riastradh | Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
|
| 1.101 | 23-Feb-2020 |
ad | Fix !DIAGNOSTIC compile
|
| 1.100 | 23-Feb-2020 |
riastradh | lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
|
| 1.99 | 23-Feb-2020 |
riastradh | Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
|
| 1.98 | 23-Feb-2020 |
riastradh | Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
|
| 1.97 | 26-Jul-2017 |
maya | branches: 1.97.4; 1.97.8; 1.97.10;
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
|
| 1.96 | 26-Jul-2017 |
maya | Deduplicate sanity check that seglock is held on segunlock
|
| 1.95 | 19-Jun-2017 |
maya | Ifdef out KDASSERT which fires on my machine.
|
| 1.94 | 10-Jun-2017 |
maya | Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.93 | 08-Jun-2017 |
chs | move some buffer cache internals declarations from buf.h to vfs_bio.c.
this is needed to avoid name conflicts with ZFS and also
makes it clearer that other code shouldn't be messing with these.
remove the LFS debug code that poked around in bufqueues and
remove the BQ_EMPTY bufqueue since nothing uses it anymore.
provide a function to let LFS and wapbl read the value of nbuf for now.
|
| 1.92 | 06-Apr-2017 |
maya | branches: 1.92.6;
don't guard lfs_sbactive or lfs_log with splbio, lfs_lock is plenty.
|
| 1.91 | 06-Apr-2017 |
maya | don't guard lfs_reshash with splbio, lfs_lock is plenty
|
| 1.90 | 06-Apr-2017 |
maya | if DEBUG panic => KDASSERT. reduces ifdefs. NFC
|
| 1.89 | 06-Apr-2017 |
maya | Provide a LFS_ENTER_LOG (__nothing) in the !DEBUG case.
so I can drop lots of #ifdef DEBUG around this macro. NFCI
|
| 1.88 | 01-Apr-2017 |
maya | Keep on holding lfs_lock when calling cv_broadcast
pointed out by skrll, thanks.
|
| 1.87 | 01-Apr-2017 |
maya | switch lfs_dirops to condvar (from mtsleep)
|
| 1.86 | 03-Oct-2015 |
dholland | branches: 1.86.2; 1.86.4;
Use IINFO in lfs_writeinode().
(both the kernel and the userland copies)
|
| 1.85 | 12-Aug-2015 |
dholland | Make 32-bit and 64-bit versions of CLEANERINFO.
XXX: while this is written to disk, it seems like much of it would
XXX: be better set up as a commpage shared with the cleaner.
|
| 1.84 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.83 | 24-Jul-2015 |
dholland | More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)
The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.
|
| 1.82 | 24-Jul-2015 |
dholland | Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)
It also gets rid of cpp abuse in the form of fake structure member
macros.
Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)
XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.
|
| 1.81 | 16-Jul-2015 |
dholland | Don't cast the return value of malloc.
|
| 1.80 | 28-Jul-2013 |
dholland | branches: 1.80.6;
Add lfs_kernel.h for declarations that don't need to be exposed to userland.
lfs currently has the following headers:
lfs.h - on-disk structures and stuff needed for userlevel tools
lfs_inode.h - additional restricted materials for userlevel tools
that operate the fs (newfs_lfs, fsck_lfs, lfs_cleanerd)
lfs_kernel.h - stuff needed only in the kernel
and the following legacy headers that are expected to be mopped up and
folded into one of the above:
lfs_extern.h - function prototypes
ulfs_bswap.h - endian-independent support
ulfs_dinode.h - now contains very little
ulfs_dirhash.h - dirhash support
ulfs_extattr.h - extattr support
ulfs_extern.h - more function prototypes
ulfs_inode.h - assorted kernel-only declarations
ulfs_quota.h - quota support
ulfs_quota1.h - more quota support
ulfs_quota2.h - more quota support
ulfs_quotacommon.h - more quota support
ulfsmount.h - legacy copy of ufsmount material
|
| 1.79 | 18-Jun-2013 |
christos | branches: 1.79.2;
Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.
|
| 1.78 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.77 | 02-Jan-2012 |
perseant | branches: 1.77.6;
* Remove PGO_RECLAIM during lfs_putpages()' call to genfs_putpages(),
to avoid a live lock in the latter when reclaiming a vnode with
dirty pages.
* Add a new segment flag, SEGM_RECLAIM, to note when a segment is
being written for vnode reclamation, and record which inode is being
reclaimed, to aid in forensic debugging.
* Add a new segment flag, SEGM_SINGLE, so that opportunistic writes
can write a single segment's worth of blocks and then stop, rather
than writing all the way up to the cleaner's reserved number of
segments.
* Add assert statements to check mutex ownership is the way it ought
to be, mostly in lfs_putpages; fix problems uncovered by this.
* Don't clear VU_DIROP until the inode actually makes its way to disk,
avoiding a problem where dirop inodes could become separated
(uncovered by a modified version of the "ckckp" forensic regression
test).
* Move the vfs_getopsbyname() call into lfs_writerd. Prepare code to
make lfs_writerd notice when there are no more LFSs, and exit losing
the reference, so that, in theory, the module can be unloaded. This
code is not enabled, since it causes a crash on exit.
* Set IN_MODIFIED on inodes flushed by lfs_flush_dirops. Really we
only need to set IN_MODIFIED if we are going to write them again
(e.g., to write pages); need to think about this more.
Finally, several changes to help avoid "no clean segments" panics:
* In lfs_bmapv, note when a vnode is loaded only to discover whether
its blocks are live, so it can immediately be recycled. Since the
cleaner will try to choose ~empty segments over full ones, this
prevents the cleaner from (1) filling the vnode cache with junk, and
(2) squeezing any unwritten writes to disk and running the fs out of
segments.
* Overestimate by half the amount of metadata that will be required
to fill the clean segments. This will make the disk appear smaller,
but should help avoid a "no clean segments" panic.
* Rearrange lfs_writerd. In particular, lfs_writerd now pays
attention to the number of clean segments available, and holds off
writing until there is room.
|
| 1.76 | 25-Jun-2010 |
hannken | branches: 1.76.8; 1.76.12;
Undo last commit and don't try to lock vnodes in lfs_unmark_dirop()
as we may deadlock trying to write the superblock.
Should fix PR #43503 Can't create device nodes on LFS.
|
| 1.75 | 24-Jun-2010 |
hannken | Clean up vnode lock operations:
- VOP_LOCK(vp, flags): Limit the set of allowed flags to LK_EXCLUSIVE,
LK_SHARED and LK_NOWAIT. LK_INTERLOCK is no longer allowed as it
makes no sense here.
- VOP_ISLOCKED(vp): Remove the for some time unused return value
LK_EXCLOTHER. Mark this operation as "diagnostic only".
Making a lock decision based on this operation is no longer allowed.
Discussed on tech-kern.
|
| 1.74 | 16-Feb-2010 |
mlelstv | branches: 1.74.2;
Three changes in a single commit.
- drop the notion of frags (LFS fragments) vs fsb (FFS fragments)
The code uses a complicated unity function that just makes the
code difficult to understand.
- support larger sector sizes. Fix disk address computations
to use DEV_BSIZE in the kernel as required by device drivers
and to use sector sizes in userland.
- Fix several locking bugs in lfs_bio.c and lfs_subr.c.
|
| 1.73 | 28-Apr-2008 |
martin | branches: 1.73.20;
Remove clause 3 and 4 from TNF licenses
|
| 1.72 | 02-Jan-2008 |
ad | branches: 1.72.6; 1.72.8; 1.72.10;
Merge vmlocking2 to head.
|
| 1.71 | 10-Oct-2007 |
ad | branches: 1.71.4; 1.71.6; 1.71.10;
Merge from vmlocking:
- Split vnode::v_flag into three fields, depending on field locking.
- simple_lock -> kmutex in a few places.
- Fix some simple locking problems.
|
| 1.70 | 15-May-2007 |
tnn | branches: 1.70.6; 1.70.8; 1.70.10;
Add missing underscore to wchan name.
|
| 1.69 | 18-Apr-2007 |
perseant | Add/change a couple of comments about locking restrictions.
|
| 1.68 | 12-Mar-2007 |
ad | branches: 1.68.2;
Pass an ipl argument to pool_init/POOL_INIT to be used when initializing
the pool's lock.
|
| 1.67 | 21-Feb-2007 |
thorpej | branches: 1.67.4;
Replace the Mach-derived boolean_t type with the C99 bool type. A
future commit will replace use of TRUE and FALSE with true and false.
|
| 1.66 | 15-Feb-2007 |
ad | branches: 1.66.2;
Replace some uses of lockmgr() / simplelocks.
|
| 1.65 | 16-Nov-2006 |
christos | branches: 1.65.2; 1.65.4;
__unused removal on arguments; approved by core.
|
| 1.64 | 12-Oct-2006 |
christos | - sprinkle __unused on function decls.
- fix a couple of unused bugs
- no more -Wno-unused for i386
|
| 1.63 | 04-Oct-2006 |
christos | fix empty if
|
| 1.62 | 15-Sep-2006 |
perseant | branches: 1.62.2;
Don't remark a locked inode with IN_MODIFIED after writing it to disk,
if we ourselves hold the lock. This prevents e.g. mknod from hanging
indefinitely.
Also, always use the return value from VOP_ISLOCKED to determine whether
we hold the lock or someone else does, rather than looking into the lock
structure ourselves.
|
| 1.61 | 01-Sep-2006 |
perseant | branches: 1.61.2;
Changes to help the roll-forward agent, to wit:
* Mark being-deleted files in the Ifile so we can finish deleting them
at fs mount time.
* Flag the Ifile with "cleaner must clean" when writers are waiting for
the cleaner, rather than relying solely on the cleaner's estimation of
whether it should clean or not.
* Note partial segments written by a user agent (in particular,
fsck_lfs) so that repeated rolls forward don't interfere with one
another.
* Add a new fcntl, LFCNPASS, that allows the log to wrap exactly once,
for better testing of the validity of checkpoints.
* Keep track of the on-disk nlink count when cleaning, so that we don't
partially complete directory operations while cleaning.
* Ensure that every single Ifile inode write represents a consistent
view of the filesystem. In particular, the accounting for the segment
we are writing the inode into must be correct, and the accounting for
the segment that inode used to reside in must be correct. Rather than
just rewriting the inode if we wrote it wrong, rewrite the necessary
ifile blocks before writing the inode so we never write it wrong.
* Don't unmark any VDIROP vnodes if we haven't written them to disk,
avoiding yet another problem with the "wait for the cleaner" error
return from lfs_putpages().
Also, move the last callback to an aiodone call, so we no longer do any
memory management from interrupt context.
|
| 1.60 | 29-Jun-2006 |
perseant | Don't wake up the cleaner if the filesystem is unwrappable, and fix the
compatibility fcntls.
Also includes one-line fixes for an MP locking bug and a zero-length FINFO
problem that manifested during testing.
|
| 1.59 | 04-May-2006 |
perseant | branches: 1.59.4;
Introduce another per-filesystem parameter, lfs_resvseg, to separate the
notion of "how many segments are reserved for the cleaner" from that of
"how many segments are not counted in lfs_bfree". The default value
used for existing filesystems is the same as the previous implicit value
of (lfs_minfreeseg / 2 + 1), modulo some sanity checking.
Count pending dirops on a per-filesystem basis, since once we start
writing them we can't stop until we're done. This seems to help stave off
the "no clean segments" panic in the case of filling the filesystem with
directories and small files (e.g. simultaneously unpacking more copies of
pkgsrc than will fit).
|
| 1.58 | 07-Apr-2006 |
perseant | Make the segment lock aware of LWPs. Fixes a (somewhat confusing)
"lockmgr: pid 3997, not exclusive lockholder 3997, unlocking" panic I
encountered while running blogbench on an LFS.
|
| 1.57 | 24-Mar-2006 |
perseant | Improvements to LFS's paging mechanism, to wit:
* Acknowledge that sometimes there are more dirty pages to be written to
disk than clean segments. When we reach the danger line,
lfs_gop_write() now returns EAGAIN. The caller of VOP_PUTPAGES(), if
it holds the segment lock, drops it and waits for the cleaner to make
room before continuing.
* Note and avoid a three-way deadlock in lfs_putpages (a writer holding
a page busy blocks on the cleaner while the cleaner blocks on the
segment lock while lfs_putpages blocks on the page).
|
| 1.56 | 14-Jan-2006 |
yamt | branches: 1.56.2; 1.56.4; 1.56.6; 1.56.8; 1.56.10;
- unify ffs_blkatoff and lfs_blkatoff.
- remove ufs_ops::uo_blkatoff.
- add directory read-ahead code. (disabled for now.)
|
| 1.55 | 11-Dec-2005 |
christos | branches: 1.55.2;
merge ktrace-lwp.
|
| 1.54 | 02-Nov-2005 |
yamt | merge yamt-vop branch. remove following VOPs.
VOP_BLKATOFF
VOP_VALLOC
VOP_BALLOC
VOP_REALLOCBLKS
VOP_VFREE
VOP_TRUNCATE
VOP_UPDATE
|
| 1.53 | 29-May-2005 |
christos | branches: 1.53.2; 1.53.4;
- sprinkle const
- avoid shadow variables.
|
| 1.52 | 16-Apr-2005 |
perseant | Use lfs_malloc() to manage the blkiov arrays that the cleaner functions use,
since the cleaner is likely to operate in a low-memory condition.
|
| 1.51 | 01-Apr-2005 |
perseant | Protect various per-fs structures with fs->lfs_interlock simple_lock, to
improve behavior in the multiprocessor case. Add debugging segment-lock
assertion statements.
|
| 1.50 | 08-Mar-2005 |
perseant | branches: 1.50.2;
Straighten out the maze of ifdefs. Instead, consolidate all the debugging
stuff under '#ifdef DEBUG', and use sysctl knobs to turn on/off particular
parts of the debugging reporting (if DEBUG is enabled). Re-enable the LFS
statistics in sysctl, while I'm there. A bit of a rototill.
|
| 1.49 | 26-Feb-2005 |
perry | nuke trailing whitespace
|
| 1.48 | 26-Feb-2005 |
perseant | Various minor LFS improvements:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statvfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
|
| 1.47 | 09-Mar-2004 |
yamt | branches: 1.47.6; 1.47.8; 1.47.10;
use correct segment size. this fixes memory corruption when using lfsv1.
|
| 1.46 | 21-Dec-2003 |
simonb | Fix usage of fifth argument to pool_init().
|
| 1.45 | 14-Oct-2003 |
dbj | add mnt_iflag field to struct mount for internal flags
mv MNT_GONE, MNT_UNMOUNT and MNT_WANTRDWR to this field
additonally add mnt_writeopcountupper and mnt_writeopcountlower fields
in preparation for pending write suspension support work
bump kernel version to 1.6ZD
|
| 1.44 | 07-Sep-2003 |
yamt | use LFS_DEBUG_COUNTLOCKED macro.
|
| 1.43 | 07-Aug-2003 |
agc | Move UCB-licensed code from 4-clause to 3-clause licence.
Patches provided by Joel Baker in PR 22364, verified by myself.
|
| 1.42 | 12-Jul-2003 |
yamt | - protect global resource counts with lfs_subsys_lock.
- clean up scattered externs a little.
|
| 1.41 | 02-Jul-2003 |
yamt | - add a new functions, lfs_writer_enter/leave, and use them instead of
duplicated code fragments.
- add an assertion.
|
| 1.40 | 23-Apr-2003 |
perseant | branches: 1.40.2;
Make LFS work better (though still not "well") as an NFS-exported
filesystem (and other things that needed to be fixed before the tests
would complete), to wit:
* Include the fs ident in the filehandle; improve stale filehandle checks.
* Change definition of blksize() to use the on-dinode size instead of
the inode's i_size, so that fsck_lfs will work properly again.
* Use b_interlock in lfs_vtruncbuf.
* Postpone dirop reclamation until after the seglock has been released,
so that lfs_truncate is not called with the segment lock held.
* Don't loop in lfs_fsync(), just write everything and wait.
* Be more careful about the interlock/uobjlock in lfs_putpages: when we
lose this lock, we have to resynchronize dirtiness of pages in each
block.
* Be sure to always write indirect blocks and update metadata in
lfs_putpages; fixes a bug that caused blocks to be accounted to the
wrong segment.
|
| 1.39 | 21-Mar-2003 |
perseant | KNF (space after keywords).
|
| 1.38 | 15-Mar-2003 |
perseant | Add simple_lock protection for lfs_seglock and lfs_subsys_pages; these will
be expanded to cover other per-fs and subsystem-wide data as well.
Fix a case of IN_MODIFIED being set without updating lfs_uinodes, resulting
in a "lfs_uinodes < 0" panic.
Fix a deadlock in lfs_putpages arising from the need to busy all pages in a
block; unbusy any that had already been busied before starting over.
|
| 1.37 | 11-Mar-2003 |
perseant | - Get rid of unused #ifdefs LFS_NO_PAGEMOVE and LFS_MALLOC_SUMMARY (both
always true) and accompanying dead code.
- When constructing write clusters in lfs_writeseg, if the block we are
about to add is itself a cluster from GOP_WRITE, don't put a cluster
in a cluster, just write the GOP_WRITE cluster on its own. This seems
to represent a slight performance gain on my test machine.
- Charge someone's rusage for writes on LFSes. It's difficult to tell
who the "right" process to charge is; just charge whoever triggered
the write.
|
| 1.36 | 08-Mar-2003 |
perseant | Add an lfs_strategy() that checks to make sure we're not trying to read
where the cleaner is trying to write, instead of tying up the "live"
buffers (or pages).
Fix a bug in the LFS_UBC case where oversized buffers would not be
checksummed correctly, causing uncleanable segments.
Make sure that wakeup(fs->lfs_iocount) is done if fs->lfs_iocount is 1
as well as 0, since we wait in some places for it to drop to 1.
Activate all pages that make it into lfs_gop_write without the segment
lock held, since they must have been dirtied very recently, even if
PG_DELWRI is not set.
|
| 1.35 | 04-Mar-2003 |
perseant | Don't add dirty blocks to the ifile in lfs_segunlock, if we're trying to
unmount the filesystem. This avoids a "dirty blocks" panic.
|
| 1.34 | 23-Feb-2003 |
perseant | Fix a buffer overflow bug in the LFS_UBC case that manifested itself
either as a mysterious UVM error or as "panic: dirty bufs". Verify
maximum size in lfs_malloc.
Teach lfs_updatemeta and lfs_shellsort about oversized cluster blocks from
lfs_gop_write.
When unwiring pages in lfs_gop_write, deactivate them, under the theory
that the pagedaemon wanted to free them last we knew.
|
| 1.33 | 20-Feb-2003 |
perseant | Tabify, and fix some comment alignment problems.
|
| 1.32 | 19-Feb-2003 |
yamt | add debug code to lfs_free.
|
| 1.31 | 17-Feb-2003 |
perseant | Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now
(there are still some details to work out) but expect that to go
away soon. To support these basic changes (creation of lfs_putpages,
lfs_gop_write, mods to lfs_balloc) several other changes were made, to
wit:
* Create a writer daemon kernel thread whose purpose is to handle page
writes for the pagedaemon, but which also takes over some of the
functions of lfs_check(). This thread is started the first time an
LFS is mounted.
* Add a "flags" parameter to GOP_SIZE. Current values are
GOP_SIZE_READ, meaning that the call should return the size of the
in-core version of the file, and GOP_SIZE_WRITE, meaning that it
should return the on-disk size. One of GOP_SIZE_READ or
GOP_SIZE_WRITE must be specified.
* Instead of using malloc(...M_WAITOK) for everything, reserve enough
resources to get by and use malloc(...M_NOWAIT), using the reserves if
necessary. Use the pool subsystem for structures small enough that
this is feasible. This also obsoletes LFS_THROTTLE.
And a few that are not strictly necessary:
* Moves the LFS inode extensions off onto a separately allocated
structure; getting closer to LFS as an LKM. "Welcome to 1.6O."
* Unified GOP_ALLOC between FFS and LFS.
* Update LFS copyright headers to correct values.
* Actually cast to unsigned in lfs_shellsort, like the comment says.
* Keep track of which segments were empty before the previous
checkpoint; any segments that pass two checkpoints both dirty and
empty can be summarily cleaned. Do this. Right now lfs_segclean
still works, but this should be turned into an effectless
compatibility syscall.
|
| 1.30 | 29-Jan-2003 |
yamt | don't use daddr_t for segment summary since it's an on-disk structure.
|
| 1.29 | 24-Jan-2003 |
fvdl | Bump daddr_t to 64 bits. Replace it with int32_t in all places where
it was used on-disk, so that on-disk formats remain the same.
Remove ufs_daddr_t and ufs_lbn_t for the time being.
|
| 1.28 | 11-Jul-2002 |
perseant | Remove lying comment on SEGM_PROT seglock.
|
| 1.27 | 06-Jul-2002 |
perseant | Deal with fragment size changes better. For each fragment that can
exist on an on-disk inode, we keep a record of its size in struct inode,
which is updated when we write the block to disk. The cleaner routines
thus have ready access to what size is the correct size for this block,
on disk.
Fixed a related bug: if a file with fragments is being cleaned
(fragments being cleaned) at the same time it is being extended beyond
NDADDR blocks, we could write a bogus FINFO record that has a frag in the
middle; when it was cleaned this would give back bogus file data. Don't
write the indirect blocks in this case, since there is no need.
lfs_fragextend and lfs_truncate no longer require the seglock, but instead
take a shared lock, which the seglock locks exclusively.
|
| 1.26 | 16-Jun-2002 |
perseant | For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.25 | 24-May-2002 |
perseant | Fix a couple of instances where reassignbuf() was not done at splbio.
Tested on i386.
|
| 1.24 | 23-May-2002 |
perseant | Back out rev 1.174 of vfs_subr.c, because the splbio() wasn't protecting
enough to be useful, and broadening it so that it did would have meant
that operations possibly requiring synchronous disk activity would have
to be done in splbio(). This clearly was not going to work.
Worked around this in the LFS case by having lfs_cluster_callback put an
extra hold on the vnode before calling biodone(), and taking the hold
off without HOLDRELE's problematic list swapping. lfs_vunref() will take
care of that---in thread context---on the next write if need be.
Also, ensure that the list walking in lfs_{writevnodes,segunlock,gather}
takes into account the possibility that the list may change
underneath it (possibly because it itself deleted an element).
Tested on i386, test-compiled on alpha.
|
| 1.23 | 17-May-2002 |
perseant | branches: 1.23.2;
use macros from <sys/queue.h>
|
| 1.22 | 14-May-2002 |
perseant | branches: 1.22.2;
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes
I found while making sure there weren't any new ones.
* Make the write clusters keep track of the buffers whose blocks they contain.
This should make it possible to (1) write clusters using a page mapping
instead of malloc, if desired, and (2) schedule blocks for rewriting
(somewhere else) if a write error occurs. Code is present to use
pagemove() to construct the clusters but that is untested and will go away
anyway in favor of page mapping.
* DEBUG now keeps a log of Ifile writes, so that any lingering instances of
the "dirty bufs" problem can be properly debugged.
* Keep track of whether the Ifile has been dirtied by various routines that
can be called by lfs_segwrite, and loop on that until it is clean, for
a checkpoint. Checkpoints need to be squeaky clean.
* Warn the user (once) if the Ifile grows larger than is reasonable for their
buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can
grow.
* If an inode is not found in a disk block, try rereading the block, under
the assumption that the block was copied to a cluster and then freed.
* Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
|
| 1.21 | 23-Nov-2001 |
chs | add spaces for KNF. confirmed to produce identical objects.
|
| 1.20 | 08-Nov-2001 |
lukem | add RCSID
|
| 1.19 | 26-Oct-2001 |
lukem | remove #include <ufs/ufs/quota.h> where it was just to appease
<ufs/ufs/inode.h>, since the latter now includes the former. leave the former
in source that obviously uses specific bits of it (for completeness.)
|
| 1.18 | 13-Jul-2001 |
perseant | branches: 1.18.4;
Merge the short-lived perseant-lfsv2 branch into the trunk.
Kernels and tools understand both v1 and v2 filesystems; newfs_lfs
generates v2 by default. Changes for the v2 layout include:
- Segments of non-PO2 size and arbitrary block offset, so these can be
matched to convenient physical characteristics of the partition (e.g.,
stripe or track size and offset).
- Address by fragment instead of by disk sector, paving the way for
non-512-byte-sector devices. In theory fragments can be as large
as you like, though in reality they must be smaller than MAXBSIZE in size.
- Use serial number and filesystem identifier to ensure that roll-forward
doesn't get old data and think it's new. Roll-forward is enabled for
v2 filesystems, though not for v1 filesystems by default.
- The inode free list is now a tailq, paving the way for undelete (undelete
is not yet implemented, but can be without further non-backwards-compatible
changes to disk structures).
- Inode atime information is kept in the Ifile, instead of on the inode;
that is, the inode is never written *just* because atime was changed.
Because of this the inodes remain near the file data on the disk, rather
than wandering all over as the disk is read repeatedly. This speeds up
repeated reads by a small but noticeable amount.
Other changes of note include:
- The ifile written by newfs_lfs can now be of arbitrary length, it is no
longer restricted to a single indirect block.
- Fixed an old bug where ctime was changed every time a vnode was created.
I need to look more closely to make sure that the times are only updated
during write(2) and friends, not after-the-fact during a segment write,
and certainly not by the cleaner.
|
| 1.17 | 09-Sep-2000 |
perseant | branches: 1.17.2; 1.17.4; 1.17.6;
Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
|
| 1.16 | 27-Jun-2000 |
perseant | Fixes associated with filling an LFS:
Change the space computation to appear to change the size of the *disk*
rather than the *bytes used* when more segment summaries and inode
blocks are written. Try to estimate the amount of space that these will
take up when more files are written, so the disk size doesn't change too
much.
Regularize error returns from lfs_valloc, lfs_balloc, lfs_truncate: they
now fail entirely, rather than succeeding half-way and leaving the fs in
an inconsistent state.
Rewrite lfs_truncate, mostly stealing from ffs_truncate. The old
lfs_truncate had difficulty truncating a large file to a non-zero size
(indirect blocks were not handled appropriately).
Unmark VDIROP on fvp after ufs_remove, ufs_rmdir, so these can be
reclaimed immediately: this vnode would not be written to disk again
anyway if the removal succeeded, and if it failed, no directory
operation occurred.
ufs_makeinode and ufs_mkdir now remove IN_ADIROP on error.
|
| 1.15 | 06-Jun-2000 |
perseant | branches: 1.15.2;
Don't try to inactivate dirop vnodes that are still in the middle of
their dirop.
|
| 1.14 | 05-May-2000 |
perseant | branches: 1.14.2;
Change the way LFS does block accounting, from trying to infer from the
buffer cache flags, to marking the inode and/or indirect blocks with a
special disk address UNWRITTEN==-2 when a block is accounted for. (This
address is never written to disk, but only used in-core. This is essentially
the same method of block accounting as on the UBC branch, where the buffer
headers don't exist.) Make sure that truncation is handled properly,
especially in the case of holey files.
Fixes PR#9994.
|
| 1.13 | 30-Mar-2000 |
augustss | Remove register declarations.
|
| 1.12 | 19-Jan-2000 |
perseant | Changes to stabilize LFS. The first two of these should also apply to the
1.4 branch.
* Use a separate per-fs lock, instead of ufs_hashlock, to protect the Inode
free list. This seems to prevent the "lockmgr: %d, not exclusive lock holder
%d, unlocking" message I was mis-attributing last night to an unlocked vnode
being passed to vrele.
* Change calling semantics of lfs_ifind, to give better error reporting:
If fed a struct buf, it can report the block number of the offending inode
block as well as the inode number.
* Back out rev 1.10 of lfs_subr.c, since the replacement code was slightly
uglier while being functionally identical.
* Make lfs_vunref use the same free list convention as vrele/vput, so that
vget does not remove vnodes from a hash list they are not on.
|
| 1.11 | 16-Jan-2000 |
perseant | Make sure that vnodes are locked when inactivated (e.g. by the cleaner)
|
| 1.10 | 16-Jan-2000 |
perseant | Fix a problem in my changes of Dec 14th, that prevents removed vnodes
from being inactivated under some conditions. Removed vnodes are now
inactivated when the VDIROP flag is cleared, and to prevent block
accounting problems this clearing has been postponed until
lfs_segunlock.
|
| 1.9 | 25-Mar-1999 |
perseant | branches: 1.9.2; 1.9.8; 1.9.14;
clean up unused/required #ifdefs
|
| 1.8 | 10-Mar-1999 |
perseant | New sources should leave the LFS in a more-or-less working state. Changes
include:
- DIROP segregation is enabled, and greater care is taken
to make sure that a checkpoint completes. Fsck is not
needed to remount the filesystem.
- Several checks to make sure that the LFS subsystem does not
overuse various resources (memory, in particular).
- The cleaner routines, lfs_markv in particular, are completely
rewritten. A buffer overflow is removed. Greater care is taken
to ensure that inodes come from where lfs_cleanerd say they come
from (so we know nothing has changed since lfs_bmapv was called).
- Fragment allocation is fixed, so that writes beyond end-of-file
do the right thing.
|
| 1.7 | 25-Aug-1998 |
thorpej | Add some braces to make egcs happy.
|
| 1.6 | 01-Mar-1998 |
fvdl | Merge with Lite2 + local changes
|
| 1.5 | 12-Oct-1996 |
christos | revert previous kprintf changes
|
| 1.4 | 10-Oct-1996 |
christos | printf -> kprintf, sprintf -> ksprintf
|
| 1.3 | 09-Feb-1996 |
christos | lfs prototypes
|
| 1.2 | 29-Jun-1994 |
cgd | New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
|
| 1.1 | 08-Jun-1994 |
mycroft | branches: 1.1.1;
Update to 4.4-Lite fs code, with local changes.
|
| 1.1.1.2 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite2
|
| 1.1.1.1 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite for reference
|
| 1.9.14.1 | 21-Dec-1999 |
wrstuden | Initial commit of recent changes to make DEV_BSIZE go away.
Runs on i386, needs work on other arch's. Main kernel routines should be
fine, but a number of the stand programs need help.
cd, fd, ccd, wd, and sd have been updated. sd has been tested with non-512
byte block devices. vnd, raidframe, and lfs need work.
Non 2**n block support is automatic for LKM's and conditional for kernels
on "options NON_PO2_BLOCKS".
|
| 1.9.8.1 | 20-Nov-2000 |
bouyer | Update thorpej_scsipi to -current as of a month ago
|
| 1.9.2.1 | 20-Jan-2000 |
he | Pull up revision 1.10 (requested by perseant):
Files removed (through unlink, rmdir) are now really removed, though the
removal is postponed until the dirop is complete to ensure validity of
the filesystem through a crash. Use a separate per-fs lock, instead of
ufs_hashlock, to protect the inode free list. Change calling semantics
of lfs_ifind, to give better error reporting: If fed a struct buf, it
can report the block number of the offending inode block as well as the
inode number.
|
| 1.14.2.1 | 22-Jun-2000 |
minoura | Sync w/ netbsd-1-5-base.
|
| 1.15.2.1 | 14-Sep-2000 |
perseant | Pull up recent LFS kernel changes (approved by thorpej):
ufs/ufs/inode.h, 1.20--1.22 (add i_lfs_effnblks extension ;
make ITIMES aware of LFS_ITIMES;
_LKM protection so userland progs
compile)
ufs/ufs/ufs_vnops.c, 1.69, 1.71 (remove IN_ADIROP;
use ITIMES instead of FFS_ITIMES)
ufs/ufs/ufs_readwrite.c, 1.27 (use lfs_reserve in lfs_write)
ufs/lfs/lfs.h, 1.26--1.32 (define LFS_EST_* macros ;
change MIN_FREE_SEGS to lfs_minfreesegs ;
add avail and bfree to CLEANERINFO ;
change lfs_uinodes to signed ;
change lfs_dmeta to signed ;
add whitespace to line up structure
members ;
explicit cast to int32_t in LFS_EST_*
macros)
ufs/lfs/lfs_alloc.c, back out 1.34.2.3 (pullups of 1.39, 1.40);
then pull up 1.38 (clean up on error)
1.39--1.43 (restore fvdl's ufs_hashlock fix ;
restore fvdl's ufs_hashlock fix ;
set i_lfs_effnblks ;
use UINO macros ;
add comments and fix long lines)
ufs/lfs/lfs_balloc.c, 1.19 (don't succeed halfway)
1.21--1.25 (use i_lfs_effnblks ;
fix i_lfs_effnblks computation and
quieten ;
fix i_ffs_blocks in unwritten fragment ;
remove useless debugging check ;
add comments and (c) 2000)
ufs/lfs/lfs_bio.c, 1.24--1.30 (cleanup and make lfs_flush_fs take
"struct lfs *" instead of "struct
mount *" ;
use lfs_minfreeseg instead of
MIN_FREE_SEGS ;
use UINO macros, and copy bfree/avail
to CLEANERINFO ;
add lfs_reserve function ;
1.28--1.30 fix printf formatting)
ufs/lfs/lfs_cksum.c, 1.13 (add (c) 2000)
ufs/lfs/lfs_debug.c, 1.11 (use btodb instead of DEV_BSIZE)
ufs/lfs/lfs_extern.h, 1.18, 1.20--1.21 (function prototype changes)
ufs/lfs/lfs_inode.c, 1.38 (rewrite lfs_truncate from
ffs_truncate)
1.40--1.44 (count written and unwritten blocks
seperately ;
use disk block units instead of bytes ;
remove unnecessary "mod" variable ;
correct B_DELWRI to avoid bawrite panic ;
use lfs_reserve)
ufs/lfs/lfs_segment.c, 1.52-1.59 (use lfs_dmeta to note used summaries ;
check for UNWRITTEN in indirect blocks ;
more debugging stuff inside #ifdef
DEBUG_LFS ;
use LK_CANRECURSE ;
don't drop dirty indirect blocks ;
use UINO macros ;
don't hose the free list ;
use btodb() instead of DEV_BSIZE ;
make it compile again (oops))
ufs/lfs/lfs_subr.c, 1.16--1.17 (check for locked inodes before
changing ;
use btodb() instead of DEV_BSIZE, (c)
2000)
ufs/lfs/lfs_syscalls.c, back out 1.41.4.2 (fvdl's ufs_hashlock fix);
then pull up 1.43 (use lfs_dmeta)
1.44--1.45 (restore fvdl's ufs_hashlock fix)
1.46--1.47 (fix lfs_avail leakage from sblock
segments ;
use UINO macros)
1.49 (bounds-check inode numbers in
lfs_markv)
ufs/lfs/lfs_vfsops.c, 1.53 (use LFS_EST_* macros in lfs_statfs)
1.56--1.58 (initialize lfs_minfreeseg, lfs_effnblk ;
initialize lfs_uinodes ;
initialize lfs_ravail)
ufs/lfs/lfs_vnops.c, 1.40 (remove VDIROP from removed files)
1.42--1.44 (move SET_ENDOP below the removal of
VDIROP ;
use UINO macros and add lfs_itimes
function ;
use lfs_reserve in dirops)
|
| 1.17.6.4 | 06-Sep-2002 |
jdolecek | sync kqueue branch with HEAD
|
| 1.17.6.3 | 23-Jun-2002 |
jdolecek | catch up with -current on kqueue branch
|
| 1.17.6.2 | 10-Jan-2002 |
thorpej | Sync kqueue branch with -current.
|
| 1.17.6.1 | 03-Aug-2001 |
lukem | update to -current
|
| 1.17.4.3 | 02-Jul-2001 |
perseant | Change disk addressing unit to be the fragment, instead of the disk sector.
All quantities in the superblock, inodes, indirect blocks, etc. refer now
to this abstract unit (called "fsb" as it is in FFS) instead of disk sectors;
as a consequence segment summary blocks have to be multiples of a fragment in
size. In v1 filesystems, compatibility code ensures that 1 fsb == 1 sector,
regardless of fragment size.
Fragments can now range in size between 512 and 32k; in the event that
LFS_LABELPAD (8k) is smaller than the disk address unit size, an extra
proto-superblock is kept at 8k from the beginning of the disk, to be used
*only* to locate the real superblocks. (Not all of the userland knows about
this yet.)
Almost all of this was done not by me, but by joff.
|
| 1.17.4.2 | 29-Jun-2001 |
perseant | Get rid of __P(), protoizing where it had not already been done
|
| 1.17.4.1 | 27-Jun-2001 |
perseant | Import of what I've been calling "LFSv2", that is, LFS with some features
added that require changes to the on-disk data structures. These include:
- 64-bit time in everything but inodes
- User-specified segment offset, and segment size no longer
restricted to PO2.
- Serial number on segment summaries in addition to timestamp, and
a new volume identifier, to make roll-forward feasible without
fear of finding old data and thinking it was new.
Although I think this version works at least as well as what's on the trunk,
we're not done yet; hence this commit is going in on a branch and not on
the trunk. Enhancements that are not here yet include fragment addressing,
like FFS does, instead of block addressing.
|
| 1.17.2.7 | 01-Aug-2002 |
nathanw | Catch up to -current.
|
| 1.17.2.6 | 24-Jun-2002 |
nathanw | Curproc->curlwp renaming.
Change uses of "curproc->l_proc" back to "curproc", which is more like the
original use. Bare uses of "curproc" are now "curlwp".
"curproc" is now #defined in proc.h as ((curlwp) ? (curlwp)->l_proc) : NULL)
so that it is always safe to reference curproc (*de*referencing curproc
is another story, but that's always been true).
|
| 1.17.2.5 | 20-Jun-2002 |
nathanw | Catch up to -current.
|
| 1.17.2.4 | 08-Jan-2002 |
nathanw | Catch up to -current.
|
| 1.17.2.3 | 14-Nov-2001 |
nathanw | Catch up to -current.
|
| 1.17.2.2 | 24-Aug-2001 |
nathanw | Catch up with -current.
|
| 1.17.2.1 | 05-Mar-2001 |
nathanw | Initial commit of scheduler activations and lightweight process support.
|
| 1.18.4.1 | 12-Nov-2001 |
thorpej | Sync the thorpej-mips-cache branch with -current.
|
| 1.22.2.3 | 15-Jul-2002 |
gehenna | catch up with -current.
|
| 1.22.2.2 | 20-Jun-2002 |
gehenna | catch up with -current.
|
| 1.22.2.1 | 30-May-2002 |
gehenna | Catch up with -current.
|
| 1.23.2.3 | 20-Jun-2002 |
lukem | Pull up revision 1.26 (requested by perseant in ticket #325):
For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.23.2.2 | 02-Jun-2002 |
tv | Pull up revision 1.25 (requested by perseant in ticket #135):
Fix a couple of instances where reassignbuf() was not done at splbio.
Tested on i386.
|
| 1.23.2.1 | 02-Jun-2002 |
tv | Pull up revision 1.24 (requested by perseant in ticket #132):
Back out rev 1.174 of vfs_subr.c, because the splbio() wasn't protecting
enough to be useful, and broadening it so that it did would have meant
that operations possibly requiring synchronous disk activity would have
to be done in splbio(). This clearly was not going to work.
Worked around this in the LFS case by having lfs_cluster_callback put an
extra hold on the vnode before calling biodone(), and taking the hold
off without HOLDRELE's problematic list swapping. lfs_vunref() will take
care of that---in thread context---on the next write if need be.
Also, ensure that the list walking in lfs_{writevnodes,segunlock,gather}
takes into account the possibility that the list may change
underneath it (possibly because it itself deleted an element).
Tested on i386, test-compiled on alpha.
|
| 1.40.2.6 | 10-Nov-2005 |
skrll | Sync with HEAD. Here we go again...
|
| 1.40.2.5 | 08-Mar-2005 |
skrll | Sync with HEAD.
|
| 1.40.2.4 | 04-Mar-2005 |
skrll | Sync with HEAD.
Hi Perry!
|
| 1.40.2.3 | 21-Sep-2004 |
skrll | Fix the sync with head I botched.
|
| 1.40.2.2 | 18-Sep-2004 |
skrll | Sync with HEAD.
|
| 1.40.2.1 | 03-Aug-2004 |
skrll | Sync with HEAD
|
| 1.47.10.1 | 19-Mar-2005 |
yamt | sync with head. xen and whitespace. xen part is not finished.
|
| 1.47.8.1 | 29-Apr-2005 |
kent | sync with -current
|
| 1.47.6.1 | 10-May-2005 |
riz | Pull up the following revisions (requested by perseant in ticket #1281):
1.8 sys/ufs/lfs/TODO
1.75 sys/ufs/lfs/lfs.h (via patch)
1.74 sys/ufs/lfs/lfs_alloc.c (via patch)
1.49, 1.51 sys/ufs/lfs/lfs_balloc.c (1.51 via patch)
1.78 sys/ufs/lfs/lfs_bio.c
1.62 sys/ufs/lfs/lfs_extern.h (via patch)
1.156 sys/ufs/lfs/lfs_segment.c (via patch)
1.48 sys/ufs/lfs/lfs_subr.c
1.101 sys/ufs/lfs/lfs_syscalls.c
1.163 sys/ufs/lfs/lfs_vfsops.c (via patch)
1.134 sys/ufs/lfs/lfs_vnops.c (via patch)
1.61 sys/ufs/ufs/ufs_readwrite.c (via patch)
1.20 libexec/lfs_cleanerd/clean.h (via patch)
1.52 libexec/lfs_cleanerd/cleanerd.c (via patch)
1.41 libexec/lfs_cleanerd/library.c (via patch)
1.4 regress/sys/fs/lfs/newfs_fsck/Makefile
1.2 regress/sys/fs/lfs/newfs_fsck/mkfs_mount
1.2 regress/sys/fs/lfs/newfs_fsck/smallfiles
1.3 sbin/fsck_lfs/bufcache.c
1.3 sbin/fsck_lfs/bufcache.h
1.3 sbin/fsck_lfs/lfs.h
1.8 sbin/fsck_lfs/lfs.c (via patch)
1.8 sbin/fsck_lfs/pass3.c (via patch)
1.18 sbin/fsck_lfs/pass0.c (via patch)
1.18 sbin/fsck_lfs/utilities.c (via patch)
1.7 sbin/fsck_lfs/segwrite.c
1.19 sbin/fsck_lfs/setup.c (via patch)
1.3 sbin/newfs_lfs/Makefile
0 sbin/newfs_lfs/lfs.c (yes, remove it)
1.1 sbin/newfs_lfs/make_lfs.c
1.15 sbin/newfs_lfs/newfs.c (via patch)
Various minor LFS improvements.
Kernel:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this. Should fix PR #29045.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
Fixes PR #26680.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
cleaner:
* Adapt lfs_cleanerd to use the fcntl call to get the Ifile filehandle,
so it need not be in the namespace.
* Make lfs_cleanerd be more careful when there are very few available
segments.
* Make lfs_cleanerd less verbose when the filesystem is unmounted.
newfs_lfs, fsck_lfs, and regression:
* Extend the lfs library from fsck_lfs(8) so that it can be used with a
not-yet-existent LFS. Make newfs_lfs(8) use this library, so it can
create LFSs whose Ifile is larger than one segment. Addresses PR #11110.
* Make newfs_lfs(8) use strsuftoi64() for its arguments, a la newfs(8).
* Make fsck_lfs(8) respect the "file system is clean" flag.
* Don't let fsck_lfs(8) think it has dirty blocks when invoked with the
-n flag.
* Remove the Ifile from the filesystem namespace. The cleaner now uses
a fcntl call on the root inode to find the Ifile filehandle. (As a
side-effect, addresses PR #29144.)
|
| 1.50.2.6 | 10-Aug-2006 |
tron | Apply patch (requested by fair in perseant #1457):
Bring LFS up to current, including a patch (1.95 lfs_alloc.c) that
should prevent the inode free list errors seen on the STABLE branch
subsequent to pullup ticket #1327.
|
| 1.50.2.5 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.92
sys/ufs/lfs/lfs.h: revision 1.105
sys/ufs/lfs/lfs_vfsops.c: revision 1.207
sys/ufs/lfs/lfs_subr.c: revision 1.59
sys/ufs/lfs/lfs_vnops.c: revision 1.173
sys/ufs/lfs/lfs_bio.c: revision 1.92
Introduce another per-filesystem parameter, lfs_resvseg, to separate the
notion of "how many segments are reserved for the cleaner" from that of
"how many segments are not counted in lfs_bfree". The default value
used for existing filesystems is the same as the previous implicit value
of (lfs_minfreeseg / 2 + 1), modulo some sanity checking.
Count pending dirops on a per-filesystem basis, since once we start
writing them we can't stop until we're done. This seems to help stave off
the "no clean segments" panic in the case of filling the filesystem with
directories and small files (e.g. simultaneously unpacking more copies of
pkgsrc than will fit).
|
| 1.50.2.4 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_subr.c: revision 1.58
sys/ufs/lfs/lfs.h: revision 1.98
Make the segment lock aware of LWPs. Fixes a (somewhat confusing)
"lockmgr: pid 3997, not exclusive lockholder 3997, unlocking" panic I
encountered while running blogbench on an LFS.
|
| 1.50.2.3 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.158
sys/ufs/lfs/lfs_subr.c: revision 1.57
sys/ufs/lfs/lfs_segment.c: revision 1.171
sys/ufs/lfs/lfs.h: revision 1.97
sys/ufs/lfs/lfs_vfsops.c: revision 1.195
sys/ufs/lfs/lfs_extern.h: revision 1.76
Improvements to LFS's paging mechanism, to wit:
* Acknowledge that sometimes there are more dirty pages to be written to
disk than clean segments. When we reach the danger line,
lfs_gop_write() now returns EAGAIN. The caller of VOP_PUTPAGES(), if
it holds the segment lock, drops it and waits for the cleaner to make
room before continuing.
* Note and avoid a three-way deadlock in lfs_putpages (a writer holding
a page busy blocks on the cleaner while the cleaner blocks on the
segment lock while lfs_putpages blocks on the page).
|
| 1.50.2.2 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.152
sys/ufs/lfs/lfs_debug.c: revision 1.31
sys/ufs/lfs/lfs_subr.c: revision 1.53
sys/ufs/lfs/lfs_extern.h: revision 1.68
sys/ufs/lfs/lfs_inode.c: revision 1.96
sys/ufs/lfs/lfs_bio.c: revision 1.86
sys/ufs/lfs/lfs_alloc.c: revision 1.83
sys/ufs/lfs/lfs_vfsops.c: revision 1.181
sys/ufs/lfs/lfs.h: revision 1.88
sys/ufs/lfs/lfs_segment.c: revision 1.164
- sprinkle const
- avoid shadow variables.
|
| 1.50.2.1 | 07-May-2005 |
tron | Apply patch (requested by perseant in ticket #242):
* fsck_lfs buffer cache fixes, including PR #29151
* Change fsck_lfs phase 0 message to reflect reality
* fsck_lfs: check phase 5 (cleanerinfo accounting) even on
roll-forward
* Keep better track of the free list during roll-forward, avoiding
a core dump
* Improve hash table use for fsck_lfs buffer and vnode cache
* Document fsck_lfs flag -f, and implement -q
* Add resize_lfs, including kernel support
* Add LFS to mountd's list of exportable filesystem types
* Make the LFS lkm work again [christos@]
* Add MP locking to the LFS kernel subsystem
* Fix pager_map deadlock in lfs_putpages()
* Avoid incomplete file extension that looks like "partial
truncation" to fsck
* Use lfs_malloc for cleaner malloc, since the cleaner often runs
in low-memory conditions.
* Use splay trees, not hash table, to track page allocation for
write.
* Fix mkdir panic on full fs
* Fix page accounting leak by counting differently.
* Use rightly named structure for lfs_getattr [skrll@]
* Cosmetic changes for readability.
|
| 1.53.4.1 | 20-Oct-2005 |
yamt | adapt ufs.
|
| 1.53.2.6 | 21-Jan-2008 |
yamt | sync with head
|
| 1.53.2.5 | 27-Oct-2007 |
yamt | sync with head.
|
| 1.53.2.4 | 03-Sep-2007 |
yamt | sync with head.
|
| 1.53.2.3 | 26-Feb-2007 |
yamt | sync with head.
|
| 1.53.2.2 | 30-Dec-2006 |
yamt | sync with head.
|
| 1.53.2.1 | 21-Jun-2006 |
yamt | sync with head.
|
| 1.55.2.1 | 15-Jan-2006 |
yamt | sync with head.
|
| 1.56.10.2 | 24-May-2006 |
tron | Merge 2006-05-24 NetBSD-current into the "peter-altq" branch.
|
| 1.56.10.1 | 28-Mar-2006 |
tron | Merge 2006-03-28 NetBSD-current into the "peter-altq" branch.
|
| 1.56.8.2 | 11-May-2006 |
elad | sync with head
|
| 1.56.8.1 | 19-Apr-2006 |
elad | sync with head.
|
| 1.56.6.5 | 03-Sep-2006 |
yamt | sync with head.
|
| 1.56.6.4 | 11-Aug-2006 |
yamt | sync with head
|
| 1.56.6.3 | 24-May-2006 |
yamt | sync with head.
|
| 1.56.6.2 | 11-Apr-2006 |
yamt | sync with head
|
| 1.56.6.1 | 01-Apr-2006 |
yamt | sync with head.
|
| 1.56.4.2 | 01-Jun-2006 |
kardel | Sync with head.
|
| 1.56.4.1 | 22-Apr-2006 |
simonb | Sync with head.
|
| 1.56.2.1 | 09-Sep-2006 |
rpaulo | sync with head
|
| 1.59.4.1 | 13-Jul-2006 |
gdamore | Merge from HEAD.
|
| 1.61.2.1 | 18-Nov-2006 |
ad | Sync with head.
|
| 1.62.2.2 | 10-Dec-2006 |
yamt | sync with head.
|
| 1.62.2.1 | 22-Oct-2006 |
yamt | sync with head
|
| 1.65.4.1 | 03-Sep-2007 |
wrstuden | Sync w/ NetBSD-4-RC_1
|
| 1.65.2.1 | 05-Jun-2007 |
bouyer | Pull up following revision(s) (requested by perseant in ticket #703):
sys/miscfs/genfs/genfs.h 1.21
sys/miscfs/genfs/genfs_vnops.c 1.151
sys/ufs/lfs/lfs.h 1.119, 1.120
sys/ufs/lfs/lfs_bio.c 1.99-101
sys/ufs/lfs/lfs_extern.h 1.89
sys/ufs/lfs/lfs_inode.c 1.108, 1.109
sys/ufs/lfs/lfs_segment.c 1.197, 1.199, 1.200
sys/ufs/lfs/lfs_subr.c 1.69, 1.70
sys/ufs/lfs/lfs_syscalls.c 1.119
sys/ufs/lfs/lfs_vfsops.c 1.234, 1.235
sys/ufs/lfs/lfs_vnops.c 1.195, 1.196, 1.200, 1.202-206
Reduce busy waiting in lfs_putpages(), and other LFS improvements.
|
| 1.66.2.4 | 17-May-2007 |
yamt | sync with head.
|
| 1.66.2.3 | 07-May-2007 |
yamt | sync with head.
|
| 1.66.2.2 | 24-Mar-2007 |
yamt | sync with head.
|
| 1.66.2.1 | 27-Feb-2007 |
yamt | - sync with head.
- move sched_changepri back to kern_synch.c as it doesn't know PPQ anymore.
|
| 1.67.4.5 | 17-Jun-2007 |
ad | - Increase the number of thread priorities from 128 to 256. How the space
is set up is to be revisited.
- Implement soft interrupts as kernel threads. A generic implementation
is provided, with hooks for fast-path MD code that can run the interrupt
threads over the top of other threads executing in the kernel.
- Split vnode::v_flag into three fields, depending on how the flag is
locked (by the interlock, by the vnode lock, by the file system).
- Miscellaneous locking fixes and improvements.
|
| 1.67.4.4 | 08-Jun-2007 |
ad | Sync with head.
|
| 1.67.4.3 | 21-Mar-2007 |
ad | GC the simplelock/spinlock debugging stuff.
|
| 1.67.4.2 | 13-Mar-2007 |
ad | Pull in the initial set of changes for the vmlocking branch.
|
| 1.67.4.1 | 13-Mar-2007 |
ad | Sync with head.
|
| 1.68.2.1 | 11-Jul-2007 |
mjf | Sync with head.
|
| 1.70.10.1 | 14-Oct-2007 |
yamt | sync with head.
|
| 1.70.8.2 | 09-Jan-2008 |
matt | sync with HEAD
|
| 1.70.8.1 | 06-Nov-2007 |
matt | sync with HEAD
|
| 1.70.6.1 | 26-Oct-2007 |
joerg | Sync with HEAD.
Follow the merge of pmap.c on i386 and amd64 and move
pmap_init_tmp_pgtbl into arch/x86/x86/pmap.c. Modify the ACPI wakeup
code to restore CR4 before jumping back into kernel space as the large
page option might cover that.
|
| 1.71.10.1 | 02-Jan-2008 |
bouyer | Sync with HEAD
|
| 1.71.6.2 | 19-Dec-2007 |
ad | Use a global lfs_lock.
|
| 1.71.6.1 | 04-Dec-2007 |
ad | Pull the vmlocking changes into a new branch.
|
| 1.71.4.1 | 18-Feb-2008 |
mjf | Sync with HEAD.
|
| 1.72.10.3 | 11-Aug-2010 |
yamt | sync with head.
|
| 1.72.10.2 | 11-Mar-2010 |
yamt | sync with head
|
| 1.72.10.1 | 16-May-2008 |
yamt | sync with head.
|
| 1.72.8.1 | 18-May-2008 |
yamt | sync with head.
|
| 1.72.6.1 | 02-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.73.20.2 | 17-Aug-2010 |
uebayasi | Sync with HEAD.
|
| 1.73.20.1 | 30-Apr-2010 |
uebayasi | Sync with HEAD.
|
| 1.74.2.1 | 03-Jul-2010 |
rmind | sync with head
|
| 1.76.12.1 | 18-Feb-2012 |
mrg | merge to -current.
|
| 1.76.8.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.76.8.1 | 17-Apr-2012 |
yamt | sync with head
|
| 1.77.6.3 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.77.6.2 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.77.6.1 | 23-Jun-2013 |
tls | resync from head
|
| 1.79.2.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.80.6.3 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.80.6.2 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.80.6.1 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.86.4.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.86.2.1 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.92.6.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.97.10.1 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.97.8.1 | 17-Aug-2020 |
martin | Pull up following revision(s) (requested by riastradh in ticket #1050):
sys/ufs/lfs/lfs_subr.c: revision 1.101
sys/ufs/lfs/lfs_subr.c: revision 1.102
sys/ufs/lfs/lfs_inode.c: revision 1.158
sys/ufs/lfs/lfs_inode.h: revision 1.25
sys/ufs/lfs/lfs_balloc.c: revision 1.95
sys/ufs/lfs/lfs_pages.c: revision 1.21
sys/ufs/lfs/lfs_vnops.c: revision 1.330
sys/ufs/lfs/lfs_alloc.c: revision 1.140 (patch)
sys/ufs/lfs/lfs_alloc.c: revision 1.141 (patch)
lib/libp2k/p2k.c: revision 1.72
sys/ufs/lfs/lfs.h: revision 1.205
sys/ufs/lfs/lfs.h: revision 1.206
sys/ufs/lfs/lfs_segment.c: revision 1.284
sys/ufs/lfs/lfs.h: revision 1.207
sys/ufs/lfs/lfs_segment.c: revision 1.285
sys/ufs/lfs/lfs_debug.c: revision 1.55
sys/ufs/lfs/lfs_rename.c: revision 1.23
usr.sbin/dumplfs/dumplfs.c: revision 1.65
sys/ufs/lfs/lfs_vfsops.c: revision 1.371
sys/arch/i386/stand/efiboot/bootx64/Makefile: revision 1.3
sys/ufs/lfs/lfs_vfsops.c: revision 1.372
sys/ufs/lfs/lfs_vfsops.c: revision 1.373
sbin/fsck_lfs/pass1.c: revision 1.46
sys/ufs/lfs/lfs_vnops.c: revision 1.326
sys/ufs/lfs/lfs_vnops.c: revision 1.327
sys/ufs/lfs/lfs_vfsops.c: revision 1.375 (patch)
sys/ufs/lfs/lfs_vnops.c: revision 1.328
sys/ufs/lfs/lfs_subr.c: revision 1.98
sys/ufs/lfs/lfs_extern.h: revision 1.116
sys/ufs/lfs/lfs_vnops.c: revision 1.329
sys/ufs/lfs/lfs_subr.c: revision 1.99
sys/ufs/lfs/lfs_extern.h: revision 1.117
sys/ufs/lfs/lfs_accessors.h: revision 1.49
sys/ufs/lfs/lfs_extern.h: revision 1.118
sys/rump/fs/lib/liblfs/Makefile: revision 1.15
sys/ufs/lfs/lfs_bio.c: revision 1.146 (patch)
sys/ufs/lfs/lfs_bio.c: revision 1.147
sys/ufs/lfs/lfs_subr.c: revision 1.100
Fix kassert in lfs by initializing vp first.
Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
Teach lfs to transition ro<->rw.
Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
Assert lfs_writer where I think we can now prove it.
Serialize access to the splay tree with lfs_lock.
Change some cheap KDASSERT into KASSERT.
Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
Call lfs_orphan in lfs_rename while we're still in the dirop.
lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
Teach LFS_ORPHAN_NEXTFREE about lfs64.
Dust off the orphan detection code and try to make it work.
Fix !DIAGNOSTIC compile
Fix userland references to LFS_ORPHAN_NEXTFREE.
Forgot to grep for these or do a full distribution build, oops!
Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
OR into bp->b_cflags; don't overwrite.
CTASSERT lfs on-disk structure sizes.
Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
Fix clang build after packed lfs64 accessor change.
Suppress spurious address-of-packed error in rump lfs too.
|
| 1.97.4.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.177 | 20-Oct-2025 |
perseant | * Generalize the partial-segment parser introduced for roll-forward,
using it to facilitate an in-kernel segment rewriter (cleaner), and a
mechanism to check whether a segment is in fact empty (only used
with DEBUG).
* Add these new fcntl calls:
- LFCNFILESTATS: For each inode given, report its number of direct
blocks, how many gaps (discontinuities) there are between direct
blocks, and how large the total gap distance is. This will be
useful for a coalescing agent.
- LFCNREWRITEFILE: For each inode given, rewrite its direct blocks,
effectively coalescing it into as compact a form as possible.
- LFCNSCRAMBLE: As above, except that it only rewrites every other
block. This causes the file to have many gaps that can be
measured with LFCNFILESTATS and addressed with LFCNREWRITEFILE,
for testing purposes.
- LFCNREWRITESEGS: Rewrite any live data in the given segments.
This is intended to simplify the cleaner API and facilitate an
in-kernel cleaner.
- LFCNCLEANERINFO: Get the most current CLEANERINFO data from the
kernel.
- LFCNSEGUSE: Retrieve segment usage data from the kernel.
* Vnodes marked IN_CLEANING now take a reference. Add a new "cleaner
lock", which must be taken by the cleaner before the segment lock,
and before marking nodes IN_CLEANING. This allows us to flush
vnodes, if necessary, before the cleaning segment is written, and
never to flush vnodes being cleaned. When the cleaner lock is
released, the vnodes are cleared of IN_CLEANING and the reference
dropped.
* Track a potential infinite loop in lfs_gatherblock.
* Pull "needs to flush" and "needs to wait for flush" into functions
instead of inlining their definitions.
|
| 1.176 | 18-Feb-2020 |
chs | remove the aiodoned thread. I originally added this to provide a thread context
for doing page cache iodone work, but since then biodone() has changed to
hand off all iodone work to a softint thread, so we no longer need the
special-purpose aiodoned thread.
|
| 1.175 | 26-Jul-2017 |
maya | branches: 1.175.4; 1.175.10;
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
|
| 1.174 | 17-Apr-2017 |
hannken | branches: 1.174.4;
Remove unused argument "nextp" from vfs_busy() and vfs_unbusy().
Remove argument "keepref" from vfs_unbusy() and add vfs_ref() where needed.
|
| 1.173 | 13-Mar-2017 |
riastradh | #if DIAGNOSTIC panic ---> KASSERT
Replace some #if DEBUG by this too. DEBUG is only for expensive
assertions; these are not.
|
| 1.172 | 15-Oct-2015 |
dholland | branches: 1.172.2; 1.172.4;
Move stuff from struct ulfsmount to struct lfs.
|
| 1.171 | 10-Oct-2015 |
dholland | Fix minor bitrot in #if 0 or otherwise disabled code.
|
| 1.170 | 01-Sep-2015 |
dholland | Use the lfs dinode accessors in place of the ufs-derived ones.
(Mostly.)
The ufs-derived ones are fake structure member macros, which are gross
and not very safe. Also, it seems that a lot of places in the lfs code
were using the ffsv1 branch of them unconditionally, and this way it's
guaranteed all those places have been updated.
Found while doing this: for non-devices, have getattr produce NODEV
in the rdev field instead of leaking the address of the first direct
block.
|
| 1.169 | 12-Aug-2015 |
dholland | Hack up dinode usage to be 64 vs. 32 as needed. Part 1.
(This part changes the native lfs code; the ufs-derived code already
has 64 vs. 32 logic, but as aspects of it are unsafe, and don't
entirely interoperate cleanly with the lfs 64/32 stuff, pass 2 will be
rehashing that.)
|
| 1.168 | 12-Aug-2015 |
dholland | Add IFILE32 and IFILE64 structures for the on-disk ifile entries.
Add and use accessors. There are also a bunch of places that cast and
I hope I've found them all...
|
| 1.167 | 12-Aug-2015 |
dholland | Make 32-bit and 64-bit versions of CLEANERINFO.
XXX: while this is written to disk, it seems like much of it would
XXX: be better set up as a commpage shared with the cleaner.
|
| 1.166 | 12-Aug-2015 |
dholland | Move the security checks for lfs_bmapv/lfs_markv into those functions.
(instead of the system call entry points)
Avoids duplication.
While touching these, pass the lwp around instead of the proc -- the
latter was there for no other reason than because once upon a time
struct proc was the first argument of all syscalls.
(For that matter, why not just use curlwp instead of passing it around
all over the place? The cost of passing it to every syscall probably
exceeds the cost of loading it from curcpu, even on machines where
it's not just kept in a register all the time.)
|
| 1.165 | 12-Aug-2015 |
dholland | Fix assorted 64->32 truncations related to BLOCK_INFO.
Also make note of a cleaner limitation: it seems that when it goes to
coalesce discontiguous files, it mallocs an array with one BLOCK_INFO
for every block in the file. Therefore, with 64-bit LFS, on a 32-bit
platform it will be possible to have files large enough to overflow
the cleaner's address space. Currently these will be skipped and cause
warnings via syslog.
At some point someone should rewrite the logic to coalesce files to
use chunks of some reasonable size, as discontinuity between such
chunks is immaterial and mallocing this much space is silly and
fragile. Also, the kernel only accepts up to 65536 blocks at a time
for bmapv and markv, so processing more than this at once probably
isn't useful and may not even work currently. I don't want to change
this around just now as it's not entirely trivial.
|
| 1.164 | 02-Aug-2015 |
dholland | Use accessor functions for the version field of the lfs superblock.
I thought at first maybe the cases that test the version should be
rolled into the accessors, but on the whole I think the conclusion on
that is no.
|
| 1.163 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.162 | 24-Jul-2015 |
dholland | More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)
The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.
|
| 1.161 | 24-Jul-2015 |
dholland | Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)
It also gets rid of cpp abuse in the form of fake structure member
macros.
Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)
XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.
|
| 1.160 | 31-May-2015 |
hannken | Change lfs from hash table to vcache.
- Change lfs_valloc() to return an inode number and version instead of
a vnode and move lfs_ialloc() and lfs_vcreate() to new lfs_init_vnode().
- Add lfs_valloc_fixed() to allocate a known inode, used by kernel
roll forward.
- Remove lfs_*ref(), these functions cannot coexist with vcache and
their commented behaviour is far away from their implementation.
- Add the cleaner lwp and blockinfo to struct ulfsmount so lfs_loadvnode()
may use hints from the cleaner.
- Remove vnode locks from ulfs_lookup() like we did with ufs_lookup().
|
| 1.159 | 31-May-2015 |
hannken | Make lfs_fastvget() private to lfs_syscalls.c, change it to take the
BLOCK_INFO and vnode lock type instead of the inode disk address and
return the vnode locked.
Change lfs_markv() and lfs_bmapv() to work on locked vnodes.
|
| 1.158 | 31-May-2015 |
hannken | Use VFS_PROTOS() for lfs.
Rename conflicting struct lfs field "lfs_start" to "lfs_s0addr".
No functional change.
|
| 1.157 | 20-Apr-2015 |
riastradh | Make vget always return vnode unlocked.
Convert callers who want locks to use vn_lock afterward.
Add extra argument so the compiler will report stragglers.
|
| 1.156 | 28-Mar-2015 |
maxv | Remove the 'cred' argument from bread(). Remove a now unused var in
ffs_snapshot.c. Update the man page accordingly.
ok hannken@
|
| 1.155 | 17-Apr-2014 |
pgoyette | branches: 1.155.4;
s/null/NULL/ to fix build break
Hello, xtos!
|
| 1.154 | 17-Apr-2014 |
christos | CID/1203190: Fix NULL deref
|
| 1.153 | 09-Apr-2014 |
riastradh | Take vp->v_interlock before vdead_check in lfs_bmapv.
XXX This code is a pile of bodge that needs a serious rototill anyway.
|
| 1.152 | 24-Mar-2014 |
hannken | branches: 1.152.2;
- Make VI_XLOCK, VI_CLEAN and VI_LOCKSHARE private to kern/vfs_*.c.
- Make vwait() static.
- Add vdead_check() to check a vnode for being or becoming dead.
Discussed on tech-kern.
Welcome to 6.99.38
|
| 1.151 | 05-Mar-2014 |
hannken | Current support for iterating over mnt_vnodelist is rudimentary. Every
caller has to care about list and vnode mutexes, reference count being zero,
intermediate vnode states like VI_CLEAN, VI_XLOCK, VI_MARKER and so on.
Add an interface to iterate over a vnode list:
void vfs_vnode_iterator_init(struct mount *mp, struct vnode_iterator **marker)
void vfs_vnode_iterator_destroy(struct vnode_iterator *marker)
bool vfs_vnode_iterator_next(struct vnode_iterator *marker, struct vnode **vpp)
vfs_vnode_iterator_next() returns either "false / *vpp == NULL" when done
or "true / *vpp != NULL" to return the next referenced vnode from the list.
To make vrecycle() work in this environment change it to
bool vrecycle(struct vnode *vp)
where "vp" is a referenced vnode to be destroyed if this is the last reference.
Discussed on tech-kern.
Welcome to 6.99.34
|
| 1.150 | 29-Oct-2013 |
hannken | Vnode API cleanup pass 1.
- Make these defines and functions private to vfs_vnode.c:
VC_MASK, VC_LOCK, DOCLOSE, VI_IANCTREDO and VI_INACTNOW
vclean() and vrelel()
- Remove the long time unused lwp argument from vrecycle().
- Remove vtryget(), it is responsible for ugly hacks and doesn't
look that effective.
Presented on tech-kern.
Welcome to 6.99.25
|
| 1.149 | 07-Oct-2013 |
dholland | Remove stray KERNEL_UNLOCK_ONE() in error path of lfs_markv().
From Wolfgang Stukenbrock in PR 44370.
This error path is only reachable if lfs_markv is handed an out of
range inode number, so it's unlikely that it gets tickled very often.
It isn't clear to me that we need the kernel lock in here at all, as
the path to lfs_markv that's actually used at this point (via fcntl)
doesn't take it. But, one thing at a time.
|
| 1.148 | 28-Jul-2013 |
dholland | Add lfs_kernel.h for declarations that don't need to be exposed to userland.
lfs currently has the following headers:
lfs.h - on-disk structures and stuff needed for userlevel tools
lfs_inode.h - additional restricted materials for userlevel tools
that operate the fs (newfs_lfs, fsck_lfs, lfs_cleanerd)
lfs_kernel.h - stuff needed only in the kernel
and the following legacy headers that are expected to be mopped up and
folded into one of the above:
lfs_extern.h - function prototypes
ulfs_bswap.h - endian-independent support
ulfs_dinode.h - now contains very little
ulfs_dirhash.h - dirhash support
ulfs_extattr.h - extattr support
ulfs_extern.h - more function prototypes
ulfs_inode.h - assorted kernel-only declarations
ulfs_quota.h - quota support
ulfs_quota1.h - more quota support
ulfs_quota2.h - more quota support
ulfs_quotacommon.h - more quota support
ulfsmount.h - legacy copy of ufsmount material
|
| 1.147 | 18-Jun-2013 |
christos | branches: 1.147.2;
Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.
|
| 1.146 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.145 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.144 | 22-Jan-2013 |
dholland | Stuff UFS_ in front of a few of ufs's symbols to reduce namespace
pollution. Specifically:
ROOTINO -> UFS_ROOTINO
WINO -> UFS_WINO
NXADDR -> UFS_NXADDR
NDADDR -> UFS_NDADDR
NIADDR -> UFS_NIADDR
MAXSYMLINKLEN -> UFS_MAXSYMLINKLEN
MAXSYMLINKLEN_UFS[12] -> UFS[12]_MAXSYMLINKLEN (for consistency)
Sort out ext2fs's misuse of NDADDR and NIADDR; fortunately, these have
the same values in ext2fs and ffs.
No functional change intended.
|
| 1.143 | 20-Dec-2012 |
hannken | Change bread() and breadn() to never return a buffer on
error and modify all callers to not brelse() on error.
Welcome to 6.99.16
PR kern/46282 (6.0_BETA crash: msdosfs_bmap -> pcbmap -> bread -> bio_doread)
|
| 1.142 | 13-Mar-2012 |
elad | branches: 1.142.2;
Replace the remaining KAUTH_GENERIC_ISSUSER authorization calls with
something meaningful. All relevant documentation has been updated or
written.
Most of these changes were brought up in the following messages:
http://mail-index.netbsd.org/tech-kern/2012/01/18/msg012490.html
http://mail-index.netbsd.org/tech-kern/2012/01/19/msg012502.html
http://mail-index.netbsd.org/tech-kern/2012/02/17/msg012728.html
Thanks to christos, manu, njoly, and jmmv for input.
Huge thanks to pgoyette for spinning these changes through some build
cycles and ATF.
|
| 1.141 | 15-Jan-2012 |
perseant | Corrections to part of rev 1.140. lfs_bmapv, not lfs_markv, marks vnodes
LFSI_BMAP and recycles them. This greatly reduces the writing leakage
occurring when the filesystem has no space available for non-cleaning
writes.
|
| 1.140 | 02-Jan-2012 |
perseant | * Remove PGO_RECLAIM during lfs_putpages()' call to genfs_putpages(),
to avoid a live lock in the latter when reclaiming a vnode with
dirty pages.
* Add a new segment flag, SEGM_RECLAIM, to note when a segment is
being written for vnode reclamation, and record which inode is being
reclaimed, to aid in forensic debugging.
* Add a new segment flag, SEGM_SINGLE, so that opportunistic writes
can write a single segment's worth of blocks and then stop, rather
than writing all the way up to the cleaner's reserved number of
segments.
* Add assert statements to check mutex ownership is the way it ought
to be, mostly in lfs_putpages; fix problems uncovered by this.
* Don't clear VU_DIROP until the inode actually makes its way to disk,
avoiding a problem where dirop inodes could become separated
(uncovered by a modified version of the "ckckp" forensic regression
test).
* Move the vfs_getopsbyname() call into lfs_writerd. Prepare code to
make lfs_writerd notice when there are no more LFSs, and exit losing
the reference, so that, in theory, the module can be unloaded. This
code is not enabled, since it causes a crash on exit.
* Set IN_MODIFIED on inodes flushed by lfs_flush_dirops. Really we
only need to set IN_MODIFIED if we are going to write them again
(e.g., to write pages); need to think about this more.
Finally, several changes to help avoid "no clean segments" panics:
* In lfs_bmapv, note when a vnode is loaded only to discover whether
its blocks are live, so it can immediately be recycled. Since the
cleaner will try to choose ~empty segments over full ones, this
prevents the cleaner from (1) filling the vnode cache with junk, and
(2) squeezing any unwritten writes to disk and running the fs out of
segments.
* Overestimate by half the amount of metadata that will be required
to fill the clean segments. This will make the disk appear smaller,
but should help avoid a "no clean segments" panic.
* Rearrange lfs_writerd. In particular, lfs_writerd now pays
attention to the number of clean segments available, and holds off
writing until there is room.
|
| 1.139 | 12-Jun-2011 |
rmind | branches: 1.139.2; 1.139.6;
Welcome to 5.99.53! Merge rmind-uvmplock branch:
- Reorganize locking in UVM and provide extra serialisation for pmap(9).
New lock order: [vmpage-owner-lock] -> pmap-lock.
- Simplify locking in some pmap(9) modules by removing P->V locking.
- Use lock object on vmobjlock (and thus vnode_t::v_interlock) to share
the locks amongst UVM objects where necessary (tmpfs, layerfs, unionfs).
- Rewrite and optimise x86 TLB shootdown code, make it simpler and cleaner.
Add TLBSTATS option for x86 to collect statistics about TLB shootdowns.
- Unify /dev/mem et al in MI code and provide required locking (removes
kernel-lock on some ports). Also, avoid cache-aliasing issues.
Thanks to Andrew Doran and Joerg Sonnenberger, as their initial patches
formed the core changes of this branch.
|
| 1.138 | 01-Jul-2010 |
hannken | branches: 1.138.6;
Remove vlockmgr(). Generic vnode lock operations now use a rwlock located
in the vnode. All LK_* flags move from sys/lock.h to sys/vnode.h. Calls
to vlockmgr() in file systems get replaced with VOP_LOCK() or VOP_UNLOCK().
Welcome to 5.99.34.
Discussed on tech-kern.
|
| 1.137 | 24-Jun-2010 |
hannken | Clean up vnode lock operations pass 2:
VOP_UNLOCK(vp, flags) -> VOP_UNLOCK(vp): Remove the unneeded flags argument.
Welcome to 5.99.32.
Discussed on tech-kern.
|
| 1.136 | 16-Feb-2010 |
mlelstv | branches: 1.136.2;
Three changes in a single commit.
- drop the notion of frags (LFS fragments) vs fsb (FFS fragments)
The code uses a complicated unity function that just makes the
code difficult to understand.
- support larger sector sizes. Fix disk address computations
to use DEV_BSIZE in the kernel as required by device drivers
and to use sector sizes in userland.
- Fix several locking bugs in lfs_bio.c and lfs_subr.c.
|
| 1.135 | 13-Sep-2009 |
tsutsui | branches: 1.135.2;
Move declaration of ufs_hashlock into <ufs/ufs_extern.h> from each c source.
|
| 1.134 | 11-Jan-2009 |
christos | merge christos-time_t
|
| 1.133 | 16-May-2008 |
hannken | branches: 1.133.6;
Make sure all cached buffers with valid, not yet written data have been
run through copy-on-write. Call fscow_run() with valid data where possible.
The LP_UFSCOW hack is no longer needed to protect ffs_copyonwrite() against
endless recursion.
- Add a flag B_MODIFY to bread(), breada() and breadn(). If set the caller
intends to modify the buffer returned.
- Always run copy-on-write on buffers returned from ffs_balloc().
- Add new function ffs_getblk() that gets a buffer, assigns a new blkno,
may clear the buffer and runs copy-on-write. Process possible errors
from getblk() or fscow_run(). Part of PR kern/38664.
Welcome to 4.99.63
Reviewed by: YAMAMOTO Takashi <yamt@netbsd.org>
|
| 1.132 | 06-May-2008 |
ad | branches: 1.132.2;
PR kern/38141 lookup/vfs_busy acquire rwlock recursively
Simplify the mount locking. Remove all the crud to deal with recursion on
the mount lock, and crud to deal with unmount as another weirdo lock.
Hopefully this will once and for all fix the deadlocks with this. With this
commit there are two locks on each mount:
- krwlock_t mnt_unmounting. This is used to prevent unmount across critical
sections like getnewvnode(). It's only ever read locked with rw_tryenter(),
and is only ever write locked in dounmount(). A write hold can't be taken
on this lock if the current LWP could hold a vnode lock.
- kmutex_t mnt_updating. This is taken by threads updating the mount, for
example when going r/o -> r/w, and is only present to serialize updates.
In order to take this lock, a read hold must first be taken on
mnt_unmounting, and the two need to be held across the operation.
One effect of this change: previously if an unmount failed, we would make a
half hearted attempt to back out of it gracefully, but that was unlikely to
work in a lot of cases. Now while an unmount that will be aborted is in
progress, new file operations within the mount will fail instead of being
delayed. That is unlikely to be a problem though, because if the admin
requests unmount of a file system then s(he) has made a decision to deny
access to the resource.
|
| 1.131 | 30-Apr-2008 |
ad | PR kern/38135 vfs_busy/vfs_trybusy confusion
The previous fix worked, but it opened a window where mounts could have
disappeared from mountlist while the caller was traversing it using
vfs_trybusy(). Fix that.
|
| 1.130 | 28-Apr-2008 |
martin | Remove clause 3 and 4 from TNF licenses
|
| 1.129 | 21-Apr-2008 |
ad | branches: 1.129.2;
Acquire kernel_lock directly in LFS syscalls.
|
| 1.128 | 30-Jan-2008 |
ad | branches: 1.128.6; 1.128.8; 1.128.10;
PR kern/37706 (forced unmount of file systems is unsafe):
- Do reference counting for 'struct mount'. Each vnode associated with a
mount takes a reference, and in turn the mount takes a reference to the
vfsops.
- Now that mounts are reference counted, replace the overcomplicated mount
locking inherited from 4.4BSD with a recursable rwlock.
|
| 1.127 | 30-Jan-2008 |
ad | Replace struct lock on vnodes with a simpler lock object built on
krwlock_t. This is a step towards removing lockmgr and simplifying
vnode locking. Discussed on tech-kern.
|
| 1.126 | 02-Jan-2008 |
ad | Merge vmlocking2 to head.
|
| 1.125 | 20-Dec-2007 |
dsl | Convert all the system call entry points from:
int foo(struct lwp *l, void *v, register_t *retval)
to:
int foo(struct lwp *l, const struct foo_args *uap, register_t *retval)
Fixup compat code to not write into 'uap' and (in some cases) to actually
pass a correctly formatted 'uap' structure with the right name to the
next routine.
A few 'compat' routines that just call standard ones have been deleted.
All the 'compat' code compiles (along with the kernels required to test
build it).
98% done by automated scripts.
|
| 1.124 | 10-Oct-2007 |
ad | branches: 1.124.4; 1.124.6; 1.124.10;
Merge from vmlocking:
- Split vnode::v_flag into three fields, depending on field locking.
- simple_lock -> kmutex in a few places.
- Fix some simple locking problems.
|
| 1.123 | 08-Oct-2007 |
ad | Merge ffs locking & brelse changes from the vmlocking branch.
|
| 1.122 | 04-Mar-2007 |
christos | branches: 1.122.2; 1.122.14; 1.122.16; 1.122.18;
Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.
|
| 1.121 | 15-Feb-2007 |
ad | branches: 1.121.2;
Replace some uses of lockmgr() / simplelocks.
|
| 1.120 | 09-Feb-2007 |
ad | Merge newlock2 to head.
|
| 1.119 | 04-Jan-2007 |
elad | Consistent usage of KAUTH_GENERIC_ISSUSER.
|
| 1.118 | 16-Nov-2006 |
christos | branches: 1.118.2; 1.118.4;
__unused removal on arguments; approved by core.
|
| 1.117 | 12-Oct-2006 |
christos | - sprinkle __unused on function decls.
- fix a couple of unused bugs
- no more -Wno-unused for i386
|
| 1.116 | 01-Sep-2006 |
perseant | branches: 1.116.2; 1.116.4;
Changes to help the roll-forward agent, to wit:
* Mark being-deleted files in the Ifile so we can finish deleting them
at fs mount time.
* Flag the Ifile with "cleaner must clean" when writers are waiting for
the cleaner, rather than relying solely on the cleaner's estimation of
whether it should clean or not.
* Note partial segments written by a user agent (in particular,
fsck_lfs) so that repeated rolls forward don't interfere with one
another.
* Add a new fcntl, LFCNPASS, that allows the log to wrap exactly once,
for better testing of the validity of checkpoints.
* Keep track of the on-disk nlink count when cleaning, so that we don't
partially complete directory operations while cleaning.
* Ensure that every single Ifile inode write represents a consistent
view of the filesystem. In particular, the accounting for the segment
we are writing the inode into must be correct, and the accounting for
the segment that inode used to reside in must be correct. Rather than
just rewriting the inode if we wrote it wrong, rewrite the necessary
ifile blocks before writing the inode so we never write it wrong.
* Don't unmark any VDIROP vnodes if we haven't written them to disk,
avoiding yet another problem with the "wait for the cleaner" error
return from lfs_putpages().
Also, move the last callback to an aiodone call, so we no longer do any
memory management from interrupt context.
|
| 1.115 | 23-Jul-2006 |
ad | Use the LWP cached credentials where sane.
|
| 1.114 | 07-Jun-2006 |
kardel | merge FreeBSD timecounters from branch simonb-timecounters
- struct timeval time is gone
time.tv_sec -> time_second
- struct timeval mono_time is gone
mono_time.tv_sec -> time_uptime
- access to time via
{get,}{micro,nano,bin}time()
get* versions are fast but less precise
- support NTP nanokernel implementation (NTP API 4)
- further reading:
Timecounter Paper: http://phk.freebsd.dk/pubs/timecounter.pdf
NTP Nanokernel: http://www.eecis.udel.edu/~mills/ntp/html/kern.html
|
| 1.113 | 14-May-2006 |
elad | branches: 1.113.2;
integrate kauth.
|
| 1.112 | 18-Apr-2006 |
perseant | Get rid of the LFS_FORCE_WRITE case. We never really used it, and it could
panic the kernel if cleaner daemon passed the right combination of arguments.
Coverity CID 2741.
|
| 1.111 | 07-Apr-2006 |
perseant | Several minor bug fixes:
* Correct (weak) segment lock assertions in lfs_fragextend and lfs_putpages.
* Keep IN_MODIFIED set if we run out of avail in lfs_putpages.
* Don't try to (re)write buffers on a VBLK vnode; fixes a panic I found
while running with an LFS root.
* Raise priority of LFCNSEGWAIT to PVFS; PUSER is way too low for
something the pagedaemon is relying on.
|
| 1.110 | 19-Mar-2006 |
rtr | init struct vnode *vp = NULL
coverity 2724 / run 6
XXX in future runs coverity may complain about deref NULL now but comment
on line 382 indicates this should not be possible
|
| 1.109 | 17-Mar-2006 |
tls | From Konrad Schroeder, in response to strange df output on anoncvs.netbsd.org:
We were returning the wrong value for free space. Now we're not.
|
| 1.108 | 11-Dec-2005 |
christos | branches: 1.108.4; 1.108.6; 1.108.8; 1.108.10; 1.108.12;
merge ktrace-lwp.
|
| 1.107 | 25-May-2005 |
perseant | branches: 1.107.2;
Don't update lfs_stats.segs_reclaimed if we're not keeping statistics.
Patch from Juan RP.
|
| 1.106 | 20-May-2005 |
perseant | Keep track of the number of segments reclaimed, since the cleaner doesn't
do this anymore (it hasn't for quite some time). Add a couple of conditional
debugging messages to indicate why segments are not cleaned, in the event
that lfs_segclean is used.
Make the LFCNSEGWAITALL fcntl work again.
|
| 1.105 | 16-Apr-2005 |
perseant | Use lfs_malloc() to manage the blkiov arrays that the cleaner functions use,
since the cleaner is likely to operate in a low-memory condition.
|
| 1.104 | 01-Apr-2005 |
perseant | Protect various per-fs structures with fs->lfs_interlock simple_lock, to
improve behavior in the multiprocessor case. Add debugging segment-lock
assertion statements.
|
| 1.103 | 08-Mar-2005 |
perseant | branches: 1.103.2;
Straighten out the maze of ifdefs. Instead, consolidate all the debugging
stuff under '#ifdef DEBUG', and use sysctl knobs to turn on/off particular
parts of the debugging reporting (if DEBUG is enabled). Re-enable the LFS
statistics in sysctl, while I'm there. A bit of a rototill.
|
| 1.102 | 26-Feb-2005 |
perry | nuke trailing whitespace
|
| 1.101 | 26-Feb-2005 |
perseant | Various minor LFS improvements:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statvfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
|
| 1.100 | 04-Dec-2003 |
yamt | branches: 1.100.6; 1.100.8; 1.100.10;
use b_private rather than b_saveaddr.
XXX LFS_USE_B_INVAL
|
| 1.99 | 07-Nov-2003 |
yamt | fix spec vnode aliasing.
|
| 1.98 | 10-Sep-2003 |
yamt | g/c CHECK_COPYIN.
|
| 1.97 | 07-Aug-2003 |
agc | Move UCB-licensed code from 4-clause to 3-clause licence.
Patches provided by Joel Baker in PR 22364, verified by myself.
|
| 1.96 | 30-Jul-2003 |
yamt | - check EROFS earlier in lfs_markv.
- remove wrong error recovery code (fake buffers are never on bufqueue)
and put a comment instead.
|
| 1.95 | 30-Jul-2003 |
yamt | remove an unused definition of LFS_VREF_THRESHOLD.
|
| 1.94 | 02-Jul-2003 |
yamt | use queue.h macros.
|
| 1.93 | 29-Jun-2003 |
fvdl | branches: 1.93.2;
Back out the lwp/ktrace changes. They contained a lot of colateral damage,
and need to be examined and discussed more.
|
| 1.92 | 29-Jun-2003 |
thorpej | Undo part of the ktrace/lwp changes. In particular:
* Remove the "lwp *" argument that was added to vget(). Turns out
that nothing actually used it!
* Remove the "lwp *" arguments that were added to VFS_ROOT(), VFS_VGET(),
and VFS_FHTOVP(); all they did was pass it to vget() (which, as noted
above, didn't use it).
* Remove all of the "lwp *" arguments to internal functions that were added
just to appease the above.
|
| 1.91 | 28-Jun-2003 |
darrenr | Pass lwp pointers throughtout the kernel, as required, so that the lwpid can
be inserted into ktrace records. The general change has been to replace
"struct proc *" with "struct lwp *" in various function prototypes, pass
the lwp through and use l_proc to get the process pointer when needed.
Bump the kernel rev up to 1.6V
|
| 1.90 | 17-May-2003 |
nakayama | Avoid comparison is always false warning in gcc 3.3 w/ 64-bit size_t.
|
| 1.89 | 02-Apr-2003 |
fvdl | Add support for UFS2. UFS2 is an enhanced FFS, adding support for
64 bit block pointers, extended attribute storage, and a few
other things.
This commit does not yet include the code to manipulate the extended
storage (for e.g. ACLs), this will be done later.
Originally written by Kirk McKusick and Network Associates Laboratories for
FreeBSD.
|
| 1.88 | 20-Mar-2003 |
yamt | fix "more than one fragment" panics;
direct and indirect block pointers are not valid in the case of shortlinks.
while i'm here, move duplicated code in lfs_vget/fastvget into a new
function, lfs_vinit.
|
| 1.87 | 15-Mar-2003 |
perseant | Add simple_lock protection for lfs_seglock and lfs_subsys_pages; these will
be expanded to cover other per-fs and subsystem-wide data as well.
Fix a case of IN_MODIFIED being set without updating lfs_uinodes, resulting
in a "lfs_uinodes < 0" panic.
Fix a deadlock in lfs_putpages arising from the need to busy all pages in a
block; unbusy any that had already been busied before starting over.
|
| 1.86 | 08-Mar-2003 |
perseant | Only #define LFS if not already defined.
|
| 1.85 | 08-Mar-2003 |
perseant | Add an lfs_strategy() that checks to make sure we're not trying to read
where the cleaner is trying to write, instead of tying up the "live"
buffers (or pages).
Fix a bug in the LFS_UBC case where oversized buffers would not be
checksummed correctly, causing uncleanable segments.
Make sure that wakeup(fs->lfs_iocount) is done if fs->lfs_iocount is 1
as well as 0, since we wait in some places for it to drop to 1.
Activate all pages that make it into lfs_gop_write without the segment
lock held, since they must have been dirtied very recently, even if
PG_DELWRI is not set.
|
| 1.84 | 24-Feb-2003 |
perseant | Add lfs_ioctl vnode op, with ioctls to take over cleaner system call
functionality (not including segment clean, since that is now done
automatically as checkpoints happen).
|
| 1.83 | 23-Feb-2003 |
simonb | Remove assigned-to but not used variable.
|
| 1.82 | 20-Feb-2003 |
perseant | Tabify, and fix some comment alignment problems.
|
| 1.81 | 18-Feb-2003 |
perseant | Make it compile again, grr....
|
| 1.80 | 17-Feb-2003 |
perseant | Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now
(there are still some details to work out) but expect that to go
away soon. To support these basic changes (creation of lfs_putpages,
lfs_gop_write, mods to lfs_balloc) several other changes were made, to
wit:
* Create a writer daemon kernel thread whose purpose is to handle page
writes for the pagedaemon, but which also takes over some of the
functions of lfs_check(). This thread is started the first time an
LFS is mounted.
* Add a "flags" parameter to GOP_SIZE. Current values are
GOP_SIZE_READ, meaning that the call should return the size of the
in-core version of the file, and GOP_SIZE_WRITE, meaning that it
should return the on-disk size. One of GOP_SIZE_READ or
GOP_SIZE_WRITE must be specified.
* Instead of using malloc(...M_WAITOK) for everything, reserve enough
resources to get by and use malloc(...M_NOWAIT), using the reserves if
necessary. Use the pool subsystem for structures small enough that
this is feasible. This also obsoletes LFS_THROTTLE.
And a few that are not strictly necessary:
* Moves the LFS inode extensions off onto a separately allocated
structure; getting closer to LFS as an LKM. "Welcome to 1.6O."
* Unified GOP_ALLOC between FFS and LFS.
* Update LFS copyright headers to correct values.
* Actually cast to unsigned in lfs_shellsort, like the comment says.
* Keep track of which segments were empty before the previous
checkpoint; any segments that pass two checkpoints both dirty and
empty can be summarily cleaned. Do this. Right now lfs_segclean
still works, but this should be turned into an effectless
compatibility syscall.
|
| 1.79 | 24-Jan-2003 |
fvdl | Bump daddr_t to 64 bits. Replace it with int32_t in all places where
it was used on-disk, so that on-disk formats remain the same.
Remove ufs_daddr_t and ufs_lbn_t for the time being.
|
| 1.78 | 18-Jan-2003 |
thorpej | Merge the nathanw_sa branch.
|
| 1.77 | 26-Dec-2002 |
yamt | don't try to write all blocks passed to lfs_markv at once
since it likely causes buf starvation.
|
| 1.76 | 21-Dec-2002 |
yamt | add a XXX comment
|
| 1.75 | 18-Dec-2002 |
yamt | correct/add assertion.
|
| 1.74 | 17-Dec-2002 |
yamt | no need for cleaner to hold vnode locks.
cleaner and normal vnode operations are synchronized enough by
seglock/fraglock and buf's B_BUSY-ness.
|
| 1.73 | 24-Nov-2002 |
yamt | in lfs_fakebuf, make corresponding buffer busy to avoid
reading blocks that isn't written yet.
it's needed because we'll update metadatas in lfs_updatemeta
before data pointed by them is actually written to disk.
XXX should be solved with fake inode/indirect blocks instead?
|
| 1.72 | 24-Nov-2002 |
yamt | blksize() macro shouldn't used for indirect blocks.
this fixes "getblk: block size invariant failed" panic.
PR 18977.
|
| 1.71 | 03-Aug-2002 |
itojun | correct range check, have overflow check, fix type mismatches,
for cmap args and some other calls. from openbsd
|
| 1.70 | 07-Jul-2002 |
briggs | Fix a printf format warning.
|
| 1.69 | 06-Jul-2002 |
perseant | Deal with fragment size changes better. For each fragment that can
exist on an on-disk inode, we keep a record of its size in struct inode,
which is updated when we write the block to disk. The cleaner routines
thus have ready access to what size is the correct size for this block,
on disk.
Fixed a related bug: if a file with fragments is being cleaned
(fragments being cleaned) at the same time it is being extended beyond
NDADDR blocks, we could write a bogus FINFO record that has a frag in the
middle; when it was cleaned this would give back bogus file data. Don't
write the indirect blocks in this case, since there is no need.
lfs_fragextend and lfs_truncate no longer require the seglock, but instead
take a shared lock, which the seglock locks exclusively.
|
| 1.68 | 20-Jun-2002 |
perseant | Don't bomb out of lfs_bmapv if the caller is requesting blocks that
live in the current segment. There's nothing wrong with this, and
it is necessary for the correct operation of the coaleascer.
|
| 1.67 | 16-Jun-2002 |
perseant | For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.66 | 06-Jun-2002 |
perseant | Let lfs_bmapv fill in the bi_size member of the BLOCK_INFO structure,
as well as bi_daddr. This lets the cleaner have an idea of what the size
of this block was at the time it was written without having to refer to
a segment header (e.g., in the file coalescing case).
Tested on i386.
|
| 1.65 | 14-May-2002 |
perseant | branches: 1.65.2; 1.65.4;
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes
I found while making sure there weren't any new ones.
* Make the write clusters keep track of the buffers whose blocks they contain.
This should make it possible to (1) write clusters using a page mapping
instead of malloc, if desired, and (2) schedule blocks for rewriting
(somewhere else) if a write error occurs. Code is present to use
pagemove() to construct the clusters but that is untested and will go away
anyway in favor of page mapping.
* DEBUG now keeps a log of Ifile writes, so that any lingering instances of
the "dirty bufs" problem can be properly debugged.
* Keep track of whether the Ifile has been dirtied by various routines that
can be called by lfs_segwrite, and loop on that until it is clean, for
a checkpoint. Checkpoints need to be squeaky clean.
* Warn the user (once) if the Ifile grows larger than is reasonable for their
buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can
grow.
* If an inode is not found in a disk block, try rereading the block, under
the assumption that the block was copied to a cluster and then freed.
* Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
|
| 1.64 | 12-May-2002 |
matt | Eliminate commons.
|
| 1.63 | 18-Dec-2001 |
chs | use the new compatibility routines to allow mmap() to work
(in the same non-coherent fashion that it worked pre-UBC)
until someone has time to do it the right way.
|
| 1.62 | 23-Nov-2001 |
chs | add spaces for KNF. confirmed to produce identical objects.
|
| 1.61 | 08-Nov-2001 |
lukem | add RCSID
|
| 1.60 | 26-Oct-2001 |
lukem | remove #include <ufs/ufs/quota.h> where it was just to appease
<ufs/ufs/inode.h>, since the latter now includes the former. leave the former
in source that obviously uses specific bits of it (for completeness.)
|
| 1.59 | 15-Sep-2001 |
chs | branches: 1.59.2;
a whole bunch of changes to improve performance and robustness under load:
- remove special treatment of pager_map mappings in pmaps. this is
required now, since I've removed the globals that expose the address range.
pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's
no longer any need to special-case it.
- eliminate struct uvm_vnode by moving its fields into struct vnode.
- rewrite the pageout path. the pager is now responsible for handling the
high-level requests instead of only getting control after a bunch of work
has already been done on its behalf. this will allow us to UBCify LFS,
which needs tighter control over its pages than other filesystems do.
writing a page to disk no longer requires making it read-only, which
allows us to write wired pages without causing all kinds of havoc.
- use a new PG_PAGEOUT flag to indicate that a page should be freed
on behalf of the pagedaemon when it's unlocked. this flag is very similar
to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the
pageout fails due to eg. an indirect-block buffer being locked.
this allows us to remove the "version" field from struct vm_page,
and together with shrinking "loan_count" from 32 bits to 16,
struct vm_page is now 4 bytes smaller.
- no longer use PG_RELEASED for swap-backed pages. if the page is busy
because it's being paged out, we can't release the swap slot to be
reallocated until that write is complete, but unlike with vnodes we
don't keep a count of in-progress writes so there's no good way to
know when the write is done. instead, when we need to free a busy
swap-backed page, just sleep until we can get it busy ourselves.
- implement a fast-path for extending writes which allows us to avoid
zeroing new pages. this substantially reduces cpu usage.
- encapsulate the data used by the genfs code in a struct genfs_node,
which must be the first element of the filesystem-specific vnode data
for filesystems which use genfs_{get,put}pages().
- eliminate many of the UVM pagerops, since they aren't needed anymore
now that the pager "put" operation is a higher-level operation.
- enhance the genfs code to allow NFS to use the genfs_{get,put}pages
instead of a modified copy.
- clean up struct vnode by removing all the fields that used to be used by
the vfs_cluster.c code (which we don't use anymore with UBC).
- remove kmem_object and mb_object since they were useless.
instead of allocating pages to these objects, we now just allocate
pages with no object. such pages are mapped in the kernel until they
are freed, so we can use the mapping to find the page to free it.
this allows us to remove splvm() protection in several places.
The sum of all these changes improves write throughput on my
decstation 5000/200 to within 1% of the rate of NetBSD 1.5
and reduces the elapsed time for "make release" of a NetBSD 1.5
source tree on my 128MB pc to 10% less than a 1.5 kernel took.
|
| 1.58 | 03-Aug-2001 |
jdolecek | branches: 1.58.2;
Constraint 'blkcnt' of lfs_markv() syscall by 64KB. Reviewed by
Konrad Schroder <perseant@NetBSD.org>.
|
| 1.57 | 13-Jul-2001 |
perseant | Merge the short-lived perseant-lfsv2 branch into the trunk.
Kernels and tools understand both v1 and v2 filesystems; newfs_lfs
generates v2 by default. Changes for the v2 layout include:
- Segments of non-PO2 size and arbitrary block offset, so these can be
matched to convenient physical characteristics of the partition (e.g.,
stripe or track size and offset).
- Address by fragment instead of by disk sector, paving the way for
non-512-byte-sector devices. In theory fragments can be as large
as you like, though in reality they must be smaller than MAXBSIZE in size.
- Use serial number and filesystem identifier to ensure that roll-forward
doesn't get old data and think it's new. Roll-forward is enabled for
v2 filesystems, though not for v1 filesystems by default.
- The inode free list is now a tailq, paving the way for undelete (undelete
is not yet implemented, but can be without further non-backwards-compatible
changes to disk structures).
- Inode atime information is kept in the Ifile, instead of on the inode;
that is, the inode is never written *just* because atime was changed.
Because of this the inodes remain near the file data on the disk, rather
than wandering all over as the disk is read repeatedly. This speeds up
repeated reads by a small but noticeable amount.
Other changes of note include:
- The ifile written by newfs_lfs can now be of arbitrary length, it is no
longer restricted to a single indirect block.
- Fixed an old bug where ctime was changed every time a vnode was created.
I need to look more closely to make sure that the times are only updated
during write(2) and friends, not after-the-fact during a segment write,
and certainly not by the cleaner.
|
| 1.56 | 03-Dec-2000 |
perseant | branches: 1.56.2; 1.56.4; 1.56.6;
Call uvm_vmp_setsize() in lfs_{fast,}vget to set initial vnode size.
|
| 1.55 | 30-Nov-2000 |
jdolecek | no need to include fs_lfs.h, define LFS directly
|
| 1.54 | 27-Nov-2000 |
perseant | If LFS_DO_ROLLFORWARD is defined, roll forward from the older checkpoint
on mount, through the newer checkpoint and on through any newer
partial-segments that may have been written but not checkpointed because
of an intervening crash.
LFS_DO_ROLLFORWARD is not defined by default.
|
| 1.53 | 22-Nov-2000 |
perseant | Protect lfs_{bmapv,markv} with vfs_{un,}busy. Fix a reference/lock leak
in an error case in lfs_markv. Change the vfs_getvfs() error to return
ENOENT, for consistency with failure of vfs_busy().
99% of this patch was from Jesse Off <joff@gci-net.com> (PR #11547).
|
| 1.52 | 17-Nov-2000 |
perseant | Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes.
(PR #11468). In the case of fragment allocation, check to see if enough
space is available before extending a fragment already scheduled for writing.
The locked_queue_* variables indicate the number of buffer headers and bytes,
respectively, that are unavailable to getnewbuf() because they are locked up
waiting for LFS to flush them; make sure that that is actually what we're
counting, i.e., never count malloced buffers, and always use b_bufsize instead
of b_bcount.
If DEBUG is defined, the periodic calls to lfs_countlocked will now complain
if either counter is incorrect. (In the future lfs_countlocked will not need
to be called at all if DEBUG is not defined.)
|
| 1.51 | 21-Oct-2000 |
toshii | In lfs_fastvget(), initialize i_lfs_effnblks correctly.
|
| 1.50 | 20-Oct-2000 |
perseant | Do not increment the clean segment counter, if a segment that the cleaner
is trying to clean is already clean (e.g., if two lfs_cleanerds are running
at once.)
|
| 1.49 | 09-Sep-2000 |
perseant | Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
|
| 1.48 | 13-Jul-2000 |
thorpej | XXX Use of hzto() return value needs to be double-checked here.
|
| 1.47 | 05-Jul-2000 |
perseant | Clean up accounting of lfs_uinodes (dirty but unwritten inodes).
Make lfs_uinodes a signed quantity for debugging purposes, and set it to
zero as fs mount time.
Enclose setting/clearing of the dirty flags (IN_MODIFIED, IN_ACCESSED,
IN_CLEANING) in macros, and use those macros everywhere. Make
LFS_ITIMES use these macros; updated the ITIMES macro in inode.h to know
about this. Make ufs_getattr use ITIMES instead of FFS_ITIMES.
|
| 1.46 | 04-Jul-2000 |
perseant | Fix errors observed while trying to fill the filesystem with yesterday's
fixes:
- Write copies of bfree and avail in the CLEANERINFO block, so the
cleaner doesn't have to guess which superblock has the current
information (if indeed any do).
- Tighten up accounting of lfs_avail (more needs to be done).
- When cleansing indirect blocks of UNWRITTEN, make sure not to mark
them clean, since they'll need to be rewritten later.
|
| 1.45 | 03-Jul-2000 |
fvdl | Correct typo in previous.
|
| 1.44 | 30-Jun-2000 |
fvdl | Rearrange code around getnewvnode as was already done for ffs, to avoid
locking against oneself because getnewvnode recycles a softdep-using vnode.
|
| 1.43 | 27-Jun-2000 |
perseant | Fixes associated with filling an LFS:
Change the space computation to appear to change the size of the *disk*
rather than the *bytes used* when more segment summaries and inode
blocks are written. Try to estimate the amount of space that these will
take up when more files are written, so the disk size doesn't change too
much.
Regularize error returns from lfs_valloc, lfs_balloc, lfs_truncate: they
now fail entirely, rather than succeeding half-way and leaving the fs in
an inconsistent state.
Rewrite lfs_truncate, mostly stealing from ffs_truncate. The old
lfs_truncate had difficulty truncating a large file to a non-zero size
(indirect blocks were not handled appropriately).
Unmark VDIROP on fvp after ufs_remove, ufs_rmdir, so these can be
reclaimed immediately: this vnode would not be written to disk again
anyway if the removal succeeded, and if it failed, no directory
operation occurred.
ufs_makeinode and ufs_mkdir now remove IN_ADIROP on error.
|
| 1.42 | 22-Jun-2000 |
perseant | Update lfs_vunref for the fact that now a vnode can be locked with no
references (locked for VOP_INACTIVE at the end of vrele) and it's okay.
Check the return value of lfs_vref where appropriate.
Fixes PR #s 10285 and 10352.
|
| 1.41 | 30-Mar-2000 |
augustss | branches: 1.41.4;
Remove register declarations.
|
| 1.40 | 19-Jan-2000 |
perseant | Changes to stabilize LFS. The first two of these should also apply to the
1.4 branch.
* Use a separate per-fs lock, instead of ufs_hashlock, to protect the Inode
free list. This seems to prevent the "lockmgr: %d, not exclusive lock holder
%d, unlocking" message I was mis-attributing last night to an unlocked vnode
being passed to vrele.
* Change calling semantics of lfs_ifind, to give better error reporting:
If fed a struct buf, it can report the block number of the offending inode
block as well as the inode number.
* Back out rev 1.10 of lfs_subr.c, since the replacement code was slightly
uglier while being functionally identical.
* Make lfs_vunref use the same free list convention as vrele/vput, so that
vget does not remove vnodes from a hash list they are not on.
|
| 1.39 | 16-Jan-2000 |
perseant | correct typo (reference uninitialized variable)
|
| 1.38 | 14-Jan-2000 |
perseant | Expand the category of "metadata" in lfs_markv to include Ifile data blocks.
This prevents a rare condition in which Ifile "ifile" blocks, that is, the
blocks of the ifile which point VOP_VGET at the inode block containing the
requested inode, from being "unwritten" when cleaning during intense disk
activity.
|
| 1.37 | 23-Nov-1999 |
fvdl | Be more careful to block bio interrupts for some data structures. There
were at least a few missed cases where vp->v_{clean,dirty}blkhd were
unprotected since the softdep/trickle sync merge.
|
| 1.36 | 21-Nov-1999 |
perseant | Initialize i_ffs_effnlink, so every file doesn't look like it's already been
deleted for the purpose of dirops (particularly create and mkdir). Addresses
PR#8815.
|
| 1.35 | 12-Nov-1999 |
perseant | Back out my patch of the 8th (to address unreferenced inode problem).
Apparently this needs more thought.
|
| 1.34 | 09-Nov-1999 |
perseant | If ifile blocks were written before dirops were complete, and then the
system crashed, inodes could be allocated that were not referenced. (Though
not a serious problem, it evidences itself in phase 4 of fsck_lfs.) Fix
this by marking if_daddr with UNASSIGNED before the inodes are actually
written; at mount time the ifile is checked for UNASSIGNED entries and
any that are found are linked back into the free list. (The latter
functionality should move into the roll-forward agent when it materializes.)
|
| 1.33 | 08-Jul-1999 |
wrstuden | branches: 1.33.2; 1.33.4; 1.33.8;
Modify file systems to deal with struct lock in struct vnode. All leaf
fs's other than nfs use genfs_lock() for locking.
Modify lookup routines to set PDIRUNLOCK when they unlock the parrent.
|
| 1.32 | 09-Jun-1999 |
drochner | complete the previous
reindent syscall args
|
| 1.31 | 09-Jun-1999 |
christos | prefix the lfs syscalls with sys_
|
| 1.30 | 14-Apr-1999 |
perseant | Fix lost lock in lfs_markv -- a typo-class bug, obvious when you look at it.
|
| 1.29 | 12-Apr-1999 |
perseant | Improve the debugging printfs in the cleaner syscalls (in particular, make
it obvious that they're coming from lfs).
|
| 1.28 | 12-Apr-1999 |
perseant | Better checking for held inode locks in lfs_fastvget, for a number of error
conditions. Also change the default setting of lfs_clean_vnhead to 0, which
seems to make the locking problems go away (although this is difficult to
test as I can't reliably reproduce them).
|
| 1.27 | 11-Apr-1999 |
perseant | Take out the `#ifdef USE_UFSHASH'; use ufs_hashlock to lock the inode free
list instead of free_lock.
|
| 1.26 | 29-Mar-1999 |
perseant | branches: 1.26.2;
Fix unit mismatch in debugging code in lfs_segclean; also put it properly
within `#ifdef DEBUG_LFS'.
|
| 1.25 | 25-Mar-1999 |
perseant | Fixes to make dirops and lfs_vflush play together well. In particular,
if we are short on vnodes, lfs_vflush from another process can grab a
vnode that lfs_markv has already processed but not yet written; but
lfs_markv holds the seglock. When lfs_vflush gets around to writing it,
the context for copyin is gone. So, now lfs_markv calls copyin itself,
rather than having lfs_writeseg do it.
|
| 1.24 | 25-Mar-1999 |
perseant | Change lfs_sb_cksum to use offsetof() instead of an inlined version.
Fix lfs_vref/lfs_vunredf to ignore VXLOCKed vnodes that are also being
flushed.
Improve the debugging messages somewhat.
|
| 1.23 | 25-Mar-1999 |
perseant | clean up unused/required #ifdefs
|
| 1.22 | 10-Mar-1999 |
perseant | New sources should leave the LFS in a more-or-less working state. Changes
include:
- DIROP segregation is enabled, and greater care is taken
to make sure that a checkpoint completes. Fsck is not
needed to remount the filesystem.
- Several checks to make sure that the LFS subsystem does not
overuse various resources (memory, in particular).
- The cleaner routines, lfs_markv in particular, are completely
rewritten. A buffer overflow is removed. Greater care is taken
to ensure that inodes come from where lfs_cleanerd say they come
from (so we know nothing has changed since lfs_bmapv was called).
- Fragment allocation is fixed, so that writes beyond end-of-file
do the right thing.
|
| 1.21 | 09-Nov-1998 |
mycroft | GC the B_CACHE bit.
|
| 1.20 | 23-Oct-1998 |
thorpej | Use DINODE_SIZE rather than sizeof(struct dinode).
|
| 1.19 | 15-Sep-1998 |
pk | Apply patch from PR#5542: buffer overflow in lfs_markv().
|
| 1.18 | 24-Jun-1998 |
sommerfe | Always include fifos; "not an option any more".
|
| 1.17 | 09-Jun-1998 |
scottr | Protect various config(8)-generated files from inclusion while
building LKMs. Fixes PR 5557.
|
| 1.16 | 01-Mar-1998 |
fvdl | Merge with Lite2 + local changes
|
| 1.15 | 19-Feb-1998 |
thorpej | Include the LFS option header.
|
| 1.14 | 13-Jan-1998 |
thorpej | Nuke spurious semicolon, from Konrad Schroder <perseant@hhhh.org>.
|
| 1.13 | 11-Jun-1997 |
bouyer | Add support for ext2fs, this needed a few modifications to ufs/ufs/inode.h:
- added an "union inode_ext" to struct inode, for the per-fs extentions.
For now only ext2fs uses it.
- i_din is now an union:
union {
struct dinode ffs_din; /* 128 bytes of the on-disk dinode. */
struct ext2fs_dinode e2fs_din; /* 128 bytes of the on-disk dinode. */
} i_din
Added a lot of #define i_ffs_* and i_e2fs_* to access the fields.
- Added two macros: FFS_ITIMES and EXT2FS_ITIMES. ITIMES calls the rigth
macro, depending on the time of the inode. ITIMES is used where necessary,
FFS_ITIMES and EXT2FS_ITIMES in other places.
|
| 1.12 | 12-Oct-1996 |
christos | revert previous kprintf changes
|
| 1.11 | 10-Oct-1996 |
christos | printf -> kprintf, sprintf -> ksprintf
|
| 1.10 | 09-Feb-1996 |
christos | lfs prototypes
|
| 1.9 | 21-Sep-1995 |
thorpej | Make system calls conform to a standard prototype and bring those
prototypes into scope.
|
| 1.8 | 21-Mar-1995 |
mycroft | Update to use timer{add,sub}().
|
| 1.7 | 14-Dec-1994 |
mycroft | Sync with CSRG.
|
| 1.6 | 11-Dec-1994 |
mycroft | Use __timeradd(), not timevaladd().
|
| 1.5 | 20-Oct-1994 |
cgd | update for new syscall args description mechanism, and deal safely
with wider types.
|
| 1.4 | 21-Aug-1994 |
cgd | C syntax fix, and syscall args style (For later.)
|
| 1.3 | 29-Jun-1994 |
cgd | New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
|
| 1.2 | 16-Jun-1994 |
mycroft | This i_flags should be i_flag.
|
| 1.1 | 08-Jun-1994 |
mycroft | branches: 1.1.1;
Update to 4.4-Lite fs code, with local changes.
|
| 1.1.1.2 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite2
|
| 1.1.1.1 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite for reference
|
| 1.26.2.4 | 20-Jan-2000 |
he | Pull up revision 1.40 (requested by perseant):
Files removed (through unlink, rmdir) are now really removed, though the
removal is postponed until the dirop is complete to ensure validity of
the filesystem through a crash. Use a separate per-fs lock, instead of
ufs_hashlock, to protect the inode free list. Change calling semantics
of lfs_ifind, to give better error reporting: If fed a struct buf, it
can report the block number of the offending inode block as well as the
inode number.
|
| 1.26.2.3 | 15-Jan-2000 |
he | Pull up revision 1.38 (requested by perseant):
Expand the category of "metadata" in lfs_markv to include Ifile
data blocks. This prevents a rare condition in which certain
Ifile blocks are "unwritten" when cleaning during intense disk
activity.
|
| 1.26.2.2 | 15-Apr-1999 |
perseant | branches: 1.26.2.2.2;
Pull up 1.29->1.30; trivial fix for the forgotten lock problem.
|
| 1.26.2.1 | 13-Apr-1999 |
perseant | Pull-up of changes made to the trunk on Sunday [1.27-1.28], to wit:
Take out the `#ifdef USE_UFSHASH'; use ufs_hashlock to lock the inode free
list instead of free_lock.
Fix inode reporting in lfs_statfs (the meaning of f_files and f_ffree was
reversed).
Fix "lfs_ifind: dinode xxx not found" panic. When inodes were freed, then
immediately reloaded, their dinodes were located in an inode block which
was not on disk at the advertized location, nor in the cache (although it
would be flushed to disk next segment write). Fix this by using getblk()
instead of lfs_newbuf() for inode blocks.
Better checking for held inode locks in lfs_fastvget, for a number of
error conditions. Also change the default setting of lfs_clean_vnhead to
0, which seems to make the locking problems go away (although this is
difficult to test as I can't reliably reproduce them).
Make sure that the wakeup occurs for vnodes that lfs_update might be
sleeping on (nodes which are not marked IN_MODIFIED/IN_CLEANING, but which
have dirty buffers), by marking them with the appropriate flag if
dirtybuffers were added while the write was in progress.
Fix block counting during file truncation, if not truncating to zero.
Disallow threshold-initiated cache flush when dirops are active. Also,
make SET_ENDOP use lfs_check instead of inlining most of it.
Improve the debugging printfs in the cleaner syscalls (in particular, make
it obvious that they're coming from lfs).
Check the superblock version field, and refuse to mount the filesystem if
the version number is higher than we know about. This allows, e.g.,
changes in the format of the ifile, segment size restrictions and
boundaries, etc., which would not affect existing fields in the
superblock, but which would drastically affect the filesystem, to be
smoothly integrated at a later date.
|
| 1.26.2.2.2.3 | 31-Aug-1999 |
perseant | Rudimentary support for LFS under UBC:
- LFS-specific VOP_BALLOC and VOP_PUTPAGES vnode ops.
- getblk VREG panic #ifdef'd out (can be reinstated when Ifile is
internalized and Ifile can be made another type from VREG)
- interface to VOP_PUTPAGES changed to pass all pager flags, not
just sync. FS putpages routines must know about the pager flags.
- new LFS magic disk address, -2 ("unwritten"), meaning accounted for
but not assigned to a fixed disk location (since LFS does these two
things separately, and the previous accounting method using buffer
headers no longer will work). Changed references to (foo == (daddr_t)-1)
to (foo < 0). Since disk drivers reject all addresses < 0, this should
not present a problem for other FSs.
|
| 1.26.2.2.2.2 | 02-Aug-1999 |
thorpej | Update from trunk.
|
| 1.26.2.2.2.1 | 21-Jun-1999 |
thorpej | Sync w/ -current.
|
| 1.33.8.2 | 27-Dec-1999 |
wrstuden | Pull up to last week's -current.
|
| 1.33.8.1 | 21-Dec-1999 |
wrstuden | Initial commit of recent changes to make DEV_BSIZE go away.
Runs on i386, needs work on other arch's. Main kernel routines should be
fine, but a number of the stand programs need help.
cd, fd, ccd, wd, and sd have been updated. sd has been tested with non-512
byte block devices. vnd, raidframe, and lfs need work.
Non 2**n block support is automatic for LKM's and conditional for kernels
on "options NON_PO2_BLOCKS".
|
| 1.33.4.1 | 15-Nov-1999 |
fvdl | Sync with -current
|
| 1.33.2.3 | 08-Dec-2000 |
bouyer | Sync with HEAD.
|
| 1.33.2.2 | 22-Nov-2000 |
bouyer | Sync with HEAD.
|
| 1.33.2.1 | 20-Nov-2000 |
bouyer | Update thorpej_scsipi to -current as of a month ago
|
| 1.41.4.7 | 16-Aug-2001 |
tv | Pullup [jdolecek]:
sys/ufs/lfs/lfs_syscalls.c 1.58 via patch
Constrain "blkcnt" of lfs_markv() syscall by 64KB.
|
| 1.41.4.6 | 03-Feb-2001 |
he | Pull up revisions 1.50,1.52-1.53 (requested by perseant):
o Close up accounting holes in LFS' accounting of immediately-
available-space, number of clean segments, and amount of dirty
space taken up by metadata (PR#11468, PR#11470, PR#11534).
o Don't clean the same segment twice.
o Fix locking and reference leaks in lfs_markv, on error (PR#11547).
|
| 1.41.4.5 | 01-Nov-2000 |
tv | Pullup 1.51 [toshii]:
In lfs_fastvget(), initialize i_lfs_effnblks correctly.
|
| 1.41.4.4 | 14-Sep-2000 |
perseant | Pull up recent LFS kernel changes (approved by thorpej):
ufs/ufs/inode.h, 1.20--1.22 (add i_lfs_effnblks extension ;
make ITIMES aware of LFS_ITIMES;
_LKM protection so userland progs
compile)
ufs/ufs/ufs_vnops.c, 1.69, 1.71 (remove IN_ADIROP;
use ITIMES instead of FFS_ITIMES)
ufs/ufs/ufs_readwrite.c, 1.27 (use lfs_reserve in lfs_write)
ufs/lfs/lfs.h, 1.26--1.32 (define LFS_EST_* macros ;
change MIN_FREE_SEGS to lfs_minfreesegs ;
add avail and bfree to CLEANERINFO ;
change lfs_uinodes to signed ;
change lfs_dmeta to signed ;
add whitespace to line up structure
members ;
explicit cast to int32_t in LFS_EST_*
macros)
ufs/lfs/lfs_alloc.c, back out 1.34.2.3 (pullups of 1.39, 1.40);
then pull up 1.38 (clean up on error)
1.39--1.43 (restore fvdl's ufs_hashlock fix ;
restore fvdl's ufs_hashlock fix ;
set i_lfs_effnblks ;
use UINO macros ;
add comments and fix long lines)
ufs/lfs/lfs_balloc.c, 1.19 (don't succeed halfway)
1.21--1.25 (use i_lfs_effnblks ;
fix i_lfs_effnblks computation and
quieten ;
fix i_ffs_blocks in unwritten fragment ;
remove useless debugging check ;
add comments and (c) 2000)
ufs/lfs/lfs_bio.c, 1.24--1.30 (cleanup and make lfs_flush_fs take
"struct lfs *" instead of "struct
mount *" ;
use lfs_minfreeseg instead of
MIN_FREE_SEGS ;
use UINO macros, and copy bfree/avail
to CLEANERINFO ;
add lfs_reserve function ;
1.28--1.30 fix printf formatting)
ufs/lfs/lfs_cksum.c, 1.13 (add (c) 2000)
ufs/lfs/lfs_debug.c, 1.11 (use btodb instead of DEV_BSIZE)
ufs/lfs/lfs_extern.h, 1.18, 1.20--1.21 (function prototype changes)
ufs/lfs/lfs_inode.c, 1.38 (rewrite lfs_truncate from
ffs_truncate)
1.40--1.44 (count written and unwritten blocks
seperately ;
use disk block units instead of bytes ;
remove unnecessary "mod" variable ;
correct B_DELWRI to avoid bawrite panic ;
use lfs_reserve)
ufs/lfs/lfs_segment.c, 1.52-1.59 (use lfs_dmeta to note used summaries ;
check for UNWRITTEN in indirect blocks ;
more debugging stuff inside #ifdef
DEBUG_LFS ;
use LK_CANRECURSE ;
don't drop dirty indirect blocks ;
use UINO macros ;
don't hose the free list ;
use btodb() instead of DEV_BSIZE ;
make it compile again (oops))
ufs/lfs/lfs_subr.c, 1.16--1.17 (check for locked inodes before
changing ;
use btodb() instead of DEV_BSIZE, (c)
2000)
ufs/lfs/lfs_syscalls.c, back out 1.41.4.2 (fvdl's ufs_hashlock fix);
then pull up 1.43 (use lfs_dmeta)
1.44--1.45 (restore fvdl's ufs_hashlock fix)
1.46--1.47 (fix lfs_avail leakage from sblock
segments ;
use UINO macros)
1.49 (bounds-check inode numbers in
lfs_markv)
ufs/lfs/lfs_vfsops.c, 1.53 (use LFS_EST_* macros in lfs_statfs)
1.56--1.58 (initialize lfs_minfreeseg, lfs_effnblk ;
initialize lfs_uinodes ;
initialize lfs_ravail)
ufs/lfs/lfs_vnops.c, 1.40 (remove VDIROP from removed files)
1.42--1.44 (move SET_ENDOP below the removal of
VDIROP ;
use UINO macros and add lfs_itimes
function ;
use lfs_reserve in dirops)
|
| 1.41.4.3 | 13-Jul-2000 |
thorpej | Pull up rev. 1.48:
XXX Use of hzto() return value needs to be double-checked here.
|
| 1.41.4.2 | 03-Jul-2000 |
fvdl | pullup the fixes from the trunk to not hold ufs_hashlock across
getnewvnode()
|
| 1.41.4.1 | 22-Jun-2000 |
perseant | Pull up lfs_vunref fix from the trunk.
|
| 1.56.6.5 | 06-Sep-2002 |
jdolecek | sync kqueue branch with HEAD
|
| 1.56.6.4 | 23-Jun-2002 |
jdolecek | catch up with -current on kqueue branch
|
| 1.56.6.3 | 10-Jan-2002 |
thorpej | Sync kqueue branch with -current.
|
| 1.56.6.2 | 25-Aug-2001 |
thorpej | Merge Aug 24 -current into the kqueue branch.
|
| 1.56.6.1 | 03-Aug-2001 |
lukem | update to -current
|
| 1.56.4.3 | 02-Jul-2001 |
perseant | Change disk addressing unit to be the fragment, instead of the disk sector.
All quantities in the superblock, inodes, indirect blocks, etc. refer now
to this abstract unit (called "fsb" as it is in FFS) instead of disk sectors;
as a consequence segment summary blocks have to be multiples of a fragment in
size. In v1 filesystems, compatibility code ensures that 1 fsb == 1 sector,
regardless of fragment size.
Fragments can now range in size between 512 and 32k; in the event that
LFS_LABELPAD (8k) is smaller than the disk address unit size, an extra
proto-superblock is kept at 8k from the beginning of the disk, to be used
*only* to locate the real superblocks. (Not all of the userland knows about
this yet.)
Almost all of this was done not by me, but by joff.
|
| 1.56.4.2 | 29-Jun-2001 |
perseant | Get rid of __P(), protoizing where it had not already been done
|
| 1.56.4.1 | 27-Jun-2001 |
perseant | Import of what I've been calling "LFSv2", that is, LFS with some features
added that require changes to the on-disk data structures. These include:
- 64-bit time in everything but inodes
- User-specified segment offset, and segment size no longer
restricted to PO2.
- Serial number on segment summaries in addition to timestamp, and
a new volume identifier, to make roll-forward feasible without
fear of finding old data and thinking it was new.
Although I think this version works at least as well as what's on the trunk,
we're not done yet; hence this commit is going in on a branch and not on
the trunk. Enhancements that are not here yet include fragment addressing,
like FFS does, instead of block addressing.
|
| 1.56.2.14 | 29-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.56.2.13 | 19-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.56.2.12 | 11-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.56.2.11 | 13-Aug-2002 |
nathanw | Catch up to -current.
|
| 1.56.2.10 | 01-Aug-2002 |
nathanw | Catch up to -current.
|
| 1.56.2.9 | 12-Jul-2002 |
nathanw | No longer need to pull in lwp.h; proc.h pulls it in for us.
|
| 1.56.2.8 | 20-Jun-2002 |
nathanw | Catch up to -current.
|
| 1.56.2.7 | 29-May-2002 |
nathanw | #include <sys/sa.h> before <sys/syscallargs.h>, to provide sa_upcall_t
now that <sys/param.h> doesn't include <sys/sa.h>.
(Behold the Power of Ed)
|
| 1.56.2.6 | 08-Jan-2002 |
nathanw | Catch up to -current.
|
| 1.56.2.5 | 14-Nov-2001 |
nathanw | Catch up to -current.
|
| 1.56.2.4 | 21-Sep-2001 |
nathanw | Catch up to -current.
|
| 1.56.2.3 | 24-Aug-2001 |
nathanw | A few files and lwp/proc conversions I missed in the last big update.
GENERIC runs again.
|
| 1.56.2.2 | 24-Aug-2001 |
nathanw | Catch up with -current.
|
| 1.56.2.1 | 05-Mar-2001 |
nathanw | Initial commit of scheduler activations and lightweight process support.
|
| 1.58.2.1 | 01-Oct-2001 |
fvdl | Catch up with -current.
|
| 1.59.2.1 | 12-Nov-2001 |
thorpej | Sync the thorpej-mips-cache branch with -current.
|
| 1.65.4.2 | 07-Aug-2002 |
lukem | Pull up revision 1.71 (requested by itojun in ticket #616):
correct range check, have overflow check, fix type mismatches,
for cmap args and some other calls. from openbsd
|
| 1.65.4.1 | 20-Jun-2002 |
lukem | Pull up revision 1.67 (requested by perseant in ticket #325):
For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.65.2.3 | 29-Aug-2002 |
gehenna | catch up with -current.
|
| 1.65.2.2 | 15-Jul-2002 |
gehenna | catch up with -current.
|
| 1.65.2.1 | 20-Jun-2002 |
gehenna | catch up with -current.
|
| 1.93.2.9 | 10-Nov-2005 |
skrll | Sync with HEAD. Here we go again...
|
| 1.93.2.8 | 08-Mar-2005 |
skrll | Sync with HEAD.
|
| 1.93.2.7 | 04-Mar-2005 |
skrll | Sync with HEAD.
Hi Perry!
|
| 1.93.2.6 | 30-Oct-2004 |
skrll | Reduced diff to HEAD by restoring the struct proc * argument to lfs_bmapv
|
| 1.93.2.5 | 21-Sep-2004 |
skrll | Fix the sync with head I botched.
|
| 1.93.2.4 | 18-Sep-2004 |
skrll | Sync with HEAD.
|
| 1.93.2.3 | 24-Aug-2004 |
skrll | Undo part of the ktrace/lwp changes. In particular:
* Remove the "lwp *" argument that was added to vget(). Turns out
that nothing actually used it!
* Remove the "lwp *" arguments that were added to VFS_ROOT(), VFS_VGET(),
and VFS_FHTOVP(); all they did was pass it to vget() (which, as noted
above, didn't use it).
* Remove all of the "lwp *" arguments to internal functions that were added
just to appease the above.
|
| 1.93.2.2 | 03-Aug-2004 |
skrll | Sync with HEAD
|
| 1.93.2.1 | 02-Jul-2003 |
darrenr | Apply the aborted ktrace-lwp changes to a specific branch. This is just for
others to review, I'm concerned that patch fuziness may have resulted in some
errant code being generated but I'll look at that later by comparing the diff
from the base to the branch with the file I attempt to apply to it. This will,
at the very least, put the changes in a better context for others to review
them and attempt to tinker with removing passing of 'struct lwp' through
the kernel.
|
| 1.100.10.1 | 19-Mar-2005 |
yamt | sync with head. xen and whitespace. xen part is not finished.
|
| 1.100.8.1 | 29-Apr-2005 |
kent | sync with -current
|
| 1.100.6.1 | 10-May-2005 |
riz | Pull up the following revisions (requested by perseant in ticket #1281):
1.8 sys/ufs/lfs/TODO
1.75 sys/ufs/lfs/lfs.h (via patch)
1.74 sys/ufs/lfs/lfs_alloc.c (via patch)
1.49, 1.51 sys/ufs/lfs/lfs_balloc.c (1.51 via patch)
1.78 sys/ufs/lfs/lfs_bio.c
1.62 sys/ufs/lfs/lfs_extern.h (via patch)
1.156 sys/ufs/lfs/lfs_segment.c (via patch)
1.48 sys/ufs/lfs/lfs_subr.c
1.101 sys/ufs/lfs/lfs_syscalls.c
1.163 sys/ufs/lfs/lfs_vfsops.c (via patch)
1.134 sys/ufs/lfs/lfs_vnops.c (via patch)
1.61 sys/ufs/ufs/ufs_readwrite.c (via patch)
1.20 libexec/lfs_cleanerd/clean.h (via patch)
1.52 libexec/lfs_cleanerd/cleanerd.c (via patch)
1.41 libexec/lfs_cleanerd/library.c (via patch)
1.4 regress/sys/fs/lfs/newfs_fsck/Makefile
1.2 regress/sys/fs/lfs/newfs_fsck/mkfs_mount
1.2 regress/sys/fs/lfs/newfs_fsck/smallfiles
1.3 sbin/fsck_lfs/bufcache.c
1.3 sbin/fsck_lfs/bufcache.h
1.3 sbin/fsck_lfs/lfs.h
1.8 sbin/fsck_lfs/lfs.c (via patch)
1.8 sbin/fsck_lfs/pass3.c (via patch)
1.18 sbin/fsck_lfs/pass0.c (via patch)
1.18 sbin/fsck_lfs/utilities.c (via patch)
1.7 sbin/fsck_lfs/segwrite.c
1.19 sbin/fsck_lfs/setup.c (via patch)
1.3 sbin/newfs_lfs/Makefile
0 sbin/newfs_lfs/lfs.c (yes, remove it)
1.1 sbin/newfs_lfs/make_lfs.c
1.15 sbin/newfs_lfs/newfs.c (via patch)
Various minor LFS improvements.
Kernel:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this. Should fix PR #29045.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
Fixes PR #26680.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
cleaner:
* Adapt lfs_cleanerd to use the fcntl call to get the Ifile filehandle,
so it need not be in the namespace.
* Make lfs_cleanerd be more careful when there are very few available
segments.
* Make lfs_cleanerd less verbose when the filesystem is unmounted.
newfs_lfs, fsck_lfs, and regression:
* Extend the lfs library from fsck_lfs(8) so that it can be used with a
not-yet-existent LFS. Make newfs_lfs(8) use this library, so it can
create LFSs whose Ifile is larger than one segment. Addresses PR #11110.
* Make newfs_lfs(8) use strsuftoi64() for its arguments, a la newfs(8).
* Make fsck_lfs(8) respect the "file system is clean" flag.
* Don't let fsck_lfs(8) think it has dirty blocks when invoked with the
-n flag.
* Remove the Ifile from the filesystem namespace. The cleaner now uses
a fcntl call on the root inode to find the Ifile filehandle. (As a
side-effect, addresses PR #29144.)
|
| 1.103.2.5 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_syscalls.c: revision 1.112
Get rid of the LFS_FORCE_WRITE case. We never really used it, and it could
panic the kernel if cleaner daemon passed the right combination of arguments.
Coverity CID 2741.
|
| 1.103.2.4 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_balloc.c: revision 1.60
sys/ufs/lfs/lfs_syscalls.c: revision 1.111
sys/ufs/lfs/lfs_segment.c: revision 1.172
sys/ufs/lfs/lfs_vnops.c: revision 1.163
Several minor bug fixes:
* Correct (weak) segment lock assertions in lfs_fragextend and lfs_putpages.
* Keep IN_MODIFIED set if we run out of avail in lfs_putpages.
* Don't try to (re)write buffers on a VBLK vnode; fixes a panic I found
while running with an LFS root.
* Raise priority of LFCNSEGWAIT to PVFS; PUSER is way too low for
something the pagedaemon is relying on.
|
| 1.103.2.3 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_segment.c: revision 1.170
sys/ufs/lfs/lfs.h: revision 1.96
sys/ufs/lfs/lfs_vfsops.c: revision 1.194
sys/ufs/lfs/lfs_syscalls.c: revision 1.109
From Konrad Schroeder, in response to strange df output on anoncvs.netbsd.org:
We were returning the wrong value for free space. Now we're not.
|
| 1.103.2.2 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.180
sys/ufs/lfs/lfs_syscalls.c: revision 1.106
sys/ufs/lfs/lfs.h: revision 1.87
Keep track of the number of segments reclaimed, since the cleaner doesn't
do this anymore (it hasn't for quite some time). Add a couple of conditional
debugging messages to indicate why segments are not cleaned, in the event
that lfs_segclean is used.
Make the LFCNSEGWAITALL fcntl work again.
sys/ufs/lfs/lfs_syscalls.c: revision 1.107
Don't update lfs_stats.segs_reclaimed if we're not keeping statistics.
Patch from Juan RP.
|
| 1.103.2.1 | 07-May-2005 |
tron | Apply patch (requested by perseant in ticket #242):
* fsck_lfs buffer cache fixes, including PR #29151
* Change fsck_lfs phase 0 message to reflect reality
* fsck_lfs: check phase 5 (cleanerinfo accounting) even on
roll-forward
* Keep better track of the free list during roll-forward, avoiding
a core dump
* Improve hash table use for fsck_lfs buffer and vnode cache
* Document fsck_lfs flag -f, and implement -q
* Add resize_lfs, including kernel support
* Add LFS to mountd's list of exportable filesystem types
* Make the LFS lkm work again [christos@]
* Add MP locking to the LFS kernel subsystem
* Fix pager_map deadlock in lfs_putpages()
* Avoid incomplete file extension that looks like "partial
truncation" to fsck
* Use lfs_malloc for cleaner malloc, since the cleaner often runs
in low-memory conditions.
* Use splay trees, not hash table, to track page allocation for
write.
* Fix mkdir panic on full fs
* Fix page accounting leak by counting differently.
* Use rightly named structure for lfs_getattr [skrll@]
* Cosmetic changes for readability.
|
| 1.107.2.7 | 04-Feb-2008 |
yamt | sync with head.
|
| 1.107.2.6 | 21-Jan-2008 |
yamt | sync with head
|
| 1.107.2.5 | 27-Oct-2007 |
yamt | sync with head.
|
| 1.107.2.4 | 03-Sep-2007 |
yamt | sync with head.
|
| 1.107.2.3 | 26-Feb-2007 |
yamt | sync with head.
|
| 1.107.2.2 | 30-Dec-2006 |
yamt | sync with head.
|
| 1.107.2.1 | 21-Jun-2006 |
yamt | sync with head.
|
| 1.108.12.2 | 24-May-2006 |
tron | Merge 2006-05-24 NetBSD-current into the "peter-altq" branch.
|
| 1.108.12.1 | 28-Mar-2006 |
tron | Merge 2006-03-28 NetBSD-current into the "peter-altq" branch.
|
| 1.108.10.5 | 11-May-2006 |
elad | sync with head
|
| 1.108.10.4 | 06-May-2006 |
christos | - Move kauth_cred_t declaration to <sys/types.h>
- Cleanup struct ucred; forward declarations that are unused.
- Don't include <sys/kauth.h> in any header, but include it in the c files
that need it.
Approved by core.
|
| 1.108.10.3 | 19-Apr-2006 |
elad | sync with head.
|
| 1.108.10.2 | 10-Mar-2006 |
elad | generic_authorize() -> kauth_authorize_generic().
|
| 1.108.10.1 | 08-Mar-2006 |
elad | Adapt to kernel authorization KPI.
|
| 1.108.8.6 | 03-Sep-2006 |
yamt | sync with head.
|
| 1.108.8.5 | 11-Aug-2006 |
yamt | sync with head
|
| 1.108.8.4 | 26-Jun-2006 |
yamt | sync with head.
|
| 1.108.8.3 | 24-May-2006 |
yamt | sync with head.
|
| 1.108.8.2 | 11-Apr-2006 |
yamt | sync with head
|
| 1.108.8.1 | 01-Apr-2006 |
yamt | sync with head.
|
| 1.108.6.3 | 01-Jun-2006 |
kardel | Sync with head.
|
| 1.108.6.2 | 22-Apr-2006 |
simonb | Sync with head.
|
| 1.108.6.1 | 04-Feb-2006 |
simonb | Adapt for timecounters: mostly use get*time() and use "time_second"
instead of "time.tv_sec".
|
| 1.108.4.1 | 09-Sep-2006 |
rpaulo | sync with head
|
| 1.113.2.1 | 19-Jun-2006 |
chap | Sync with head.
|
| 1.116.4.2 | 10-Dec-2006 |
yamt | sync with head.
|
| 1.116.4.1 | 22-Oct-2006 |
yamt | sync with head
|
| 1.116.2.3 | 30-Jan-2007 |
ad | Remove support for SA. Ok core@.
|
| 1.116.2.2 | 12-Jan-2007 |
ad | Sync with head.
|
| 1.116.2.1 | 18-Nov-2006 |
ad | Sync with head.
|
| 1.118.4.1 | 03-Sep-2007 |
wrstuden | Sync w/ NetBSD-4-RC_1
|
| 1.118.2.1 | 05-Jun-2007 |
bouyer | Pull up following revision(s) (requested by perseant in ticket #703):
sys/miscfs/genfs/genfs.h 1.21
sys/miscfs/genfs/genfs_vnops.c 1.151
sys/ufs/lfs/lfs.h 1.119, 1.120
sys/ufs/lfs/lfs_bio.c 1.99-101
sys/ufs/lfs/lfs_extern.h 1.89
sys/ufs/lfs/lfs_inode.c 1.108, 1.109
sys/ufs/lfs/lfs_segment.c 1.197, 1.199, 1.200
sys/ufs/lfs/lfs_subr.c 1.69, 1.70
sys/ufs/lfs/lfs_syscalls.c 1.119
sys/ufs/lfs/lfs_vfsops.c 1.234, 1.235
sys/ufs/lfs/lfs_vnops.c 1.195, 1.196, 1.200, 1.202-206
Reduce busy waiting in lfs_putpages(), and other LFS improvements.
|
| 1.121.2.1 | 12-Mar-2007 |
rmind | Sync with HEAD.
|
| 1.122.18.1 | 14-Oct-2007 |
yamt | sync with head.
|
| 1.122.16.3 | 23-Mar-2008 |
matt | sync with HEAD
|
| 1.122.16.2 | 09-Jan-2008 |
matt | sync with HEAD
|
| 1.122.16.1 | 06-Nov-2007 |
matt | sync with HEAD
|
| 1.122.14.1 | 26-Oct-2007 |
joerg | Sync with HEAD.
Follow the merge of pmap.c on i386 and amd64 and move
pmap_init_tmp_pgtbl into arch/x86/x86/pmap.c. Modify the ACPI wakeup
code to restore CR4 before jumping back into kernel space as the large
page option might cover that.
|
| 1.122.2.4 | 24-Aug-2007 |
ad | Sync with buffer cache locking changes. See buf.h/vfs_bio.c for details.
Some minor portions are incomplete and needs to be verified as a whole.
|
| 1.122.2.3 | 17-Jun-2007 |
ad | - Increase the number of thread priorities from 128 to 256. How the space
is set up is to be revisited.
- Implement soft interrupts as kernel threads. A generic implementation
is provided, with hooks for fast-path MD code that can run the interrupt
threads over the top of other threads executing in the kernel.
- Split vnode::v_flag into three fields, depending on how the flag is
locked (by the interlock, by the vnode lock, by the file system).
- Miscellaneous locking fixes and improvements.
|
| 1.122.2.2 | 13-May-2007 |
ad | - Pass the error number and residual count to biodone(), and let it handle
setting error indicators. Prepare to eliminate B_ERROR.
- Add a flag argument to brelse() to be set into the buf's flags, instead
of doing it directly. Typically used to set B_INVAL.
- Add a "struct cpu_info *" argument to kthread_create(), to be used to
create bound threads. Change "bool mpsafe" to "int flags".
- Allow exit of LWPs in the IDL state when (l != curlwp).
- More locking fixes & conversion to the new API.
|
| 1.122.2.1 | 13-Mar-2007 |
ad | Pull in the initial set of changes for the vmlocking branch.
|
| 1.124.10.1 | 02-Jan-2008 |
bouyer | Sync with HEAD
|
| 1.124.6.5 | 26-Dec-2007 |
ad | Sync with head.
|
| 1.124.6.4 | 19-Dec-2007 |
ad | Use a global lfs_lock.
|
| 1.124.6.3 | 19-Dec-2007 |
ad | Fix some more problems w/lfs on this branch.
|
| 1.124.6.2 | 19-Dec-2007 |
ad | Get lfs mostly working.
|
| 1.124.6.1 | 04-Dec-2007 |
ad | Pull the vmlocking changes into a new branch.
|
| 1.124.4.1 | 18-Feb-2008 |
mjf | Sync with HEAD.
|
| 1.128.10.1 | 18-May-2008 |
yamt | sync with head.
|
| 1.128.8.2 | 01-Nov-2008 |
christos | Sync with head.
|
| 1.128.8.1 | 29-Mar-2008 |
christos | Welcome to the time_t=long long dev_t=uint64_t branch.
|
| 1.128.6.2 | 17-Jan-2009 |
mjf | Sync with HEAD.
|
| 1.128.6.1 | 02-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.129.2.5 | 11-Aug-2010 |
yamt | sync with head.
|
| 1.129.2.4 | 11-Mar-2010 |
yamt | sync with head
|
| 1.129.2.3 | 16-Sep-2009 |
yamt | sync with head
|
| 1.129.2.2 | 04-May-2009 |
yamt | sync with head.
|
| 1.129.2.1 | 16-May-2008 |
yamt | sync with head.
|
| 1.132.2.3 | 23-Jun-2008 |
wrstuden | Sync w/ -current. 34 merge conflicts to follow.
|
| 1.132.2.2 | 14-May-2008 |
wrstuden | Per discussion with ad, remove most of the #include <sys/sa.h> lines
as they were including sa.h just for the type(s) needed for syscallargs.h.
Instead, create a new file, sys/satypes.h, which contains just the
types needed for syscallargs.h. Yes, there's only one now, but that
may change and it's probably more likely to change if it'd be difficult
to handle. :-)
Per discussion with matt at n dot o, add an include of satypes.h to
sigtypes.h. Upcall handlers are kinda signal handlers, and signalling
is the header file that's already included for syscallargs.h that
closest matches SA.
This shaves about 3000 lines off of the diff of the branch relative
to the base. That also represents about 18% of the total before this
checkin.
I think this reduction is very good thing.
|
| 1.132.2.1 | 10-May-2008 |
wrstuden | Initial checkin of re-adding SA. Everything except kern_sa.c
compiles in GENERIC for i386. This is still a work-in-progress, but
this checkin covers most of the mechanical work (changing signalling
to be able to accomidate SA's process-wide signalling and re-adding
includes of sys/sa.h and savar.h). Subsequent changes will be much
more interesting.
Also, kern_sa.c has received partial cleanup. There's still more
to do, though.
|
| 1.133.6.1 | 19-Jan-2009 |
skrll | Sync with HEAD.
|
| 1.135.2.2 | 17-Aug-2010 |
uebayasi | Sync with HEAD.
|
| 1.135.2.1 | 30-Apr-2010 |
uebayasi | Sync with HEAD.
|
| 1.136.2.3 | 19-May-2011 |
rmind | Implement sharing of vnode_t::v_interlock amongst vnodes:
- Lock is shared amongst UVM objects using uvm_obj_setlock() or getnewvnode().
- Adjust vnode cache to handle unsharing, add VI_LOCKSHARE flag for that.
- Use sharing in tmpfs and layerfs for underlying object.
- Simplify locking in ubc_fault().
- Sprinkle some asserts.
Discussed with ad@.
|
| 1.136.2.2 | 03-Jul-2010 |
rmind | sync with head
|
| 1.136.2.1 | 16-Mar-2010 |
rmind | Change struct uvm_object::vmobjlock to be dynamically allocated with
mutex_obj_alloc(). It allows us to share the locks among UVM objects.
|
| 1.138.6.1 | 23-Jun-2011 |
cherry | Catchup with rmind-uvmplock merge.
|
| 1.139.6.2 | 05-Apr-2012 |
mrg | sync to latest -current.
|
| 1.139.6.1 | 18-Feb-2012 |
mrg | merge to -current.
|
| 1.139.2.3 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.139.2.2 | 23-Jan-2013 |
yamt | sync with head
|
| 1.139.2.1 | 17-Apr-2012 |
yamt | sync with head
|
| 1.142.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.142.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.142.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.142.2.1 | 25-Feb-2013 |
tls | resync with head
|
| 1.147.2.2 | 18-May-2014 |
rmind | sync with head
|
| 1.147.2.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.152.2.1 | 10-Aug-2014 |
tls | Rebase.
|
| 1.155.4.5 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.155.4.4 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.155.4.3 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.155.4.2 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.155.4.1 | 06-Apr-2015 |
skrll | Sync with HEAD
|
| 1.172.4.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.172.2.2 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.172.2.1 | 20-Mar-2017 |
pgoyette | Sync with HEAD
|
| 1.174.4.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.175.10.1 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.175.4.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.391 | 20-Oct-2025 |
perseant | Correct handling of B_MODIFY in lfs_resize_fs to avoid leaving some old
file-entry data in the segment table.
|
| 1.390 | 20-Oct-2025 |
perseant | * Generalize the partial-segment parser introduced for roll-forward,
using it to facilitate an in-kernel segment rewriter (cleaner), and a
mechanism to check whether a segment is in fact empty (only used
with DEBUG).
* Add these new fcntl calls:
- LFCNFILESTATS: For each inode given, report its number of direct
blocks, how many gaps (discontinuities) there are between direct
blocks, and how large the total gap distance is. This will be
useful for a coalescing agent.
- LFCNREWRITEFILE: For each inode given, rewrite its direct blocks,
effectively coalescing it into as compact a form as possible.
- LFCNSCRAMBLE: As above, except that it only rewrites every other
block. This causes the file to have many gaps that can be
measured with LFCNFILESTATS and addressed with LFCNREWRITEFILE,
for testing purposes.
- LFCNREWRITESEGS: Rewrite any live data in the given segments.
This is intended to simplify the cleaner API and facilitate an
in-kernel cleaner.
- LFCNCLEANERINFO: Get the most current CLEANERINFO data from the
kernel.
- LFCNSEGUSE: Retrieve segment usage data from the kernel.
* Vnodes marked IN_CLEANING now take a reference. Add a new "cleaner
lock", which must be taken by the cleaner before the segment lock,
and before marking nodes IN_CLEANING. This allows us to flush
vnodes, if necessary, before the cleaning segment is written, and
never to flush vnodes being cleaned. When the cleaner lock is
released, the vnodes are cleared of IN_CLEANING and the reference
dropped.
* Track a potential infinite loop in lfs_gatherblock.
* Pull "needs to flush" and "needs to wait for flush" into functions
instead of inlining their definitions.
|
| 1.389 | 29-Sep-2025 |
perseant | Use the symbolic name MNT_WAIT when calling VFS_SYNC. No functional change.
|
| 1.388 | 19-Sep-2025 |
perseant | Interpret the "waitfor" argument to lfs_sync to match the passed values.
Not every sync needs to be a synchronous checkpoint.
|
| 1.387 | 17-Sep-2025 |
perseant | Add working in-kernel roll forward.
|
| 1.386 | 17-Sep-2025 |
perseant | Use a workqueue to handle the superblock callback.
|
| 1.385 | 17-Sep-2025 |
perseant | Add routines to check freelist consistency if compiled with DEBUG and
conditional on a kernel variable manipulated via sysctl.
Add checks before and after each routine that modifies the free list.
#if 0 a section of lfs_vfree() that was intended to keep the free list ordered
but instead corrupted it.
|
| 1.384 | 02-Sep-2025 |
perseant | Use a workqueue to handle cluster iodone, rather than doing it in interrupt context.
|
| 1.383 | 30-Dec-2024 |
hannken | emove comment "we are always called with the filesystem marked `MPBUSY'."
above some xxx_sync() operations. These operations get called without
any exclusive lock.
This comment appeared with "add quota support" on 1990-05-02.
On 1998/02/18 MNT_MPBUSY disappeared when vfs_busy() was changed from
an exclusive lock to a shared lock.
PR kern/58837 "ffs: Missing locking around fs_fmod/time"
|
| 1.382 | 19-Mar-2022 |
hannken | branches: 1.382.4; 1.382.10;
Remove now unused VV_LOCKSWORK, all file systems support locking.
Remove unused predicates vn_locked() and vn_anylocked().
Welcome to 9.99.95
|
| 1.381 | 31-Jul-2021 |
andvar | s/threshhold/threshold
|
| 1.380 | 05-Sep-2020 |
riastradh | branches: 1.380.6;
Round of uvm.h cleanup.
The poorly named uvm.h is generally supposed to be for uvm-internal
users only.
- Narrow it to files that actually need it -- mostly files that need
to query whether curlwp is the pagedaemon, which should maybe be
exposed by an external header.
- Use uvm_extern.h where feasible and uvm_*.h for things not exposed
by it. We should split up uvm_extern.h but this will serve for now
to reduce the uvm.h dependencies.
- Use uvm_stat.h and #ifdef UVMHIST uvm.h for files that use
UVMHIST(ubchist), since ubchist is declared in uvm.h but the
reference evaporates if UVMHIST is not defined, so we reduce header
file dependencies.
- Make uvm_device.h and uvm_swap.h independently includable while
here.
ok chs@
|
| 1.379 | 04-Aug-2020 |
riastradh | Mark lfs vnodes with VV_LOCKSWORK, same as ffs.
|
| 1.378 | 04-Apr-2020 |
ad | Merge the remaining changes from the ad-namecache branch, affecting namei()
and getcwd():
- push vnode locking back as far as possible.
- do most lookups directly in the namecache, avoiding vnode locks & refs.
- don't block new refs to vnodes across VOP_INACTIVE().
- get shared locks for VOP_LOOKUP() if the file system supports it.
- correct lock types for VOP_ACCESS() / VOP_GETATTR() in a few places.
Possible future enhancements:
- make the lookups lockless.
- support dotdot lookups by being lockless and inferring absence of chroot.
- maybe make it work for layered file systems.
- avoid vnode references at the root & cwd.
|
| 1.377 | 16-Mar-2020 |
pgoyette | Use the module subsystem's ability to process SYSCTL_SETUP() entries to
automate installation of sysctl nodes.
Note that there are still a number of device and pseudo-device modules
that create entries tied to individual device units, rather than to the
module itself. These are not changed.
|
| 1.376 | 14-Mar-2020 |
ad | Make page waits (WANTED vs BUSY) interlocked by pg->interlock. Gets RW
locks out of the equation for sleep/wakeup, and allows observing+waiting
for busy pages when holding only a read lock. Proposed on tech-kern.
|
| 1.375 | 14-Mar-2020 |
ad | OR into bp->b_cflags; don't overwrite.
|
| 1.374 | 23-Feb-2020 |
ad | UVM locking changes, proposed on tech-kern:
- Change the lock on uvm_object, vm_amap and vm_anon to be a RW lock.
- Break v_interlock and vmobjlock apart. v_interlock remains a mutex.
- Do partial PV list locking in the x86 pmap. Others to follow later.
|
| 1.373 | 23-Feb-2020 |
riastradh | Dust off the orphan detection code and try to make it work.
|
| 1.372 | 23-Feb-2020 |
riastradh | Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
|
| 1.371 | 23-Feb-2020 |
riastradh | Teach lfs to transition ro<->rw.
|
| 1.370 | 18-Feb-2020 |
chs | remove the aiodoned thread. I originally added this to provide a thread context
for doing page cache iodone work, but since then biodone() has changed to
hand off all iodone work to a softint thread, so we no longer need the
special-purpose aiodoned thread.
|
| 1.369 | 17-Jan-2020 |
ad | VFS_VGET(), VFS_ROOT(), VFS_FHTOVP(): give them a "int lktype" argument, to
allow us to get shared locks (or no lock) on the returned vnode. Matches
FreeBSD.
|
| 1.368 | 15-Jan-2020 |
ad | Merge from yamt-pagecache (after much testing):
- Reduce unnecessary page scan in putpages esp. when an object has a ton of
pages cached but only a few of them are dirty.
- Reduce the number of pmap operations by tracking page dirtiness more
precisely in uvm layer.
|
| 1.367 | 31-Dec-2019 |
ad | branches: 1.367.2;
- Add and use wrapper functions that take and acquire page interlocks, and pairs
of page interlocks. Require that the page interlock be held over calls to
uvm_pageactivate(), uvm_pagewire() and similar.
- Solve the concurrency problem with page replacement state. Rather than
updating the global state synchronously, set an intended state on
individual pages (active, inactive, enqueued, dequeued) while holding the
page interlock. After the interlock is released put the pages on a 128
entry per-CPU queue for their state changes to be made real in batch.
This results in in a ~400 fold decrease in contention on my test system.
Proposed on tech-kern but modified to use the page interlock rather than
atomics to synchronise as it's much easier to maintain that way, and
cheaper.
|
| 1.366 | 13-Dec-2019 |
ad | Break the global uvm_pageqlock into a per-page identity lock and a private
lock for use of the pagedaemon policy code. Discussed on tech-kern.
PR kern/54209: NetBSD 8 large memory performance extremely low
PR kern/54210: NetBSD-8 processes presumably not exiting
PR kern/54727: writing a large file causes unreasonable system behaviour
|
| 1.365 | 28-May-2019 |
msaitoh | branches: 1.365.2;
s/recieve/receive/
|
| 1.364 | 01-Jan-2019 |
hannken | Add "void *extra" argument to vcache_new() so a file system may
pass more information about the file to create.
Welcome to 8.99.30
|
| 1.363 | 10-Dec-2018 |
maxv | Remove unused mbuf.h includes.
|
| 1.362 | 28-May-2018 |
chs | branches: 1.362.2;
add a genfs method to allow a file system to limit the range of pages
that are given to a single GOP_WRITE() call. needed by ZFS.
|
| 1.361 | 28-Oct-2017 |
pgoyette | branches: 1.361.2;
Update the kernhist(9) kernel history code to address issues identified
in PR kern/52639, as well as some general cleaning-up...
(As proposed on tech-kern@ with additional changes and enhancements.)
Details of changes:
* All history arguments are now stored as uintmax_t values[1], both in
the kernel and in the structures used for exporting the history data
to userland via sysctl(9). This avoids problems on some architectures
where passing a 64-bit (or larger) value to printf(3) can cause it to
process the value as multiple arguments. (This can be particularly
problematic when printf()'s format string is not a literal, since in
that case the compiler cannot know how large each argument should be.)
* Update the data structures used for exporting kernel history data to
include a version number as well as the length of history arguments.
* All [2] existing users of kernhist(9) have had their format strings
updated. Each format specifier now includes an explicit length
modifier 'j' to refer to numeric values of the size of uintmax_t.
* All [2] existing users of kernhist(9) have had their format strings
updated to replace uses of "%p" with "%#jx", and the pointer
arguments are now cast to (uintptr_t) before being subsequently cast
to (uintmax_t). This is needed to avoid compiler warnings about
casting "pointer to integer of a different size."
* All [2] existing users of kernhist(9) have had instances of "%s" or
"%c" format strings replaced with numeric formats; several instances
of mis-match between format string and argument list have been fixed.
* vmstat(1) has been modified to handle the new size of arguments in the
history data as exported by sysctl(9).
* vmstat(1) now provides a warning message if the history requested with
the -u option does not exist (previously, this condition was silently
ignored, with only a single blank line being printed).
* vmstat(1) now checks the version and argument length included in the
data exported via sysctl(9) and exits if they do not match the values
with which vmstat was built.
* The kernhist(9) man-page has been updated to note the additional
requirements imposed on the format strings, along with several other
minor changes and enhancements.
[1] It would have been possible to use an explicit length (for example,
uint64_t) for the history arguments. But that would require another
"rototill" of all the users in the future when we add support for an
architecture that supports a larger size. Also, the printf(3) format
specifiers for explicitly-sized values, such as "%"PRIu64, are much
more verbose (and less aesthetically appealing, IMHO) than simply
using "%ju".
[2] I've tried very hard to find "all [the] existing users of kernhist(9)"
but it is possible that I've missed some of them. I would be glad to
update any stragglers that anyone identifies.
|
| 1.360 | 26-Jul-2017 |
maya | change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
|
| 1.359 | 17-Apr-2017 |
hannken | branches: 1.359.2; 1.359.4;
Remove unused argument "nextp" from vfs_busy() and vfs_unbusy().
Remove argument "keepref" from vfs_unbusy() and add vfs_ref() where needed.
|
| 1.358 | 17-Apr-2017 |
hannken | Add vfs_ref(mp) and vfs_rele(mp) to add or remove a reference to
struct mount. Rename vfs_destroy(mp) to vfs_rele(mp) and replace
incrementing mp->mnt_refcnt with vfs_ref(mp).
|
| 1.357 | 13-Apr-2017 |
hannken | Switch lfs_flush() and lfs_writerd() to mountlist iterator.
|
| 1.356 | 01-Apr-2017 |
maya | Switch lfs_writer_daemon to use condvar instead of mtsleep.
track thread existence with struct lwp instead of pid + lid,
it's more useful from ddb.
|
| 1.355 | 01-Apr-2017 |
maya | switch lfs_dirops to condvar (from mtsleep)
|
| 1.354 | 01-Apr-2017 |
maya | switch lfs_sleepers to condvar (from mtsleep)
|
| 1.353 | 13-Mar-2017 |
riastradh | #if DIAGNOSTIC panic ---> KASSERT
Replace some #if DEBUG by this too. DEBUG is only for expensive
assertions; these are not.
|
| 1.352 | 17-Feb-2017 |
hannken | Add generic genfs_suspendctl() and use it for all file systems.
Layered file systems need work.
|
| 1.351 | 07-Jul-2016 |
msaitoh | branches: 1.351.2; 1.351.4;
KNF. Remove extra spaces. No functional change.
|
| 1.350 | 20-Jun-2016 |
dholland | Merge -r1.44 of ufs_extattr.c and related change -r1.302 of ffs_vfops.c:
fix use-after-free on failed unmount with extended attributes enabled.
|
| 1.349 | 19-Oct-2015 |
dholland | Set the legacy ulfs fstype field to ULFS2 when mounting lfs64. Oops.
|
| 1.348 | 15-Oct-2015 |
dholland | Enable mounting lfs64 volumes.
|
| 1.347 | 15-Oct-2015 |
dholland | Move stuff from struct ulfsmount to struct lfs.
|
| 1.346 | 10-Oct-2015 |
dholland | Remove no longer needed explicit 32->64 sign extension.
This is the last 32-bit-on-disk item among those that were either
already tagged or readily discoverable.
|
| 1.345 | 01-Sep-2015 |
dholland | Add new accessors for the d_type and d_namlen fields of struct lfs_direct.
Napalm the old byteswap access logic for these.
|
| 1.344 | 01-Sep-2015 |
dholland | Make the inode fields in the 64-bit superblock 64 bits wide.
Reasoning as before.
Note that I am not going through and checking for 64->32 truncations
in inode numbers; I'm sure there are quite a few, but that's a project
for later.
|
| 1.343 | 01-Sep-2015 |
dholland | Add byteswapping to the dinode accessors.
This prevents regressions in the ulfs code when switching to the new
accessors. Note that while adding byteswapping to the other accessors
is straightforward, I haven't done it yet; and that also is not enough
to make LFS_EI work, because there are places lying around that bypass
the accessors for one reason and another and all of them need to be
updated. That is going to have to wait for a later day as LFS_EI is
not on the critical path right now.
|
| 1.342 | 01-Sep-2015 |
dholland | Use the lfs dinode accessors in place of the ufs-derived ones.
(Mostly.)
The ufs-derived ones are fake structure member macros, which are gross
and not very safe. Also, it seems that a lot of places in the lfs code
were using the ffsv1 branch of them unconditionally, and this way it's
guaranteed all those places have been updated.
Found while doing this: for non-devices, have getattr produce NODEV
in the rdev field instead of leaking the address of the first direct
block.
|
| 1.341 | 19-Aug-2015 |
dholland | Part two of dinodes; use the same union everywhere.
(previously the ufs-derived code had things set up slightly different)
Remove a bunch of associated mess.
|
| 1.340 | 12-Aug-2015 |
dholland | Hack up dinode usage to be 64 vs. 32 as needed. Part 1.
(This part changes the native lfs code; the ufs-derived code already
has 64 vs. 32 logic, but as aspects of it are unsafe, and don't
entirely interoperate cleanly with the lfs 64/32 stuff, pass 2 will be
rehashing that.)
|
| 1.339 | 12-Aug-2015 |
dholland | Provide 32-bit and 64-bit versions of FINFO.
This also entailed sorting out part of struct segment, as that
contains a pointer into the current FINFO data.
|
| 1.338 | 12-Aug-2015 |
dholland | Make 32-bit and 64-bit versions of SEGSUM.
Also fix some of the FINFO handling as it's closely entangled.
|
| 1.337 | 12-Aug-2015 |
dholland | Add IFILE32 and IFILE64 structures for the on-disk ifile entries.
Add and use accessors. There are also a bunch of places that cast and
I hope I've found them all...
|
| 1.336 | 12-Aug-2015 |
dholland | Make 32-bit and 64-bit versions of CLEANERINFO.
XXX: while this is written to disk, it seems like much of it would
XXX: be better set up as a commpage shared with the cleaner.
|
| 1.335 | 12-Aug-2015 |
dholland | Fix botched syscall_package. HI CHRISTOS
|
| 1.334 | 02-Aug-2015 |
dholland | Pass the fs object to LFS_MAX_DADDR so it can check lfs_is64.
Remove some hackish intentional 64->32 truncations next to the checks
using LFS_MAX_DADDR, and tackle the problem they handled in bmap
instead.
The problem: the magic block pointer value UNWRITTEN has magic value
-2, and if it's not handled specifically, uint32 -> uint64 promotion
turns it into 4294967294, which then causes consternation and
monkeyhouse downstream.
What's here is still kind of a hack, but it's a step forward.
|
| 1.333 | 02-Aug-2015 |
dholland | Add a (draft) 64-bit superblock. Make things build again.
Add pieces of support for using both superblock types where
convenient, and specifically to the superblock accessors, but don't
actually enable it anywhere.
First substantive step on PR 50000.
|
| 1.332 | 02-Aug-2015 |
dholland | Use accessor functions for the version field of the lfs superblock.
I thought at first maybe the cases that test the version should be
rolled into the accessors, but on the whole I think the conclusion on
that is no.
|
| 1.331 | 02-Aug-2015 |
dholland | Second batch of 64 -> 32 truncations in lfs, along with more minor
tidyups and corrections in passing.
|
| 1.330 | 02-Aug-2015 |
dholland | Fix assorted 64 -> 32 truncations in lfs. Also, some minor tidyups and
corrections in passing.
|
| 1.329 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.328 | 24-Jul-2015 |
dholland | More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)
The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.
|
| 1.327 | 24-Jul-2015 |
dholland | Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)
It also gets rid of cpp abuse in the form of fake structure member
macros.
Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)
XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.
|
| 1.326 | 16-Jul-2015 |
dholland | Don't cast the return value of malloc.
|
| 1.325 | 07-Jun-2015 |
hannken | Fix copy and paste errors from last commits.
- Kernel i386/ALL and amd64/ALL compile again.
- Resolves CID 1304138 (DEADCODE) and 1304139 (IDENTICAL_BRANCHES).
|
| 1.324 | 31-May-2015 |
hannken | Change lfs from hash table to vcache.
- Change lfs_valloc() to return an inode number and version instead of
a vnode and move lfs_ialloc() and lfs_vcreate() to new lfs_init_vnode().
- Add lfs_valloc_fixed() to allocate a known inode, used by kernel
roll forward.
- Remove lfs_*ref(), these functions cannot coexist with vcache and
their commented behaviour is far away from their implementation.
- Add the cleaner lwp and blockinfo to struct ulfsmount so lfs_loadvnode()
may use hints from the cleaner.
- Remove vnode locks from ulfs_lookup() like we did with ufs_lookup().
|
| 1.323 | 31-May-2015 |
hannken | Use VFS_PROTOS() for lfs.
Rename conflicting struct lfs field "lfs_start" to "lfs_s0addr".
No functional change.
|
| 1.322 | 28-Mar-2015 |
maxv | Remove the 'cred' argument from bread(). Remove a now unused var in
ffs_snapshot.c. Update the man page accordingly.
ok hannken@
|
| 1.321 | 16-Apr-2014 |
maxv | branches: 1.321.4;
An (un)privileged user can easily make the kernel dereference a NULL
pointer.
The kernel allows 'data' to be NULL; it's the fs's responsibility to
ensure that it isn't NULL (if the fs actually needs data).
ok christos@
|
| 1.320 | 24-Mar-2014 |
hannken | branches: 1.320.2;
- Make VI_XLOCK, VI_CLEAN and VI_LOCKSHARE private to kern/vfs_*.c.
- Make vwait() static.
- Add vdead_check() to check a vnode for being or becoming dead.
Discussed on tech-kern.
Welcome to 6.99.38
|
| 1.319 | 23-Mar-2014 |
hannken | Change all vfsops to use C99 designated initializers.
No functional changes intended.
|
| 1.318 | 25-Feb-2014 |
pooka | Ensure that the top level sysctl nodes (kern, vfs, net, ...) exist before
the sysctl link sets are processed, and remove redundancy.
Shaves >13kB off of an amd64 GENERIC, not to mention >1k duplicate
lines of code.
|
| 1.317 | 27-Nov-2013 |
christos | Change the queue.3 *_END(&head) macros to NULL. Since we don't have CIRCLEQ
anymore, all the macros expand to NULL anyway, so this improves readability.
Requested by rmind@
|
| 1.316 | 23-Nov-2013 |
christos | change the mountlist CIRCLEQ into a TAILQ
|
| 1.315 | 17-Oct-2013 |
christos | - remove unused variables
- add debug ifdefs for debugging variables
- __USE() where appropriate.
|
| 1.314 | 30-Sep-2013 |
hannken | Replace macro v_specmountpoint with two functions spec_node_getmountedfs()
and spec_node_setmountedfs() to manage the file system mounted on a device.
Assert the device is a block device.
Welcome to 6.99.24
Discussed on tech-kern@ some time ago.
Reviewed by: David Holland <dholland@netbsd.org>
|
| 1.313 | 28-Jul-2013 |
dholland | Merge in some of the stuff for supporting the extended attributes code.
|
| 1.312 | 28-Jul-2013 |
dholland | Add more of the bits for supporting quotas.
|
| 1.311 | 28-Jul-2013 |
dholland | Bring in a copy of ffs_quota2_mount() for reference.
Add stuff to struct lfs that it needs to initialize.
Clear these fields in mount as there's no on-disk support for quota2;
but this increases the chances of being able to add it (or something
like it) in the future.
|
| 1.310 | 28-Jul-2013 |
dholland | Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.309 | 28-Jul-2013 |
dholland | Add lfs_kernel.h for declarations that don't need to be exposed to userland.
lfs currently has the following headers:
lfs.h - on-disk structures and stuff needed for userlevel tools
lfs_inode.h - additional restricted materials for userlevel tools
that operate the fs (newfs_lfs, fsck_lfs, lfs_cleanerd)
lfs_kernel.h - stuff needed only in the kernel
and the following legacy headers that are expected to be mopped up and
folded into one of the above:
lfs_extern.h - function prototypes
ulfs_bswap.h - endian-independent support
ulfs_dinode.h - now contains very little
ulfs_dirhash.h - dirhash support
ulfs_extattr.h - extattr support
ulfs_extern.h - more function prototypes
ulfs_inode.h - assorted kernel-only declarations
ulfs_quota.h - quota support
ulfs_quota1.h - more quota support
ulfs_quota2.h - more quota support
ulfs_quotacommon.h - more quota support
ulfsmount.h - legacy copy of ufsmount material
|
| 1.308 | 28-Jul-2013 |
dholland | Get rid of the ulfs_ops table as we only have one fs in here now.
|
| 1.307 | 18-Jun-2013 |
christos | branches: 1.307.2;
Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.
|
| 1.306 | 17-Jun-2013 |
christos | LFS module does not depend on FFS anymore. (NAKAJIMA Yoshihiro)
|
| 1.305 | 10-Jun-2013 |
hannken | Make DEBUG kernel compile: di_u.inumber -> di_inumber
|
| 1.304 | 08-Jun-2013 |
dholland | DIRBLKSIZ -> LFS_DIRBLKSIZ
DIRECTSIZ -> LFS_DIRECTSIZ
DIRSIZ -> LFS_DIRSIZ
OLDDIRFMT -> LFS_OLDDIRFMT
NEWDIRFMT -> LFS_NEWDIRFMT
IFTODT -> LFS_IFTODT
DTTOIF -> LFS_DTTOIF
|
| 1.303 | 08-Jun-2013 |
dholland | Stick LFS_ in front of IFMT, IFIFO, IFREG, etc. so as not to conflict
with the UFS copies of these symbols. (Which themselves ought to have
UFS_ stuck on.)
|
| 1.302 | 06-Jun-2013 |
dholland | Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.301 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.300 | 06-Jun-2013 |
dholland | Split lfs from ufs step 3: rearrange config stuff.
Add new options:
LFS_EI
LFS_DIRHASH
LFS_EXTATTR
LFS_EXTATTR_AUTOSTART
LFS_QUOTA
LFS_QUOTA2
and update code referring to the corresponding FFS and UFS config
symbols to use the LFS versions. Disable the one extant reference
to APPLE_UFS in the ulfs files. Use opt_lfs.h only, not opt_ffs.h.
|
| 1.299 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.298 | 22-Jan-2013 |
dholland | Stuff UFS_ in front of a few of ufs's symbols to reduce namespace
pollution. Specifically:
ROOTINO -> UFS_ROOTINO
WINO -> UFS_WINO
NXADDR -> UFS_NXADDR
NDADDR -> UFS_NDADDR
NIADDR -> UFS_NIADDR
MAXSYMLINKLEN -> UFS_MAXSYMLINKLEN
MAXSYMLINKLEN_UFS[12] -> UFS[12]_MAXSYMLINKLEN (for consistency)
Sort out ext2fs's misuse of NDADDR and NIADDR; fortunately, these have
the same values in ext2fs and ffs.
No functional change intended.
|
| 1.297 | 20-Dec-2012 |
hannken | Change bread() and breadn() to never return a buffer on
error and modify all callers to not brelse() on error.
Welcome to 6.99.16
PR kern/46282 (6.0_BETA crash: msdosfs_bmap -> pcbmap -> bread -> bio_doread)
|
| 1.296 | 30-Apr-2012 |
rmind | branches: 1.296.2;
- Replace some malloc(9) uses with kmem(9).
- G/C M_IPMOPTS, M_IPMADDR and M_BWMETER.
|
| 1.295 | 13-Mar-2012 |
elad | Replace the remaining KAUTH_GENERIC_ISSUSER authorization calls with
something meaningful. All relevant documentation has been updated or
written.
Most of these changes were brought up in the following messages:
http://mail-index.netbsd.org/tech-kern/2012/01/18/msg012490.html
http://mail-index.netbsd.org/tech-kern/2012/01/19/msg012502.html
http://mail-index.netbsd.org/tech-kern/2012/02/17/msg012728.html
Thanks to christos, manu, njoly, and jmmv for input.
Huge thanks to pgoyette for spinning these changes through some build
cycles and ATF.
|
| 1.294 | 16-Feb-2012 |
perseant | Pass t_renamerace and t_rmdirrace tests.
Adapt dholland@'s fix to ufs_rename to fix PR kern/43582. Address several
other MP locking issues discovered during the course of investigating the
same problem.
Removed extraneous vn_lock() calls on the Ifile, since the Ifile writes
are controlled by the segment lock.
Fix PR kern/45982 by deemphasizing the estimate of how much metadata
will fill the empty space on disk when the disk is nearly empty
(t_renamerace crates a lot of inode blocks on a tiny empty disk).
|
| 1.293 | 04-Jan-2012 |
perseant | branches: 1.293.2;
lfs_writerd thread exits when no more LFSs are mounted.
|
| 1.292 | 02-Jan-2012 |
perseant | * Remove PGO_RECLAIM during lfs_putpages()' call to genfs_putpages(),
to avoid a live lock in the latter when reclaiming a vnode with
dirty pages.
* Add a new segment flag, SEGM_RECLAIM, to note when a segment is
being written for vnode reclamation, and record which inode is being
reclaimed, to aid in forensic debugging.
* Add a new segment flag, SEGM_SINGLE, so that opportunistic writes
can write a single segment's worth of blocks and then stop, rather
than writing all the way up to the cleaner's reserved number of
segments.
* Add assert statements to check mutex ownership is the way it ought
to be, mostly in lfs_putpages; fix problems uncovered by this.
* Don't clear VU_DIROP until the inode actually makes its way to disk,
avoiding a problem where dirop inodes could become separated
(uncovered by a modified version of the "ckckp" forensic regression
test).
* Move the vfs_getopsbyname() call into lfs_writerd. Prepare code to
make lfs_writerd notice when there are no more LFSs, and exit losing
the reference, so that, in theory, the module can be unloaded. This
code is not enabled, since it causes a crash on exit.
* Set IN_MODIFIED on inodes flushed by lfs_flush_dirops. Really we
only need to set IN_MODIFIED if we are going to write them again
(e.g., to write pages); need to think about this more.
Finally, several changes to help avoid "no clean segments" panics:
* In lfs_bmapv, note when a vnode is loaded only to discover whether
its blocks are live, so it can immediately be recycled. Since the
cleaner will try to choose ~empty segments over full ones, this
prevents the cleaner from (1) filling the vnode cache with junk, and
(2) squeezing any unwritten writes to disk and running the fs out of
segments.
* Overestimate by half the amount of metadata that will be required
to fill the clean segments. This will make the disk appear smaller,
but should help avoid a "no clean segments" panic.
* Rearrange lfs_writerd. In particular, lfs_writerd now pays
attention to the number of clean segments available, and holds off
writing until there is room.
|
| 1.291 | 14-Nov-2011 |
hannken | branches: 1.291.4;
VOP_OPEN() needs a locked vnode. All these copy-and-pasted xxxfs_mount()
implementations need more review.
|
| 1.290 | 11-Jul-2011 |
hannken | branches: 1.290.2;
Change VOP_BWRITE() to take a vnode as its first argument like all other
VOPs do. Layered file systems no longer have to modify bp->b_vp and run
into trouble when an async VOP_BWRITE() uses the wrong vnode.
- change all occurences of VOP_BWRITE(bp) to VOP_BWRITE(bp->b_vp, bp).
- remove layer_bwrite().
- welcome to 5.99.55
Adresses PR kern/38762 panic: vwakeup: neg numoutput
No objections from tech-kern@.
|
| 1.289 | 12-Jun-2011 |
rmind | Welcome to 5.99.53! Merge rmind-uvmplock branch:
- Reorganize locking in UVM and provide extra serialisation for pmap(9).
New lock order: [vmpage-owner-lock] -> pmap-lock.
- Simplify locking in some pmap(9) modules by removing P->V locking.
- Use lock object on vmobjlock (and thus vnode_t::v_interlock) to share
the locks amongst UVM objects where necessary (tmpfs, layerfs, unionfs).
- Rewrite and optimise x86 TLB shootdown code, make it simpler and cleaner.
Add TLBSTATS option for x86 to collect statistics about TLB shootdowns.
- Unify /dev/mem et al in MI code and provide required locking (removes
kernel-lock on some ports). Also, avoid cache-aliasing issues.
Thanks to Andrew Doran and Joerg Sonnenberger, as their initial patches
formed the core changes of this branch.
|
| 1.288 | 06-Mar-2011 |
bouyer | branches: 1.288.2;
merge the bouyer-quota2 branch. This adds a new on-disk format
to store disk quota usage and limits, integrated with ffs
metadata. Usage is checked by fsck_ffs (no more quotacheck)
and is covered by the WAPBL journal. Enabled with kernel
option QUOTA2 (added where QUOTA was enabled in kernel config files),
turned on with tunefs(8) on a per-filesystem
basis. mount_mfs(8) can also turn quotas on.
See http://mail-index.netbsd.org/tech-kern/2011/02/19/msg010025.html
for details.
|
| 1.287 | 24-Jun-2010 |
hannken | branches: 1.287.2; 1.287.4;
Clean up vnode lock operations pass 2:
VOP_UNLOCK(vp, flags) -> VOP_UNLOCK(vp): Remove the unneeded flags argument.
Welcome to 5.99.32.
Discussed on tech-kern.
|
| 1.286 | 02-Mar-2010 |
pooka | branches: 1.286.2;
load lfs syscalls in modload
|
| 1.285 | 02-Mar-2010 |
pooka | /*
* XXX: Get extra reference to LFS vfsops. This prevents unload,
* but also prevents kernel panic due to text being unloaded
* from below lfs_writerd. When lfs_writerd can exit, remove
* this!!!
*/
|
| 1.284 | 18-Feb-2010 |
eeh | Fix root filesystem support.
|
| 1.283 | 16-Feb-2010 |
mlelstv | Three changes in a single commit.
- drop the notion of frags (LFS fragments) vs fsb (FFS fragments)
The code uses a complicated unity function that just makes the
code difficult to understand.
- support larger sector sizes. Fix disk address computations
to use DEV_BSIZE in the kernel as required by device drivers
and to use sector sizes in userland.
- Fix several locking bugs in lfs_bio.c and lfs_subr.c.
|
| 1.282 | 08-Jan-2010 |
pooka | branches: 1.282.2;
The VATTR_NULL/VREF/VHOLD/HOLDRELE() macros lost their will to live
years ago when the kernel was modified to not alter ABI based on
DIAGNOSTIC, and now just call the respective function interfaces
(in lowercase). Plenty of mix'n match upper/lowercase has creeped
into the tree since then. Nuke the macros and convert all callsites
to lowercase.
no functional change
|
| 1.281 | 07-Dec-2009 |
eeh | Fix some more hangs and deadlocks.
|
| 1.280 | 17-Nov-2009 |
pooka | Create unwind log in global variable instead of automatic variable.
memory leak spotted by njoly's valgrind run
|
| 1.279 | 29-Oct-2009 |
eeh | Fix up numoutput accounting.
|
| 1.278 | 13-Sep-2009 |
tsutsui | Move declaration of ufs_hashlock into <ufs/ufs_extern.h> from each c source.
|
| 1.277 | 05-Aug-2009 |
pooka | Compensate v_numoutput & nestbuf for lfs's rather peculiar I/O habits.
|
| 1.276 | 05-Aug-2009 |
pooka | remember to nestiobuf_done() too
|
| 1.275 | 05-Aug-2009 |
pooka | Use nestiobuf instead of homerolled equivalent.
|
| 1.274 | 29-Jun-2009 |
dholland | Convert 67 namei call sites to use namei_simple, in these functions:
check_console, veriexecclose, veriexec_delete, veriexec_file_add,
emul_find_root, coff_load_shlib (sh3 version), coff_load_shlib,
compat_20_sys_statfs, compat_20_netbsd32_statfs,
ELFNAME2(netbsd32,probe_noteless), darwin_sys_statfs,
ibcs2_sys_statfs, ibcs2_sys_statvfs, linux_sys_uselib,
osf1_sys_statfs, sunos_sys_statfs, sunos32_sys_statfs,
ultrix_sys_statfs, do_sys_mount, fss_create_files (3 of 4),
adosfs_mount, cd9660_mount, coda_ioctl, coda_mount, ext2fs_mount,
ffs_mount, filecore_mount, hfs_mount, lfs_mount, msdosfs_mount,
ntfs_mount, sysvbfs_mount, udf_mount, union_mount, sys_chflags,
sys_lchflags, sys_chmod, sys_lchmod, sys_chown, sys_lchown,
sys___posix_chown, sys___posix_lchown, sys_link, do_sys_pstatvfs,
sys_quotactl, sys_revoke, sys_truncate, do_sys_utimes, sys_extattrctl,
sys_extattr_set_file, sys_extattr_set_link, sys_extattr_get_file,
sys_extattr_get_link, sys_extattr_delete_file,
sys_extattr_delete_link, sys_extattr_list_file, sys_extattr_list_link,
sys_setxattr, sys_lsetxattr, sys_getxattr, sys_lgetxattr,
sys_listxattr, sys_llistxattr, sys_removexattr, sys_lremovexattr
All have been scrutinized (several times, in fact) and compile-tested,
but not all have been explicitly tested in action.
XXX: While I haven't (intentionally) changed the use or nonuse of
XXX: TRYEMULROOT in any of these places, I'm not convinced all the
XXX: uses are correct; an audit might be desirable.
|
| 1.273 | 07-May-2009 |
elad | Use genfs_can_mount().
|
| 1.272 | 04-Apr-2009 |
ad | Turn up the volume on the warning message a bit.
|
| 1.271 | 15-Mar-2009 |
cegger | ansify function definitions
|
| 1.270 | 22-Feb-2009 |
ad | PR kern/26878 FFSv2 + softdep = livelock (no free ram)
PR kern/16942 panic with softdep and quotas
PR kern/19565 panic: softdep_write_inodeblock: indirect pointer #1 mismatch
PR kern/26274 softdep panic: allocdirect_merge: ...
PR kern/26374 Long delay before non-root users can write to softdep partitions
PR kern/28621 1.6.x "vp != NULL" panic in ffs_softdep.c:4653 while unmounting a softdep (+quota) filesystem
PR kern/29513 FFS+Softdep panic with unfsck-able file-corruption
PR kern/31544 The ffs softdep code appears to fail to write dirty bits to disk
PR kern/31981 stopping scsi disk can cause panic (softdep)
PR kern/32116 kernel panic in softdep (assertion failure)
PR kern/32532 softdep_trackbufs deadlock
PR kern/37191 softdep: locking against myself
PR kern/40474 Kernel panic after remounting raid root with softdep
Retire softdep, pass 2. As discussed and later formally announced on the
mailing lists.
|
| 1.269 | 13-Nov-2008 |
ad | branches: 1.269.4;
These depend on ffs.
|
| 1.268 | 13-Nov-2008 |
ad | Remove #ifdef LFS from the ufs code.
|
| 1.267 | 28-Jun-2008 |
rumble | branches: 1.267.2; 1.267.4; 1.267.6;
Fix lkm fallout from previous sysctl changes. This largely duplicates
sysctl creation code, but lkms are going away soon(ish) anyway.
Spotted by Chris Gilbert.
|
| 1.266 | 28-Jun-2008 |
rumble | Create sysctl entries during module initialisation and destroy them
appropriately.
Many of these file systems are now ready for modularisation.
|
| 1.265 | 24-May-2008 |
nakayama | branches: 1.265.2;
s/log file system/log-structured file system/
|
| 1.264 | 20-May-2008 |
ad | Don't moan about LFS unless the mount succeeds.
|
| 1.263 | 18-May-2008 |
ad | Until these get fixed or replaced:
WARNING: the foo file system is experimental and may be unstable
|
| 1.262 | 16-May-2008 |
hannken | Make sure all cached buffers with valid, not yet written data have been
run through copy-on-write. Call fscow_run() with valid data where possible.
The LP_UFSCOW hack is no longer needed to protect ffs_copyonwrite() against
endless recursion.
- Add a flag B_MODIFY to bread(), breada() and breadn(). If set the caller
intends to modify the buffer returned.
- Always run copy-on-write on buffers returned from ffs_balloc().
- Add new function ffs_getblk() that gets a buffer, assigns a new blkno,
may clear the buffer and runs copy-on-write. Process possible errors
from getblk() or fscow_run(). Part of PR kern/38664.
Welcome to 4.99.63
Reviewed by: YAMAMOTO Takashi <yamt@netbsd.org>
|
| 1.261 | 10-May-2008 |
rumble | Convert file systems to dynamically attach with the new module interface.
Make VFS hooks dynamic while we're here and say farewell to VFS_ATTACH and
VFS_HOOKS_ATTACH linksets.
As a consequence, most of the file systems can now be loaded as new style
modules.
Quick sanity check by ad@.
|
| 1.260 | 06-May-2008 |
ad | branches: 1.260.2;
PR kern/38141 lookup/vfs_busy acquire rwlock recursively
Simplify the mount locking. Remove all the crud to deal with recursion on
the mount lock, and crud to deal with unmount as another weirdo lock.
Hopefully this will once and for all fix the deadlocks with this. With this
commit there are two locks on each mount:
- krwlock_t mnt_unmounting. This is used to prevent unmount across critical
sections like getnewvnode(). It's only ever read locked with rw_tryenter(),
and is only ever write locked in dounmount(). A write hold can't be taken
on this lock if the current LWP could hold a vnode lock.
- kmutex_t mnt_updating. This is taken by threads updating the mount, for
example when going r/o -> r/w, and is only present to serialize updates.
In order to take this lock, a read hold must first be taken on
mnt_unmounting, and the two need to be held across the operation.
One effect of this change: previously if an unmount failed, we would make a
half hearted attempt to back out of it gracefully, but that was unlikely to
work in a lot of cases. Now while an unmount that will be aborted is in
progress, new file operations within the mount will fail instead of being
delayed. That is unlikely to be a problem though, because if the admin
requests unmount of a file system then s(he) has made a decision to deny
access to the resource.
|
| 1.259 | 30-Apr-2008 |
ad | PR kern/38135 vfs_busy/vfs_trybusy confusion
The previous fix worked, but it opened a window where mounts could have
disappeared from mountlist while the caller was traversing it using
vfs_trybusy(). Fix that.
|
| 1.258 | 29-Apr-2008 |
ad | kern/38135 vfs_busy/vfs_trybusy confusion
The symptom was that sometimes file systems would occasionally not appear
in output from 'df' or 'mount' if the system was busy. Resolution:
- Make mount locks work somewhat like vm_map locks.
- vfs_trybusy() now only fails if the mount is gone, or if someone is
unmounting the file system. Simple contention on mnt_lock doesn't
cause it to fail.
- vfs_busy() will wait even if the file system is being unmounted.
|
| 1.257 | 29-Apr-2008 |
ad | PR kern/38057 ffs makes assuptions about devvp file system
PR kern/33406 softdeps get stuck in endless loop
Introduce VFS_FSYNC() and call it when syncing a block device, if it
has a mounted file system.
|
| 1.256 | 28-Apr-2008 |
martin | Remove clause 3 and 4 from TNF licenses
|
| 1.255 | 30-Jan-2008 |
ad | branches: 1.255.6; 1.255.8; 1.255.10;
PR kern/37706 (forced unmount of file systems is unsafe):
- Do reference counting for 'struct mount'. Each vnode associated with a
mount takes a reference, and in turn the mount takes a reference to the
vfsops.
- Now that mounts are reference counted, replace the overcomplicated mount
locking inherited from 4.4BSD with a recursable rwlock.
|
| 1.254 | 28-Jan-2008 |
dholland | Fix some race conditions in rename.
Introduce a per-FS rename lock and new vfsops to manipulate it.
Get this lock while renaming. Also add another relookup() in do_sys_rename,
which is a hack to kludge around some of the worst deficiencies of
ufs_rename.
reviewed-by: pooka (and an earlier rev by ad)
posted on tech-kern with no objections.
|
| 1.253 | 24-Jan-2008 |
ad | specfs changes for PR kern/37717 (raidclose() is no longer called on
shutdown). There are still problems with device access and a PR will be
filed.
- Kill checkalias(). Allow multiple vnodes to reference a single device.
- Don't play dangerous tricks with block vnodes to ensure that only one
vnode can describe a block device. Instead, prohibit concurrent opens of
block devices. As a bonus remove the unreliable code that prevents
multiple file system mounts on the same device. It's no longer needed.
- Track opens by vnode and by device. Issue cdev_close() when the last open
goes away, instead of abusing vnode::v_usecount to tell if the device is
open.
|
| 1.252 | 02-Jan-2008 |
ad | Merge vmlocking2 to head.
|
| 1.251 | 12-Dec-2007 |
lukem | defflag LFS_KERNEL_RFW (in opt_lfs.h).
Note: lfs_rfw.c doesn't compile if you define the option; locking API fallout?
|
| 1.250 | 08-Dec-2007 |
pooka | branches: 1.250.2; 1.250.4;
Remove cn_lwp from struct componentname. curlwp should be used
from on. The NDINIT() macro no longer takes the lwp parameter and
associates the credentials of the calling thread with the namei
structure.
|
| 1.249 | 26-Nov-2007 |
pooka | branches: 1.249.2;
Remove the "struct lwp *" argument from all VFS and VOP interfaces.
The general trend is to remove it from all kernel interfaces and
this is a start. In case the calling lwp is desired, curlwp should
be used.
quick consensus on tech-kern
|
| 1.248 | 22-Nov-2007 |
yamt | lfs_mountroot: use vfs_destroy.
|
| 1.247 | 10-Nov-2007 |
rmind | Use PRI_BIO for kthreads instead of PINOD. Fixes a missed case of priority
inversion, which caused LFS to fire some assertions.
Reported by Kurt Schreiner on <current-users>.
|
| 1.246 | 10-Oct-2007 |
ad | branches: 1.246.2; 1.246.4;
Merge from vmlocking:
- Split vnode::v_flag into three fields, depending on field locking.
- simple_lock -> kmutex in a few places.
- Fix some simple locking problems.
|
| 1.245 | 08-Oct-2007 |
ad | Merge ffs locking & brelse changes from the vmlocking branch.
|
| 1.244 | 31-Jul-2007 |
pooka | branches: 1.244.2; 1.244.4; 1.244.6; 1.244.8;
* nuke the nameidata parameter from VFS_MOUNT(). Nobody on tech-kern
knew what it was supposed to be used for and wrstuden gave a go-ahead
* while rototilling, convert file systems which went easily to
use VFS_PROTOS() instead of manually prototyping the methods
|
| 1.243 | 29-Jul-2007 |
ad | It's not a good idea for device drivers to modify b_flags, as they don't
need to understand the locking around that field. Instead of setting
B_ERROR, set b_error instead. b_error is 'owned' by whoever completes
the I/O request.
|
| 1.242 | 26-Jul-2007 |
pooka | Use eopnotsupp() instead of vfs_stdsuspendctl() and retire the latter.
|
| 1.241 | 23-Jul-2007 |
ad | Workaround the ufs_haslock/ufs_ihash_lock deadlock. From a patch
posted by Blair Sadewitz.
|
| 1.240 | 17-Jul-2007 |
christos | branches: 1.240.2;
Eliminate MFSNAMELEN
|
| 1.239 | 17-Jul-2007 |
pooka | Make set_statvfs_info() take a parameter for the vfs name instead
of always retrieving it from mp->mnt_op->vfs_name
christos ok
|
| 1.238 | 12-Jul-2007 |
dsl | Change the VFS_MOUNT() interface so that the 'data' buffer passed to the
fs code is a kernel buffer, pass though the length of the buffer as well.
Since the length of the userspace buffer isn'it (yet) passed through the mount
system call, add a field to the vfsops structure containing the default length.
Split sys_mount() for calls from compat code.
Ride one of the recent kernel version changes - old fs LKMs will load, but
sys_mount() will reject any attempt to use them.
|
| 1.237 | 09-Jul-2007 |
ad | Merge some of the less invasive changes from the vmlocking branch:
- kthread, callout, devsw API changes
- select()/poll() improvements
- miscellaneous MT safety improvements
|
| 1.236 | 30-Jun-2007 |
pooka | Using POOL_INIT here makes no sense, since file systems always have
an init method. So get rid of it and #ifdef _LKM and just always
init in the init method. Give malloc types the same treatment.
Makes file systems nicer to work with in linksetless environments
and fixes a few LKM discrepancies.
|
| 1.235 | 16-May-2007 |
perseant | Change references to SEGM_W_DIROPS to SEGM_CKP, and replace the logic that
formerly used SEGM_W_DIROPS in lfs_segwrite() appropriately. This prevents
a problem in which processes could get stuck in "buffers" sleep forever.
|
| 1.234 | 17-Apr-2007 |
perseant | Install a new sysctl, vfs.lfs.ignore_lazy_sync, which causes LFS to ignore
the "smooth" syncer, as if vfs.sync.*delay = 0, but only for LFS. The
default is "on", i.e., ignore lazy sync.
Reduce the amount of polling/busy-waiting done by lfs_putpages(). To
accomplish this, copied genfs_putpages() and modified it to indicate which
page it was that caused it to return with EDEADLK. fsync()/fdatasync()
should no longer ever fail with EAGAIN, and should not consume huge
quantities of cpu.
Also, try to make dirops less likely to be written as the result of a
VOP_PUTPAGES(), while ensuring that they are written regularly.
|
| 1.233 | 13-Mar-2007 |
ad | Pass an ipl argument to pool_init/POOL_INIT to be used when initializing
the pool's lock.
|
| 1.232 | 12-Mar-2007 |
ad | branches: 1.232.2;
Pass an ipl argument to pool_init/POOL_INIT to be used when initializing
the pool's lock.
|
| 1.231 | 22-Feb-2007 |
thorpej | branches: 1.231.4;
TRUE -> true, FALSE -> false
|
| 1.230 | 21-Feb-2007 |
thorpej | Replace the Mach-derived boolean_t type with the C99 bool type. A
future commit will replace use of TRUE and FALSE with true and false.
|
| 1.229 | 18-Feb-2007 |
ad | Release ufs_hashlock before calling ungetnewvnode().
|
| 1.228 | 15-Feb-2007 |
ad | branches: 1.228.2;
Destroy the fraglock on unmount.
|
| 1.227 | 15-Feb-2007 |
ad | Replace some uses of lockmgr() / simplelocks.
|
| 1.226 | 19-Jan-2007 |
hannken | New file system suspension API to replace vn_start_write and vn_finished_write.
The suspension helpers are now put into file system specific operations.
This means every file system not supporting these helpers cannot be suspended
and therefore snapshots are no longer possible.
Implemented for file systems of type ffs.
The new API is enabled on a kernel option NEWVNGATE. This option is
not enabled by default in any kernel config.
Presented and discussed on tech-kern with much input from
Bill Studenmund <wrstuden@netbsd.org> and YAMAMOTO Takashi <yamt@netbsd.org>.
Welcome to 4.99.9 (new vfs op vfs_suspendctl).
|
| 1.225 | 04-Jan-2007 |
elad | Consistent usage of KAUTH_GENERIC_ISSUSER.
|
| 1.224 | 16-Nov-2006 |
christos | branches: 1.224.2; 1.224.4;
__unused removal on arguments; approved by core.
|
| 1.223 | 12-Oct-2006 |
christos | - sprinkle __unused on function decls.
- fix a couple of unused bugs
- no more -Wno-unused for i386
|
| 1.222 | 04-Oct-2006 |
christos | fix empty if
|
| 1.221 | 28-Sep-2006 |
perseant | Use lockstatus instead of a homebrewed locking system to control
LFCNWRAPSTOP and LFCNWRAPGO.
Be less verbose about the various looping checks: use log() rather than
printf(), and only log anything if we are really looping ("count = 2" is
not an error condition).
Allow dirops sleeping on available space to be interruptible.
|
| 1.220 | 02-Sep-2006 |
christos | branches: 1.220.2; 1.220.4;
- add missing initializers
- comment out impossible code
|
| 1.219 | 01-Sep-2006 |
perseant | Changes to help the roll-forward agent, to wit:
* Mark being-deleted files in the Ifile so we can finish deleting them
at fs mount time.
* Flag the Ifile with "cleaner must clean" when writers are waiting for
the cleaner, rather than relying solely on the cleaner's estimation of
whether it should clean or not.
* Note partial segments written by a user agent (in particular,
fsck_lfs) so that repeated rolls forward don't interfere with one
another.
* Add a new fcntl, LFCNPASS, that allows the log to wrap exactly once,
for better testing of the validity of checkpoints.
* Keep track of the on-disk nlink count when cleaning, so that we don't
partially complete directory operations while cleaning.
* Ensure that every single Ifile inode write represents a consistent
view of the filesystem. In particular, the accounting for the segment
we are writing the inode into must be correct, and the accounting for
the segment that inode used to reside in must be correct. Rather than
just rewriting the inode if we wrote it wrong, rewrite the necessary
ifile blocks before writing the inode so we never write it wrong.
* Don't unmark any VDIROP vnodes if we haven't written them to disk,
avoiding yet another problem with the "wait for the cleaner" error
return from lfs_putpages().
Also, move the last callback to an aiodone call, so we no longer do any
memory management from interrupt context.
|
| 1.218 | 23-Jul-2006 |
ad | Use the LWP cached credentials where sane.
|
| 1.217 | 20-Jul-2006 |
perseant | Separate the (non-working) LFS kernel roll-forward code into its own file,
lfs_rfw.c.
|
| 1.216 | 13-Jul-2006 |
martin | Fix alignement problems for fhandle_t, exposed by gcc4.1.
While touching all vptofh/fhtovp functions, get rid of VFS_MAXFIDSIZ,
version the getfh(2) syscall and explicitly pass the size available in
the filehandle from userland.
Discussed on tech-kern, with lots of help from yamt (thanks!).
|
| 1.215 | 06-Jul-2006 |
perseant | Fix a typo that caused a "multiple free" panic on unmounting a resized lfs.
|
| 1.214 | 29-Jun-2006 |
perseant | Don't wake up the cleaner if the filesystem is unwrappable, and fix the
compatibility fcntls.
Also includes one-line fixes for an MP locking bug and a zero-length FINFO
problem that manifested during testing.
|
| 1.213 | 24-May-2006 |
perseant | branches: 1.213.2;
Read the inode version number fro a more reliable source, quelling a
diagnostic assertion panic.
|
| 1.212 | 18-May-2006 |
perseant | branches: 1.212.2;
Break out the finfo array manipulation code into two new functions,
lfs_acquire_finfo() and lfs_release_finfo(). Add a debugging check
for zero-length finfo arrays in the segment summary to avoid future
regressions.
|
| 1.211 | 18-May-2006 |
perseant | Don't duplicate the LFS_STARVED_FOR_SEGS check (an oversight that came
in with rev 1.210).
|
| 1.210 | 14-May-2006 |
elad | integrate kauth.
|
| 1.209 | 12-May-2006 |
perseant | Fixes to address the "vinvalbuf: dirty blocks" panic that can occur when
many inodes are cleaned at once. Make sure that we write all the pages
on vnodes that are being flushed, even if we don't think there's room;
drain v_numoutput before lfs_vflush() completes.
Also, don't allow a vnode that is in the process of being cleaned to be
chosen by getnewvnode(); this avoids a segment accounting panic in the case
that a large number of inodes are fed to lfs_markv() all at once.
|
| 1.208 | 10-May-2006 |
mrg | quell GCC 4.1 uninitialised variable warnings.
XXX: we should audit the tree for which old ones are no longer needed
after getting the older compilers out of the tree..
|
| 1.207 | 04-May-2006 |
perseant | Introduce another per-filesystem parameter, lfs_resvseg, to separate the
notion of "how many segments are reserved for the cleaner" from that of
"how many segments are not counted in lfs_bfree". The default value
used for existing filesystems is the same as the previous implicit value
of (lfs_minfreeseg / 2 + 1), modulo some sanity checking.
Count pending dirops on a per-filesystem basis, since once we start
writing them we can't stop until we're done. This seems to help stave off
the "no clean segments" panic in the case of filling the filesystem with
directories and small files (e.g. simultaneously unpacking more copies of
pkgsrc than will fit).
|
| 1.206 | 30-Apr-2006 |
perseant | Postpone the segment accounting changes coming from truncation until the
inode that makes those changes valid is either written to disk by
lfs_writeinode() or discarded by lfs_vfree().
A couple of locking fixes are also included as well.
|
| 1.205 | 18-Apr-2006 |
perseant | Don't roll forward if we aren't given a process context. Coverity CID 1076.
|
| 1.204 | 15-Apr-2006 |
christos | Coverity CID 2499: Fix uninitialize variable use.
|
| 1.203 | 10-Apr-2006 |
perseant | Remove mostly useless BUFPAGES warning message from lfs_{un,}mount.
|
| 1.202 | 10-Apr-2006 |
perseant | Optimize the free list search a little more; in particular use words
instead of bytes for the index, and never search below fs->lfs_freehd.
Fix a bug in the previous version of the search (an erroneous assumption
that ino_t was signed).
Free the bitmap when we unmount the filesystem.
|
| 1.201 | 10-Apr-2006 |
perseant | Correct a locking bug in the recent pager optimization.
|
| 1.200 | 08-Apr-2006 |
perseant | Implement a somewhat finer-grained mechanism for paging LFS-backed pages.
The writer daemon, if it does not need to flush the whole filesystem,
now only writes the vnodes for which the pagedaemon has requested pageouts
(although it does not pay attention to the page ranges the pagedaemon
supplies).
|
| 1.199 | 08-Apr-2006 |
perseant | Keep the free list ordered. This solves a problem first pointed out to me
by Michel Oey, in which an aged LFS writes up to an extra Ifile block for
every file created; and paves the way for the truncation of the Ifile when
many files are deleted.
|
| 1.198 | 31-Mar-2006 |
perseant | Handle the "filesystem is clean" flag correctly when upgrading from
read-only to read-write mount. This makes "root on lfs" work for me,
although it looks like a different traceback from PR#32667.
|
| 1.197 | 30-Mar-2006 |
yamt | some cleanups after the introduction of GOP_SIZE_MEM flag.
- remove GOP_SIZE_READ/GOP_SIZE_WRITE flags.
they have not been used since the change.
- ufs_balloc_range: remove code which has been no-op since the change.
thanks Konrad Schroder for explaining the original intention of the code.
- ffs_gop_size: don't extend past eof, in the case of GOP_SIZE_MEM.
otherwise genfs_getpages end up to allocate pages past eof unnecessarily.
|
| 1.196 | 28-Mar-2006 |
perseant | Double-checkpoint on unmount. This ensures that vnodes belonging to removed
files are really freed, preventing occasional spurious EBUSY returns from
vflush().
|
| 1.195 | 24-Mar-2006 |
perseant | Improvements to LFS's paging mechanism, to wit:
* Acknowledge that sometimes there are more dirty pages to be written to
disk than clean segments. When we reach the danger line,
lfs_gop_write() now returns EAGAIN. The caller of VOP_PUTPAGES(), if
it holds the segment lock, drops it and waits for the cleaner to make
room before continuing.
* Note and avoid a three-way deadlock in lfs_putpages (a writer holding
a page busy blocks on the cleaner while the cleaner blocks on the
segment lock while lfs_putpages blocks on the page).
|
| 1.194 | 17-Mar-2006 |
tls | From Konrad Schroeder, in response to strange df output on anoncvs.netbsd.org:
We were returning the wrong value for free space. Now we're not.
|
| 1.193 | 21-Feb-2006 |
thorpej | branches: 1.193.2; 1.193.4; 1.193.6;
Use device_class() instead of accessing dv_class directly.
|
| 1.192 | 14-Jan-2006 |
yamt | branches: 1.192.2; 1.192.4;
- unify ffs_blkatoff and lfs_blkatoff.
- remove ufs_ops::uo_blkatoff.
- add directory read-ahead code. (disabled for now.)
|
| 1.191 | 04-Jan-2006 |
yamt | - add simple functions to allocate/free a buffer for i/o.
- make bufpool static.
|
| 1.190 | 11-Dec-2005 |
christos | branches: 1.190.2;
merge ktrace-lwp.
|
| 1.189 | 02-Nov-2005 |
yamt | merge yamt-vop branch. remove following VOPs.
VOP_BLKATOFF
VOP_VALLOC
VOP_BALLOC
VOP_REALLOCBLKS
VOP_VFREE
VOP_TRUNCATE
VOP_UPDATE
|
| 1.188 | 27-Sep-2005 |
yamt | branches: 1.188.2;
introduce "ufs_ops" and use it for ITIMES.
|
| 1.187 | 23-Sep-2005 |
jmmv | Apply the NFS exports list rototill patch:
- Remove all NFS related stuff from file system specific code.
- Drop the vfs_checkexp hook and generalize it in the new nfs_check_export
function, thus removing redundancy from all file systems.
- Move all NFS export-related stuff from kern/vfs_subr.c to the new
file sys/nfs/nfs_export.c. The former was becoming large and its code
is always compiled, regardless of the build options. Using the latter,
the code is only compiled in when NFSSERVER is enabled. While doing this,
also make some functions in nfs_subs.c conditional to NFSSERVER.
- Add a new command in nfssvc(2), called NFSSVC_SETEXPORTSLIST, that takes a
path and a set of export entries. At the moment it can only clear the
exports list or append entries, one by one, but it is done in a way that
allows setting the whole set of entries atomically in the future (see the
comment in mountd_set_exports_list or in doc/TODO).
- Change mountd(8) to use the nfssvc(2) system call instead of mount(2) so
that it becomes file system agnostic. In fact, all this whole thing was
done to remove a 'XXX' block from this utility!
- Change the mount*, newfs and fsck* userland utilities to not deal with NFS
exports initialization; done internally by the kernel when initializing
the NFS support for each file system.
- Implement an interface for VFS (called VFS hooks) so that several kernel
subsystems can run arbitrary code upon receipt of specific VFS events.
At the moment, this only provides support for unmount and is used to
destroy NFS exports lists from the file systems being unmounted, though it
has room for extension.
Thanks go to yamt@, chs@, thorpej@, wrstuden@ and others for their comments
and advice in the development of this patch.
|
| 1.186 | 23-Aug-2005 |
christos | Don't overload MAXNAMLEN, use a separate constant for each filesystem type.
|
| 1.185 | 19-Aug-2005 |
christos | 64 bit inode changes.
|
| 1.184 | 23-Jul-2005 |
yamt | update file timestamps for nfsd loaned-read and mmap.
PR/25279. discussed on tech-kern@.
|
| 1.183 | 28-Jun-2005 |
yamt | branches: 1.183.2;
- constify genfs_ops.
- use member designators.
|
| 1.182 | 09-Jun-2005 |
atatat | Properly fix the constipated lossage wrt -Wcast-qual and the sysctl
code. I know it's not the prettiest code, but it seems to work rather
well in spite of itself.
|
| 1.181 | 29-May-2005 |
christos | - sprinkle const
- avoid shadow variables.
|
| 1.180 | 20-May-2005 |
perseant | Keep track of the number of segments reclaimed, since the cleaner doesn't
do this anymore (it hasn't for quite some time). Add a couple of conditional
debugging messages to indicate why segments are not cleaned, in the event
that lfs_segclean is used.
Make the LFCNSEGWAITALL fcntl work again.
|
| 1.179 | 20-May-2005 |
perseant | Fill in the lfs_fsmnt field in the superblock when we mount the filesystem,
so fsck(8) can tell where it was last mounted.
|
| 1.178 | 04-May-2005 |
perseant | Don't let the pager_map deadlock avoidance code in lfs_putpages() write
segments containing zero-block FINFO records. These records cause segments
to become uncleanable, which would eventually result in a "no clean segments"
panic.
|
| 1.177 | 23-Apr-2005 |
perseant | Provide a resize_lfs(8), including kernel and cleaner support. The current
implementation requires the fs to be mounted while resizing. Tested in both
directions, and everything appears to work happily, but ymmv.
|
| 1.176 | 19-Apr-2005 |
perseant | Keep per-inode, per-fs, and subsystem-wide counts of blocks allocated through
lfs_balloc(), and use that to estimate the number of dirty pages belonging
to LFS (subsystem or filesystem). This is almost certainly wrong for
the case of a large mmap()ed region, but the accounting is tighter than
what we had before, and performs much better in the typical case of pages
dirtied through write().
|
| 1.175 | 16-Apr-2005 |
perseant | Remove left-over reference to "lfs_blist", for _LKM case.
|
| 1.174 | 16-Apr-2005 |
perseant | Use splay trees, rather than a hash table, to manage the accounting of
blocks allocated through VOP_BALLOC() for pages to be written to disk.
This accounting no longer takes a noticeable fraction of the system CPU.
|
| 1.173 | 14-Apr-2005 |
perseant | Consolidate the hash table we use to maintain the integrity of lfs_avail
into a single, system-wide table, rather than having a separate hash table
per inode. Significantly reduces the "system" cpu usage of your average
file write.
|
| 1.172 | 14-Apr-2005 |
perseant | Keep track of the highest block held by an LFS inode, so that we can
be assured that the last byte of a file is always allocated. Previously
a file extension could cause the filesystem to be flushed, writing an
inconsistent inode to disk. Although this condition would be corrected
the next time blocks were written to disk, an intervening crash would leave
the filesystem in an inconsistent state, leaving fsck_lfs to complain
of an inode "partially truncated".
|
| 1.171 | 08-Apr-2005 |
perseant | Clean up the handling of the pager_map deadlock in lfs_putpages, after
realizing that it is safe to sleep the second time through the loop.
|
| 1.170 | 06-Apr-2005 |
perseant | Fix some locking issues that appeared with the simple_lock work.
Address a "pager_map" deadlock in lfs_putpages().
|
| 1.169 | 01-Apr-2005 |
perseant | Protect various per-fs structures with fs->lfs_interlock simple_lock, to
improve behavior in the multiprocessor case. Add debugging segment-lock
assertion statements.
|
| 1.168 | 29-Mar-2005 |
thorpej | - Define a VFS_ATTACH() macro that places a reference to a vfsops structure
into the "vfsops" link set.
- Use VFS_ATTACH() where vfsops are declared for individual file systems.
- In vfsinit(), traverse the "vfsops" link set, rather than vfs_list_initial[].
|
| 1.167 | 08-Mar-2005 |
simonb | branches: 1.167.2;
Tab Police.
|
| 1.166 | 08-Mar-2005 |
perseant | Straighten out the maze of ifdefs. Instead, consolidate all the debugging
stuff under '#ifdef DEBUG', and use sysctl knobs to turn on/off particular
parts of the debugging reporting (if DEBUG is enabled). Re-enable the LFS
statistics in sysctl, while I'm there. A bit of a rototill.
|
| 1.165 | 04-Mar-2005 |
perseant | Move "ifile is too large for your NBUFS/BUFPAGES" messages into a function.
Use log(9) to warn the user instead of printf(9). Since the theory is that
the Ifile is "always in cache", but the greater performance risk is
when the inode entries can't be held in cache, note these two cases
separately, at different log levels (notice and warning, respectively).
|
| 1.164 | 26-Feb-2005 |
perry | nuke trailing whitespace
|
| 1.163 | 26-Feb-2005 |
perseant | Various minor LFS improvements:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statvfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
|
| 1.162 | 11-Jan-2005 |
mycroft | branches: 1.162.2; 1.162.4;
Rearrange some code slightly to avoid uninitialized variable warnings.
|
| 1.161 | 09-Jan-2005 |
mycroft | Whoops -- move the location of the VOP_OPEN()/VOP_CLOSE(), et al, from
foo_mountfs() to foo_mount(), to match the new mountroot API.
Also, for ext2fs and lfs, copy some restructuring from ffs to allow changing
file system parameters without specifying the device name.
(ntfs could use some more work.)
|
| 1.160 | 09-Jan-2005 |
mycroft | Rework the mountroot interface so that vfs_mountroot() opens the root device
and just passes it on to the file system functions. This avoids opening and
closing the device several times.
Mentioned on tech-kern some time ago, IIRC. I've been running this for a
long time.
|
| 1.159 | 02-Jan-2005 |
thorpej | Add the system call and VFS infrastructure for file system extended
attributes.
From FreeBSD.
|
| 1.158 | 16-Aug-2004 |
mycroft | Make sure to set IMNT_DTYPE here...
|
| 1.157 | 15-Aug-2004 |
mycroft | Need to set um_dirblksiz here...
|
| 1.156 | 15-Aug-2004 |
mycroft | Fixing age old cruft:
* Rather than using mnt_maxsymlinklen to indicate that a file systems returns
d_type fields(!), add a new internal flag, IMNT_DTYPE.
Add 3 new elements to ufsmount:
* um_maxsymlinklen, replaces mnt_maxsymlinklen (which never should have existed
in the first place).
* um_dirblksiz, which tracks the current directory block size, eliminating the
FS-specific checks littered throughout the code. This may be used later to
make the block size variable.
* um_maxfilesize, which is the maximum file size, possibly adjusted lower due
to implementation issues.
Sync some bug fixes from FFS into ext2fs, particularly:
* ffs_lookup.c 1.21, 1.28, 1.33, 1.48
* ffs_inode.c 1.43, 1.44, 1.45, 1.66, 1.67
* ffs_vnops.c 1.84, 1.85, 1.86
Clean up some crappy pointer frobnication.
|
| 1.155 | 14-Aug-2004 |
mycroft | Add a new flag, IN_MODIFY. This is like IN_UPDATE|IN_CHANGE, but unlike
setting those flags, it does not cause the inode to be written in the periodic
sync. This is used for writes to special files (devices and named pipes) and
FIFOs.
Do not preemptively sync updates to access times and modification times. They
are now updated in the inode only opportunistically, or when the file or device
is closed. (Really, it should be delayed beyond close, but this is enough to
help substantially with device nodes.)
And the most amusing part:
Trickle sync was broken on both FFS and ext2fs, in different ways. In FFS, the
periodic call to VFS_SYNC(MNT_LAZY) was still causing all file data to be
synced. In ext2fs, it was causing the metadata to *not* be synced. We now
only call VOP_UPDATE() on the node if we're doing MNT_LAZY. I've confirmed
that we do in fact trickle correctly now.
|
| 1.154 | 05-Jul-2004 |
pk | Call inittodr() from main(). Let file system code set the recorded `last
update' time (if any) through the new function setrootfstime().
|
| 1.153 | 30-May-2004 |
yamt | lfs_gop_write: assert that ifile never come here.
|
| 1.152 | 25-May-2004 |
hannken | Add ffs internal snapshots. Written by Marshall Kirk McKusick for FreeBSD.
- Not enabled by default. Needs kernel option FFS_SNAPSHOT.
- Change parameters of ffs_blkfree.
- Let the copy-on-write functions return an error so spec_strategy
may fail if the copy-on-write fails.
- Change genfs_*lock*() to use vp->v_vnlock instead of &vp->v_lock.
- Add flag B_METAONLY to VOP_BALLOC to return indirect block buffer.
- Add a function ffs_checkfreefile needed for snapshot creation.
- Add special handling of snapshot files:
Snapshots may not be opened for writing and the attributes are read-only.
Use the mtime as the time this snapshot was taken.
Deny mtime updates for snapshot files.
- Add function transferlockers to transfer any waiting processes from
one lock to another.
- Add vfsop VFS_SNAPSHOT to take a snapshot and make it accessible through
a vnode.
- Add snapshot support to ls, fsck_ffs and dump.
Welcome to 2.0F.
Approved by: Jason R. Thorpe <thorpej@netbsd.org>
|
| 1.151 | 25-May-2004 |
atatat | Sysctl descriptions under vfs subtree
|
| 1.150 | 20-May-2004 |
atatat | Explicitly call pool_init() (and pool_destroy()) when being built as
an _LKM.
This adds pools to the list of things that lkms must do manually
because they're set up with link sets. Not that there's anything
wrong with link sets, but that we need to try harder to remember that
lkms are second class citizens. Of a sort.
|
| 1.149 | 25-Apr-2004 |
simonb | Initialise (most) pools from a link set instead of explicit calls
to pool_init. Untouched pools are ones that either in arch-specific
code, or aren't initialiased during initial system startup.
Convert struct session, ucred and lockf to pools.
|
| 1.148 | 22-Apr-2004 |
yamt | lfs_statvfs: report f_frsize correctly.
|
| 1.147 | 21-Apr-2004 |
christos | Replace the statfs() family of system calls with statvfs().
Retain binary compatibility.
|
| 1.146 | 27-Mar-2004 |
atatat | branches: 1.146.2;
Manually attach malloc types when being built as an lkm.
|
| 1.145 | 24-Mar-2004 |
atatat | Tango on sysctl_createv() and flags. The flags have all been renamed,
and sysctl_createv() now uses more arguments.
|
| 1.144 | 26-Feb-2004 |
oster | Add a missing:
pool_destroy(&lfs_dinode_pool);
to lfs_done().
Approved-by: yamt
|
| 1.143 | 28-Jan-2004 |
he | Let the cast to (long long) for using the result as a printf argument
apply to the whole expression, not just the first factor.
|
| 1.142 | 28-Jan-2004 |
yamt | use bufmem instead of bufpages to make lfs a little less broken.
|
| 1.141 | 04-Dec-2003 |
atatat | Dynamic sysctl.
Gone are the old kern_sysctl(), cpu_sysctl(), hw_sysctl(),
vfs_sysctl(), etc, routines, along with sysctl_int() et al. Now all
nodes are registered with the tree, and nodes can be added (or
removed) easily, and I/O to and from the tree is handled generically.
Since the nodes are registered with the tree, the mapping from name to
number (and back again) can now be discovered, instead of having to be
hard coded. Adding new nodes to the tree is likewise much simpler --
the new infrastructure handles almost all the work for simple types,
and just about anything else can be done with a small helper function.
All existing nodes are where they were before (numerically speaking),
so all existing consumers of sysctl information should notice no
difference.
PS - I'm sorry, but there's a distinct lack of documentation at the
moment. I'm working on sysctl(3/8/9) right now, and I promise to
watch out for buses.
|
| 1.140 | 07-Nov-2003 |
yamt | - tweak lfs_update_single()'s prototype so that it can be used by
roll-forward code.
- reduce code duplication using the above in update_meta()
this also fixes fragment accounting.
|
| 1.139 | 07-Nov-2003 |
yamt | fix spec vnode aliasing.
|
| 1.138 | 07-Nov-2003 |
yamt | - tell filesize changes to vm when roll-forwarding data blocks.
- handle fragment extension better during roll-forward.
- related assertions.
|
| 1.137 | 30-Oct-2003 |
simonb | Remove some assigned-to but otherwise unused variables.
|
| 1.136 | 25-Oct-2003 |
christos | Fix uninitialized variable warnings.
|
| 1.135 | 14-Oct-2003 |
dbj | add mnt_iflag field to struct mount for internal flags
mv MNT_GONE, MNT_UNMOUNT and MNT_WANTRDWR to this field
additonally add mnt_writeopcountupper and mnt_writeopcountlower fields
in preparation for pending write suspension support work
bump kernel version to 1.6ZD
|
| 1.134 | 14-Oct-2003 |
yamt | add a prototype of check_segsum().
|
| 1.133 | 14-Oct-2003 |
yamt | when roll-forwarding, check segment serial numbers correctly.
|
| 1.132 | 14-Oct-2003 |
yamt | add a missing fsbtodb() to read a correct block for roll-forwarding.
|
| 1.131 | 07-Sep-2003 |
yamt | comments on lfs_issequential_hole.
|
| 1.130 | 07-Aug-2003 |
agc | Move UCB-licensed code from 4-clause to 3-clause licence.
Patches provided by Joel Baker in PR 22364, verified by myself.
|
| 1.129 | 23-Jul-2003 |
yamt | add parenthesis missed in rev.1.127.
|
| 1.128 | 23-Jul-2003 |
yamt | whitespace
|
| 1.127 | 23-Jul-2003 |
yamt | add KASSERTs in lfs_issequential_hole.
|
| 1.126 | 12-Jul-2003 |
yamt | more MP locks.
|
| 1.125 | 12-Jul-2003 |
yamt | - protect global resource counts with lfs_subsys_lock.
- clean up scattered externs a little.
|
| 1.124 | 02-Jul-2003 |
yamt | use queue.h macros.
|
| 1.123 | 02-Jul-2003 |
yamt | use VFSTOUFS macro.
|
| 1.122 | 02-Jul-2003 |
yamt | - add a new functions, lfs_writer_enter/leave, and use them instead of
duplicated code fragments.
- add an assertion.
|
| 1.121 | 29-Jun-2003 |
fvdl | branches: 1.121.2;
Back out the lwp/ktrace changes. They contained a lot of colateral damage,
and need to be examined and discussed more.
|
| 1.120 | 29-Jun-2003 |
thorpej | Undo part of the ktrace/lwp changes. In particular:
* Remove the "lwp *" argument that was added to vget(). Turns out
that nothing actually used it!
* Remove the "lwp *" arguments that were added to VFS_ROOT(), VFS_VGET(),
and VFS_FHTOVP(); all they did was pass it to vget() (which, as noted
above, didn't use it).
* Remove all of the "lwp *" arguments to internal functions that were added
just to appease the above.
|
| 1.119 | 28-Jun-2003 |
bouyer | Adapt for struct proc* -> struct lwp* changes.
|
| 1.118 | 28-Jun-2003 |
darrenr | Pass lwp pointers throughtout the kernel, as required, so that the lwpid can
be inserted into ktrace records. The general change has been to replace
"struct proc *" with "struct lwp *" in various function prototypes, pass
the lwp through and use l_proc to get the process pointer when needed.
Bump the kernel rev up to 1.6V
|
| 1.117 | 18-May-2003 |
yamt | make is_sequential a callback in order to achieve better lfs write clustering.
since lfs always rewrite blocks into the new segment,
current on-disk place of the block doesn't affect to write clustering.
ok'ed by Konrad Schroder.
|
| 1.116 | 29-Apr-2003 |
perseant | Restrict the run of cluster blocks to on-disk contiguous blocks (back out
part of rev 1.115), to avoid writing over holes. This is the lesser of two
evils, to be replaced soon.
|
| 1.115 | 23-Apr-2003 |
perseant | Make LFS work better (though still not "well") as an NFS-exported
filesystem (and other things that needed to be fixed before the tests
would complete), to wit:
* Include the fs ident in the filehandle; improve stale filehandle checks.
* Change definition of blksize() to use the on-dinode size instead of
the inode's i_size, so that fsck_lfs will work properly again.
* Use b_interlock in lfs_vtruncbuf.
* Postpone dirop reclamation until after the seglock has been released,
so that lfs_truncate is not called with the segment lock held.
* Don't loop in lfs_fsync(), just write everything and wait.
* Be more careful about the interlock/uobjlock in lfs_putpages: when we
lose this lock, we have to resynchronize dirtiness of pages in each
block.
* Be sure to always write indirect blocks and update metadata in
lfs_putpages; fixes a bug that caused blocks to be accounted to the
wrong segment.
|
| 1.114 | 16-Apr-2003 |
christos | PR/1796: John Kohl: statfs misbehaves under chrooted environments.
- Under chroot it displays only the visible filesystems with appropriate paths.
- The statfs f_mntonname gets adjusted to contain the real path from root.
- While was there, fixed a bug in ext2fs, locking problems with vfs_getfsstat(),
and factored out some of the vfsop statfs() code to copy_statfs_info(). This
fixes the problem where some filesystems forgot to set fsid.
- Made coda look more like a normal fs.
|
| 1.113 | 02-Apr-2003 |
fvdl | Add support for UFS2. UFS2 is an enhanced FFS, adding support for
64 bit block pointers, extended attribute storage, and a few
other things.
This commit does not yet include the code to manipulate the extended
storage (for e.g. ACLs), this will be done later.
Originally written by Kirk McKusick and Network Associates Laboratories for
FreeBSD.
|
| 1.112 | 28-Mar-2003 |
perseant | Add a sleeper count, to prevent the cleaner from panicing the kernel
when the filesystem is unmounted, relocking the Ifile when its lock is
draining. (We can't use vfs_busy() since the process is sleeping for a
good long time.) Clean up / organize lfs.h, while I'm here.
In lfs_update_single, assert that disk addresses are either negative, or
are still positive when converted to int32_t, to prevent recurrence of a
negative/positive block problem.
|
| 1.111 | 21-Mar-2003 |
dsl | Use 'void *' instead of 'caddr_t' in prototypes of VOP_IOCTL, VOP_FCNTL
and VOP_ADVLOCK, delete casts from callers (and some to copyin/out).
|
| 1.110 | 21-Mar-2003 |
perseant | KNF (space after keywords).
|
| 1.109 | 21-Mar-2003 |
perseant | Use VONWORKLST as a heuristic for vnode emptiness, rather than exhaustively
checking the memq.
Take greater care not to dirty the Ifile vnode when unmounting the filesystem.
This should fix a "(vp->v_flag & VONWORKLST) == 0" assertion panic in vgonel
that could occur when unmounting.
Do not allow the Ifile to be mapped for writing.
|
| 1.108 | 21-Mar-2003 |
yamt | make this compilable with DIAGNOSTIC and without DEBUG.
fix PR 20827 from FUKAUMI Naoki.
|
| 1.107 | 20-Mar-2003 |
yamt | fix "more than one fragment" panics;
direct and indirect block pointers are not valid in the case of shortlinks.
while i'm here, move duplicated code in lfs_vget/fastvget into a new
function, lfs_vinit.
|
| 1.106 | 18-Mar-2003 |
perseant | Remember to destroy lfs_inoext_pool when closing up the LFS subsystem.
|
| 1.105 | 15-Mar-2003 |
perseant | Add simple_lock protection for lfs_seglock and lfs_subsys_pages; these will
be expanded to cover other per-fs and subsystem-wide data as well.
Fix a case of IN_MODIFIED being set without updating lfs_uinodes, resulting
in a "lfs_uinodes < 0" panic.
Fix a deadlock in lfs_putpages arising from the need to busy all pages in a
block; unbusy any that had already been busied before starting over.
|
| 1.104 | 08-Mar-2003 |
perseant | Take away "#ifdef LFS_UBC".
|
| 1.103 | 08-Mar-2003 |
perseant | Add an lfs_strategy() that checks to make sure we're not trying to read
where the cleaner is trying to write, instead of tying up the "live"
buffers (or pages).
Fix a bug in the LFS_UBC case where oversized buffers would not be
checksummed correctly, causing uncleanable segments.
Make sure that wakeup(fs->lfs_iocount) is done if fs->lfs_iocount is 1
as well as 0, since we wait in some places for it to drop to 1.
Activate all pages that make it into lfs_gop_write without the segment
lock held, since they must have been dirtied very recently, even if
PG_DELWRI is not set.
|
| 1.102 | 02-Mar-2003 |
perseant | Account SEGUSE_ACTIVE correctly so that the automatic segment cleaning
actually happens.
Add a new fcntl call that will write the minimum necessary to checkpoint
(i.e., for on-disk directory structure to be consistent, not including
updates to file data) so that the cleaner can clean segments more quickly
without sacrificing three-way commit for cleaning.
|
| 1.101 | 01-Mar-2003 |
yamt | use pid_t for pid.
|
| 1.100 | 01-Mar-2003 |
perseant | Be careful to always zero pages on truncation/fragment extension,
in the case where the filesystem block size is larger than PAGE_SIZE.
|
| 1.99 | 25-Feb-2003 |
thorpej | Add a new BUF_INIT() macro which initializes b_dep and b_interlock, and
use it. This fixes a few places where either b_dep or b_interlock were
not properly initialized.
|
| 1.98 | 25-Feb-2003 |
yamt | fix simplelocks
|
| 1.97 | 23-Feb-2003 |
perseant | Fix a buffer overflow bug in the LFS_UBC case that manifested itself
either as a mysterious UVM error or as "panic: dirty bufs". Verify
maximum size in lfs_malloc.
Teach lfs_updatemeta and lfs_shellsort about oversized cluster blocks from
lfs_gop_write.
When unwiring pages in lfs_gop_write, deactivate them, under the theory
that the pagedaemon wanted to free them last we knew.
|
| 1.96 | 20-Feb-2003 |
perseant | Tabify, and fix some comment alignment problems.
|
| 1.95 | 19-Feb-2003 |
yamt | workaround for "another flush is..." infinity loop in writerd.
if we're writerd, sleep in lfs_flush until another writer goes away
instead of busy loop in writed.
|
| 1.94 | 19-Feb-2003 |
yamt | wire the pages instead of just dequeue'ing them.
advised by Chuck Silvers.
|
| 1.93 | 19-Feb-2003 |
yamt | init b_interlock.
|
| 1.92 | 19-Feb-2003 |
yamt | init b_interlock.
|
| 1.91 | 17-Feb-2003 |
perseant | Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now
(there are still some details to work out) but expect that to go
away soon. To support these basic changes (creation of lfs_putpages,
lfs_gop_write, mods to lfs_balloc) several other changes were made, to
wit:
* Create a writer daemon kernel thread whose purpose is to handle page
writes for the pagedaemon, but which also takes over some of the
functions of lfs_check(). This thread is started the first time an
LFS is mounted.
* Add a "flags" parameter to GOP_SIZE. Current values are
GOP_SIZE_READ, meaning that the call should return the size of the
in-core version of the file, and GOP_SIZE_WRITE, meaning that it
should return the on-disk size. One of GOP_SIZE_READ or
GOP_SIZE_WRITE must be specified.
* Instead of using malloc(...M_WAITOK) for everything, reserve enough
resources to get by and use malloc(...M_NOWAIT), using the reserves if
necessary. Use the pool subsystem for structures small enough that
this is feasible. This also obsoletes LFS_THROTTLE.
And a few that are not strictly necessary:
* Moves the LFS inode extensions off onto a separately allocated
structure; getting closer to LFS as an LKM. "Welcome to 1.6O."
* Unified GOP_ALLOC between FFS and LFS.
* Update LFS copyright headers to correct values.
* Actually cast to unsigned in lfs_shellsort, like the comment says.
* Keep track of which segments were empty before the previous
checkpoint; any segments that pass two checkpoints both dirty and
empty can be summarily cleaned. Do this. Right now lfs_segclean
still works, but this should be turned into an effectless
compatibility syscall.
|
| 1.90 | 29-Jan-2003 |
yamt | don't use daddr_t for segment summary since it's an on-disk structure.
|
| 1.89 | 27-Jan-2003 |
yamt | make these compilable with lfs debug options.
(follow daddr_t change)
XXX maybe segment number should be 64bit.
|
| 1.88 | 25-Jan-2003 |
kleink | Fix further printf format warnings for DEBUG, in the wake of daddr_t
having changed.
|
| 1.87 | 25-Jan-2003 |
tron | Use PRId64 instead of hard coding "%lld" to fix build problems under
LP64 ports.
|
| 1.86 | 25-Jan-2003 |
tron | Fix printf() format strings problems caused by "daddr_t" change.
|
| 1.85 | 24-Jan-2003 |
fvdl | Bump daddr_t to 64 bits. Replace it with int32_t in all places where
it was used on-disk, so that on-disk formats remain the same.
Remove ufs_daddr_t and ufs_lbn_t for the time being.
|
| 1.84 | 12-Jan-2003 |
yamt | - zerofill struct lfs when allocating it.
- use M_ZERO instead of memset after malloc.
|
| 1.83 | 24-Nov-2002 |
yamt | lfs_sync should wait at lfs_writer, not lfs_dirops.
PR 18973.
|
| 1.82 | 27-Sep-2002 |
provos | remove trailing \n in panic(). approved perry.
|
| 1.81 | 21-Sep-2002 |
christos | MNT_GETARGS support
|
| 1.80 | 06-Sep-2002 |
gehenna | Merge the gehenna-devsw branch into the trunk.
This merge changes the device switch tables from static array to
dynamically generated by config(8).
- All device switches is defined as a constant structure in device drivers.
- The new grammer ``device-major'' is introduced to ``files''.
device-major <prefix> char <num> [block <num>] [<rules>]
- All device major numbers must be listed up in port dependent majors.<arch>
by using this grammer.
- Added the new naming convention.
The name of the device switch must be <prefix>_[bc]devsw for auto-generation
of device switch tables.
- The backward compatibility of loading block/character device
switch by LKM framework is broken. This is necessary to convert
from block/character device major to device name in runtime and vice versa.
- The restriction to assign device major by LKM is completely removed.
We don't need to reserve LKM entries for dynamic loading of device switch.
- In compile time, device major numbers list is packed into the kernel and
the LKM framework will refer it to assign device major number dynamically.
|
| 1.79 | 30-Jul-2002 |
soren | Die, qaddr_t, die! - mnt_data in struct mount is already effectively
a void *, so stop pretending otherwise.
|
| 1.78 | 06-Jul-2002 |
perseant | Deal with fragment size changes better. For each fragment that can
exist on an on-disk inode, we keep a record of its size in struct inode,
which is updated when we write the block to disk. The cleaner routines
thus have ready access to what size is the correct size for this block,
on disk.
Fixed a related bug: if a file with fragments is being cleaned
(fragments being cleaned) at the same time it is being extended beyond
NDADDR blocks, we could write a bogus FINFO record that has a frag in the
middle; when it was cleaned this would give back bogus file data. Don't
write the indirect blocks in this case, since there is no need.
lfs_fragextend and lfs_truncate no longer require the seglock, but instead
take a shared lock, which the seglock locks exclusively.
|
| 1.77 | 16-Jun-2002 |
perseant | For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.76 | 17-May-2002 |
perseant | branches: 1.76.2;
use macros from <sys/queue.h>
|
| 1.75 | 16-May-2002 |
thorpej | Fix LP64 printf format warning.
|
| 1.74 | 14-May-2002 |
perseant | branches: 1.74.2;
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes
I found while making sure there weren't any new ones.
* Make the write clusters keep track of the buffers whose blocks they contain.
This should make it possible to (1) write clusters using a page mapping
instead of malloc, if desired, and (2) schedule blocks for rewriting
(somewhere else) if a write error occurs. Code is present to use
pagemove() to construct the clusters but that is untested and will go away
anyway in favor of page mapping.
* DEBUG now keeps a log of Ifile writes, so that any lingering instances of
the "dirty bufs" problem can be properly debugged.
* Keep track of whether the Ifile has been dirtied by various routines that
can be called by lfs_segwrite, and loop on that until it is clean, for
a checkpoint. Checkpoints need to be squeaky clean.
* Warn the user (once) if the Ifile grows larger than is reasonable for their
buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can
grow.
* If an inode is not found in a disk block, try rereading the block, under
the assumption that the block was copied to a cluster and then freed.
* Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
|
| 1.73 | 12-May-2002 |
matt | Eliminate commons.
|
| 1.72 | 08-Mar-2002 |
thorpej | Pool deals fairly well with physical memory shortage, but it doesn't
deal with shortages of the VM maps where the backing pages are mapped
(usually kmem_map). Try to deal with this:
* Group all information about the backend allocator for a pool in a
separate structure. The pool references this structure, rather than
the individual fields.
* Change the pool_init() API accordingly, and adjust all callers.
* Link all pools using the same backend allocator on a list.
* The backend allocator is responsible for waiting for physical memory
to become available, but will still fail if it cannot callocate KVA
space for the pages. If this happens, carefully drain all pools using
the same backend allocator, so that some KVA space can be freed.
* Change pool_reclaim() to indicate if it actually succeeded in freeing
some pages, and use that information to make draining easier and more
efficient.
* Get rid of PR_URGENT. There was only one use of it, and it could be
dealt with by the caller.
From art@openbsd.org.
|
| 1.71 | 18-Dec-2001 |
chs | use the new compatibility routines to allow mmap() to work
(in the same non-coherent fashion that it worked pre-UBC)
until someone has time to do it the right way.
|
| 1.70 | 23-Nov-2001 |
chs | add spaces for KNF. confirmed to produce identical objects.
|
| 1.69 | 08-Nov-2001 |
lukem | add RCSID
|
| 1.68 | 15-Sep-2001 |
chs | branches: 1.68.2;
a whole bunch of changes to improve performance and robustness under load:
- remove special treatment of pager_map mappings in pmaps. this is
required now, since I've removed the globals that expose the address range.
pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's
no longer any need to special-case it.
- eliminate struct uvm_vnode by moving its fields into struct vnode.
- rewrite the pageout path. the pager is now responsible for handling the
high-level requests instead of only getting control after a bunch of work
has already been done on its behalf. this will allow us to UBCify LFS,
which needs tighter control over its pages than other filesystems do.
writing a page to disk no longer requires making it read-only, which
allows us to write wired pages without causing all kinds of havoc.
- use a new PG_PAGEOUT flag to indicate that a page should be freed
on behalf of the pagedaemon when it's unlocked. this flag is very similar
to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the
pageout fails due to eg. an indirect-block buffer being locked.
this allows us to remove the "version" field from struct vm_page,
and together with shrinking "loan_count" from 32 bits to 16,
struct vm_page is now 4 bytes smaller.
- no longer use PG_RELEASED for swap-backed pages. if the page is busy
because it's being paged out, we can't release the swap slot to be
reallocated until that write is complete, but unlike with vnodes we
don't keep a count of in-progress writes so there's no good way to
know when the write is done. instead, when we need to free a busy
swap-backed page, just sleep until we can get it busy ourselves.
- implement a fast-path for extending writes which allows us to avoid
zeroing new pages. this substantially reduces cpu usage.
- encapsulate the data used by the genfs code in a struct genfs_node,
which must be the first element of the filesystem-specific vnode data
for filesystems which use genfs_{get,put}pages().
- eliminate many of the UVM pagerops, since they aren't needed anymore
now that the pager "put" operation is a higher-level operation.
- enhance the genfs code to allow NFS to use the genfs_{get,put}pages
instead of a modified copy.
- clean up struct vnode by removing all the fields that used to be used by
the vfs_cluster.c code (which we don't use anymore with UBC).
- remove kmem_object and mb_object since they were useless.
instead of allocating pages to these objects, we now just allocate
pages with no object. such pages are mapped in the kernel until they
are freed, so we can use the mapping to find the page to free it.
this allows us to remove splvm() protection in several places.
The sum of all these changes improves write throughput on my
decstation 5000/200 to within 1% of the rate of NetBSD 1.5
and reduces the elapsed time for "make release" of a NetBSD 1.5
source tree on my 128MB pc to 10% less than a 1.5 kernel took.
|
| 1.67 | 15-Sep-2001 |
chs | add a new VFS op, vfs_reinit, which is called when desiredvnodes is
adjusted via sysctl. file systems that have hash tables which are
sized based on the value of this variable now resize those hash tables
using the new value. the max number of FFS softdeps is also recalculated.
convert various file systems to use the <sys/queue.h> macros for
their hash tables.
|
| 1.66 | 13-Jul-2001 |
perseant | branches: 1.66.2;
Merge the short-lived perseant-lfsv2 branch into the trunk.
Kernels and tools understand both v1 and v2 filesystems; newfs_lfs
generates v2 by default. Changes for the v2 layout include:
- Segments of non-PO2 size and arbitrary block offset, so these can be
matched to convenient physical characteristics of the partition (e.g.,
stripe or track size and offset).
- Address by fragment instead of by disk sector, paving the way for
non-512-byte-sector devices. In theory fragments can be as large
as you like, though in reality they must be smaller than MAXBSIZE in size.
- Use serial number and filesystem identifier to ensure that roll-forward
doesn't get old data and think it's new. Roll-forward is enabled for
v2 filesystems, though not for v1 filesystems by default.
- The inode free list is now a tailq, paving the way for undelete (undelete
is not yet implemented, but can be without further non-backwards-compatible
changes to disk structures).
- Inode atime information is kept in the Ifile, instead of on the inode;
that is, the inode is never written *just* because atime was changed.
Because of this the inodes remain near the file data on the disk, rather
than wandering all over as the disk is read repeatedly. This speeds up
repeated reads by a small but noticeable amount.
Other changes of note include:
- The ifile written by newfs_lfs can now be of arbitrary length, it is no
longer restricted to a single indirect block.
- Fixed an old bug where ctime was changed every time a vnode was created.
I need to look more closely to make sure that the times are only updated
during write(2) and friends, not after-the-fact during a segment write,
and certainly not by the cleaner.
|
| 1.65 | 30-May-2001 |
mrg | branches: 1.65.2; 1.65.4;
use _KERNEL_OPT
|
| 1.64 | 26-Jan-2001 |
itohy | branches: 1.64.2;
Call inittodr() from lfs_mountroot() so that the system time is set properly
when booted from LFS.
|
| 1.63 | 22-Jan-2001 |
jdolecek | make filesystem vnodeop, specop, fifoop and vnodeopv_* arrays const
|
| 1.62 | 03-Dec-2000 |
perseant | Call uvm_vmp_setsize() in lfs_{fast,}vget to set initial vnode size.
|
| 1.61 | 03-Dec-2000 |
chs | don't forget to set um_lognindir (now required by ufs_bmaparray()).
|
| 1.60 | 27-Nov-2000 |
perseant | If LFS_DO_ROLLFORWARD is defined, roll forward from the older checkpoint
on mount, through the newer checkpoint and on through any newer
partial-segments that may have been written but not checkpointed because
of an intervening crash.
LFS_DO_ROLLFORWARD is not defined by default.
|
| 1.59 | 14-Nov-2000 |
perseant | Initialize the cleaner information in the Ifile from the same info from
the superblock at fs mount time, enabling the previous patch to fsck_lfs.
Patch from Jesse Off <joff@gci-net.com> (Closes PR #11470).
|
| 1.58 | 09-Sep-2000 |
perseant | Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
|
| 1.57 | 05-Jul-2000 |
perseant | Clean up accounting of lfs_uinodes (dirty but unwritten inodes).
Make lfs_uinodes a signed quantity for debugging purposes, and set it to
zero as fs mount time.
Enclose setting/clearing of the dirty flags (IN_MODIFIED, IN_ACCESSED,
IN_CLEANING) in macros, and use those macros everywhere. Make
LFS_ITIMES use these macros; updated the ITIMES macro in inode.h to know
about this. Make ufs_getattr use ITIMES instead of FFS_ITIMES.
|
| 1.56 | 03-Jul-2000 |
perseant | Allow the number of free segments reserved for the cleaner to be
parametrized in the filesystem, defaulting to MIN_FREE_SEGS = 2 but set
to something more reasonable at newfs_lfs time.
Note the number of blocks that have been scheduled for writing but which
are not yet on disk in an inode extension, i_lfs_effnblks. Move
i_ffs_effnlink out of the ffs extension and onto the main inode, since
it's used all over the shared code and the lfs extension would clobber
it.
At inode write time, indirect blocks and inode-held blocks of inodes
that have i_lfs_effnblks != i_ffs_blocks are cleansed of UNWRITTEN disk
addresses, so that these never make it to disk.
|
| 1.55 | 30-Jun-2000 |
fvdl | Rearrange code around getnewvnode as was already done for ffs, to avoid
locking against oneself because getnewvnode recycles a softdep-using vnode.
|
| 1.54 | 28-Jun-2000 |
mrg | <vm/vm.h> -> <uvm/uvm_extern.h>
|
| 1.53 | 27-Jun-2000 |
perseant | Fixes associated with filling an LFS:
Change the space computation to appear to change the size of the *disk*
rather than the *bytes used* when more segment summaries and inode
blocks are written. Try to estimate the amount of space that these will
take up when more files are written, so the disk size doesn't change too
much.
Regularize error returns from lfs_valloc, lfs_balloc, lfs_truncate: they
now fail entirely, rather than succeeding half-way and leaving the fs in
an inconsistent state.
Rewrite lfs_truncate, mostly stealing from ffs_truncate. The old
lfs_truncate had difficulty truncating a large file to a non-zero size
(indirect blocks were not handled appropriately).
Unmark VDIROP on fvp after ufs_remove, ufs_rmdir, so these can be
reclaimed immediately: this vnode would not be written to disk again
anyway if the removal succeeded, and if it failed, no directory
operation occurred.
ufs_makeinode and ufs_mkdir now remove IN_ADIROP on error.
|
| 1.52 | 27-May-2000 |
perseant | branches: 1.52.4;
Prevent dirops from getting around lfs_check and wedging the buffer cache.
All the dirop vnops now mark the inodes with a new flag, IN_ADIROP, which
is removed as soon as the dirop is done (as opposed to VDIROP which stays
until the file is written). To address one issue raised in PR#9357.
|
| 1.51 | 19-May-2000 |
thorpej | NULL != 0
|
| 1.50 | 29-Apr-2000 |
perseant | Test whether the filesystem is an LFS before trying to read the alternate
superblock (whose disk address is stored in the primary superblock). Also,
refuse to mount a filesystem whose superblocks overlap or where the alt.
superblock has a lower disk address than the primary superblock.
Solves PR#10001.
|
| 1.49 | 23-Apr-2000 |
perseant | Fix problems outlined in PR#9926:
- lfs_truncate extends the file if called with length > i_ffs_size;
- lfs_truncate errors out if called with length < 0;
- lfs_balloc block accounting corrected for the case of blocks read
into the cache before they exist on disk;
- mp->mnt_stat.f_iosize is initialized in lfs_mountfs.
|
| 1.48 | 30-Mar-2000 |
augustss | Remove register declarations.
|
| 1.47 | 16-Mar-2000 |
jdolecek | Add new VFS op routine - vfs_done and call it on filesystem detach
in vfs_detach(). vfs_done may free global filesystem's resources,
typically those allocated in respective filesystem's init function.
Needed so those filesystems which went in via LKM have a chance to
clean after themselves before unloading. This fixes random panics
when LKM for filesystem using pools was loaded and unloaded several
times.
For each leaf filesystem, add appropriate vfs_done routine.
|
| 1.46 | 19-Jan-2000 |
perseant | Changes to stabilize LFS. The first two of these should also apply to the
1.4 branch.
* Use a separate per-fs lock, instead of ufs_hashlock, to protect the Inode
free list. This seems to prevent the "lockmgr: %d, not exclusive lock holder
%d, unlocking" message I was mis-attributing last night to an unlocked vnode
being passed to vrele.
* Change calling semantics of lfs_ifind, to give better error reporting:
If fed a struct buf, it can report the block number of the offending inode
block as well as the inode number.
* Back out rev 1.10 of lfs_subr.c, since the replacement code was slightly
uglier while being functionally identical.
* Make lfs_vunref use the same free list convention as vrele/vput, so that
vget does not remove vnodes from a hash list they are not on.
|
| 1.45 | 21-Nov-1999 |
perseant | Initialize i_ffs_effnlink, so every file doesn't look like it's already been
deleted for the purpose of dirops (particularly create and mkdir). Addresses
PR#8815.
|
| 1.44 | 15-Nov-1999 |
fvdl | Add Kirk McKusick's soft updates code to the trunk. Not enabled by
default, as the copyright on the main file (ffs_softdep.c) is such
that is has been put into gnusrc. options SOFTDEP will pull this
in. This code also contains the trickle syncer.
Bump version number to 1.4O
|
| 1.43 | 12-Nov-1999 |
perseant | Back out my patch of the 8th (to address unreferenced inode problem).
Apparently this needs more thought.
|
| 1.42 | 09-Nov-1999 |
perseant | If ifile blocks were written before dirops were complete, and then the
system crashed, inodes could be allocated that were not referenced. (Though
not a serious problem, it evidences itself in phase 4 of fsck_lfs.) Fix
this by marking if_daddr with UNASSIGNED before the inodes are actually
written; at mount time the ifile is checked for UNASSIGNED entries and
any that are found are linked back into the free list. (The latter
functionality should move into the roll-forward agent when it materializes.)
|
| 1.41 | 06-Nov-1999 |
perseant | branches: 1.41.2;
Address ufs_hashlock/ufs_ihashins protocol bug, discovered while doing a
post-mortem of a production machine. Also, take the active dirop
count off of the fs and make it global (since it is measuring a global
resource) and tie the threshold value LFS_MAXDIROP to desiredvnodes.
|
| 1.40 | 20-Oct-1999 |
enami | Check if the type of device node isn't VBAD before touching v_specinfo. If
the device vnode is revoked, the field is NULL and touching it causes null
pointer derefercence.
|
| 1.39 | 18-Oct-1999 |
wrstuden | branches: 1.39.2; 1.39.4;
Catch a few cases missed earlier where we need to lock the vnode before
calling VOP_CLOSE().
|
| 1.38 | 08-Sep-1999 |
augustss | branches: 1.38.2;
Add #include <sys/device.h> so this file compiles again.
|
| 1.37 | 08-Sep-1999 |
sommerfeld | Avoid dereferencing NULL rootvp if booting diskless.
|
| 1.36 | 03-Sep-1999 |
perseant | Make changes that will allow an LFS filesystem to be used as the root
filesystem. In particular,
- Fix mknod deadlock, described in PR 8172.
- Enable lfs_mountroot.
- Make lfs_writevnodes treat filesystems mounted on lfs device nodes properly,
by flushing that device rather than trying to add blocks to the device inode.
This, in combination with lfs boot blocks, will allow operation of an all-lfs
system.
|
| 1.35 | 17-Jul-1999 |
wrstuden | Adjust mountroot routines to vrele rootvp in case of mount error. Closes
PR 7977 by Neil Carson, <neil@brini.com>.
|
| 1.34 | 01-Jun-1999 |
perseant | Fixed lfs_update (and related functions) so that calls from lfs_fsync
will DTRT with vnodes marked VDIROP. In particular, the message
"flushing VDIROP" will no longer appear, and the filesystem will remain
stable in the event of a crash.
This was particularly a problem with NFS-exported LFSes, since fsync
was called on every file close.
|
| 1.33 | 04-May-1999 |
scottr | Include opt_ddb.h so we will get the Debugger() prototype.
|
| 1.32 | 12-Apr-1999 |
perseant | Check the superblock version field, and refuse to mount the filesystem
if the version number is higher than we know about. This allows, e.g.,
changes in the format of the ifile, segment size restrictions and boundaries,
etc., which would not affect existing fields in the superblock, but which
would drastically affect the filesystem, to be smoothly integrated at a
later date.
|
| 1.31 | 11-Apr-1999 |
perseant | Fix inode reporting in lfs_statfs (the meaning of f_files and f_ffree was
reversed).
|
| 1.30 | 11-Apr-1999 |
perseant | Mark the current segment with SEGUSE_ACTIVE at mount time, rather than waiting
for the first write. If this is not done, the cleaner may try to clean the
current segment out from under the writer if the filesystem is mounted after
a crash (or any other time that the dirty:clean segment ration is high enough).
|
| 1.29 | 04-Apr-1999 |
mycroft | Fix obvious bugs:
* The MNT_UPDATE case had a null pointer dereference. (This is a good example
of why blindly adding bogus initializiers is a FUNDAMENTALLY BAD IDEA!)
* Make sure the whole ufsmount is zeroed, as the export code relies on this.
* If we decided to use the second/alternate superblock, make sure to copy the
in-core version from the right buffer.
Also, reenable NFS exporting.
|
| 1.28 | 25-Mar-1999 |
perseant | branches: 1.28.2;
clean up unused/required #ifdefs
|
| 1.27 | 24-Mar-1999 |
tron | Don't include "opt_uvm.h" any more.
|
| 1.26 | 10-Mar-1999 |
perseant | New sources should leave the LFS in a more-or-less working state. Changes
include:
- DIROP segregation is enabled, and greater care is taken
to make sure that a checkpoint completes. Fsck is not
needed to remount the filesystem.
- Several checks to make sure that the LFS subsystem does not
overuse various resources (memory, in particular).
- The cleaner routines, lfs_markv in particular, are completely
rewritten. A buffer overflow is removed. Greater care is taken
to ensure that inodes come from where lfs_cleanerd say they come
from (so we know nothing has changed since lfs_bmapv was called).
- Fragment allocation is fixed, so that writes beyond end-of-file
do the right thing.
|
| 1.25 | 26-Feb-1999 |
wrstuden | Modify vfsops to seperate vfs_fhtovp() into two routines. vfs_fhtovp() now
only handles the file handle to vnode conversion, and a new call,
vfs_checkexp(), performs the export verification.
|
| 1.24 | 11-Sep-1998 |
pk | PR#6032: define fixed sized on-disk superblock structure.
|
| 1.23 | 01-Sep-1998 |
thorpej | Use the pool allocator and the "nointr" pool page allocator for LFS inodes.
|
| 1.22 | 24-Jun-1998 |
sommerfe | Always include fifos; "not an option any more".
|
| 1.21 | 22-Jun-1998 |
sommerfe | defopt for options FIFO
|
| 1.20 | 09-Jun-1998 |
scottr | Protect various config(8)-generated files from inclusion while
building LKMs. Fixes PR 5557.
|
| 1.19 | 08-Jun-1998 |
scottr | Use the newly-defined opt_quota.h.
|
| 1.18 | 18-Mar-1998 |
bouyer | Add support for reading/writing FFS in non-native byte order, conditioned
to "options FFS_EI". The superblock and inodes (without blk addr) are
byteswapped at disk read/write time, other metadatas are byteswapped
when used (as they are acceeded directly in the buffer cache).
This required the addition of a "um_flags" field to struct ufsmount.
ffs_bswap.c contains superblock and inode byteswap routines also used
by userland utilities.
|
| 1.17 | 01-Mar-1998 |
fvdl | Remove accidentally enabled lfs_mountroot from vfsops struct.
|
| 1.16 | 01-Mar-1998 |
fvdl | Merge with Lite2 + local changes
|
| 1.15 | 18-Feb-1998 |
thorpej | Place a pointer to an array of our vnodeopv_desc *'s in our vfsops
structure, for use by vfs_attach().
|
| 1.14 | 16-Oct-1997 |
mjacob | In calculating the f_bavail field, don't take 32 bit quantities and
multiply them by 90 (to be divided by 100) and expect them to be sane
for very large values (I was getting a negative 'avail' count).
|
| 1.13 | 11-Jun-1997 |
bouyer | Add support for ext2fs, this needed a few modifications to ufs/ufs/inode.h:
- added an "union inode_ext" to struct inode, for the per-fs extentions.
For now only ext2fs uses it.
- i_din is now an union:
union {
struct dinode ffs_din; /* 128 bytes of the on-disk dinode. */
struct ext2fs_dinode e2fs_din; /* 128 bytes of the on-disk dinode. */
} i_din
Added a lot of #define i_ffs_* and i_e2fs_* to access the fields.
- Added two macros: FFS_ITIMES and EXT2FS_ITIMES. ITIMES calls the rigth
macro, depending on the time of the inode. ITIMES is used where necessary,
FFS_ITIMES and EXT2FS_ITIMES in other places.
|
| 1.12 | 22-Dec-1996 |
cgd | Change the second and third args to struct vfsops' (*vfs_mount)() to
'const char *', and 'void *', respectively. The second arg is taken directly
from user arguments, and is const there, so must be const in the prototypes
and functions. The third arg is also taken directly from user arguments.
It doesn't have to be changed, but since it's cleaner to keep the type
the same as the user arg's type, and I'm already making the 'const char *'
change...
|
| 1.11 | 25-Mar-1996 |
pk | Appease gcc: unused variables if !QUOTA
|
| 1.10 | 09-Feb-1996 |
christos | lfs prototypes
|
| 1.9 | 18-Jun-1995 |
cgd | don't assume the f_fsnamelen is nul-truncated or longer than MFSNAMELEN
|
| 1.8 | 09-Mar-1995 |
mycroft | copy*str() should use size_t.
|
| 1.7 | 08-Mar-1995 |
cgd | size for copyinstr should be u_long
|
| 1.6 | 18-Jan-1995 |
mycroft | Clean up the code to frob mnt_stat a bit.
|
| 1.5 | 18-Jan-1995 |
mycroft | Turn mountlist into a CIRCLEQ, and handle setting and checking of MNT_ROOTFS
differently.
|
| 1.4 | 15-Dec-1994 |
mycroft | Call foo_statfs() from a common place when mounting.
|
| 1.3 | 14-Dec-1994 |
mycroft | Sync with CSRG.
|
| 1.2 | 29-Jun-1994 |
cgd | New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
|
| 1.1 | 08-Jun-1994 |
mycroft | branches: 1.1.1;
Update to 4.4-Lite fs code, with local changes.
|
| 1.1.1.2 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite2
|
| 1.1.1.1 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite for reference
|
| 1.28.2.9 | 05-May-2000 |
he | Pull up revision 1.50 (requested by perseant):
Sanity check the superblock before trying to use it to find the
alt superblock; sanity check the disk address of the alt superblock
to avoid deadlocking when trying to read it with the primary
superblock buffer still busy. Fixes PR#10001.
|
| 1.28.2.8 | 29-Mar-2000 |
he | Pull up revision 1.38 (requested by simonb):
Prevent lfs_mountroot() from attempting to use a network device
as root.
(This revision is needed on some NetBSD platforms.)
|
| 1.28.2.7 | 29-Mar-2000 |
he | Pull up revision 1.37 (requested by pk):
Prevent lfs_mountroot() from attempting to use a network device
as root.
|
| 1.28.2.6 | 20-Jan-2000 |
he | Pull up revision 1.46 (via patch, requested by perseant):
Files removed (through unlink, rmdir) are now really removed, though the
removal is postponed until the dirop is complete to ensure validity of
the filesystem through a crash. Use a separate per-fs lock, instead of
ufs_hashlock, to protect the inode free list. Change calling semantics
of lfs_ifind, to give better error reporting: If fed a struct buf, it
can report the block number of the offending inode block as well as the
inode number.
|
| 1.28.2.5 | 15-Jan-2000 |
he | Pull up revision 1.36 (requested by perseant):
Address problems related to using an LFS filesystem as the root
filesystem, including mknod hangs. Fixes PR#8172 and PR#9072.
|
| 1.28.2.4 | 17-Dec-1999 |
he | Pull up revision 1.41 (requested by perseant):
Address locking protocol error for inode hash, and make the
maximum number of active dirops a global quantity.
|
| 1.28.2.3 | 17-Dec-1999 |
he | Pull up revision 1.34 (via patch, requested by perseant):
Avoid flushing vnodes involved in a dirop, making lfs' promise
of "no fsck needed, even in the event of a crash" closer to
reality.
|
| 1.28.2.2 | 19-Oct-1999 |
he | Pull up revision 1.39 (requested by wrstuden):
Catch a few cases missed earlier where we need to lock the vnode before
calling VOP_CLOSE().
|
| 1.28.2.1 | 13-Apr-1999 |
perseant | branches: 1.28.2.1.2;
Pull-up of changes made to the trunk on Sunday [1.30->1.32], to wit:
Take out the `#ifdef USE_UFSHASH'; use ufs_hashlock to lock the inode free
list instead of free_lock.
Fix inode reporting in lfs_statfs (the meaning of f_files and f_ffree was
reversed).
Fix "lfs_ifind: dinode xxx not found" panic. When inodes were freed, then
immediately reloaded, their dinodes were located in an inode block which
was not on disk at the advertized location, nor in the cache (although it
would be flushed to disk next segment write). Fix this by using getblk()
instead of lfs_newbuf() for inode blocks.
Better checking for held inode locks in lfs_fastvget, for a number of
error conditions. Also change the default setting of lfs_clean_vnhead to
0, which seems to make the locking problems go away (although this is
difficult to test as I can't reliably reproduce them).
Make sure that the wakeup occurs for vnodes that lfs_update might be
sleeping on (nodes which are not marked IN_MODIFIED/IN_CLEANING, but which
have dirty buffers), by marking them with the appropriate flag if
dirtybuffers were added while the write was in progress.
Fix block counting during file truncation, if not truncating to zero.
Disallow threshold-initiated cache flush when dirops are active. Also,
make SET_ENDOP use lfs_check instead of inlining most of it.
Improve the debugging printfs in the cleaner syscalls (in particular, make
it obvious that they're coming from lfs).
Check the superblock version field, and refuse to mount the filesystem if
the version number is higher than we know about. This allows, e.g.,
changes in the format of the ifile, segment size restrictions and
boundaries, etc., which would not affect existing fields in the
superblock, but which would drastically affect the filesystem, to be
smoothly integrated at a later date.
|
| 1.28.2.1.2.3 | 31-Aug-1999 |
perseant | Rudimentary support for LFS under UBC:
- LFS-specific VOP_BALLOC and VOP_PUTPAGES vnode ops.
- getblk VREG panic #ifdef'd out (can be reinstated when Ifile is
internalized and Ifile can be made another type from VREG)
- interface to VOP_PUTPAGES changed to pass all pager flags, not
just sync. FS putpages routines must know about the pager flags.
- new LFS magic disk address, -2 ("unwritten"), meaning accounted for
but not assigned to a fixed disk location (since LFS does these two
things separately, and the previous accounting method using buffer
headers no longer will work). Changed references to (foo == (daddr_t)-1)
to (foo < 0). Since disk drivers reject all addresses < 0, this should
not present a problem for other FSs.
|
| 1.28.2.1.2.2 | 02-Aug-1999 |
thorpej | Update from trunk.
|
| 1.28.2.1.2.1 | 21-Jun-1999 |
thorpej | Sync w/ -current.
|
| 1.38.2.2 | 27-Dec-1999 |
wrstuden | Pull up to last week's -current.
|
| 1.38.2.1 | 21-Dec-1999 |
wrstuden | Initial commit of recent changes to make DEV_BSIZE go away.
Runs on i386, needs work on other arch's. Main kernel routines should be
fine, but a number of the stand programs need help.
cd, fd, ccd, wd, and sd have been updated. sd has been tested with non-512
byte block devices. vnd, raidframe, and lfs need work.
Non 2**n block support is automatic for LKM's and conditional for kernels
on "options NON_PO2_BLOCKS".
|
| 1.39.4.3 | 15-Nov-1999 |
fvdl | Sync with -current
|
| 1.39.4.2 | 03-Nov-1999 |
fvdl | Give ufs_ihashget an extra argument: the flags passed to vget() for
locking. This way we can avoid locking against ourselves when
ufs_ihashget is called during the flushing of metadata. XXX
Also, comment out a VOP_FSYNC call that I think is now unneeded, and
put a diagnostic printf there to check if this still happens.
|
| 1.39.4.1 | 19-Oct-1999 |
fvdl | Bring in Kirk McKusick's FFS softdep code on a branch.
|
| 1.39.2.5 | 11-Feb-2001 |
bouyer | Sync with HEAD.
|
| 1.39.2.4 | 08-Dec-2000 |
bouyer | Sync with HEAD.
|
| 1.39.2.3 | 22-Nov-2000 |
bouyer | Sync with HEAD.
|
| 1.39.2.2 | 20-Nov-2000 |
bouyer | Update thorpej_scsipi to -current as of a month ago
|
| 1.39.2.1 | 20-Oct-1999 |
thorpej | Sync w/ trunk.
|
| 1.41.2.2 | 06-Nov-1999 |
perseant | Address ufs_hashlock/ufs_ihashins protocol bug, discovered while doing a
post-mortem of a production machine. Also, take the active dirop
count off of the fs and make it global (since it is measuring a global
resource) and tie the threshold value LFS_MAXDIROP to desiredvnodes.
|
| 1.41.2.1 | 06-Nov-1999 |
perseant | file lfs_vfsops.c was added on branch comdex-fall-1999 on 1999-11-06 20:33:07 +0000
|
| 1.52.4.3 | 03-Feb-2001 |
he | Pull up revision 1.59 (requested by perseant):
o Initialize cleaner info from superblock, making fsck_lfs'
accounting of lfs_nclean work.
|
| 1.52.4.2 | 14-Sep-2000 |
perseant | Pull up recent LFS kernel changes (approved by thorpej):
ufs/ufs/inode.h, 1.20--1.22 (add i_lfs_effnblks extension ;
make ITIMES aware of LFS_ITIMES;
_LKM protection so userland progs
compile)
ufs/ufs/ufs_vnops.c, 1.69, 1.71 (remove IN_ADIROP;
use ITIMES instead of FFS_ITIMES)
ufs/ufs/ufs_readwrite.c, 1.27 (use lfs_reserve in lfs_write)
ufs/lfs/lfs.h, 1.26--1.32 (define LFS_EST_* macros ;
change MIN_FREE_SEGS to lfs_minfreesegs ;
add avail and bfree to CLEANERINFO ;
change lfs_uinodes to signed ;
change lfs_dmeta to signed ;
add whitespace to line up structure
members ;
explicit cast to int32_t in LFS_EST_*
macros)
ufs/lfs/lfs_alloc.c, back out 1.34.2.3 (pullups of 1.39, 1.40);
then pull up 1.38 (clean up on error)
1.39--1.43 (restore fvdl's ufs_hashlock fix ;
restore fvdl's ufs_hashlock fix ;
set i_lfs_effnblks ;
use UINO macros ;
add comments and fix long lines)
ufs/lfs/lfs_balloc.c, 1.19 (don't succeed halfway)
1.21--1.25 (use i_lfs_effnblks ;
fix i_lfs_effnblks computation and
quieten ;
fix i_ffs_blocks in unwritten fragment ;
remove useless debugging check ;
add comments and (c) 2000)
ufs/lfs/lfs_bio.c, 1.24--1.30 (cleanup and make lfs_flush_fs take
"struct lfs *" instead of "struct
mount *" ;
use lfs_minfreeseg instead of
MIN_FREE_SEGS ;
use UINO macros, and copy bfree/avail
to CLEANERINFO ;
add lfs_reserve function ;
1.28--1.30 fix printf formatting)
ufs/lfs/lfs_cksum.c, 1.13 (add (c) 2000)
ufs/lfs/lfs_debug.c, 1.11 (use btodb instead of DEV_BSIZE)
ufs/lfs/lfs_extern.h, 1.18, 1.20--1.21 (function prototype changes)
ufs/lfs/lfs_inode.c, 1.38 (rewrite lfs_truncate from
ffs_truncate)
1.40--1.44 (count written and unwritten blocks
seperately ;
use disk block units instead of bytes ;
remove unnecessary "mod" variable ;
correct B_DELWRI to avoid bawrite panic ;
use lfs_reserve)
ufs/lfs/lfs_segment.c, 1.52-1.59 (use lfs_dmeta to note used summaries ;
check for UNWRITTEN in indirect blocks ;
more debugging stuff inside #ifdef
DEBUG_LFS ;
use LK_CANRECURSE ;
don't drop dirty indirect blocks ;
use UINO macros ;
don't hose the free list ;
use btodb() instead of DEV_BSIZE ;
make it compile again (oops))
ufs/lfs/lfs_subr.c, 1.16--1.17 (check for locked inodes before
changing ;
use btodb() instead of DEV_BSIZE, (c)
2000)
ufs/lfs/lfs_syscalls.c, back out 1.41.4.2 (fvdl's ufs_hashlock fix);
then pull up 1.43 (use lfs_dmeta)
1.44--1.45 (restore fvdl's ufs_hashlock fix)
1.46--1.47 (fix lfs_avail leakage from sblock
segments ;
use UINO macros)
1.49 (bounds-check inode numbers in
lfs_markv)
ufs/lfs/lfs_vfsops.c, 1.53 (use LFS_EST_* macros in lfs_statfs)
1.56--1.58 (initialize lfs_minfreeseg, lfs_effnblk ;
initialize lfs_uinodes ;
initialize lfs_ravail)
ufs/lfs/lfs_vnops.c, 1.40 (remove VDIROP from removed files)
1.42--1.44 (move SET_ENDOP below the removal of
VDIROP ;
use UINO macros and add lfs_itimes
function ;
use lfs_reserve in dirops)
|
| 1.52.4.1 | 03-Jul-2000 |
fvdl | pullup the fixes from the trunk to not hold ufs_hashlock across
getnewvnode()
|
| 1.64.2.15 | 15-Jan-2003 |
thorpej | Sync with HEAD.
|
| 1.64.2.14 | 11-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.64.2.13 | 18-Oct-2002 |
nathanw | Catch up to -current.
|
| 1.64.2.12 | 17-Sep-2002 |
nathanw | Catch up to -current.
|
| 1.64.2.11 | 01-Aug-2002 |
nathanw | Catch up to -current.
|
| 1.64.2.10 | 15-Jul-2002 |
nathanw | Whitespace.
|
| 1.64.2.9 | 24-Jun-2002 |
nathanw | Curproc->curlwp renaming.
Change uses of "curproc->l_proc" back to "curproc", which is more like the
original use. Bare uses of "curproc" are now "curlwp".
"curproc" is now #defined in proc.h as ((curlwp) ? (curlwp)->l_proc) : NULL)
so that it is always safe to reference curproc (*de*referencing curproc
is another story, but that's always been true).
|
| 1.64.2.8 | 20-Jun-2002 |
nathanw | Catch up to -current.
|
| 1.64.2.7 | 01-Apr-2002 |
nathanw | Catch up to -current.
(CVS: It's not just a program. It's an adventure!)
|
| 1.64.2.6 | 08-Jan-2002 |
nathanw | Catch up to -current.
|
| 1.64.2.5 | 14-Nov-2001 |
nathanw | Catch up to -current.
|
| 1.64.2.4 | 21-Sep-2001 |
nathanw | Catch up to -current.
|
| 1.64.2.3 | 24-Aug-2001 |
nathanw | Catch up with -current.
|
| 1.64.2.2 | 21-Jun-2001 |
nathanw | Catch up to -current.
|
| 1.64.2.1 | 05-Mar-2001 |
nathanw | Initial commit of scheduler activations and lightweight process support.
|
| 1.65.4.6 | 10-Oct-2002 |
jdolecek | sync kqueue with -current; this includes merge of gehenna-devsw branch,
merge of i386 MP branch, and part of autoconf rototil work
|
| 1.65.4.5 | 06-Sep-2002 |
jdolecek | sync kqueue branch with HEAD
|
| 1.65.4.4 | 23-Jun-2002 |
jdolecek | catch up with -current on kqueue branch
|
| 1.65.4.3 | 16-Mar-2002 |
jdolecek | Catch up with -current.
|
| 1.65.4.2 | 10-Jan-2002 |
thorpej | Sync kqueue branch with -current.
|
| 1.65.4.1 | 03-Aug-2001 |
lukem | update to -current
|
| 1.65.2.4 | 02-Jul-2001 |
perseant | Change disk addressing unit to be the fragment, instead of the disk sector.
All quantities in the superblock, inodes, indirect blocks, etc. refer now
to this abstract unit (called "fsb" as it is in FFS) instead of disk sectors;
as a consequence segment summary blocks have to be multiples of a fragment in
size. In v1 filesystems, compatibility code ensures that 1 fsb == 1 sector,
regardless of fragment size.
Fragments can now range in size between 512 and 32k; in the event that
LFS_LABELPAD (8k) is smaller than the disk address unit size, an extra
proto-superblock is kept at 8k from the beginning of the disk, to be used
*only* to locate the real superblocks. (Not all of the userland knows about
this yet.)
Almost all of this was done not by me, but by joff.
|
| 1.65.2.3 | 29-Jun-2001 |
perseant | fix comment in light of roll_id
|
| 1.65.2.2 | 29-Jun-2001 |
perseant | Get rid of __P(), protoizing where it had not already been done
|
| 1.65.2.1 | 27-Jun-2001 |
perseant | Import of what I've been calling "LFSv2", that is, LFS with some features
added that require changes to the on-disk data structures. These include:
- 64-bit time in everything but inodes
- User-specified segment offset, and segment size no longer
restricted to PO2.
- Serial number on segment summaries in addition to timestamp, and
a new volume identifier, to make roll-forward feasible without
fear of finding old data and thinking it was new.
Although I think this version works at least as well as what's on the trunk,
we're not done yet; hence this commit is going in on a branch and not on
the trunk. Enhancements that are not here yet include fragment addressing,
like FFS does, instead of block addressing.
|
| 1.66.2.3 | 01-Oct-2001 |
fvdl | Catch up with -current.
|
| 1.66.2.2 | 26-Sep-2001 |
fvdl | * add a VCLONED vnode flag that indicates a vnode representing a cloned
device.
* rename REVOKEALL to REVOKEALIAS, and add a REVOKECLONE flag, to pass
to VOP_REVOKE
* the revoke system call will revoke all aliases, as before, but not the
clones
* vdevgone is called when detaching a device, so make it use REVOKECLONE
to get rid of all clones as well
* clean up all uses of VOP_OPEN wrt. locking.
* add a few VOPS to spec_vnops that need to do something when it's a
clone vnode (access and getattr)
* add a copy of the vnode vattr structure of the original 'master' vnode
to the specinfo of a cloned vnode. could possibly redirect getattr to
the 'master' vnode, but this has issues with revoke
* add a vdev_reassignvp function that disassociates a vnode from its
original device, and reassociates it with the specified dev_t. to be
used by cloning devices only, in case a new minor is allocated.
* change all direct references in drivers to v_devcookie and v_rdev
to vdev_privdata(vp) and vdev_rdev(vp). for diagnostic purposes
when debugging race conditions that still exist wrt. locking and
revoking vnodes.
* make the locking state of a vnode consistent when passed to
d_open and d_close (unlocked). locked would be better, but has
some deadlock issues
|
| 1.66.2.1 | 18-Sep-2001 |
fvdl | Various changes to make cloning devices possible:
* Add an extra argument (struct vnode **) to VOP_OPEN. If it is
not NULL, specfs will create a cloned (aliased) vnode during
the call, and return it there. The caller should release and
unlock the original vnode if a new vnode was returned. The
new vnode is returned locked.
* Add a flag field to the cdevsw and bdevsw structures.
DF_CLONING indicates that it wants a new vnode for each
open (XXX is there a better way? devprop?)
* If a device is cloning, always call the close entry
point for a VOP_CLOSE.
Also, rewrite cons.c to do the right thing with vnodes. Use VOPs
rather then direct device entry calls. Suggested by mycroft@
Light to moderate testing done an i386 system (arch doesn't matter
though, these are MI changes).
|
| 1.68.2.1 | 12-Nov-2001 |
thorpej | Sync the thorpej-mips-cache branch with -current.
|
| 1.74.2.5 | 29-Aug-2002 |
gehenna | catch up with -current.
|
| 1.74.2.4 | 15-Jul-2002 |
gehenna | catch up with -current.
|
| 1.74.2.3 | 20-Jun-2002 |
gehenna | catch up with -current.
|
| 1.74.2.2 | 30-May-2002 |
gehenna | Catch up with -current.
|
| 1.74.2.1 | 16-May-2002 |
gehenna | Use devsw APIs for checking validity of major numbers.
|
| 1.76.2.1 | 20-Jun-2002 |
lukem | Pull up revision 1.77 (requested by perseant in ticket #325):
For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.121.2.13 | 10-Nov-2005 |
skrll | Sync with HEAD. Here we go again...
|
| 1.121.2.12 | 01-Apr-2005 |
skrll | Sync with HEAD.
|
| 1.121.2.11 | 08-Mar-2005 |
skrll | Sync with HEAD.
|
| 1.121.2.10 | 04-Mar-2005 |
skrll | Sync with HEAD.
Hi Perry!
|
| 1.121.2.9 | 17-Jan-2005 |
skrll | Sync with HEAD.
|
| 1.121.2.8 | 27-Oct-2004 |
skrll | Remove the struct lwp * arguments from qsync and ufs_checkpath that are
no longer (read: were never) required.
|
| 1.121.2.7 | 21-Sep-2004 |
skrll | Fix the sync with head I botched.
|
| 1.121.2.6 | 18-Sep-2004 |
skrll | Sync with HEAD.
|
| 1.121.2.5 | 25-Aug-2004 |
skrll | Sync with HEAD.
|
| 1.121.2.4 | 24-Aug-2004 |
skrll | Undo part of the ktrace/lwp changes. In particular:
* Remove the "lwp *" argument that was added to vget(). Turns out
that nothing actually used it!
* Remove the "lwp *" arguments that were added to VFS_ROOT(), VFS_VGET(),
and VFS_FHTOVP(); all they did was pass it to vget() (which, as noted
above, didn't use it).
* Remove all of the "lwp *" arguments to internal functions that were added
just to appease the above.
|
| 1.121.2.3 | 03-Aug-2004 |
skrll | Sync with HEAD
|
| 1.121.2.2 | 19-Aug-2003 |
skrll | LWPify
|
| 1.121.2.1 | 02-Jul-2003 |
darrenr | Apply the aborted ktrace-lwp changes to a specific branch. This is just for
others to review, I'm concerned that patch fuziness may have resulted in some
errant code being generated but I'll look at that later by comparing the diff
from the base to the branch with the file I attempt to apply to it. This will,
at the very least, put the changes in a better context for others to review
them and attempt to tinker with removing passing of 'struct lwp' through
the kernel.
|
| 1.146.2.1 | 29-May-2004 |
tron | branches: 1.146.2.1.2;
Pull up revision 1.151 (requested by atatat in ticket #393):
Sysctl descriptions under vfs subtree
|
| 1.146.2.1.2.1 | 10-May-2005 |
riz | Pull up the following revisions (requested by perseant in ticket #1281):
1.8 sys/ufs/lfs/TODO
1.75 sys/ufs/lfs/lfs.h (via patch)
1.74 sys/ufs/lfs/lfs_alloc.c (via patch)
1.49, 1.51 sys/ufs/lfs/lfs_balloc.c (1.51 via patch)
1.78 sys/ufs/lfs/lfs_bio.c
1.62 sys/ufs/lfs/lfs_extern.h (via patch)
1.156 sys/ufs/lfs/lfs_segment.c (via patch)
1.48 sys/ufs/lfs/lfs_subr.c
1.101 sys/ufs/lfs/lfs_syscalls.c
1.163 sys/ufs/lfs/lfs_vfsops.c (via patch)
1.134 sys/ufs/lfs/lfs_vnops.c (via patch)
1.61 sys/ufs/ufs/ufs_readwrite.c (via patch)
1.20 libexec/lfs_cleanerd/clean.h (via patch)
1.52 libexec/lfs_cleanerd/cleanerd.c (via patch)
1.41 libexec/lfs_cleanerd/library.c (via patch)
1.4 regress/sys/fs/lfs/newfs_fsck/Makefile
1.2 regress/sys/fs/lfs/newfs_fsck/mkfs_mount
1.2 regress/sys/fs/lfs/newfs_fsck/smallfiles
1.3 sbin/fsck_lfs/bufcache.c
1.3 sbin/fsck_lfs/bufcache.h
1.3 sbin/fsck_lfs/lfs.h
1.8 sbin/fsck_lfs/lfs.c (via patch)
1.8 sbin/fsck_lfs/pass3.c (via patch)
1.18 sbin/fsck_lfs/pass0.c (via patch)
1.18 sbin/fsck_lfs/utilities.c (via patch)
1.7 sbin/fsck_lfs/segwrite.c
1.19 sbin/fsck_lfs/setup.c (via patch)
1.3 sbin/newfs_lfs/Makefile
0 sbin/newfs_lfs/lfs.c (yes, remove it)
1.1 sbin/newfs_lfs/make_lfs.c
1.15 sbin/newfs_lfs/newfs.c (via patch)
Various minor LFS improvements.
Kernel:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this. Should fix PR #29045.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
Fixes PR #26680.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
cleaner:
* Adapt lfs_cleanerd to use the fcntl call to get the Ifile filehandle,
so it need not be in the namespace.
* Make lfs_cleanerd be more careful when there are very few available
segments.
* Make lfs_cleanerd less verbose when the filesystem is unmounted.
newfs_lfs, fsck_lfs, and regression:
* Extend the lfs library from fsck_lfs(8) so that it can be used with a
not-yet-existent LFS. Make newfs_lfs(8) use this library, so it can
create LFSs whose Ifile is larger than one segment. Addresses PR #11110.
* Make newfs_lfs(8) use strsuftoi64() for its arguments, a la newfs(8).
* Make fsck_lfs(8) respect the "file system is clean" flag.
* Don't let fsck_lfs(8) think it has dirty blocks when invoked with the
-n flag.
* Remove the Ifile from the filesystem namespace. The cleaner now uses
a fcntl call on the root inode to find the Ifile filehandle. (As a
side-effect, addresses PR #29144.)
|
| 1.162.4.1 | 19-Mar-2005 |
yamt | sync with head. xen and whitespace. xen part is not finished.
|
| 1.162.2.1 | 29-Apr-2005 |
kent | sync with -current
|
| 1.167.2.22 | 10-Aug-2006 |
tron | Apply patch (requested by fair in perseant #1457):
Bring LFS up to current, including a patch (1.95 lfs_alloc.c) that
should prevent the inode free list errors seen on the STABLE branch
subsequent to pullup ticket #1327.
|
| 1.167.2.21 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.93
sys/ufs/lfs/lfs.h: revision 1.106
sys/ufs/lfs/lfs_vfsops.c: revision 1.209
sys/ufs/lfs/lfs_vnops.c: revision 1.175
sys/ufs/lfs/lfs_segment.c: revision 1.178
Fixes to address the "vinvalbuf: dirty blocks" panic that can occur when
many inodes are cleaned at once. Make sure that we write all the pages
on vnodes that are being flushed, even if we don't think there's room;
drain v_numoutput before lfs_vflush() completes.
Also, don't allow a vnode that is in the process of being cleaned to be
chosen by getnewvnode(); this avoids a segment accounting panic in the case
that a large number of inodes are fed to lfs_markv() all at once.
|
| 1.167.2.20 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.92
sys/ufs/lfs/lfs.h: revision 1.105
sys/ufs/lfs/lfs_vfsops.c: revision 1.207
sys/ufs/lfs/lfs_subr.c: revision 1.59
sys/ufs/lfs/lfs_vnops.c: revision 1.173
sys/ufs/lfs/lfs_bio.c: revision 1.92
Introduce another per-filesystem parameter, lfs_resvseg, to separate the
notion of "how many segments are reserved for the cleaner" from that of
"how many segments are not counted in lfs_bfree". The default value
used for existing filesystems is the same as the previous implicit value
of (lfs_minfreeseg / 2 + 1), modulo some sanity checking.
Count pending dirops on a per-filesystem basis, since once we start
writing them we can't stop until we're done. This seems to help stave off
the "no clean segments" panic in the case of filling the filesystem with
directories and small files (e.g. simultaneously unpacking more copies of
pkgsrc than will fit).
|
| 1.167.2.19 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.104
sys/ufs/lfs/lfs_vfsops.c: revision 1.206
sys/ufs/lfs/lfs_vnops.c: revision 1.170
sys/ufs/lfs/lfs_extern.h: revision 1.80
sys/ufs/lfs/lfs_segment.c: revision 1.176
sys/ufs/lfs/lfs_inode.c: revision 1.103 via patch
sys/ufs/lfs/lfs_alloc.c: revision 1.90
Postpone the segment accounting changes coming from truncation until the
inode that makes those changes valid is either written to disk by
lfs_writeinode() or discarded by lfs_vfree().
A couple of locking fixes are also included as well.
|
| 1.167.2.18 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.205
Don't roll forward if we aren't given a process context. Coverity CID 1076.
|
| 1.167.2.17 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.204 via patch
Coverity CID 2499: Fix uninitialize variable use.
|
| 1.167.2.16 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.203 via patch
Remove mostly useless BUFPAGES warning message from lfs_{un,}mount.
|
| 1.167.2.15 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.101
sys/ufs/lfs/lfs_vfsops.c: revision 1.202
sys/ufs/lfs/lfs_alloc.c: revision 1.88
Optimize the free list search a little more; in particular use words
instead of bytes for the index, and never search below fs->lfs_freehd.
Fix a bug in the previous version of the search (an erroneous assumption
that ino_t was signed).
Free the bitmap when we unmount the filesystem.
|
| 1.167.2.14 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.201
Correct a locking bug in the recent pager optimization.
|
| 1.167.2.13 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.200
sys/ufs/lfs/lfs_vnops.c: revision 1.164
sys/ufs/lfs/lfs_inode.c: revision 1.101
sys/ufs/lfs/lfs_extern.h: revision 1.78
sys/ufs/lfs/lfs.h: revision 1.100
Implement a somewhat finer-grained mechanism for paging LFS-backed pages.
The writer daemon, if it does not need to flush the whole filesystem,
now only writes the vnodes for which the pagedaemon has requested pageouts
(although it does not pay attention to the page ranges the pagedaemon
supplies).
|
| 1.167.2.12 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.87
sys/ufs/lfs/lfs.h: revision 1.99
sys/ufs/lfs/lfs_vfsops.c: revision 1.199
sys/ufs/lfs/lfs_extern.h: revision 1.77 via patch
Keep the free list ordered. This solves a problem first pointed out to me
by Michel Oey, in which an aged LFS writes up to an extra Ifile block for
every file created; and paves the way for the truncation of the Ifile when
many files are deleted.
|
| 1.167.2.11 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.198
sys/ufs/lfs/lfs_vnops.c: revision 1.161
Handle the "filesystem is clean" flag correctly when upgrading from
read-only to read-write mount. This makes "root on lfs" work for me,
although it looks like a different traceback from PR#32667.
|
| 1.167.2.10 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.196
Double-checkpoint on unmount. This ensures that vnodes belonging to removed
files are really freed, preventing occasional spurious EBUSY returns from
vflush().
|
| 1.167.2.9 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.158
sys/ufs/lfs/lfs_subr.c: revision 1.57
sys/ufs/lfs/lfs_segment.c: revision 1.171
sys/ufs/lfs/lfs.h: revision 1.97
sys/ufs/lfs/lfs_vfsops.c: revision 1.195
sys/ufs/lfs/lfs_extern.h: revision 1.76
Improvements to LFS's paging mechanism, to wit:
* Acknowledge that sometimes there are more dirty pages to be written to
disk than clean segments. When we reach the danger line,
lfs_gop_write() now returns EAGAIN. The caller of VOP_PUTPAGES(), if
it holds the segment lock, drops it and waits for the cleaner to make
room before continuing.
* Note and avoid a three-way deadlock in lfs_putpages (a writer holding
a page busy blocks on the cleaner while the cleaner blocks on the
segment lock while lfs_putpages blocks on the page).
|
| 1.167.2.8 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_segment.c: revision 1.170
sys/ufs/lfs/lfs.h: revision 1.96
sys/ufs/lfs/lfs_vfsops.c: revision 1.194
sys/ufs/lfs/lfs_syscalls.c: revision 1.109
From Konrad Schroeder, in response to strange df output on anoncvs.netbsd.org:
We were returning the wrong value for free space. Now we're not.
|
| 1.167.2.7 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.153
sys/ufs/lfs/lfs_debug.c: revision 1.32
sys/ufs/lfs/lfs_alloc.c: revision 1.84
sys/ufs/lfs/lfs_vfsops.c: revision 1.185
sys/ufs/lfs/lfs_segment.c: revision 1.165
64 bit inode changes.
|
| 1.167.2.6 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.152
sys/ufs/lfs/lfs_debug.c: revision 1.31
sys/ufs/lfs/lfs_subr.c: revision 1.53
sys/ufs/lfs/lfs_extern.h: revision 1.68
sys/ufs/lfs/lfs_inode.c: revision 1.96
sys/ufs/lfs/lfs_bio.c: revision 1.86
sys/ufs/lfs/lfs_alloc.c: revision 1.83
sys/ufs/lfs/lfs_vfsops.c: revision 1.181
sys/ufs/lfs/lfs.h: revision 1.88
sys/ufs/lfs/lfs_segment.c: revision 1.164
- sprinkle const
- avoid shadow variables.
|
| 1.167.2.5 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.180
sys/ufs/lfs/lfs_syscalls.c: revision 1.106
sys/ufs/lfs/lfs.h: revision 1.87
Keep track of the number of segments reclaimed, since the cleaner doesn't
do this anymore (it hasn't for quite some time). Add a couple of conditional
debugging messages to indicate why segments are not cleaned, in the event
that lfs_segclean is used.
Make the LFCNSEGWAITALL fcntl work again.
|
| 1.167.2.4 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.179
Fill in the lfs_fsmnt field in the superblock when we mount the filesystem,
so fsck(8) can tell where it was last mounted.
|
| 1.167.2.3 | 24-Aug-2005 |
riz | Pull up following revision(s) (requested by yamt in ticket #688):
sys/miscfs/genfs/genfs_vnops.c: revision 1.98 via patch
sys/ufs/ffs/ffs_vfsops.c: revision 1.165
sys/ufs/lfs/lfs_extern.h: revision 1.69
sys/fs/filecorefs/filecore_vfsops.c: revision 1.20
sys/nfs/nfs_node.c: revision 1.80
sys/fs/smbfs/smbfs_node.c: revision 1.24
sys/fs/cd9660/cd9660_vfsops.c: revision 1.24
sys/fs/msdosfs/msdosfs_denode.c: revision 1.8
sys/miscfs/genfs/genfs_node.h: revision 1.6
sys/ufs/lfs/lfs_vfsops.c: revision 1.183
sys/ufs/ext2fs/ext2fs_vfsops.c: revision 1.86
sys/fs/adosfs/advfsops.c: revision 1.23
sys/fs/ntfs/ntfs_vfsops.c: revision 1.31
- constify genfs_ops.
- use member designators.
sys/miscfs/genfs/genfs_vnops.c: revision 1.99 via patch
genfs_getpages: don't forget to put the vnode onto the syncer's work que
ue
even in the case of PGO_LOCKED.
sys/uvm/uvm_bio.c: revision 1.40
sys/uvm/uvm_pager.h: revision 1.29
sys/miscfs/genfs/genfs_vnops.c: revision 1.100 via patch
sys/ufs/ufs/ufs_inode.c: revision 1.50
- introduce PGO_NOBLOCKALLOC and use it for ubc mapping
to prevent unnecessary block allocations in the case that
page size > block size.
- ufs_balloc_range: use VM_PROT_WRITE+PGO_NOBLOCKALLOC rather than
VM_PROT_READ.
sys/uvm/uvm_fault.c: revision 1.96
sys/miscfs/genfs/genfs_vnops.c: revision 1.101 via patch
sys/uvm/uvm_object.h: revision 1.19
sys/miscfs/genfs/genfs_node.h: revision 1.7
ensure that vnodes with dirty pages are always on syncer's queue.
- genfs_putpages: wait for i/o completion of PG_RELEASED/PG_PAGEOUT pages by
setting "wasclean" false when encountering them.
suggested by Stephan Uphoff in PR/24596 (1).
- genfs_putpages: write protect pages when cleaning out, if
we're going to take the vnode off the syncer's queue.
uvm_fault: don't write-map pages unless its vnode is already on
the syncer's queue.
fix PR/24596 (3) but in the different way from the suggested fix.
(to keep our current behaviour, ie. not to require explicit msync.
discussed on tech-kern@.)
- genfs_putpages: don't mistakenly take a vnode off the queue
by introducing a generation number in genfs_node.
genfs_getpages: increment the generation number.
suggested by Stephan Uphoff in PR/24596 (2).
- add some assertions.
sys/miscfs/genfs/genfs_vnops.c: revision 1.102 via patch
genfs_putpages: don't bother to clean the vnode unless VONWORKLST.
sys/ufs/ffs/ffs_vnops.c: revision 1.71
ffs_full_fsync: because VBLK/VCHR can be mmap'ed,
do VOP_PUTPAGES for them as well.
sys/uvm/uvm_fault.c: revision 1.97
uvm_fault: check a correct object in the case of layered filesystems.
fix PR/30811 from Jukka Salmi.
sys/uvm/uvm_object.h: revision 1.20
sys/ufs/ffs/ffs_vfsops.c: revision 1.167
sys/uvm/uvm_bio.c: revision 1.41
sys/ufs/ufs/ufs_vnops.c: revision 1.129
sys/uvm/uvm_mmap.c: revision 1.92
sys/uvm/uvm_fault.c: revision 1.98
sys/kern/vfs_subr.c: revision 1.252
sys/fs/msdosfs/denode.h: revision 1.5
sys/miscfs/genfs/genfs_vnops.c: revision 1.103 via patch
sys/fs/msdosfs/msdosfs_denode.c: revision 1.9
sys/sys/vnode.h: revision 1.141
sys/ufs/ufs/ufs_inode.c: revision 1.51
sys/ufs/ufs/ufs_extern.h: revision 1.45 via patch
sys/miscfs/genfs/genfs_node.h: revision 1.8
sys/ufs/lfs/lfs_vfsops.c: revision 1.184
sys/uvm/uvm_pager.h: revision 1.30
sys/ufs/ext2fs/ext2fs_vfsops.c: revision 1.87
update file timestamps for nfsd loaned-read and mmap.
PR/25279. discussed on tech-kern@.
sys/miscfs/genfs/genfs_vnops.c: revision 1.104 via patch
don't write-protect wired pages. pointed by Chuck Silvers.
for now, leave a vnode on the syncer's queue, as suggested by him.
sys/ufs/ffs/ffs_vnops.c: revision 1.72
revert VCHR part of ffs_vnops.c 1.71.
as VCHR uses the device pager, no point to call VOP_PUTPAGES here.
pointed by Chuck Silvers.
|
| 1.167.2.2 | 18-May-2005 |
snj | Pull up revision 1.178 (requested by perseant in ticket #311):
Don't let the pager_map deadlock avoidance code in lfs_putpages() write
segments containing zero-block FINFO records. These records cause segments
to become uncleanable, which would eventually result in a "no clean segments"
panic.
|
| 1.167.2.1 | 07-May-2005 |
tron | Apply patch (requested by perseant in ticket #242):
* fsck_lfs buffer cache fixes, including PR #29151
* Change fsck_lfs phase 0 message to reflect reality
* fsck_lfs: check phase 5 (cleanerinfo accounting) even on
roll-forward
* Keep better track of the free list during roll-forward, avoiding
a core dump
* Improve hash table use for fsck_lfs buffer and vnode cache
* Document fsck_lfs flag -f, and implement -q
* Add resize_lfs, including kernel support
* Add LFS to mountd's list of exportable filesystem types
* Make the LFS lkm work again [christos@]
* Add MP locking to the LFS kernel subsystem
* Fix pager_map deadlock in lfs_putpages()
* Avoid incomplete file extension that looks like "partial
truncation" to fsck
* Use lfs_malloc for cleaner malloc, since the cleaner often runs
in low-memory conditions.
* Use splay trees, not hash table, to track page allocation for
write.
* Fix mkdir panic on full fs
* Fix page accounting leak by counting differently.
* Use rightly named structure for lfs_getattr [skrll@]
* Cosmetic changes for readability.
|
| 1.183.2.9 | 04-Feb-2008 |
yamt | sync with head.
|
| 1.183.2.8 | 21-Jan-2008 |
yamt | sync with head
|
| 1.183.2.7 | 07-Dec-2007 |
yamt | sync with head
|
| 1.183.2.6 | 15-Nov-2007 |
yamt | sync with head.
|
| 1.183.2.5 | 27-Oct-2007 |
yamt | sync with head.
|
| 1.183.2.4 | 03-Sep-2007 |
yamt | sync with head.
|
| 1.183.2.3 | 26-Feb-2007 |
yamt | sync with head.
|
| 1.183.2.2 | 30-Dec-2006 |
yamt | sync with head.
|
| 1.183.2.1 | 21-Jun-2006 |
yamt | sync with head.
|
| 1.188.2.2 | 29-Oct-2005 |
yamt | use lfs_* directly rather than via ufs_ops.
suggested by Chuck Silvers.
|
| 1.188.2.1 | 20-Oct-2005 |
yamt | adapt ufs.
|
| 1.190.2.2 | 01-Mar-2006 |
yamt | sync with head.
|
| 1.190.2.1 | 15-Jan-2006 |
yamt | sync with head.
|
| 1.192.4.2 | 01-Jun-2006 |
kardel | Sync with head.
|
| 1.192.4.1 | 22-Apr-2006 |
simonb | Sync with head.
|
| 1.192.2.1 | 09-Sep-2006 |
rpaulo | sync with head
|
| 1.193.6.3 | 24-May-2006 |
tron | Merge 2006-05-24 NetBSD-current into the "peter-altq" branch.
|
| 1.193.6.2 | 31-Mar-2006 |
tron | Merge 2006-03-31 NetBSD-current into the "peter-altq" branch.
|
| 1.193.6.1 | 28-Mar-2006 |
tron | Merge 2006-03-28 NetBSD-current into the "peter-altq" branch.
|
| 1.193.4.4 | 11-May-2006 |
elad | sync with head
|
| 1.193.4.3 | 06-May-2006 |
christos | - Move kauth_cred_t declaration to <sys/types.h>
- Cleanup struct ucred; forward declarations that are unused.
- Don't include <sys/kauth.h> in any header, but include it in the c files
that need it.
Approved by core.
|
| 1.193.4.2 | 19-Apr-2006 |
elad | sync with head.
|
| 1.193.4.1 | 08-Mar-2006 |
elad | Adapt to kernel authorization KPI.
|
| 1.193.2.6 | 03-Sep-2006 |
yamt | sync with head.
|
| 1.193.2.5 | 11-Aug-2006 |
yamt | sync with head
|
| 1.193.2.4 | 26-Jun-2006 |
yamt | sync with head.
|
| 1.193.2.3 | 24-May-2006 |
yamt | sync with head.
|
| 1.193.2.2 | 11-Apr-2006 |
yamt | sync with head
|
| 1.193.2.1 | 01-Apr-2006 |
yamt | sync with head.
|
| 1.212.2.1 | 19-Jun-2006 |
chap | Sync with head.
|
| 1.213.2.1 | 13-Jul-2006 |
gdamore | Merge from HEAD.
|
| 1.220.4.2 | 10-Dec-2006 |
yamt | sync with head.
|
| 1.220.4.1 | 22-Oct-2006 |
yamt | sync with head
|
| 1.220.2.3 | 01-Feb-2007 |
ad | Sync with head.
|
| 1.220.2.2 | 12-Jan-2007 |
ad | Sync with head.
|
| 1.220.2.1 | 18-Nov-2006 |
ad | Sync with head.
|
| 1.224.4.1 | 03-Sep-2007 |
wrstuden | Sync w/ NetBSD-4-RC_1
|
| 1.224.2.1 | 05-Jun-2007 |
bouyer | Pull up following revision(s) (requested by perseant in ticket #703):
sys/miscfs/genfs/genfs.h 1.21
sys/miscfs/genfs/genfs_vnops.c 1.151
sys/ufs/lfs/lfs.h 1.119, 1.120
sys/ufs/lfs/lfs_bio.c 1.99-101
sys/ufs/lfs/lfs_extern.h 1.89
sys/ufs/lfs/lfs_inode.c 1.108, 1.109
sys/ufs/lfs/lfs_segment.c 1.197, 1.199, 1.200
sys/ufs/lfs/lfs_subr.c 1.69, 1.70
sys/ufs/lfs/lfs_syscalls.c 1.119
sys/ufs/lfs/lfs_vfsops.c 1.234, 1.235
sys/ufs/lfs/lfs_vnops.c 1.195, 1.196, 1.200, 1.202-206
Reduce busy waiting in lfs_putpages(), and other LFS improvements.
|
| 1.228.2.4 | 17-May-2007 |
yamt | sync with head.
|
| 1.228.2.3 | 07-May-2007 |
yamt | sync with head.
|
| 1.228.2.2 | 24-Mar-2007 |
yamt | sync with head.
|
| 1.228.2.1 | 27-Feb-2007 |
yamt | - sync with head.
- move sched_changepri back to kern_synch.c as it doesn't know PPQ anymore.
|
| 1.231.4.13 | 16-Sep-2007 |
ad | - Checkpoint work in progress on the vnode lifecycle and reference counting
stuff. This makes it work properly without kernel_lock and fixes a few
quite old bugs. See vfs_subr.c 1.283.2.17 for details.
- Fix some problems with softdep. Unfortunately our softdep code appears
to have some longstanding bugs that cause it fail under stress test.
|
| 1.231.4.12 | 24-Aug-2007 |
ad | Sync with buffer cache locking changes. See buf.h/vfs_bio.c for details.
Some minor portions are incomplete and needs to be verified as a whole.
|
| 1.231.4.11 | 21-Aug-2007 |
yamt | fix some races around pagedaemon and uvm_wait. ok'ed by Andrew Doran.
|
| 1.231.4.10 | 20-Aug-2007 |
ad | Sync with HEAD.
|
| 1.231.4.9 | 19-Aug-2007 |
ad | - Back out the biodone() changes.
- Eliminate B_ERROR (from HEAD).
|
| 1.231.4.8 | 15-Jul-2007 |
ad | Sync with head.
|
| 1.231.4.7 | 23-Jun-2007 |
ad | - Lock v_cleanblkhd, v_dirtyblkhd, v_numoutput with the vnode's interlock.
Get rid of global_v_numoutput_lock. Partially incomplete as the buffer
cache locking doesn't work very well and needs an overhaul.
- Some changes to try and make softdep MP safe. Untested.
|
| 1.231.4.6 | 08-Jun-2007 |
ad | Sync with head.
|
| 1.231.4.5 | 13-May-2007 |
ad | - Pass the error number and residual count to biodone(), and let it handle
setting error indicators. Prepare to eliminate B_ERROR.
- Add a flag argument to brelse() to be set into the buf's flags, instead
of doing it directly. Typically used to set B_INVAL.
- Add a "struct cpu_info *" argument to kthread_create(), to be used to
create bound threads. Change "bool mpsafe" to "int flags".
- Allow exit of LWPs in the IDL state when (l != curlwp).
- More locking fixes & conversion to the new API.
|
| 1.231.4.4 | 10-Apr-2007 |
ad | Nuke the deferred kthread creation stuff, as it's no longer needed.
Pointed out by thorpej@.
|
| 1.231.4.3 | 09-Apr-2007 |
ad | - Add two new arguments to kthread_create1: pri_t pri, bool mpsafe.
- Fork kthreads off proc0 as new LWPs, not new processes.
|
| 1.231.4.2 | 13-Mar-2007 |
ad | Pull in the initial set of changes for the vmlocking branch.
|
| 1.231.4.1 | 13-Mar-2007 |
ad | Sync with head.
|
| 1.232.2.1 | 11-Jul-2007 |
mjf | Sync with head.
|
| 1.240.2.1 | 15-Aug-2007 |
skrll | Sync with HEAD.
|
| 1.244.8.2 | 31-Jul-2007 |
pooka | * nuke the nameidata parameter from VFS_MOUNT(). Nobody on tech-kern
knew what it was supposed to be used for and wrstuden gave a go-ahead
* while rototilling, convert file systems which went easily to
use VFS_PROTOS() instead of manually prototyping the methods
|
| 1.244.8.1 | 31-Jul-2007 |
pooka | file lfs_vfsops.c was added on branch matt-mips64 on 2007-07-31 21:14:21 +0000
|
| 1.244.6.1 | 14-Oct-2007 |
yamt | sync with head.
|
| 1.244.4.3 | 23-Mar-2008 |
matt | sync with HEAD
|
| 1.244.4.2 | 09-Jan-2008 |
matt | sync with HEAD
|
| 1.244.4.1 | 06-Nov-2007 |
matt | sync with HEAD
|
| 1.244.2.4 | 09-Dec-2007 |
jmcneill | Sync with HEAD.
|
| 1.244.2.3 | 27-Nov-2007 |
joerg | Sync with HEAD. amd64 Xen support needs testing.
|
| 1.244.2.2 | 11-Nov-2007 |
joerg | Sync with HEAD.
|
| 1.244.2.1 | 26-Oct-2007 |
joerg | Sync with HEAD.
Follow the merge of pmap.c on i386 and amd64 and move
pmap_init_tmp_pgtbl into arch/x86/x86/pmap.c. Modify the ACPI wakeup
code to restore CR4 before jumping back into kernel space as the large
page option might cover that.
|
| 1.246.4.4 | 18-Feb-2008 |
mjf | Sync with HEAD.
|
| 1.246.4.3 | 27-Dec-2007 |
mjf | Sync with HEAD.
|
| 1.246.4.2 | 08-Dec-2007 |
mjf | Sync with HEAD.
|
| 1.246.4.1 | 19-Nov-2007 |
mjf | Sync with HEAD.
|
| 1.246.2.2 | 22-Nov-2007 |
bouyer | Sync with HEAD
|
| 1.246.2.1 | 13-Nov-2007 |
bouyer | Sync with HEAD
|
| 1.249.2.5 | 26-Dec-2007 |
ad | Sync with head.
|
| 1.249.2.4 | 19-Dec-2007 |
ad | Use a global lfs_lock.
|
| 1.249.2.3 | 19-Dec-2007 |
ad | Fix some more problems w/lfs on this branch.
|
| 1.249.2.2 | 19-Dec-2007 |
ad | Get lfs mostly working.
|
| 1.249.2.1 | 04-Dec-2007 |
ad | Pull the vmlocking changes into a new branch.
|
| 1.250.4.2 | 02-Jan-2008 |
bouyer | Sync with HEAD
|
| 1.250.4.1 | 13-Dec-2007 |
bouyer | Sync with HEAD
|
| 1.250.2.1 | 13-Dec-2007 |
yamt | sync with head.
|
| 1.255.10.8 | 11-Aug-2010 |
yamt | sync with head.
|
| 1.255.10.7 | 11-Mar-2010 |
yamt | sync with head
|
| 1.255.10.6 | 16-Sep-2009 |
yamt | sync with head
|
| 1.255.10.5 | 19-Aug-2009 |
yamt | sync with head.
|
| 1.255.10.4 | 18-Jul-2009 |
yamt | sync with head.
|
| 1.255.10.3 | 16-May-2009 |
yamt | sync with head
|
| 1.255.10.2 | 04-May-2009 |
yamt | sync with head.
|
| 1.255.10.1 | 16-May-2008 |
yamt | sync with head.
|
| 1.255.8.2 | 04-Jun-2008 |
yamt | sync with head
|
| 1.255.8.1 | 18-May-2008 |
yamt | sync with head.
|
| 1.255.6.3 | 17-Jan-2009 |
mjf | Sync with HEAD.
|
| 1.255.6.2 | 29-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.255.6.1 | 02-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.260.2.2 | 18-Sep-2008 |
wrstuden | Sync with wrstuden-revivesa-base-2.
|
| 1.260.2.1 | 23-Jun-2008 |
wrstuden | Sync w/ -current. 34 merge conflicts to follow.
|
| 1.265.2.1 | 03-Jul-2008 |
simonb | Sync with head.
|
| 1.267.6.2 | 25-Apr-2014 |
sborrill | Pull up the following revisions(s) (requested by maxv in ticket #1901):
sys/kern/vfs_syscalls.c: revision 1.478, 1.480 via patch
sys/coda/coda_vfsops.c: revision 1.81
sys/fs/adosfs/advfsops.c: revision 1.70
sys/fs/cd9660/cd9660_vfsops.c: revision 1.84
sys/fs/efs/efs_vfsops.c: revision 1.25
sys/fs/filecorefs/filecore_vfsops.c: revision 1.76
sys/fs/hfs/hfs_vfsops.c: revision 1.31
sys/fs/msdosfs/msdosfs_vfsops.c: revision 1.107
sys/fs/ntfs/ntfs_vfsops.c: revision 1.94
sys/fs/ptyfs/ptyfs_vfsops.c: revision 1.50 via patch
sys/fs/puffs/puffs_vfsops.c: revision 1.110 via patch
sys/fs/smbfs/smbfs_vfsops.c: revision 1.100
sys/fs/sysvbfs/sysvbfs_vfsops.c: revision 1.43
sys/fs/tmpfs/tmpfs_vfsops.c: revision 1.59 via patch
sys/fs/udf/udf_vfsops.c: revision 1.67
sys/fs/union/union_vfsops.c: revision 1.72
sys/fs/unionfs/unionfs_vfsops.c: revision 1.13
sys/kern/vfs_syscalls.c: revision 1.479
sys/miscfs/nullfs/null_vfsops.c: revision 1.88 via patch
sys/miscfs/overlay/overlay_vfsops.c: revision 1.61
sys/miscfs/procfs/procfs_vfsops.c: revision 1.91
sys/miscfs/umapfs/umap_vfsops.c: revision 1.92
sys/nfs/nfs_vfsops.c: revision 1.227
sys/ufs/ext2fs/ext2fs_vfsops.c: revision 1.180
sys/ufs/ffs/ffs_vfsops.c: revision 1.297
sys/ufs/lfs/lfs_vfsops.c: revision 1.321
sys/ufs/mfs/mfs_vfsops.c: revision 1.107
Due to missing checks in the mount syscall, and a wrong assumption on the
file systems side, the kernel could allocate an unbounded or zero-sized
memory buffer, and could dereference a NULL pointer when particular
arguments are given by a user.
|
| 1.267.6.1 | 04-Apr-2009 |
snj | branches: 1.267.6.1.4; 1.267.6.1.6; 1.267.6.1.10;
Pull up following revision(s) (requested by ad in ticket #662):
sys/ufs/lfs/lfs_vfsops.c: revision 1.272
Turn up the volume on the warning message a bit.
|
| 1.267.6.1.10.1 | 28-Apr-2014 |
sborrill | Pull up the following revisions(s) (requested by maxv in ticket #1901):
sys/kern/vfs_syscalls.c: revision 1.478, 1.480 via patch
sys/coda/coda_vfsops.c: revision 1.81
sys/fs/adosfs/advfsops.c: revision 1.70
sys/fs/cd9660/cd9660_vfsops.c: revision 1.84
sys/fs/efs/efs_vfsops.c: revision 1.25
sys/fs/filecorefs/filecore_vfsops.c: revision 1.76
sys/fs/hfs/hfs_vfsops.c: revision 1.31
sys/fs/msdosfs/msdosfs_vfsops.c: revision 1.107
sys/fs/ntfs/ntfs_vfsops.c: revision 1.94
sys/fs/ptyfs/ptyfs_vfsops.c: revision 1.50 via patch
sys/fs/puffs/puffs_vfsops.c: revision 1.110 via patch
sys/fs/smbfs/smbfs_vfsops.c: revision 1.100
sys/fs/sysvbfs/sysvbfs_vfsops.c: revision 1.43
sys/fs/tmpfs/tmpfs_vfsops.c: revision 1.59 via patch
sys/fs/udf/udf_vfsops.c: revision 1.67
sys/fs/union/union_vfsops.c: revision 1.72
sys/fs/unionfs/unionfs_vfsops.c: revision 1.13
sys/kern/vfs_syscalls.c: revision 1.479
sys/miscfs/nullfs/null_vfsops.c: revision 1.88 via patch
sys/miscfs/overlay/overlay_vfsops.c: revision 1.61
sys/miscfs/procfs/procfs_vfsops.c: revision 1.91
sys/miscfs/umapfs/umap_vfsops.c: revision 1.92
sys/nfs/nfs_vfsops.c: revision 1.227
sys/ufs/ext2fs/ext2fs_vfsops.c: revision 1.180
sys/ufs/ffs/ffs_vfsops.c: revision 1.297
sys/ufs/lfs/lfs_vfsops.c: revision 1.321
sys/ufs/mfs/mfs_vfsops.c: revision 1.107
Due to missing checks in the mount syscall, and a wrong assumption on the
file systems side, the kernel could allocate an unbounded or zero-sized
memory buffer, and could dereference a NULL pointer when particular
arguments are given by a user.
|
| 1.267.6.1.6.1 | 28-Apr-2014 |
sborrill | Pull up the following revisions(s) (requested by maxv in ticket #1901):
sys/kern/vfs_syscalls.c: revision 1.478, 1.480 via patch
sys/coda/coda_vfsops.c: revision 1.81
sys/fs/adosfs/advfsops.c: revision 1.70
sys/fs/cd9660/cd9660_vfsops.c: revision 1.84
sys/fs/efs/efs_vfsops.c: revision 1.25
sys/fs/filecorefs/filecore_vfsops.c: revision 1.76
sys/fs/hfs/hfs_vfsops.c: revision 1.31
sys/fs/msdosfs/msdosfs_vfsops.c: revision 1.107
sys/fs/ntfs/ntfs_vfsops.c: revision 1.94
sys/fs/ptyfs/ptyfs_vfsops.c: revision 1.50 via patch
sys/fs/puffs/puffs_vfsops.c: revision 1.110 via patch
sys/fs/smbfs/smbfs_vfsops.c: revision 1.100
sys/fs/sysvbfs/sysvbfs_vfsops.c: revision 1.43
sys/fs/tmpfs/tmpfs_vfsops.c: revision 1.59 via patch
sys/fs/udf/udf_vfsops.c: revision 1.67
sys/fs/union/union_vfsops.c: revision 1.72
sys/fs/unionfs/unionfs_vfsops.c: revision 1.13
sys/kern/vfs_syscalls.c: revision 1.479
sys/miscfs/nullfs/null_vfsops.c: revision 1.88 via patch
sys/miscfs/overlay/overlay_vfsops.c: revision 1.61
sys/miscfs/procfs/procfs_vfsops.c: revision 1.91
sys/miscfs/umapfs/umap_vfsops.c: revision 1.92
sys/nfs/nfs_vfsops.c: revision 1.227
sys/ufs/ext2fs/ext2fs_vfsops.c: revision 1.180
sys/ufs/ffs/ffs_vfsops.c: revision 1.297
sys/ufs/lfs/lfs_vfsops.c: revision 1.321
sys/ufs/mfs/mfs_vfsops.c: revision 1.107
Due to missing checks in the mount syscall, and a wrong assumption on the
file systems side, the kernel could allocate an unbounded or zero-sized
memory buffer, and could dereference a NULL pointer when particular
arguments are given by a user.
|
| 1.267.6.1.4.1 | 09-Feb-2012 |
matt | Change to use the updated uvm_pageout_* signature.
|
| 1.267.4.3 | 28-Apr-2009 |
skrll | Sync with HEAD.
|
| 1.267.4.2 | 03-Mar-2009 |
skrll | Sync with HEAD.
|
| 1.267.4.1 | 19-Jan-2009 |
skrll | Sync with HEAD.
|
| 1.267.2.1 | 13-Dec-2008 |
haad | Update haad-dm branch to haad-dm-base2.
|
| 1.269.4.2 | 23-Jul-2009 |
jym | Sync with HEAD.
|
| 1.269.4.1 | 13-May-2009 |
jym | Sync with HEAD.
Commit is split, to avoid a "too many arguments" protocol error.
|
| 1.282.2.2 | 17-Aug-2010 |
uebayasi | Sync with HEAD.
|
| 1.282.2.1 | 30-Apr-2010 |
uebayasi | Sync with HEAD.
|
| 1.286.2.4 | 19-May-2011 |
rmind | Implement sharing of vnode_t::v_interlock amongst vnodes:
- Lock is shared amongst UVM objects using uvm_obj_setlock() or getnewvnode().
- Adjust vnode cache to handle unsharing, add VI_LOCKSHARE flag for that.
- Use sharing in tmpfs and layerfs for underlying object.
- Simplify locking in ubc_fault().
- Sprinkle some asserts.
Discussed with ad@.
|
| 1.286.2.3 | 21-Apr-2011 |
rmind | sync with head
|
| 1.286.2.2 | 03-Jul-2010 |
rmind | sync with head
|
| 1.286.2.1 | 16-Mar-2010 |
rmind | Change struct uvm_object::vmobjlock to be dynamically allocated with
mutex_obj_alloc(). It allows us to share the locks among UVM objects.
|
| 1.287.4.1 | 09-Feb-2011 |
bouyer | Various build fixes
|
| 1.287.2.1 | 06-Jun-2011 |
jruoho | Sync with HEAD.
|
| 1.288.2.1 | 23-Jun-2011 |
cherry | Catchup with rmind-uvmplock merge.
|
| 1.290.2.5 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.290.2.4 | 23-Jan-2013 |
yamt | sync with head
|
| 1.290.2.3 | 23-May-2012 |
yamt | sync with head.
|
| 1.290.2.2 | 17-Apr-2012 |
yamt | sync with head
|
| 1.290.2.1 | 02-Nov-2011 |
yamt | page cache related changes
- maintain object pages in radix tree rather than rb tree.
- reduce unnecessary page scan in putpages. esp. when an object has a ton of
pages cached but only a few of them are dirty.
- reduce the number of pmap operations by tracking page dirtiness more
precisely in uvm layer.
- fix nfs commit range tracking.
- fix nfs write clustering. XXX hack
|
| 1.291.4.3 | 02-Jun-2012 |
mrg | sync to latest -current.
|
| 1.291.4.2 | 05-Apr-2012 |
mrg | sync to latest -current.
|
| 1.291.4.1 | 18-Feb-2012 |
mrg | merge to -current.
|
| 1.293.2.2 | 21-Apr-2014 |
bouyer | Pull up following revision(s) (requested by maxv in ticket #1050):
sys/ufs/chfs/chfs_vfsops.c: revision 1.11
sys/fs/unionfs/unionfs_vfsops.c: revision 1.13
sys/fs/nilfs/nilfs_vfsops.c: revision 1.16
sys/ufs/mfs/mfs_vfsops.c: revision 1.107
sys/fs/sysvbfs/sysvbfs_vfsops.c: revision 1.43
sys/ufs/ffs/ffs_vfsops.c: revision 1.297
sys/kern/vfs_syscalls.c: revision 1.478
sys/kern/vfs_syscalls.c: revision 1.479
sys/fs/puffs/puffs_vfsops.c: revision 1.110
sys/fs/cd9660/cd9660_vfsops.c: revision 1.84
sys/nfs/nfs_vfsops.c: revision 1.227
sys/fs/v7fs/v7fs_vfsops.c: revision 1.10
sys/ufs/ext2fs/ext2fs_vfsops.c: revision 1.180
sys/miscfs/umapfs/umap_vfsops.c: revision 1.92
sys/fs/filecorefs/filecore_vfsops.c: revision 1.76
sys/miscfs/nullfs/null_vfsops.c: revision 1.88
sys/fs/ptyfs/ptyfs_vfsops.c: revision 1.50
sys/coda/coda_vfsops.c: revision 1.81
sys/ufs/lfs/lfs_vfsops.c: revision 1.321
sys/fs/tmpfs/tmpfs_vfsops.c: revision 1.59
sys/fs/hfs/hfs_vfsops.c: revision 1.31
sys/miscfs/overlay/overlay_vfsops.c: revision 1.61
sys/fs/union/union_vfsops.c: revision 1.72
sys/fs/ntfs/ntfs_vfsops.c: revision 1.94
sys/kern/vfs_syscalls.c: revision 1.480
sys/fs/efs/efs_vfsops.c: revision 1.25
sys/kern/vfs_syscalls.c: revision 1.482
sys/fs/msdosfs/msdosfs_vfsops.c: revision 1.107
external/cddl/osnet/dist/uts/common/fs/zfs/zfs_vfsops.c: revision 1.12
sys/miscfs/procfs/procfs_vfsops.c: revision 1.91
sys/fs/smbfs/smbfs_vfsops.c: revision 1.100
sys/fs/adosfs/advfsops.c: revision 1.70
sys/fs/udf/udf_vfsops.c: revision 1.67
Limit check for 'data_len'. Otherwise a (un)privileged user can easily
panic the system by passing a huge size.
ok christos@
An (un)privileged user can easily make the kernel dereference a NULL
pointer.
The kernel allows 'data' to be NULL; it's the fs's responsibility to
ensure that it isn't NULL (if the fs actually needs data).
ok christos@
Some fs's - like kernfs - set their vfs_min_mount_data to zero. Add a check
to prevent an (un)privileged user from requesting a zero-sized allocation
(and thus a panic).
This thing is totally buggy: 'data_len' is modified by the fs, so calling
kmem_free with it while its value has changed since the kmem_alloc is far
from being a good idea.
If the kernel figures out that something mismatches, it will panic
(typically with kernfs).
|
| 1.293.2.1 | 17-Mar-2012 |
bouyer | branches: 1.293.2.1.4; 1.293.2.1.6;
Pull up following revision(s) (requested by perseant in ticket #116):
sys/ufs/lfs/lfs_alloc.c: revision 1.112
tests/fs/vfs/t_rmdirrace.c: revision 1.9
tests/fs/vfs/t_renamerace.c: revision 1.25
sys/ufs/lfs/lfs_vnops.c: revision 1.240
sys/ufs/lfs/lfs_segment.c: revision 1.224
sys/ufs/lfs/lfs_bio.c: revision 1.122
sys/ufs/lfs/lfs_vfsops.c: revision 1.294
sbin/newfs_lfs/make_lfs.c: revision 1.19
sys/ufs/lfs/lfs.h: revision 1.136
Pass t_renamerace and t_rmdirrace tests.
Adapt dholland@'s fix to ufs_rename to fix PR kern/43582. Address several
other MP locking issues discovered during the course of investigating the
same problem.
Removed extraneous vn_lock() calls on the Ifile, since the Ifile writes
are controlled by the segment lock.
Fix PR kern/45982 by deemphasizing the estimate of how much metadata
will fill the empty space on disk when the disk is nearly empty
(t_renamerace crates a lot of inode blocks on a tiny empty disk).
|
| 1.293.2.1.6.1 | 21-Apr-2014 |
bouyer | Pull up following revision(s) (requested by maxv in ticket #1050):
sys/ufs/chfs/chfs_vfsops.c: revision 1.11
sys/fs/unionfs/unionfs_vfsops.c: revision 1.13
sys/fs/nilfs/nilfs_vfsops.c: revision 1.16
sys/ufs/mfs/mfs_vfsops.c: revision 1.107
sys/fs/sysvbfs/sysvbfs_vfsops.c: revision 1.43
sys/ufs/ffs/ffs_vfsops.c: revision 1.297
sys/kern/vfs_syscalls.c: revision 1.478
sys/kern/vfs_syscalls.c: revision 1.479
sys/fs/puffs/puffs_vfsops.c: revision 1.110
sys/fs/cd9660/cd9660_vfsops.c: revision 1.84
sys/nfs/nfs_vfsops.c: revision 1.227
sys/fs/v7fs/v7fs_vfsops.c: revision 1.10
sys/ufs/ext2fs/ext2fs_vfsops.c: revision 1.180
sys/miscfs/umapfs/umap_vfsops.c: revision 1.92
sys/fs/filecorefs/filecore_vfsops.c: revision 1.76
sys/miscfs/nullfs/null_vfsops.c: revision 1.88
sys/fs/ptyfs/ptyfs_vfsops.c: revision 1.50
sys/coda/coda_vfsops.c: revision 1.81
sys/ufs/lfs/lfs_vfsops.c: revision 1.321
sys/fs/tmpfs/tmpfs_vfsops.c: revision 1.59
sys/fs/hfs/hfs_vfsops.c: revision 1.31
sys/miscfs/overlay/overlay_vfsops.c: revision 1.61
sys/fs/union/union_vfsops.c: revision 1.72
sys/fs/ntfs/ntfs_vfsops.c: revision 1.94
sys/kern/vfs_syscalls.c: revision 1.480
sys/fs/efs/efs_vfsops.c: revision 1.25
sys/kern/vfs_syscalls.c: revision 1.482
sys/fs/msdosfs/msdosfs_vfsops.c: revision 1.107
external/cddl/osnet/dist/uts/common/fs/zfs/zfs_vfsops.c: revision 1.12
sys/miscfs/procfs/procfs_vfsops.c: revision 1.91
sys/fs/smbfs/smbfs_vfsops.c: revision 1.100
sys/fs/adosfs/advfsops.c: revision 1.70
sys/fs/udf/udf_vfsops.c: revision 1.67
Limit check for 'data_len'. Otherwise a (un)privileged user can easily
panic the system by passing a huge size.
ok christos@
An (un)privileged user can easily make the kernel dereference a NULL
pointer.
The kernel allows 'data' to be NULL; it's the fs's responsibility to
ensure that it isn't NULL (if the fs actually needs data).
ok christos@
Some fs's - like kernfs - set their vfs_min_mount_data to zero. Add a check
to prevent an (un)privileged user from requesting a zero-sized allocation
(and thus a panic).
This thing is totally buggy: 'data_len' is modified by the fs, so calling
kmem_free with it while its value has changed since the kmem_alloc is far
from being a good idea.
If the kernel figures out that something mismatches, it will panic
(typically with kernfs).
|
| 1.293.2.1.4.1 | 21-Apr-2014 |
bouyer | Pull up following revision(s) (requested by maxv in ticket #1050):
sys/ufs/chfs/chfs_vfsops.c: revision 1.11
sys/fs/unionfs/unionfs_vfsops.c: revision 1.13
sys/fs/nilfs/nilfs_vfsops.c: revision 1.16
sys/ufs/mfs/mfs_vfsops.c: revision 1.107
sys/fs/sysvbfs/sysvbfs_vfsops.c: revision 1.43
sys/ufs/ffs/ffs_vfsops.c: revision 1.297
sys/kern/vfs_syscalls.c: revision 1.478
sys/kern/vfs_syscalls.c: revision 1.479
sys/fs/puffs/puffs_vfsops.c: revision 1.110
sys/fs/cd9660/cd9660_vfsops.c: revision 1.84
sys/nfs/nfs_vfsops.c: revision 1.227
sys/fs/v7fs/v7fs_vfsops.c: revision 1.10
sys/ufs/ext2fs/ext2fs_vfsops.c: revision 1.180
sys/miscfs/umapfs/umap_vfsops.c: revision 1.92
sys/fs/filecorefs/filecore_vfsops.c: revision 1.76
sys/miscfs/nullfs/null_vfsops.c: revision 1.88
sys/fs/ptyfs/ptyfs_vfsops.c: revision 1.50
sys/coda/coda_vfsops.c: revision 1.81
sys/ufs/lfs/lfs_vfsops.c: revision 1.321
sys/fs/tmpfs/tmpfs_vfsops.c: revision 1.59
sys/fs/hfs/hfs_vfsops.c: revision 1.31
sys/miscfs/overlay/overlay_vfsops.c: revision 1.61
sys/fs/union/union_vfsops.c: revision 1.72
sys/fs/ntfs/ntfs_vfsops.c: revision 1.94
sys/kern/vfs_syscalls.c: revision 1.480
sys/fs/efs/efs_vfsops.c: revision 1.25
sys/kern/vfs_syscalls.c: revision 1.482
sys/fs/msdosfs/msdosfs_vfsops.c: revision 1.107
external/cddl/osnet/dist/uts/common/fs/zfs/zfs_vfsops.c: revision 1.12
sys/miscfs/procfs/procfs_vfsops.c: revision 1.91
sys/fs/smbfs/smbfs_vfsops.c: revision 1.100
sys/fs/adosfs/advfsops.c: revision 1.70
sys/fs/udf/udf_vfsops.c: revision 1.67
Limit check for 'data_len'. Otherwise a (un)privileged user can easily
panic the system by passing a huge size.
ok christos@
An (un)privileged user can easily make the kernel dereference a NULL
pointer.
The kernel allows 'data' to be NULL; it's the fs's responsibility to
ensure that it isn't NULL (if the fs actually needs data).
ok christos@
Some fs's - like kernfs - set their vfs_min_mount_data to zero. Add a check
to prevent an (un)privileged user from requesting a zero-sized allocation
(and thus a panic).
This thing is totally buggy: 'data_len' is modified by the fs, so calling
kmem_free with it while its value has changed since the kmem_alloc is far
from being a good idea.
If the kernel figures out that something mismatches, it will panic
(typically with kernfs).
|
| 1.296.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.296.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.296.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.296.2.1 | 25-Feb-2013 |
tls | resync with head
|
| 1.307.2.2 | 18-May-2014 |
rmind | sync with head
|
| 1.307.2.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.320.2.1 | 10-Aug-2014 |
tls | Rebase.
|
| 1.321.4.6 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.321.4.5 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.321.4.4 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.321.4.3 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.321.4.2 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.321.4.1 | 06-Apr-2015 |
skrll | Sync with HEAD
|
| 1.351.4.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.351.2.3 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.351.2.2 | 20-Mar-2017 |
pgoyette | Sync with HEAD
|
| 1.351.2.1 | 20-Jul-2016 |
pgoyette | Adapt machine-independant code to the new {b,c}devsw reference-counting
(using localcount(9)). All callers of {b,c}devsw_lookup() now call
{b,c}devsw_lookup_acquire() which retains a reference on the 'struct
{b,c}devsw'. This reference must be released by the caller once it is
finished with the structure's content (or other data that would disappear
if the 'struct {b,c}devsw' were to disappear).
|
| 1.359.4.2 | 02-Nov-2017 |
snj | Pull up following revision(s) (requested by pgoyette in ticket #335):
share/man/man9/kernhist.9: 1.5-1.8
sys/arch/acorn26/acorn26/pmap.c: 1.39
sys/arch/arm/arm32/fault.c: 1.105 via patch
sys/arch/arm/arm32/pmap.c: 1.350, 1.359
sys/arch/arm/broadcom/bcm2835_bsc.c: 1.7
sys/arch/arm/omap/if_cpsw.c: 1.20
sys/arch/arm/omap/tiotg.c: 1.7
sys/arch/evbarm/conf/RPI2_INSTALL: 1.3
sys/dev/ic/sl811hs.c: 1.98
sys/dev/usb/ehci.c: 1.256
sys/dev/usb/if_axe.c: 1.83
sys/dev/usb/motg.c: 1.18
sys/dev/usb/ohci.c: 1.274
sys/dev/usb/ucom.c: 1.119
sys/dev/usb/uhci.c: 1.277
sys/dev/usb/uhub.c: 1.137
sys/dev/usb/umass.c: 1.160-1.162
sys/dev/usb/umass_quirks.c: 1.100
sys/dev/usb/umass_scsipi.c: 1.55
sys/dev/usb/usb.c: 1.168
sys/dev/usb/usb_mem.c: 1.70
sys/dev/usb/usb_subr.c: 1.221
sys/dev/usb/usbdi.c: 1.175
sys/dev/usb/usbdi_util.c: 1.67-1.70
sys/dev/usb/usbroothub.c: 1.3
sys/dev/usb/xhci.c: 1.75
sys/external/bsd/drm2/dist/drm/i915/i915_gem.c: 1.34
sys/kern/kern_history.c: 1.15
sys/kern/kern_xxx.c: 1.74
sys/kern/vfs_bio.c: 1.275-1.276
sys/miscfs/genfs/genfs_io.c: 1.71
sys/sys/kernhist.h: 1.21
sys/ufs/ffs/ffs_balloc.c: 1.63
sys/ufs/lfs/lfs_vfsops.c: 1.361
sys/ufs/lfs/ulfs_inode.c: 1.21
sys/ufs/lfs/ulfs_vnops.c: 1.52
sys/ufs/ufs/ufs_inode.c: 1.102
sys/ufs/ufs/ufs_vnops.c: 1.239
sys/uvm/pmap/pmap.c: 1.37-1.39
sys/uvm/pmap/pmap_tlb.c: 1.22
sys/uvm/uvm_amap.c: 1.108
sys/uvm/uvm_anon.c: 1.64
sys/uvm/uvm_aobj.c: 1.126
sys/uvm/uvm_bio.c: 1.91
sys/uvm/uvm_device.c: 1.66
sys/uvm/uvm_fault.c: 1.201
sys/uvm/uvm_km.c: 1.144
sys/uvm/uvm_loan.c: 1.85
sys/uvm/uvm_map.c: 1.353
sys/uvm/uvm_page.c: 1.194
sys/uvm/uvm_pager.c: 1.111
sys/uvm/uvm_pdaemon.c: 1.109
sys/uvm/uvm_swap.c: 1.175
sys/uvm/uvm_vnode.c: 1.103
usr.bin/vmstat/vmstat.c: 1.219
Reorder to test for null before null deref in debug code
--
Reorder to test for null before null deref in debug code
--
KNF
--
No need for '\n' in UVMHIST_LOG
--
normalise a BIOHIST log message
--
Update the kernhist(9) kernel history code to address issues identified
in PR kern/52639, as well as some general cleaning-up...
(As proposed on tech-kern@ with additional changes and enhancements.)
Details of changes:
* All history arguments are now stored as uintmax_t values[1], both in
the kernel and in the structures used for exporting the history data
to userland via sysctl(9). This avoids problems on some architectures
where passing a 64-bit (or larger) value to printf(3) can cause it to
process the value as multiple arguments. (This can be particularly
problematic when printf()'s format string is not a literal, since in
that case the compiler cannot know how large each argument should be.)
* Update the data structures used for exporting kernel history data to
include a version number as well as the length of history arguments.
* All [2] existing users of kernhist(9) have had their format strings
updated. Each format specifier now includes an explicit length
modifier 'j' to refer to numeric values of the size of uintmax_t.
* All [2] existing users of kernhist(9) have had their format strings
updated to replace uses of "%p" with "%#jx", and the pointer
arguments are now cast to (uintptr_t) before being subsequently cast
to (uintmax_t). This is needed to avoid compiler warnings about
casting "pointer to integer of a different size."
* All [2] existing users of kernhist(9) have had instances of "%s" or
"%c" format strings replaced with numeric formats; several instances
of mis-match between format string and argument list have been fixed.
* vmstat(1) has been modified to handle the new size of arguments in the
history data as exported by sysctl(9).
* vmstat(1) now provides a warning message if the history requested with
the -u option does not exist (previously, this condition was silently
ignored, with only a single blank line being printed).
* vmstat(1) now checks the version and argument length included in the
data exported via sysctl(9) and exits if they do not match the values
with which vmstat was built.
* The kernhist(9) man-page has been updated to note the additional
requirements imposed on the format strings, along with several other
minor changes and enhancements.
[1] It would have been possible to use an explicit length (for example,
uint64_t) for the history arguments. But that would require another
"rototill" of all the users in the future when we add support for an
architecture that supports a larger size. Also, the printf(3)
format
specifiers for explicitly-sized values, such as "%"PRIu64, are much
more verbose (and less aesthetically appealing, IMHO) than simply
using "%ju".
[2] I've tried very hard to find "all [the] existing users of
kernhist(9)"
but it is possible that I've missed some of them. I would be glad
to
update any stragglers that anyone identifies.
--
For some reason this single kernel seems to have outgrown its declared
size as a result of the kernhist(9) changes. Bump the size.
XXX The amount of increase may be excessive - anyone with more detailed
XXX knowledge please feel free to further adjust the value
appropriately.
--
Misssed one cast of pointer --> uintptr_t in previous kernhist(9) commit
--
And yet another one. :(
--
Use correct mark-up for NetBSD version.
--
More improvements in grammar and readability.
--
Remove a stray '"' (obvious typo) and add a couple of casts that are
probably needed.
--
And replace an instance of "%p" conversion with "%#jx"
--
Whitespace fix. Give Bl tag table a width. Fix Xr.
|
| 1.359.4.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.359.2.1 | 27-Apr-2017 |
pgoyette | Restore all work from the former pgoyette-localcount branch (which is
now abandoned doe to cvs merge botch).
The branch now builds, and installs via anita. There are still some
problems (cgd is non-functional and all atf tests time-out) but they
will get resolved soon.
|
| 1.361.2.3 | 18-Jan-2019 |
pgoyette | Synch with HEAD
|
| 1.361.2.2 | 26-Dec-2018 |
pgoyette | Sync with HEAD, resolve a few conflicts
|
| 1.361.2.1 | 25-Jun-2018 |
pgoyette | Sync with HEAD
|
| 1.362.2.2 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.362.2.1 | 10-Jun-2019 |
christos | Sync with HEAD
|
| 1.365.2.2 | 07-Jan-2025 |
martin | Pull up following revision(s) (requested by hannken in ticket #1934):
sys/ufs/ext2fs/ext2fs_vfsops.c: revision 1.228
sys/ufs/lfs/lfs_vfsops.c: revision 1.383
sys/ufs/ffs/ffs_wapbl.c: revision 1.50
sys/ufs/ffs/ffs_vfsops.c: revision 1.383 (patch)
sys/ufs/ffs/ffs_vfsops.c: revision 1.384 (patch)
Remove comment "we are always called with the filesystem marked `MPBUSY'."
above some xxx_sync() operations. These operations get called without
any exclusive lock.
This comment appeared with "add quota support" on 1990-05-02.
On 1998/02/18 MNT_MPBUSY disappeared when vfs_busy() was changed from
an exclusive lock to a shared lock.
PR kern/58837 "ffs: Missing locking around fs_fmod/time"
Protect test/clear fs->fs_fmod with um_lock like it is already
protected in ffs_alloc.c.
When writing to disk protect moving superblock to buffer with um_lock.
Set/clear fs->fmod while mounting, updating a mount or unmounting
is safe as these operations run exclusive, either mounting creates
a new file system or the file system is suspended. Assert suspension
for update and unmount.
PR kern/58837 "ffs: Missing locking around fs_fmod/time"
|
| 1.365.2.1 | 17-Aug-2020 |
martin | Pull up following revision(s) (requested by riastradh in ticket #1050):
sys/ufs/lfs/lfs_subr.c: revision 1.101
sys/ufs/lfs/lfs_subr.c: revision 1.102
sys/ufs/lfs/lfs_inode.c: revision 1.158
sys/ufs/lfs/lfs_inode.h: revision 1.25
sys/ufs/lfs/lfs_balloc.c: revision 1.95
sys/ufs/lfs/lfs_pages.c: revision 1.21
sys/ufs/lfs/lfs_vnops.c: revision 1.330
sys/ufs/lfs/lfs_alloc.c: revision 1.140 (patch)
sys/ufs/lfs/lfs_alloc.c: revision 1.141 (patch)
lib/libp2k/p2k.c: revision 1.72
sys/ufs/lfs/lfs.h: revision 1.205
sys/ufs/lfs/lfs.h: revision 1.206
sys/ufs/lfs/lfs_segment.c: revision 1.284
sys/ufs/lfs/lfs.h: revision 1.207
sys/ufs/lfs/lfs_segment.c: revision 1.285
sys/ufs/lfs/lfs_debug.c: revision 1.55
sys/ufs/lfs/lfs_rename.c: revision 1.23
usr.sbin/dumplfs/dumplfs.c: revision 1.65
sys/ufs/lfs/lfs_vfsops.c: revision 1.371
sys/arch/i386/stand/efiboot/bootx64/Makefile: revision 1.3
sys/ufs/lfs/lfs_vfsops.c: revision 1.372
sys/ufs/lfs/lfs_vfsops.c: revision 1.373
sbin/fsck_lfs/pass1.c: revision 1.46
sys/ufs/lfs/lfs_vnops.c: revision 1.326
sys/ufs/lfs/lfs_vnops.c: revision 1.327
sys/ufs/lfs/lfs_vfsops.c: revision 1.375 (patch)
sys/ufs/lfs/lfs_vnops.c: revision 1.328
sys/ufs/lfs/lfs_subr.c: revision 1.98
sys/ufs/lfs/lfs_extern.h: revision 1.116
sys/ufs/lfs/lfs_vnops.c: revision 1.329
sys/ufs/lfs/lfs_subr.c: revision 1.99
sys/ufs/lfs/lfs_extern.h: revision 1.117
sys/ufs/lfs/lfs_accessors.h: revision 1.49
sys/ufs/lfs/lfs_extern.h: revision 1.118
sys/rump/fs/lib/liblfs/Makefile: revision 1.15
sys/ufs/lfs/lfs_bio.c: revision 1.146 (patch)
sys/ufs/lfs/lfs_bio.c: revision 1.147
sys/ufs/lfs/lfs_subr.c: revision 1.100
Fix kassert in lfs by initializing vp first.
Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
Teach lfs to transition ro<->rw.
Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
Assert lfs_writer where I think we can now prove it.
Serialize access to the splay tree with lfs_lock.
Change some cheap KDASSERT into KASSERT.
Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
Call lfs_orphan in lfs_rename while we're still in the dirop.
lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
Teach LFS_ORPHAN_NEXTFREE about lfs64.
Dust off the orphan detection code and try to make it work.
Fix !DIAGNOSTIC compile
Fix userland references to LFS_ORPHAN_NEXTFREE.
Forgot to grep for these or do a full distribution build, oops!
Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
OR into bp->b_cflags; don't overwrite.
CTASSERT lfs on-disk structure sizes.
Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
Fix clang build after packed lfs64 accessor change.
Suppress spurious address-of-packed error in rump lfs too.
|
| 1.367.2.3 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.367.2.2 | 19-Jan-2020 |
ad | Set IMNT_SHRLOOKUP and use it for the in-cache case. Need to check what
more can be done with tmpfs though, it can probably do the whole lookup.
|
| 1.367.2.1 | 17-Jan-2020 |
ad | Sync with head.
|
| 1.380.6.1 | 01-Aug-2021 |
thorpej | Sync with HEAD.
|
| 1.382.10.1 | 02-Aug-2025 |
perseant | Sync with HEAD
|
| 1.382.4.1 | 07-Jan-2025 |
martin | Pull up following revision(s) (requested by hannken in ticket #1037):
sys/ufs/ext2fs/ext2fs_vfsops.c: revision 1.228
sys/ufs/lfs/lfs_vfsops.c: revision 1.383
sys/ufs/ffs/ffs_wapbl.c: revision 1.50
sys/ufs/ffs/ffs_vfsops.c: revision 1.383
sys/ufs/ffs/ffs_vfsops.c: revision 1.384
Remove comment "we are always called with the filesystem marked `MPBUSY'."
above some xxx_sync() operations. These operations get called without
any exclusive lock.
This comment appeared with "add quota support" on 1990-05-02.
On 1998/02/18 MNT_MPBUSY disappeared when vfs_busy() was changed from
an exclusive lock to a shared lock.
PR kern/58837 "ffs: Missing locking around fs_fmod/time"
Protect test/clear fs->fs_fmod with um_lock like it is already
protected in ffs_alloc.c.
When writing to disk protect moving superblock to buffer with um_lock.
Set/clear fs->fmod while mounting, updating a mount or unmounting
is safe as these operations run exclusive, either mounting creates
a new file system or the file system is suspended. Assert suspension
for update and unmount.
PR kern/58837 "ffs: Missing locking around fs_fmod/time"
|
| 1.345 | 20-Oct-2025 |
perseant | * Generalize the partial-segment parser introduced for roll-forward,
using it to facilitate an in-kernel segment rewriter (cleaner), and a
mechanism to check whether a segment is in fact empty (only used
with DEBUG).
* Add these new fcntl calls:
- LFCNFILESTATS: For each inode given, report its number of direct
blocks, how many gaps (discontinuities) there are between direct
blocks, and how large the total gap distance is. This will be
useful for a coalescing agent.
- LFCNREWRITEFILE: For each inode given, rewrite its direct blocks,
effectively coalescing it into as compact a form as possible.
- LFCNSCRAMBLE: As above, except that it only rewrites every other
block. This causes the file to have many gaps that can be
measured with LFCNFILESTATS and addressed with LFCNREWRITEFILE,
for testing purposes.
- LFCNREWRITESEGS: Rewrite any live data in the given segments.
This is intended to simplify the cleaner API and facilitate an
in-kernel cleaner.
- LFCNCLEANERINFO: Get the most current CLEANERINFO data from the
kernel.
- LFCNSEGUSE: Retrieve segment usage data from the kernel.
* Vnodes marked IN_CLEANING now take a reference. Add a new "cleaner
lock", which must be taken by the cleaner before the segment lock,
and before marking nodes IN_CLEANING. This allows us to flush
vnodes, if necessary, before the cleaning segment is written, and
never to flush vnodes being cleaned. When the cleaner lock is
released, the vnodes are cleared of IN_CLEANING and the reference
dropped.
* Track a potential infinite loop in lfs_gatherblock.
* Pull "needs to flush" and "needs to wait for flush" into functions
instead of inlining their definitions.
|
| 1.344 | 01-Oct-2025 |
perseant | Align case labels with 8-character tab stops. No functional change.
|
| 1.343 | 17-Sep-2025 |
perseant | Add working in-kernel roll forward.
|
| 1.342 | 06-Sep-2025 |
perseant | Lock the vnode before calling lfs_set_dirop, to meet the conditions of
the assertion. Fixes a regression introduced in rev 1.341.
|
| 1.341 | 05-Sep-2025 |
perseant | Protect the changed link count of the linked vnode with {,UN}MARK_DIROP
in lfs_link(). Necessary for roll-forward.
|
| 1.340 | 20-Oct-2021 |
thorpej | Overhaul of the EVFILT_VNODE kevent(2) filter:
- Centralize vnode kevent handling in the VOP_*() wrappers, rather than
forcing each individual file system to deal with it (except VOP_RENAME(),
because VOP_RENAME() is a mess and we currently have 2 different ways
of handling it; at least it's reasonably well-centralized in the "new"
way).
- Add support for NOTE_OPEN, NOTE_CLOSE, NOTE_CLOSE_WRITE, and NOTE_READ,
compatible with the same events in FreeBSD.
- Track which kevent notifications clients are interested in receiving
to avoid doing work for events no one cares about (avoiding, e.g.
taking locks and traversing the klist to send a NOTE_WRITE when
someone is merely watching for a file to be deleted, for example).
In support of the above:
- Add support in vnode_if.sh for specifying PRE- and POST-op handlers,
to be invoked before and after vop_pre() and vop_post(), respectively.
Basic idea from FreeBSD, but implemented differently.
- Add support in vnode_if.sh for specifying CONTEXT fields in the
vop_*_args structures. These context fields are used to convey information
between the file system VOP function and the VOP wrapper, but do not
occupy an argument slot in the VOP_*() call itself. These context fields
are initialized and subsequently interpreted by PRE- and POST-op handlers.
- Version VOP_REMOVE(), uses the a context field for the file system to report
back the resulting link count of the target vnode. Return this in tmpfs,
udf, nfs, chfs, ext2fs, lfs, and ufs.
NetBSD 9.99.92.
|
| 1.339 | 18-Jul-2021 |
dholland | Abolish all the silly indirection macros for initializing vnode ops tables.
These are things of the form #define foofs_op genfs_op, or #define
foofs_op genfs_eopnotsupp, or similar. They serve no purpose besides
obfuscation, and have gotten cutpasted all over everywhere.
|
| 1.338 | 18-Jul-2021 |
dholland | Use macros for the canned parts of device and fifo vnode op tables.
Add GENFS_SPECOP_ENTRIES and GENFS_FIFOOP_ENTRIES macros that contain
the portion of the vnode ops table declaration that is
(conservatively) the same in every fs. Use these in every fs that
supports devices and/or fifos with separate ops tables.
Note that ptyfs works differently (it has one type of vnode with
open-coded dispatch to the specfs code, which I haven't changed in
this commit) and rump/librump/rumpvfs/rumpfs.c has an indirect dynamic
dispatch that already does more or less the same thing, which I also
haven't changed.
Also note that this anticipates a few bits in the next changeset here
and there, and adds missing but unreachable calls in some cases (e.g.
most fses weren't defining whiteout on devices and fifos, but it isn't
reachable there), and it changes parsepath on devices and fifos to
genfs_badop from genfs_parsepath (but it's not reachable there
either).
It appears that devices in kernfs were missing kqfilter, so it's
possible that if you try to use kqueue on /kern/rootdev that it'll
explode.
And finally note that the ops declaration tables aren't
order-dependent. (Other than vop_default_desc has to come first.)
Otherwise this wouldn't work.
|
| 1.337 | 29-Jun-2021 |
dholland | - Add a new vnode op: VOP_PARSEPATH.
- Move namei_getcomponent to genfs_vnops.c and call it genfs_parsepath.
- Add a parsepath entry to every vnode ops table.
VOP_PARSEPATH takes a directory vnode to be searched and a complete
following path and chooses how much of that path to consume. To begin
with, all parsepath calls are genfs_parsepath, which locates the first
'/' as always.
Note that the call doesn't take the whole struct componentname, only
the string. The other bits of struct componentname should not be
needed and there's no reason to cause potential complications by
exposing them.
|
| 1.336 | 05-Sep-2020 |
riastradh | branches: 1.336.6;
Round of uvm.h cleanup.
The poorly named uvm.h is generally supposed to be for uvm-internal
users only.
- Narrow it to files that actually need it -- mostly files that need
to query whether curlwp is the pagedaemon, which should maybe be
exposed by an external header.
- Use uvm_extern.h where feasible and uvm_*.h for things not exposed
by it. We should split up uvm_extern.h but this will serve for now
to reduce the uvm.h dependencies.
- Use uvm_stat.h and #ifdef UVMHIST uvm.h for files that use
UVMHIST(ubchist), since ubchist is declared in uvm.h but the
reference evaporates if UVMHIST is not defined, so we reduce header
file dependencies.
- Make uvm_device.h and uvm_swap.h independently includable while
here.
ok chs@
|
| 1.335 | 05-Sep-2020 |
riastradh | Revert "ufs: Prevent mkdir from choking on deleted directories."
This change made no sense and should not have been committed.
|
| 1.334 | 05-Sep-2020 |
riastradh | ufs: Prevent mkdir from choking on deleted directories.
Fix some missing uvm_vnp_setsize in screw cases while here.
|
| 1.333 | 16-May-2020 |
christos | Add ACL support for FFS. From FreeBSD.
|
| 1.332 | 13-Apr-2020 |
ad | Replace most uses of vp->v_usecount with a call to vrefcnt(vp), a function
that hides the details and does atomic_load_relaxed(). Signature matches
FreeBSD.
|
| 1.331 | 23-Feb-2020 |
ad | branches: 1.331.4;
UVM locking changes, proposed on tech-kern:
- Change the lock on uvm_object, vm_amap and vm_anon to be a RW lock.
- Break v_interlock and vmobjlock apart. v_interlock remains a mutex.
- Do partial PV list locking in the x86 pmap. Others to follow later.
|
| 1.330 | 23-Feb-2020 |
riastradh | Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
|
| 1.329 | 23-Feb-2020 |
riastradh | Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
|
| 1.328 | 23-Feb-2020 |
riastradh | Change some cheap KDASSERT into KASSERT.
|
| 1.327 | 23-Feb-2020 |
riastradh | Assert lfs_writer where I think we can now prove it.
|
| 1.326 | 23-Feb-2020 |
riastradh | Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
|
| 1.325 | 18-Sep-2019 |
christos | branches: 1.325.2;
Add newly created vnodes to the namei cache. The rest of the filesystems
already did that (or they don't support writing). Discussed in tech-kern.
|
| 1.324 | 20-Jun-2019 |
christos | branches: 1.324.2;
unifdef -DLFS_READWRITE ulfs_readwrite.c
|
| 1.323 | 01-Jan-2019 |
hannken | Add "void *extra" argument to vcache_new() so a file system may
pass more information about the file to create.
Welcome to 8.99.30
|
| 1.322 | 11-Aug-2018 |
zafer | In lfs_mkdir fix wrong return path in case of EMLINK which causes a panic. Also, check earlier before setting up dirop.
|
| 1.321 | 20-Aug-2017 |
maya | branches: 1.321.2; 1.321.4;
Fix typo in comment
|
| 1.320 | 19-Aug-2017 |
maya | Not much point doing anything after a panic call
|
| 1.319 | 19-Aug-2017 |
maya | Consistently use {,UN}MARK_VNODE macros rather than function calls.
|
| 1.318 | 26-Jul-2017 |
maya | change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
|
| 1.317 | 10-Jun-2017 |
maya | Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.316 | 05-Jun-2017 |
maya | Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
|
| 1.315 | 26-May-2017 |
riastradh | branches: 1.315.2;
Make VOP_RECLAIM do the last unlock of the vnode.
VOP_RECLAIM naturally has exclusive access to the vnode, so having it
locked on entry is not strictly necessary -- but it means if there
are any final operations that must be done on the vnode, such as
ffs_update, requiring exclusive access to it, we can now kassert that
the vnode is locked in those operations.
We can't just have the caller release the last lock because some file
systems don't use genfs_lock, and require the vnode to remain valid
for VOP_UNLOCK to work, notably unionfs.
|
| 1.314 | 26-Apr-2017 |
riastradh | Change VOP_REMOVE and VOP_RMDIR to preserve lock/ref on dvp.
No change to vp -- the plan is to replace the node by the
componentname in the vop parameters, and let all directory vops do
lookups internally.
Proposed on tech-kern with no objections:
https://mail-index.netbsd.org/tech-kern/2017/04/17/msg021825.html
|
| 1.313 | 11-Apr-2017 |
riastradh | Make VOP_INACTIVE preserve vnode lock on return.
Discussed on tech-kern:
https://mail-index.netbsd.org/tech-kern/2017/04/01/msg021751.html
Ride 7.99.68, a bumpy bus of incremental vfs improvements!
|
| 1.312 | 11-Apr-2017 |
riastradh | Fix non-DIAGNOSTIC build by using vp outside KASSERT too.
|
| 1.311 | 11-Apr-2017 |
riastradh | Sprinkle lock ownership assertions.
|
| 1.310 | 01-Apr-2017 |
maya | Switch lfs_writer_daemon to use condvar instead of mtsleep.
track thread existence with struct lwp instead of pid + lid,
it's more useful from ddb.
|
| 1.309 | 01-Apr-2017 |
maya | switch lfs_dirops to condvar (from mtsleep)
|
| 1.308 | 01-Apr-2017 |
maya | switch lfs_sleepers to condvar (from mtsleep)
|
| 1.307 | 30-Mar-2017 |
hannken | Remove now redundant calls to fstrans_start()/fstrans_done().
Add fstrans_start()/fstrans_done() to lfs_putpages().
|
| 1.306 | 16-Mar-2017 |
maya | actually cast to unsigned long long and use %llu. certainly not use hex (oops)
suggested by dh
|
| 1.305 | 15-Mar-2017 |
maya | print inode number in an assert I keep hitting and the adjacent one.
use PRIx64 for printing inode number elsewhere.
|
| 1.304 | 13-Jul-2016 |
maya | branches: 1.304.2; 1.304.4;
Fix a deadlock
ok dholland@
|
| 1.303 | 20-Jun-2016 |
dholland | In lfs_mknod, don't release dvp until done with it. This was exposed a
while back when I removed a sketchy preprocessor macro scheme, but I'd
left it the way it was at the time and marked it for later. Now I
guess it's later.
Also don't randomly use both dvp and ap->a_dvp; they're the same, so
pick one and stick to it.
|
| 1.302 | 20-Jun-2016 |
dholland | One more batch of already-synced ufs changes:
ufs_extern.h 1.79 is equivalent to ulfs_extern.h 1.14
ufsmount.h 1.43 is (roughly) equivalent to lfs_extern.h 1.102
ufs_inode.c 1.94 does not apply to lfs
ufs_inode.c 1.95 does not apply to lfs either
ufs_readwrite.c 1.108 is equivalent to ulfs_readwrite.c 1.8
ufs_readwrite.c 1.109 is equivalent to ulfs_readwrite.c 1.9
ufs_readwrite.c 1.110 is equivalent to ulfs_readwrite.c 1.10
ufs_readwrite.c 1.111 does not apply to lfs
ufs_readwrite.c 1.112 is equivalent to ulfs_readwrite.c 1.11
ufs_readwrite.c 1.113 is equivalent to ulfs_readwrite.c 1.13
ufs_readwrite.c 1.114 is equivalent to ulfs_readwrite.c 1.14
ufs_readwrite.c 1.115 is equivalent to ulfs_readwrite.c 1.15
ufs_readwrite.c 1.116-1.118 does not apply to lfs
ufs_readwrite.c 1.119-1.120 are equivalent to ulfs_readwrite.c 1.16
ufs_rename.c 1.12 is equivalent to lfs_rename.c 1.8
ufs_vnops.c 1.226 is equivalent to ulfs_vnops.c 1.22 and lfs_vnops.c 1.270
ufs_vnops.c 1.227 is equivalent to ulfs_vnops.c 1.23
ufs_vnops.c 1.228-1.229 are equivalent to ulfs_vnops.c 1.24
ufs_vnops.c 1.230 is equivalent to ulfs_vnops.c 1.25 and lfs_vnops.c 1.271
ufs_vnops.c 1.231 originated in lfs
ufs_vnops.c 1.232 does not apply to lfs
|
| 1.301 | 20-Jun-2016 |
dholland | With the previous we seem to have the changes from -r1.225 of ufs_vnops.c.
(as that was stuff from moving ffs to the new vcache and lfs has also been
moved, this is not surprising)
|
| 1.300 | 20-Jun-2016 |
dholland | ulfs_makeinode -> lfs_makeinode
|
| 1.299 | 20-Jun-2016 |
dholland | Merge (effectively) -r1.78 of ufs_extern.h: shift ulfs_makeinode to
lfs_vnops.c and make it file-static there, as that's the only place
it's used.
|
| 1.298 | 20-Jun-2016 |
dholland | Note more already-merged versions:
inode.h 1.68 is subsumed by ulfs_inode.h 1.19
inode.h 1.69-1.72 do not apply to lfs
ufs_extern.h 1.74 was covered when lfs was moved to the new vnode cache
ufs_extern.h 1.75 is equivalent to ulfs_extern.h 1.13
ufs_extern.h 1.76-1.77 do not apply to lfs
ufsmount.h 1.42 does not apply to lfs
ufs_inode.c 1.90 is subsumed by ulfs_inode.c 1.10
ufs_inode.c 1.91-1.92 do not apply to lfs
ufs_lookup.c 1.130 is subsumed by ulfs_lookup.c 1.24
ufs_lookup.c 1.131 is equivalent to ulfs_lookup.c 1.20
ufs_lookup.c 1.132 is equivalent to ulfs_lookup.c 1.21
ufs_lookup.c 1.133 is equivalent to ulfs_lookup.c 1.22
ufs_lookup.c 1.134 is equivalent to ulfs_lookup.c 1.23
ufs_lookup.c 1.135 is equivalent to ulfs_lookup.c 1.25
ufs_quota2.c 1.38 is equivalent to ulfs_quota2.c 1.17
ufs_quota2.c 1.39 is equivalent to ulfs_quota2.c 1.16
ufs_quota2.c 1.40 is equivalent to ulfs_quota2.c 1.18
ufs_vfsops.c 1.53 is subsumed by lfs_vfsops.c 1.324
ufs_vfsops.c 1.54 is subsumed by lfs_vfsops.c 1.324
ufs_vnops.c 1.223-1.224 do not apply to lfs
|
| 1.297 | 20-Jun-2016 |
dholland | More already-merged or equivalent changes:
ufs_dirhash.c 1.36 corresponds to ulfs_dirhash.c 1.8
ufs_extattr.c 1.43 corresponds to ulfs_extattr.c 1.7
ufs_lookup.c 1.126 does not apply to lfs
ufs_lookup.c 1.127 we already have
ufs_lookup.c 1.128 does not apply to lfs
ufs_lookup.c 1.129 corresponds to ulfs_lookup.c 1.19
ufs_quota1.c 1.19 corresponds to ulfs_quota1.c 1.7
ufs_quota1.c 1.20 corresponds to ulfs_quota1.c 1.8
ufs_quota2.c 1.36 we have equivalent changes for
ufs_rename.c 1.9 corresponds to lfs_rename.c 1.5
ufs_rename.c 1.10 corresponds to lfs_rename.c 1.6
ufs_vnops.c 1.219 corresponds to lfs_vnops.c 1.260 and ulfs_vnops.c 1.19
ufs_vnops.c 1.220 corresponds to lfs_vnops.c 1.261 and ulfs_vnops.c 1.20
ufs_vnops.c 1.221 was superseded by later changes
ufs_vnops.c 1.222 got fixed independently in lfs
|
| 1.296 | 19-Jun-2016 |
dholland | we already have ufs_lookup.c 1.125 and ufs_vnops.c 1.218.
|
| 1.295 | 19-Jun-2016 |
dholland | missed one
(probably this should be tracked in some way other than pasting rcsid
comments, but it's what we've got)
|
| 1.294 | 19-Jun-2016 |
dholland | Merge -r1.216 of ufs_vnops.c: comments about maxsymlinklen handling
|
| 1.293 | 21-Sep-2015 |
dholland | Add 64-bit directory entry structures, and adjust accessors accordingly.
The LFS64 directory entry has a 64-bit inode number. This is stored as
two 32-bit values to avoid inducing 64-bit alignment requirements.
The exposed type for manipulating directory entries is now
LFS_DIRHEADER, following the same convention as e.g. IFILE and SEGUSE.
(But with LFS_ on it, because.)
|
| 1.292 | 21-Sep-2015 |
dholland | Oops; LFS_DIRECTSIZ() is going to need the fs as an argument.
Also, it turns out that dirhash needs a compile-time-constant version
of LFS_DIRECTSIZ(LFS_MAXNAMLEN+1), independent of 64-vs-32, so create
LFS_MAXDIRENTRYSIZE for this. Sigh.
|
| 1.291 | 20-Sep-2015 |
dholland | Clean up struct lfs_dirtemplate.
|
| 1.290 | 15-Sep-2015 |
dholland | Kill off ulfs_makedirentry; just pass the data to ulfs_direnter instead.
For now, move one copy of the code that allocates and fills in a
temporary struct lfs_direct to the top of ulfs_direnter; but it should
go away shortly.
|
| 1.289 | 01-Sep-2015 |
dholland | Add new accessors for the d_type and d_namlen fields of struct lfs_direct.
Napalm the old byteswap access logic for these.
|
| 1.288 | 01-Sep-2015 |
dholland | Use the lfs dinode accessors in place of the ufs-derived ones.
(Mostly.)
The ufs-derived ones are fake structure member macros, which are gross
and not very safe. Also, it seems that a lot of places in the lfs code
were using the ffsv1 branch of them unconditionally, and this way it's
guaranteed all those places have been updated.
Found while doing this: for non-devices, have getattr produce NODEV
in the rdev field instead of leaking the address of the first direct
block.
|
| 1.287 | 19-Aug-2015 |
dholland | Part two of dinodes; use the same union everywhere.
(previously the ufs-derived code had things set up slightly different)
Remove a bunch of associated mess.
|
| 1.286 | 12-Aug-2015 |
dholland | Hack up dinode usage to be 64 vs. 32 as needed. Part 1.
(This part changes the native lfs code; the ufs-derived code already
has 64 vs. 32 logic, but as aspects of it are unsafe, and don't
entirely interoperate cleanly with the lfs 64/32 stuff, pass 2 will be
rehashing that.)
|
| 1.285 | 12-Aug-2015 |
dholland | Make 32-bit and 64-bit versions of SEGSUM.
Also fix some of the FINFO handling as it's closely entangled.
|
| 1.284 | 12-Aug-2015 |
dholland | Make 32-bit and 64-bit versions of CLEANERINFO.
XXX: while this is written to disk, it seems like much of it would
XXX: be better set up as a commpage shared with the cleaner.
|
| 1.283 | 12-Aug-2015 |
dholland | Widen several of the fields of BLOCK_INFO to 64 bits.
Keep the old BLOCK_INFO as BLOCK_INFO_70, and version the fcntls that
use it.
Note that BLOCK_INFO_70 has 64-bit padding issues so that it's
different on 32-bit and 64-bit machines. This has been fixed. However,
BLOCK_INFO also contains a pointer, so compat32 stuff for 32-on-64 is
still needed and doesn't currently exist.
|
| 1.282 | 12-Aug-2015 |
dholland | Move the security checks for lfs_bmapv/lfs_markv into those functions.
(instead of the system call entry points)
Avoids duplication.
While touching these, pass the lwp around instead of the proc -- the
latter was there for no other reason than because once upon a time
struct proc was the first argument of all syscalls.
(For that matter, why not just use curlwp instead of passing it around
all over the place? The cost of passing it to every syscall probably
exceeds the cost of loading it from curcpu, even on machines where
it's not just kept in a register all the time.)
|
| 1.281 | 03-Aug-2015 |
dholland | Simplify some leftover code and remove some old assertions.
Last year when I killed off some evil dirop-related macros, I added
these assertions because if the things they asserted weren't true we'd
be leaking vnodes. Well, it seems that the code at the time did leak
vnodes, so certain failure cases (e.g. mkdir with disk full) would
assert. Nobody apparently tripped on this in the past fourteen months,
until I broke balloc so it always failed (unrelatedly) while working
on some LFS64 changes.
However, the vnode leak has since been removed by hannken@ as part of
the vnode cache changes, so the assertions are now superfluous;
instead, just make sure *vpp gets nulled on failure, and don't worry
about whether or not VU_DIROP is set as it shouldn't matter any more.
XXX: there's still a lot of gratuitous pointer aliasing in here that
should be tidied away.
|
| 1.280 | 02-Aug-2015 |
dholland | lfs_cleanint[] in the in-memory superblock needs to have 64-bit entries.
|
| 1.279 | 02-Aug-2015 |
dholland | Make i_eff_nblks in the in-memory inode 64 bits wide.
|
| 1.278 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.277 | 26-Jul-2015 |
hannken | lfs_flush_pchain: replace vget() with vcache_get().
|
| 1.276 | 25-Jul-2015 |
martin | Use accessors in DEBUG and DIAGNOSTIC code as well
|
| 1.275 | 24-Jul-2015 |
dholland | More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)
The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.
|
| 1.274 | 24-Jul-2015 |
dholland | Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)
It also gets rid of cpp abuse in the form of fake structure member
macros.
Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)
XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.
|
| 1.273 | 07-Jun-2015 |
hannken | Fix copy and paste errors from last commits.
- Kernel i386/ALL and amd64/ALL compile again.
- Resolves CID 1304138 (DEADCODE) and 1304139 (IDENTICAL_BRANCHES).
|
| 1.272 | 31-May-2015 |
hannken | Change lfs from hash table to vcache.
- Change lfs_valloc() to return an inode number and version instead of
a vnode and move lfs_ialloc() and lfs_vcreate() to new lfs_init_vnode().
- Add lfs_valloc_fixed() to allocate a known inode, used by kernel
roll forward.
- Remove lfs_*ref(), these functions cannot coexist with vcache and
their commented behaviour is far away from their implementation.
- Add the cleaner lwp and blockinfo to struct ulfsmount so lfs_loadvnode()
may use hints from the cleaner.
- Remove vnode locks from ulfs_lookup() like we did with ufs_lookup().
|
| 1.271 | 20-Apr-2015 |
riastradh | Make VOP_LINK return directory still locked and referenced.
Ride 7.99.10 bump.
|
| 1.270 | 27-Mar-2015 |
riastradh | Disentangle buffer-cached I/O from page-cached I/O in UFS.
Page-cached I/O is used for regular files, and is initiated by VFS
users such as userland and NFS.
Buffer-cached I/O is used for directories and symlinks, and is issued
only internally by UFS.
New UFS routine ufs_bufio replaces vn_rdwr for internal use.
ufs_bufio is implemented by new UFS operations uo_bufrd/uo_bufwr,
which sit in ufs_readwrite.c alongside the VOP_READ/VOP_WRITE
implementations.
I preserved the code as much as possible and will leave further
simplification for future commits. I kept the ulfs_readwrite.c
copypasta close to ufs_readwrite.c in case we ever want to merge them
back; likewise ext2fs_readwrite.c.
No externally visible semantic change. All atf fs tests still pass.
|
| 1.269 | 25-Jul-2014 |
dholland | branches: 1.269.2; 1.269.4;
Add VOP_FALLOCATE and VOP_FDISCARD to every vnode ops table I can
find.
The filesystem ones all call genfs_eopnotsupp - right now I am only
implementing the plumbing and we can implement fallocate and/or
fdiscard for files later.
The device ones call spec_fallocate (which is also genfs_eopnotsupp)
and spec_fdiscard, which dispatches to the device-level op.
The fifo ones all call vn_fifo_bypass, which also ends up being
EOPNOTSUPP.
|
| 1.268 | 17-May-2014 |
dholland | Merge ulfs_create into lfs_create.
|
| 1.267 | 17-May-2014 |
dholland | Merge ulfs_mkdir into lfs_mkdir.
|
| 1.266 | 17-May-2014 |
dholland | Merge ulfs_symlink into lfs_symlink.
|
| 1.265 | 17-May-2014 |
dholland | Move the ulfs-level (copy of ufs) vnops for symlink, create, and mkdir
into lfs_vnops.c preparatory to folding them into the lfs entry points.
(lfs_vnops.c now has four licenses. sigh.)
|
| 1.264 | 17-May-2014 |
dholland | Remove the DIROP macros. They are evil, especially the CREATE ones.
This results in some duplicate logic in the creation vnops (symlink,
mknod, create, mkdir) but we will probably be able to factor it out in
a more sensible way later.
Now the creation vnops call getnewvnode explicitly instead of under
multiple layers of obscure gunk. Then we explicitly do lfs_set_dirop,
and afterwards lfs_unset_dirop.
|
| 1.263 | 16-May-2014 |
dholland | Move lfs_getpages and lfs_putpages to their own file.
|
| 1.262 | 24-Mar-2014 |
hannken | branches: 1.262.2;
- Make VI_XLOCK, VI_CLEAN and VI_LOCKSHARE private to kern/vfs_*.c.
- Make vwait() static.
- Add vdead_check() to check a vnode for being or becoming dead.
Discussed on tech-kern.
Welcome to 6.99.38
|
| 1.261 | 23-Jan-2014 |
hannken | Change vnode operations create, mknod, mkdir and symlink to return
the resulting vnode *vpp unlocked.
Discussed on tech-kern@
Welcome to 6.99.30
|
| 1.260 | 17-Jan-2014 |
hannken | Change vnode operations create, mknod, mkdir and symlink to keep the
directory node dvp locked on return.
Discussed on tech-kern@
Welcome to 6.99.29
|
| 1.259 | 18-Oct-2013 |
christos | use __USE() in the right place, instead of (void)var.
|
| 1.258 | 17-Oct-2013 |
christos | - remove unused variables
- add debug ifdefs for debugging variables
- __USE() where appropriate.
|
| 1.257 | 29-Jul-2013 |
dholland | Fix build both with and without options LFS_EI.
|
| 1.256 | 29-Jul-2013 |
dholland | Revert previous; it is wrong.
|
| 1.255 | 28-Jul-2013 |
pgoyette | Remove unused variable to fix the build.
|
| 1.254 | 28-Jul-2013 |
dholland | Merge the extattr VOPs from ffs.
As these do nothing besides dispatch to ulfs_extattr.c it wasn't
exactly hard.
This might just make extended attributes work on lfs...
|
| 1.253 | 28-Jul-2013 |
dholland | Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.252 | 28-Jul-2013 |
dholland | Add lfs_kernel.h for declarations that don't need to be exposed to userland.
lfs currently has the following headers:
lfs.h - on-disk structures and stuff needed for userlevel tools
lfs_inode.h - additional restricted materials for userlevel tools
that operate the fs (newfs_lfs, fsck_lfs, lfs_cleanerd)
lfs_kernel.h - stuff needed only in the kernel
and the following legacy headers that are expected to be mopped up and
folded into one of the above:
lfs_extern.h - function prototypes
ulfs_bswap.h - endian-independent support
ulfs_dinode.h - now contains very little
ulfs_dirhash.h - dirhash support
ulfs_extattr.h - extattr support
ulfs_extern.h - more function prototypes
ulfs_inode.h - assorted kernel-only declarations
ulfs_quota.h - quota support
ulfs_quota1.h - more quota support
ulfs_quota2.h - more quota support
ulfs_quotacommon.h - more quota support
ulfsmount.h - legacy copy of ufsmount material
|
| 1.251 | 21-Jul-2013 |
dholland | Merge logic from ulfs_close(), ulfs_getattr(), and ulfs_strategy()
into the preexisting lfs_*() versions of these functions, and delete
the unused ulfs copies.
|
| 1.250 | 20-Jul-2013 |
dholland | Merge ulfs_mknod into lfs_mknod, which was missing some bits.
|
| 1.249 | 20-Jul-2013 |
dholland | Collect the pieces of lfs rename into lfs_rename.c, and sprinkle static.
|
| 1.248 | 18-Jun-2013 |
christos | branches: 1.248.2; 1.248.4;
Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.
|
| 1.247 | 08-Jun-2013 |
dholland | ulfs_dir.h has been emptied; remove it.
|
| 1.246 | 08-Jun-2013 |
dholland | Stick LFS_ in front of IFMT, IFIFO, IFREG, etc. so as not to conflict
with the UFS copies of these symbols. (Which themselves ought to have
UFS_ stuck on.)
|
| 1.245 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.244 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.243 | 22-Jan-2013 |
dholland | Stuff UFS_ in front of a few of ufs's symbols to reduce namespace
pollution. Specifically:
ROOTINO -> UFS_ROOTINO
WINO -> UFS_WINO
NXADDR -> UFS_NXADDR
NDADDR -> UFS_NDADDR
NIADDR -> UFS_NIADDR
MAXSYMLINKLEN -> UFS_MAXSYMLINKLEN
MAXSYMLINKLEN_UFS[12] -> UFS[12]_MAXSYMLINKLEN (for consistency)
Sort out ext2fs's misuse of NDADDR and NIADDR; fortunately, these have
the same values in ext2fs and ffs.
No functional change intended.
|
| 1.242 | 09-May-2012 |
riastradh | branches: 1.242.2;
Adapt ffs, lfs, and ext2fs to use genfs_rename.
ok dholland, rmind
|
| 1.241 | 13-Mar-2012 |
elad | Replace the remaining KAUTH_GENERIC_ISSUSER authorization calls with
something meaningful. All relevant documentation has been updated or
written.
Most of these changes were brought up in the following messages:
http://mail-index.netbsd.org/tech-kern/2012/01/18/msg012490.html
http://mail-index.netbsd.org/tech-kern/2012/01/19/msg012502.html
http://mail-index.netbsd.org/tech-kern/2012/02/17/msg012728.html
Thanks to christos, manu, njoly, and jmmv for input.
Huge thanks to pgoyette for spinning these changes through some build
cycles and ATF.
|
| 1.240 | 16-Feb-2012 |
perseant | Pass t_renamerace and t_rmdirrace tests.
Adapt dholland@'s fix to ufs_rename to fix PR kern/43582. Address several
other MP locking issues discovered during the course of investigating the
same problem.
Removed extraneous vn_lock() calls on the Ifile, since the Ifile writes
are controlled by the segment lock.
Fix PR kern/45982 by deemphasizing the estimate of how much metadata
will fill the empty space on disk when the disk is nearly empty
(t_renamerace crates a lot of inode blocks on a tiny empty disk).
|
| 1.239 | 02-Jan-2012 |
perseant | branches: 1.239.2;
* Remove PGO_RECLAIM during lfs_putpages()' call to genfs_putpages(),
to avoid a live lock in the latter when reclaiming a vnode with
dirty pages.
* Add a new segment flag, SEGM_RECLAIM, to note when a segment is
being written for vnode reclamation, and record which inode is being
reclaimed, to aid in forensic debugging.
* Add a new segment flag, SEGM_SINGLE, so that opportunistic writes
can write a single segment's worth of blocks and then stop, rather
than writing all the way up to the cleaner's reserved number of
segments.
* Add assert statements to check mutex ownership is the way it ought
to be, mostly in lfs_putpages; fix problems uncovered by this.
* Don't clear VU_DIROP until the inode actually makes its way to disk,
avoiding a problem where dirop inodes could become separated
(uncovered by a modified version of the "ckckp" forensic regression
test).
* Move the vfs_getopsbyname() call into lfs_writerd. Prepare code to
make lfs_writerd notice when there are no more LFSs, and exit losing
the reference, so that, in theory, the module can be unloaded. This
code is not enabled, since it causes a crash on exit.
* Set IN_MODIFIED on inodes flushed by lfs_flush_dirops. Really we
only need to set IN_MODIFIED if we are going to write them again
(e.g., to write pages); need to think about this more.
Finally, several changes to help avoid "no clean segments" panics:
* In lfs_bmapv, note when a vnode is loaded only to discover whether
its blocks are live, so it can immediately be recycled. Since the
cleaner will try to choose ~empty segments over full ones, this
prevents the cleaner from (1) filling the vnode cache with junk, and
(2) squeezing any unwritten writes to disk and running the fs out of
segments.
* Overestimate by half the amount of metadata that will be required
to fill the clean segments. This will make the disk appear smaller,
but should help avoid a "no clean segments" panic.
* Rearrange lfs_writerd. In particular, lfs_writerd now pays
attention to the number of clean segments available, and holds off
writing until there is room.
|
| 1.238 | 20-Sep-2011 |
chs | branches: 1.238.2; 1.238.6;
strengthen the assertions about pages existing during block allocation,
which were incorrectly relaxed last year. add some comments so that
the intent of these is hopefully clearer.
in ufs_balloc_range(), don't free pages or mark them dirty if
allocating their backing store failed. this fixes PR 45369.
|
| 1.237 | 12-Jul-2011 |
dholland | Pass the ufs_lookup_results pointer around instead of fetching it from
the inode in the guts of ufs. Now, in VOPs where i_crap is used it is
used (directly) only immediately on entry to the VOP call and then
passed around by reference.
Except for rename, which needs explicit sorting out. The code in
ufs_wapbl_rename is unchanged in behavior but I'm increasingly
inclined to think it's wrong.
|
| 1.236 | 11-Jul-2011 |
hannken | Change VOP_BWRITE() to take a vnode as its first argument like all other
VOPs do. Layered file systems no longer have to modify bp->b_vp and run
into trouble when an async VOP_BWRITE() uses the wrong vnode.
- change all occurences of VOP_BWRITE(bp) to VOP_BWRITE(bp->b_vp, bp).
- remove layer_bwrite().
- welcome to 5.99.55
Adresses PR kern/38762 panic: vwakeup: neg numoutput
No objections from tech-kern@.
|
| 1.235 | 12-Jun-2011 |
rmind | Welcome to 5.99.53! Merge rmind-uvmplock branch:
- Reorganize locking in UVM and provide extra serialisation for pmap(9).
New lock order: [vmpage-owner-lock] -> pmap-lock.
- Simplify locking in some pmap(9) modules by removing P->V locking.
- Use lock object on vmobjlock (and thus vnode_t::v_interlock) to share
the locks amongst UVM objects where necessary (tmpfs, layerfs, unionfs).
- Rewrite and optimise x86 TLB shootdown code, make it simpler and cleaner.
Add TLBSTATS option for x86 to collect statistics about TLB shootdowns.
- Unify /dev/mem et al in MI code and provide required locking (removes
kernel-lock on some ports). Also, avoid cache-aliasing issues.
Thanks to Andrew Doran and Joerg Sonnenberger, as their initial patches
formed the core changes of this branch.
|
| 1.234 | 05-Jan-2011 |
martin | branches: 1.234.6;
Avoid NULL deref inside a KASSERT, as discussed on tech-kern.
|
| 1.233 | 02-Jan-2011 |
dholland | Remove the special refcount behavior (adding an extra reference to the
parent dir) associated with SAVESTART in relookup().
Check all call sites to make sure that SAVESTART wasn't set while
calling relookup(); if it was, adjust the refcount behavior. Remove
related references to SAVESTART.
The only code that was reaching the extra ref was msdosfs_rename,
where the refcount behavior was already fairly broken and/or gross;
repair it.
Add a dummy 4th argument to relookup to make sure code that hasn't
been inspected won't compile. (This will go away next time the
relookup semantics change, which they will.)
|
| 1.232 | 18-Dec-2010 |
eeh | Byebye deadlock.
|
| 1.231 | 04-Aug-2010 |
hannken | Free the on disk inodes in the reclaim routine.
|
| 1.230 | 29-Jul-2010 |
hannken | Add vm page flag PG_MARKER and use it to tag dummy marker pages
in genfs_do_putpages() and uao_put().
Use 'v_uobj.uo_npages' to check for an empty memq.
Put some assertions where these marker pages may not appear.
Ok: YAMAMOTO Takashi <yamt@netbsd.org>
|
| 1.229 | 24-Jun-2010 |
hannken | Clean up vnode lock operations pass 2:
VOP_UNLOCK(vp, flags) -> VOP_UNLOCK(vp): Remove the unneeded flags argument.
Welcome to 5.99.32.
Discussed on tech-kern.
|
| 1.228 | 24-Jun-2010 |
hannken | Clean up vnode lock operations:
- VOP_LOCK(vp, flags): Limit the set of allowed flags to LK_EXCLUSIVE,
LK_SHARED and LK_NOWAIT. LK_INTERLOCK is no longer allowed as it
makes no sense here.
- VOP_ISLOCKED(vp): Remove the for some time unused return value
LK_EXCLOTHER. Mark this operation as "diagnostic only".
Making a lock decision based on this operation is no longer allowed.
Discussed on tech-kern.
|
| 1.227 | 29-Mar-2010 |
pooka | Stop exposing fifofs internals and leave only fifo_vnodeop_p visible.
|
| 1.226 | 07-Dec-2009 |
eeh | branches: 1.226.2; 1.226.4;
Fix some more hangs and deadlocks.
|
| 1.225 | 17-Nov-2009 |
eeh | This should fix a deadlock.
|
| 1.224 | 05-Nov-2009 |
pooka | Include compat code by default.
|
| 1.223 | 30-Oct-2009 |
christos | compile without COMPAT_50
|
| 1.222 | 29-Oct-2009 |
christos | PR/42246: NAKAJIMA Yoshihiro: provide COMPAT_50 for LFS
|
| 1.221 | 07-May-2009 |
elad | Replace KAUTH_GENERIC_ISSUSER with a better alternative.
|
| 1.220 | 22-Feb-2009 |
ad | PR kern/26878 FFSv2 + softdep = livelock (no free ram)
PR kern/16942 panic with softdep and quotas
PR kern/19565 panic: softdep_write_inodeblock: indirect pointer #1 mismatch
PR kern/26274 softdep panic: allocdirect_merge: ...
PR kern/26374 Long delay before non-root users can write to softdep partitions
PR kern/28621 1.6.x "vp != NULL" panic in ffs_softdep.c:4653 while unmounting a softdep (+quota) filesystem
PR kern/29513 FFS+Softdep panic with unfsck-able file-corruption
PR kern/31544 The ffs softdep code appears to fail to write dirty bits to disk
PR kern/31981 stopping scsi disk can cause panic (softdep)
PR kern/32116 kernel panic in softdep (assertion failure)
PR kern/32532 softdep_trackbufs deadlock
PR kern/37191 softdep: locking against myself
PR kern/40474 Kernel panic after remounting raid root with softdep
Retire softdep, pass 2. As discussed and later formally announced on the
mailing lists.
|
| 1.219 | 16-Jan-2009 |
yamt | branches: 1.219.2;
one more change which i forgot to commit with
UVM_PAGE_HASH_PENALTY -> UVM_PAGE_TREE_PENALTY rename.
noticed by Andreas Wrede.
|
| 1.218 | 24-Jun-2008 |
gmcgarry | branches: 1.218.4; 1.218.6; 1.218.12;
fcntl(4) says the command is type int. lfs_fcntl() comment says u_long. The implementation says int. Synchronise comment with documentation and cast to int before comparison.
|
| 1.217 | 04-Jun-2008 |
ad | branches: 1.217.2;
vm_page: put TAILQ_ENTRY into a union with LIST_ENTRY, so we can use both.
|
| 1.216 | 28-Apr-2008 |
martin | branches: 1.216.2;
Remove clause 3 and 4 from TNF licenses
|
| 1.215 | 25-Jan-2008 |
ad | branches: 1.215.6; 1.215.8; 1.215.10;
Remove VOP_LEASE. Discussed on tech-kern.
|
| 1.214 | 02-Jan-2008 |
ad | Merge vmlocking2 to head.
|
| 1.213 | 26-Nov-2007 |
pooka | branches: 1.213.2; 1.213.6;
Remove the "struct lwp *" argument from all VFS and VOP interfaces.
The general trend is to remove it from all kernel interfaces and
this is a start. In case the calling lwp is desired, curlwp should
be used.
quick consensus on tech-kern
|
| 1.212 | 10-Oct-2007 |
ad | branches: 1.212.4;
Merge from vmlocking:
- Split vnode::v_flag into three fields, depending on field locking.
- simple_lock -> kmutex in a few places.
- Fix some simple locking problems.
|
| 1.211 | 08-Oct-2007 |
ad | Merge ffs locking & brelse changes from the vmlocking branch.
|
| 1.210 | 29-Jul-2007 |
ad | branches: 1.210.4; 1.210.6; 1.210.8; 1.210.10;
It's not a good idea for device drivers to modify b_flags, as they don't
need to understand the locking around that field. Instead of setting
B_ERROR, set b_error instead. b_error is 'owned' by whoever completes
the I/O request.
|
| 1.209 | 27-Jul-2007 |
pooka | Change unused fflags parameter in VOP_MMAP to prot and pass in
desired vm protection.
|
| 1.208 | 10-Jul-2007 |
perseant | branches: 1.208.2;
Move the "vp = NULL" assignment after the code that requires vp != NULL.
Reported by Chris Ross on current-users.
|
| 1.207 | 09-Jul-2007 |
ad | Merge some of the less invasive changes from the vmlocking branch:
- kthread, callout, devsw API changes
- select()/poll() improvements
- miscellaneous MT safety improvements
|
| 1.206 | 24-Apr-2007 |
perseant | Get rid of our own private copy of genfs_putpages, having adapted the real
genfs_putpages to suit our purposes.
|
| 1.205 | 17-Apr-2007 |
perseant | Fix another locking protocol error in lfs_fsync().
|
| 1.204 | 17-Apr-2007 |
perseant | Fix MP locking protocol violations introduced in my previous commit.
|
| 1.203 | 17-Apr-2007 |
perseant | Install a new sysctl, vfs.lfs.ignore_lazy_sync, which causes LFS to ignore
the "smooth" syncer, as if vfs.sync.*delay = 0, but only for LFS. The
default is "on", i.e., ignore lazy sync.
Reduce the amount of polling/busy-waiting done by lfs_putpages(). To
accomplish this, copied genfs_putpages() and modified it to indicate which
page it was that caused it to return with EDEADLK. fsync()/fdatasync()
should no longer ever fail with EAGAIN, and should not consume huge
quantities of cpu.
Also, try to make dirops less likely to be written as the result of a
VOP_PUTPAGES(), while ensuring that they are written regularly.
|
| 1.202 | 05-Apr-2007 |
perseant | correct comment for lfs_putpages
|
| 1.201 | 04-Mar-2007 |
christos | branches: 1.201.2; 1.201.4;
Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.
|
| 1.200 | 21-Feb-2007 |
thorpej | Replace the Mach-derived boolean_t type with the C99 bool type. A
future commit will replace use of TRUE and FALSE with true and false.
|
| 1.199 | 20-Feb-2007 |
ad | Call genfs_node_destroy() where appropriate.
|
| 1.198 | 09-Feb-2007 |
ad | branches: 1.198.2;
Merge newlock2 to head.
|
| 1.197 | 19-Jan-2007 |
hannken | New file system suspension API to replace vn_start_write and vn_finished_write.
The suspension helpers are now put into file system specific operations.
This means every file system not supporting these helpers cannot be suspended
and therefore snapshots are no longer possible.
Implemented for file systems of type ffs.
The new API is enabled on a kernel option NEWVNGATE. This option is
not enabled by default in any kernel config.
Presented and discussed on tech-kern with much input from
Bill Studenmund <wrstuden@netbsd.org> and YAMAMOTO Takashi <yamt@netbsd.org>.
Welcome to 4.99.9 (new vfs op vfs_suspendctl).
|
| 1.196 | 04-Jan-2007 |
elad | Consistent usage of KAUTH_GENERIC_ISSUSER.
|
| 1.195 | 03-Jan-2007 |
perseant | Change VONWORKLST handling to better match its other uses; in particular,
check memq and clear VWRITEMAPDIRTY at the same time.
|
| 1.194 | 09-Dec-2006 |
chs | a smorgasbord of improvements to vnode locking and path lookup:
- LOCKPARENT is no longer relevant for lookup(), relookup() or VOP_LOOKUP().
these now always return the parent vnode locked. namei() works as before.
lookup() and various other paths no longer acquire vnode locks in the
wrong order via vrele(). fixes PR 32535.
as a nice side effect, path lookup is also up to 25% faster.
- the above allows us to get rid of PDIRUNLOCK.
- also get rid of WANTPARENT (just use LOCKPARENT and unlock it).
- remove an assumption in layer_node_find() that all file systems implement
a recursive VOP_LOCK() (unionfs doesn't).
- require that all file systems supply vfs_vptofh and vfs_fhtovp routines.
fill in eopnotsupp() for file systems that don't support being exported
and remove the checks for NULL. (layerfs calls these without checking.)
- in union_lookup1(), don't change refcounts in the ISDOTDOT case, just
adjust which vnode is locked. fixes PR 33374.
- apply fixes for ufs_rename() from ufs_vnops.c rev. 1.61 to ext2fs_rename().
|
| 1.193 | 16-Nov-2006 |
christos | branches: 1.193.2;
__unused removal on arguments; approved by core.
|
| 1.192 | 20-Oct-2006 |
reinoud | Replace the LIST structure mp->mnt_vnodelist to a TAILQ structure since all
vnodes were synced and processed backwards. This meant that the last
accessed node was processed first and the earlierst last.
An extra benefit is the removal of the ugly hack from the Berkly days on
LFS.
In the proces, i've also replaced the various variations hand written loops
by the TAILQ_FOREACH() macro's.
|
| 1.191 | 12-Oct-2006 |
christos | - sprinkle __unused on function decls.
- fix a couple of unused bugs
- no more -Wno-unused for i386
|
| 1.190 | 28-Sep-2006 |
perseant | Use lockstatus instead of a homebrewed locking system to control
LFCNWRAPSTOP and LFCNWRAPGO.
Be less verbose about the various looping checks: use log() rather than
printf(), and only log anything if we are really looping ("count = 2" is
not an error condition).
Allow dirops sleeping on available space to be interruptible.
|
| 1.189 | 15-Sep-2006 |
perseant | branches: 1.189.2;
Don't remark a locked inode with IN_MODIFIED after writing it to disk,
if we ourselves hold the lock. This prevents e.g. mknod from hanging
indefinitely.
Also, always use the return value from VOP_ISLOCKED to determine whether
we hold the lock or someone else does, rather than looking into the lock
structure ourselves.
|
| 1.188 | 01-Sep-2006 |
perseant | branches: 1.188.2;
Changes to help the roll-forward agent, to wit:
* Mark being-deleted files in the Ifile so we can finish deleting them
at fs mount time.
* Flag the Ifile with "cleaner must clean" when writers are waiting for
the cleaner, rather than relying solely on the cleaner's estimation of
whether it should clean or not.
* Note partial segments written by a user agent (in particular,
fsck_lfs) so that repeated rolls forward don't interfere with one
another.
* Add a new fcntl, LFCNPASS, that allows the log to wrap exactly once,
for better testing of the validity of checkpoints.
* Keep track of the on-disk nlink count when cleaning, so that we don't
partially complete directory operations while cleaning.
* Ensure that every single Ifile inode write represents a consistent
view of the filesystem. In particular, the accounting for the segment
we are writing the inode into must be correct, and the accounting for
the segment that inode used to reside in must be correct. Rather than
just rewriting the inode if we wrote it wrong, rewrite the necessary
ifile blocks before writing the inode so we never write it wrong.
* Don't unmark any VDIROP vnodes if we haven't written them to disk,
avoiding yet another problem with the "wait for the cleaner" error
return from lfs_putpages().
Also, move the last callback to an aiodone call, so we no longer do any
memory management from interrupt context.
|
| 1.187 | 06-Aug-2006 |
martin | Fix size confusion with lfs_fhandle - and as it now turns out to be the same
as the lfs compat_30_fhandle, g/c the latter.
Add an alias for the LFCNIFILEFH fcntl, so that binaries compiled in the
meantime (with too large lfs_fhandle) continue to work.
This makes vfs_cleanerd work again after the kernel checks filehandle size
more strictly (problem reported by Kurt Schreiner on current-users).
|
| 1.186 | 31-Jul-2006 |
martin | Make filehandles opaque to userland
|
| 1.185 | 23-Jul-2006 |
ad | Use the LWP cached credentials where sane.
|
| 1.184 | 20-Jul-2006 |
perseant | Move the kauth checks up front, so that all new LFS fcntl calls are subject
to the check for superuser privilege.
|
| 1.183 | 13-Jul-2006 |
martin | Apply _KERNEL_OPT
|
| 1.182 | 13-Jul-2006 |
martin | Version the lfs_cleanerd internal fcntl() for filehandles too,
so old cleaners should work with newer kernels.
|
| 1.181 | 13-Jul-2006 |
martin | Fix alignement problems for fhandle_t, exposed by gcc4.1.
While touching all vptofh/fhtovp functions, get rid of VFS_MAXFIDSIZ,
version the getfh(2) syscall and explicitly pass the size available in
the filehandle from userland.
Discussed on tech-kern, with lots of help from yamt (thanks!).
|
| 1.180 | 29-Jun-2006 |
perseant | Don't wake up the cleaner if the filesystem is unwrappable, and fix the
compatibility fcntls.
Also includes one-line fixes for an MP locking bug and a zero-length FINFO
problem that manifested during testing.
|
| 1.179 | 24-Jun-2006 |
perseant | Change LFCNWRAP{STOP,GO} to make them more suitable for snapshotting; in
particular, the caller can now choose whether to wait for the condition
to be met, and if the caller of LFCNWRAPSTOP dies or otherwise closes
the descriptor, the filesystem is started again. Updated the ckckp
regression test to use the new semantics.
dump_lfs(8) now uses the fcntls to implement LFS-style snapshotting through
the -X flag, addressing PR#33457 albeit not using fss(4). Fixed a couple
other problems with dump_lfs that manifested themselves during testing.
|
| 1.178 | 18-May-2006 |
perseant | branches: 1.178.4;
Break out the finfo array manipulation code into two new functions,
lfs_acquire_finfo() and lfs_release_finfo(). Add a debugging check
for zero-length finfo arrays in the segment summary to avoid future
regressions.
|
| 1.177 | 17-May-2006 |
perseant | Don't be quite so eager to error out from lfs_putpages() when pages are
busy; if we've sensed a possible 3-way deadlock and are not the pagedaemon,
relock and try again.
|
| 1.176 | 14-May-2006 |
elad | integrate kauth.
|
| 1.175 | 12-May-2006 |
perseant | Fixes to address the "vinvalbuf: dirty blocks" panic that can occur when
many inodes are cleaned at once. Make sure that we write all the pages
on vnodes that are being flushed, even if we don't think there's room;
drain v_numoutput before lfs_vflush() completes.
Also, don't allow a vnode that is in the process of being cleaned to be
chosen by getnewvnode(); this avoids a segment accounting panic in the case
that a large number of inodes are fed to lfs_markv() all at once.
|
| 1.174 | 04-May-2006 |
perseant | Change VOP_FCNTL to take an unlocked vnode. Approved by wrstuden@.
|
| 1.173 | 04-May-2006 |
perseant | Introduce another per-filesystem parameter, lfs_resvseg, to separate the
notion of "how many segments are reserved for the cleaner" from that of
"how many segments are not counted in lfs_bfree". The default value
used for existing filesystems is the same as the previous implicit value
of (lfs_minfreeseg / 2 + 1), modulo some sanity checking.
Count pending dirops on a per-filesystem basis, since once we start
writing them we can't stop until we're done. This seems to help stave off
the "no clean segments" panic in the case of filling the filesystem with
directories and small files (e.g. simultaneously unpacking more copies of
pkgsrc than will fit).
|
| 1.172 | 02-May-2006 |
perseant | Fix a "locking against myself": lfs_flush_dirops() doesn't need to lock the
vnodes to write their blocks, since it holds the segment lock.
|
| 1.171 | 01-May-2006 |
perseant | Don't ever partially write dirops, even if we need the cleaner to run.
This increases the chances of the "no clean segments" panic slightly,
but allows us to run the ckckp regression test successfully to completion.
|
| 1.170 | 30-Apr-2006 |
perseant | Postpone the segment accounting changes coming from truncation until the
inode that makes those changes valid is either written to disk by
lfs_writeinode() or discarded by lfs_vfree().
A couple of locking fixes are also included as well.
|
| 1.169 | 18-Apr-2006 |
perseant | Yet another MP locking issue.
|
| 1.168 | 17-Apr-2006 |
perseant | Introduce two fcntl calls that freeze the filesystem right at the point
where segment 0 is being considered for writing. This allows for automated
checkpoint vailidity scanning, and could be used (in conjunction with the
existing LFCNREWIND) for e.g. snapshot dumps as well.
Include a regression test that does such scanning.
When writing the Ifile, loop through the dirty block list three times to
make sure that the checkpoint is always consistent (the first and second
times the Ifile blocks can cross a segment boundary; not so the third time
unless the segments are very small). Discovered by using the aforementioned
regression test.
|
| 1.167 | 13-Apr-2006 |
perseant | Make lfs_vref/lfs_vunref not need to know about VXLOCK and VFREEING
explicitly (especially since we didn't know about VFREEING at all before),
but notice the EBUSY return from vget() instead.
Fix some more MP locking protocol issues, most of which were pointed out by
Christian Ehrhardt this morning on tech-kern.
|
| 1.166 | 11-Apr-2006 |
perseant | Another MP locking fix.
|
| 1.165 | 10-Apr-2006 |
perseant | Don't leak vnode references if we fail to lock a vnode in lfs_flush_pchain().
Also fix another (probably only academic) simple_lock protocol error.
|
| 1.164 | 08-Apr-2006 |
perseant | Implement a somewhat finer-grained mechanism for paging LFS-backed pages.
The writer daemon, if it does not need to flush the whole filesystem,
now only writes the vnodes for which the pagedaemon has requested pageouts
(although it does not pay attention to the page ranges the pagedaemon
supplies).
|
| 1.163 | 07-Apr-2006 |
perseant | Several minor bug fixes:
* Correct (weak) segment lock assertions in lfs_fragextend and lfs_putpages.
* Keep IN_MODIFIED set if we run out of avail in lfs_putpages.
* Don't try to (re)write buffers on a VBLK vnode; fixes a panic I found
while running with an LFS root.
* Raise priority of LFCNSEGWAIT to PVFS; PUSER is way too low for
something the pagedaemon is relying on.
|
| 1.162 | 01-Apr-2006 |
perseant | Make sure we unlock to zero when avoiding 3-way deadlock; otherwise we
simply have a different form of deadlock.
|
| 1.161 | 31-Mar-2006 |
perseant | Handle the "filesystem is clean" flag correctly when upgrading from
read-only to read-write mount. This makes "root on lfs" work for me,
although it looks like a different traceback from PR#32667.
|
| 1.160 | 30-Mar-2006 |
yamt | some cleanups after the introduction of GOP_SIZE_MEM flag.
- remove GOP_SIZE_READ/GOP_SIZE_WRITE flags.
they have not been used since the change.
- ufs_balloc_range: remove code which has been no-op since the change.
thanks Konrad Schroder for explaining the original intention of the code.
- ffs_gop_size: don't extend past eof, in the case of GOP_SIZE_MEM.
otherwise genfs_getpages end up to allocate pages past eof unnecessarily.
|
| 1.159 | 28-Mar-2006 |
perseant | Don't let the pagedaemon wait for pages, since that is just asking for
a deadlock.
|
| 1.158 | 24-Mar-2006 |
perseant | Improvements to LFS's paging mechanism, to wit:
* Acknowledge that sometimes there are more dirty pages to be written to
disk than clean segments. When we reach the danger line,
lfs_gop_write() now returns EAGAIN. The caller of VOP_PUTPAGES(), if
it holds the segment lock, drops it and waits for the cleaner to make
room before continuing.
* Note and avoid a three-way deadlock in lfs_putpages (a writer holding
a page busy blocks on the cleaner while the cleaner blocks on the
segment lock while lfs_putpages blocks on the page).
|
| 1.157 | 11-Dec-2005 |
christos | branches: 1.157.4; 1.157.6; 1.157.8; 1.157.10; 1.157.12;
merge ktrace-lwp.
|
| 1.156 | 02-Nov-2005 |
yamt | merge yamt-vop branch. remove following VOPs.
VOP_BLKATOFF
VOP_VALLOC
VOP_BALLOC
VOP_REALLOCBLKS
VOP_VFREE
VOP_TRUNCATE
VOP_UPDATE
|
| 1.155 | 13-Sep-2005 |
christos | branches: 1.155.2;
split out lfs_itimes(). It is used in fsck_lfs.
|
| 1.154 | 12-Sep-2005 |
christos | Use nanotime() to update the time fields in filesystems. Convert the code
from macros to real functions. Original patch and review from chuq.
Note: ext2fs only keeps seconds in the on-disk inode, and msdosfs does not
have enough precision for all fields, so this is not very useful for those
two.
|
| 1.153 | 19-Aug-2005 |
christos | 64 bit inode changes.
|
| 1.152 | 29-May-2005 |
christos | branches: 1.152.2;
- sprinkle const
- avoid shadow variables.
|
| 1.151 | 20-May-2005 |
perseant | VOP_LOCK drops the interlock; pick it up again to avoid an "already unlocked"
panic in lfs_putpages.
|
| 1.150 | 27-Apr-2005 |
perseant | Recognize that we hold the v_interlock when relocking after a flush in
lfs_putpages.
|
| 1.149 | 25-Apr-2005 |
skrll | Use the right arg structure for lfs_setattr, i.e. s/getattr/setattr/.
|
| 1.148 | 23-Apr-2005 |
perseant | Provide a resize_lfs(8), including kernel and cleaner support. The current
implementation requires the fs to be mounted while resizing. Tested in both
directions, and everything appears to work happily, but ymmv.
|
| 1.147 | 19-Apr-2005 |
perseant | Keep per-inode, per-fs, and subsystem-wide counts of blocks allocated through
lfs_balloc(), and use that to estimate the number of dirty pages belonging
to LFS (subsystem or filesystem). This is almost certainly wrong for
the case of a large mmap()ed region, but the accounting is tighter than
what we had before, and performs much better in the typical case of pages
dirtied through write().
|
| 1.146 | 18-Apr-2005 |
perseant | Check for the inode having been previously freed, in UNMARK_VNODE().
Avoids a panic when calling mkdir() on a full filesystem.
|
| 1.145 | 16-Apr-2005 |
perseant | Use splay trees, rather than a hash table, to manage the accounting of
blocks allocated through VOP_BALLOC() for pages to be written to disk.
This accounting no longer takes a noticeable fraction of the system CPU.
|
| 1.144 | 16-Apr-2005 |
perseant | Use lfs_malloc() to manage the blkiov arrays that the cleaner functions use,
since the cleaner is likely to operate in a low-memory condition.
|
| 1.143 | 14-Apr-2005 |
perseant | Tabify leading whitespace
|
| 1.142 | 14-Apr-2005 |
perseant | Consolidate the hash table we use to maintain the integrity of lfs_avail
into a single, system-wide table, rather than having a separate hash table
per inode. Significantly reduces the "system" cpu usage of your average
file write.
|
| 1.141 | 01-Apr-2005 |
perseant | Protect various per-fs structures with fs->lfs_interlock simple_lock, to
improve behavior in the multiprocessor case. Add debugging segment-lock
assertion statements.
|
| 1.140 | 25-Mar-2005 |
perseant | Don't sleep while holding the vnode interlock. Should take care of the
first panic case in PR #26043.
|
| 1.139 | 24-Mar-2005 |
chs | avoid the need for recursive locking lfs_flush_dirops() by unlocking
the vnode around the call to this in the caller.
|
| 1.138 | 23-Mar-2005 |
perseant | Make LFS dirops get their vnode first, before incrementing the dirop count,
to prevent a deadlock trying to call VOP_PUTPAGES() on a VDIROP vnode.
This can happen when a stacked filesystem is mounted on top of an LFS: an
LFS dirop needs to get a vnode, which is available from the upper layer.
The corresponding lower layer vnode, however, is VDIROP, so the upper layer
can't be cleaned out since its VOP_PUTPAGES() is passed through to the lower
layer, which waits for dirops to drain before it can proceed. Deadlock.
Tweak ufs_makeinode() and ufs_mkdir() to pass the a_vpp argument through
to VOP_VALLOC().
Partially addresses PR # 26043, though it probably does not completely fix
the problem described there.
|
| 1.137 | 08-Mar-2005 |
simonb | branches: 1.137.2;
Tab Police.
|
| 1.136 | 08-Mar-2005 |
perseant | Straighten out the maze of ifdefs. Instead, consolidate all the debugging
stuff under '#ifdef DEBUG', and use sysctl knobs to turn on/off particular
parts of the debugging reporting (if DEBUG is enabled). Re-enable the LFS
statistics in sysctl, while I'm there. A bit of a rototill.
|
| 1.135 | 26-Feb-2005 |
perry | nuke trailing whitespace
|
| 1.134 | 26-Feb-2005 |
perseant | Various minor LFS improvements:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statvfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
|
| 1.133 | 25-Jan-2005 |
wrstuden | Extend fsync_range(2) to support the FDISKSYNC flag, which requests
that the sync be propogated out through the disk drive caches.
|
| 1.132 | 22-Apr-2004 |
yamt | branches: 1.132.4; 1.132.6;
check_dirty: fix another PHOLD leak. ("goto top" path)
|
| 1.131 | 21-Apr-2004 |
christos | Replace the statfs() family of system calls with statvfs().
Retain binary compatibility.
|
| 1.130 | 20-Apr-2004 |
yamt | check_dirty: plug a PHOLD leak. from Greg Oster.
|
| 1.129 | 26-Feb-2004 |
yamt | branches: 1.129.4;
lfs_putpages: fix a simple_lock mismatch.
|
| 1.128 | 26-Jan-2004 |
hannken | Fix xxx_strategy() to use the vnode arg instead of bp->b_vp.
|
| 1.127 | 25-Jan-2004 |
hannken | Make VOP_STRATEGY(bp) a real VOP as discussed on tech-kern.
VOP_STRATEGY(bp) is replaced by one of two new functions:
- VOP_STRATEGY(vp, bp) Call the strategy routine of vp for bp.
- DEV_STRATEGY(bp) Call the d_strategy routine of bp->b_dev for bp.
DEV_STRATEGY(bp) is used only for block-to-block device situations.
|
| 1.126 | 16-Dec-2003 |
yamt | - reduce code duplication.
- use boolean_t where appropriate.
|
| 1.125 | 16-Dec-2003 |
yamt | g/c lfs_no_inactive.
|
| 1.124 | 25-Nov-2003 |
yamt | use FINFOSIZE macro.
|
| 1.123 | 30-Oct-2003 |
simonb | Remove some assigned-to but otherwise unused variables.
|
| 1.122 | 25-Oct-2003 |
christos | Fix uninitialized variable warnings.
|
| 1.121 | 21-Oct-2003 |
fvdl | Correct preempt() calls.
|
| 1.120 | 18-Oct-2003 |
yamt | be more strict about sa->vp.
(make sure the last lfs_updatemata in lfs_putpages takes effect.)
|
| 1.119 | 14-Oct-2003 |
dbj | add mnt_iflag field to struct mount for internal flags
mv MNT_GONE, MNT_UNMOUNT and MNT_WANTRDWR to this field
additonally add mnt_writeopcountupper and mnt_writeopcountlower fields
in preparation for pending write suspension support work
bump kernel version to 1.6ZD
|
| 1.118 | 24-Sep-2003 |
yamt | fix a bug of lfs.
genfs_getpages() can read in more blocks than it should due to faked filesize
of lfs_gop_size(). it's a security problem and it makes gcc3 "internal error"
to fix this,
- in genfs_getpages(), always calculate diskeof and memeof separately
so that filesystems (in this case, lfs) can use different strategies
for them.
- introduce GOP_SIZE_MEM flag and use it to request in-core filesize.
(it was an intention of GOP_SIZE_READ,
but after the above change _READ is not a straightforward name)
after this, no one uses GOP_SIZE_{READ,WRITE} anymore but leave them for now.
|
| 1.117 | 23-Sep-2003 |
yamt | cleanup IN_ADIROP/VDIROP handling a little.
|
| 1.116 | 23-Sep-2003 |
yamt | remove unnecessary externs of lfs_do_flush.
|
| 1.115 | 20-Sep-2003 |
yamt | some comments
|
| 1.114 | 07-Aug-2003 |
agc | Move UCB-licensed code from 4-clause to 3-clause licence.
Patches provided by Joel Baker in PR 22364, verified by myself.
|
| 1.113 | 12-Jul-2003 |
yamt | more MP locks.
|
| 1.112 | 12-Jul-2003 |
yamt | - protect global resource counts with lfs_subsys_lock.
- clean up scattered externs a little.
|
| 1.111 | 02-Jul-2003 |
yamt | - add a new functions, lfs_writer_enter/leave, and use them instead of
duplicated code fragments.
- add an assertion.
|
| 1.110 | 02-Jul-2003 |
yamt | drain dirops before aqcuiring seglock. otherwise it might deadlocks.
PR/20676 (Karl Knutsson)
|
| 1.109 | 29-Jun-2003 |
fvdl | branches: 1.109.2;
Back out the lwp/ktrace changes. They contained a lot of colateral damage,
and need to be examined and discussed more.
|
| 1.108 | 29-Jun-2003 |
thorpej | Undo part of the ktrace/lwp changes. In particular:
* Remove the "lwp *" argument that was added to vget(). Turns out
that nothing actually used it!
* Remove the "lwp *" arguments that were added to VFS_ROOT(), VFS_VGET(),
and VFS_FHTOVP(); all they did was pass it to vget() (which, as noted
above, didn't use it).
* Remove all of the "lwp *" arguments to internal functions that were added
just to appease the above.
|
| 1.107 | 28-Jun-2003 |
darrenr | Pass lwp pointers throughtout the kernel, as required, so that the lwpid can
be inserted into ktrace records. The general change has been to replace
"struct proc *" with "struct lwp *" in various function prototypes, pass
the lwp through and use l_proc to get the process pointer when needed.
Bump the kernel rev up to 1.6V
|
| 1.106 | 07-May-2003 |
ragge | Add a missing ifdef DDB.
|
| 1.105 | 02-May-2003 |
perseant | Correct arguments to check_dirty, ensuring that all pages in a block are
written if any of them are dirty. Pointed out by yamt.
|
| 1.104 | 27-Apr-2003 |
yamt | fix a comment.
|
| 1.103 | 23-Apr-2003 |
perseant | Make LFS work better (though still not "well") as an NFS-exported
filesystem (and other things that needed to be fixed before the tests
would complete), to wit:
* Include the fs ident in the filehandle; improve stale filehandle checks.
* Change definition of blksize() to use the on-dinode size instead of
the inode's i_size, so that fsck_lfs will work properly again.
* Use b_interlock in lfs_vtruncbuf.
* Postpone dirop reclamation until after the seglock has been released,
so that lfs_truncate is not called with the segment lock held.
* Don't loop in lfs_fsync(), just write everything and wait.
* Be more careful about the interlock/uobjlock in lfs_putpages: when we
lose this lock, we have to resynchronize dirtiness of pages in each
block.
* Be sure to always write indirect blocks and update metadata in
lfs_putpages; fixes a bug that caused blocks to be accounted to the
wrong segment.
|
| 1.102 | 02-Apr-2003 |
fvdl | Add support for UFS2. UFS2 is an enhanced FFS, adding support for
64 bit block pointers, extended attribute storage, and a few
other things.
This commit does not yet include the code to manipulate the extended
storage (for e.g. ACLs), this will be done later.
Originally written by Kirk McKusick and Network Associates Laboratories for
FreeBSD.
|
| 1.101 | 01-Apr-2003 |
yamt | lfs_strategy is used only for read.
|
| 1.100 | 28-Mar-2003 |
perseant | Add a sleeper count, to prevent the cleaner from panicing the kernel
when the filesystem is unmounted, relocking the Ifile when its lock is
draining. (We can't use vfs_busy() since the process is sleeping for a
good long time.) Clean up / organize lfs.h, while I'm here.
In lfs_update_single, assert that disk addresses are either negative, or
are still positive when converted to int32_t, to prevent recurrence of a
negative/positive block problem.
|
| 1.99 | 22-Mar-2003 |
perseant | Unlock ifile inode during streamlined VOP_INACTIVE.
|
| 1.98 | 21-Mar-2003 |
perseant | KNF (space after keywords).
|
| 1.97 | 21-Mar-2003 |
perseant | Use VONWORKLST as a heuristic for vnode emptiness, rather than exhaustively
checking the memq.
Take greater care not to dirty the Ifile vnode when unmounting the filesystem.
This should fix a "(vp->v_flag & VONWORKLST) == 0" assertion panic in vgonel
that could occur when unmounting.
Do not allow the Ifile to be mapped for writing.
|
| 1.96 | 15-Mar-2003 |
perseant | Add simple_lock protection for lfs_seglock and lfs_subsys_pages; these will
be expanded to cover other per-fs and subsystem-wide data as well.
Fix a case of IN_MODIFIED being set without updating lfs_uinodes, resulting
in a "lfs_uinodes < 0" panic.
Fix a deadlock in lfs_putpages arising from the need to busy all pages in a
block; unbusy any that had already been busied before starting over.
|
| 1.95 | 08-Mar-2003 |
perseant | Take away "#ifdef LFS_UBC".
|
| 1.94 | 08-Mar-2003 |
perseant | Add an lfs_strategy() that checks to make sure we're not trying to read
where the cleaner is trying to write, instead of tying up the "live"
buffers (or pages).
Fix a bug in the LFS_UBC case where oversized buffers would not be
checksummed correctly, causing uncleanable segments.
Make sure that wakeup(fs->lfs_iocount) is done if fs->lfs_iocount is 1
as well as 0, since we wait in some places for it to drop to 1.
Activate all pages that make it into lfs_gop_write without the segment
lock held, since they must have been dirtied very recently, even if
PG_DELWRI is not set.
|
| 1.93 | 04-Mar-2003 |
perseant | Make sure we hold the uobjlock when checking for dirty pages, in lfs_vflush.
Note that pages can become dirty without our knowing it, anyway; don't
panic if that happens.
|
| 1.92 | 02-Mar-2003 |
perseant | Account SEGUSE_ACTIVE correctly so that the automatic segment cleaning
actually happens.
Add a new fcntl call that will write the minimum necessary to checkpoint
(i.e., for on-disk directory structure to be consistent, not including
updates to file data) so that the cleaner can clean segments more quickly
without sacrificing three-way commit for cleaning.
|
| 1.91 | 01-Mar-2003 |
yamt | use pid_t for pid.
|
| 1.90 | 25-Feb-2003 |
perseant | Make fs-specific fcntl macros take three arguments (approved wrstuden).
Let LFS use fcntl for cleaner functions.
|
| 1.89 | 24-Feb-2003 |
perseant | Add lfs_ioctl vnode op, with ioctls to take over cleaner system call
functionality (not including segment clean, since that is now done
automatically as checkpoints happen).
|
| 1.88 | 23-Feb-2003 |
perseant | Fix a buffer overflow bug in the LFS_UBC case that manifested itself
either as a mysterious UVM error or as "panic: dirty bufs". Verify
maximum size in lfs_malloc.
Teach lfs_updatemeta and lfs_shellsort about oversized cluster blocks from
lfs_gop_write.
When unwiring pages in lfs_gop_write, deactivate them, under the theory
that the pagedaemon wanted to free them last we knew.
|
| 1.87 | 22-Feb-2003 |
yamt | fix simple_lock/unlock mismatches.
|
| 1.86 | 20-Feb-2003 |
perseant | Tabify, and fix some comment alignment problems.
|
| 1.85 | 19-Feb-2003 |
yamt | wire the pages instead of just dequeue'ing them.
advised by Chuck Silvers.
|
| 1.84 | 17-Feb-2003 |
perseant | Add code to UBCify LFS. This is still behind "#ifdef LFS_UBC" for now
(there are still some details to work out) but expect that to go
away soon. To support these basic changes (creation of lfs_putpages,
lfs_gop_write, mods to lfs_balloc) several other changes were made, to
wit:
* Create a writer daemon kernel thread whose purpose is to handle page
writes for the pagedaemon, but which also takes over some of the
functions of lfs_check(). This thread is started the first time an
LFS is mounted.
* Add a "flags" parameter to GOP_SIZE. Current values are
GOP_SIZE_READ, meaning that the call should return the size of the
in-core version of the file, and GOP_SIZE_WRITE, meaning that it
should return the on-disk size. One of GOP_SIZE_READ or
GOP_SIZE_WRITE must be specified.
* Instead of using malloc(...M_WAITOK) for everything, reserve enough
resources to get by and use malloc(...M_NOWAIT), using the reserves if
necessary. Use the pool subsystem for structures small enough that
this is feasible. This also obsoletes LFS_THROTTLE.
And a few that are not strictly necessary:
* Moves the LFS inode extensions off onto a separately allocated
structure; getting closer to LFS as an LKM. "Welcome to 1.6O."
* Unified GOP_ALLOC between FFS and LFS.
* Update LFS copyright headers to correct values.
* Actually cast to unsigned in lfs_shellsort, like the comment says.
* Keep track of which segments were empty before the previous
checkpoint; any segments that pass two checkpoints both dirty and
empty can be summarily cleaned. Do this. Right now lfs_segclean
still works, but this should be turned into an effectless
compatibility syscall.
|
| 1.83 | 03-Feb-2003 |
perseant | Don't call a dirop within a dirop: if lfs_rename is actually deleting
a link, call lfs_remove directly before starting dirop rather than
having ufs_rename do it.
|
| 1.82 | 30-Jan-2003 |
yamt | there's no need to treat VOP_WHITEOUT as dirop
because it modifies only one inode.
|
| 1.81 | 25-Jan-2003 |
kleink | Fix further printf format warnings for DEBUG, in the wake of daddr_t
having changed.
|
| 1.80 | 18-Jan-2003 |
thorpej | Merge the nathanw_sa branch.
|
| 1.79 | 08-Jan-2003 |
yamt | for lfs_remove/lfs_rmdir, keep removed vnodes marked VDIROP.
(backout parts of rev.1.40)
otherwise, directory structures can be corrupted because checkpoints can
occur via eg. lfs_vflush before parent directory is written.
|
| 1.78 | 08-Jan-2003 |
yamt | in set_dirop/endop, use normal vref/vrele instead of lfs versions
so that we don't miss lfs_inactivate.
|
| 1.77 | 08-Jan-2003 |
yamt | add assertions.
|
| 1.76 | 08-Jan-2003 |
yamt | use lfs_unmark_vnode instead of duplicated code fragments.
|
| 1.75 | 29-Dec-2002 |
yamt | backout assertions in lfs_inactive.
they can be false when unmounting forcibly.
|
| 1.74 | 28-Dec-2002 |
christos | fix compile problem.
|
| 1.73 | 28-Dec-2002 |
yamt | avoid warnings without DIAGNOSTIC.
pointed by Andreas Wrede.
|
| 1.72 | 28-Dec-2002 |
yamt | dirop inode can't be passed to lfs_inactivate.
|
| 1.71 | 28-Dec-2002 |
yamt | - in lfs_reserve, vref vnodes that we're locking so that cleaner doesn't
try to reclaim them.
(workaround for deadlock noted in the comment in lfs_reserveavail)
- in lfs_rename, mark vnodes which are being moved as well as directry vnodes.
|
| 1.70 | 26-Dec-2002 |
yamt | - in lfs_reserve, reserve locked buffer count as well.
- don't wait for locking buf in lfs_bwrite_ext to avoid deadlocks.
- skip lfs_reserve when we're doing dirop.
reserve more (for lfs_truncate) in set_dirop instead.
this mostly solves PR 18972. (and hopefully PR 19196)
|
| 1.69 | 24-Nov-2002 |
yamt | correct locking for lfs_rmdir. PR 18976.
|
| 1.68 | 23-Oct-2002 |
jdolecek | merge kqueue branch into -current
kqueue provides a stateful and efficient event notification framework
currently supported events include socket, file, directory, fifo,
pipe, tty and device changes, and monitoring of processes and signals
kqueue is supported by all writable filesystems in NetBSD tree
(with exception of Coda) and all device drivers supporting poll(2)
based on work done by Jonathan Lemon for FreeBSD
initial NetBSD port done by Luke Mewburn and Jason Thorpe
|
| 1.67 | 27-Sep-2002 |
provos | remove trailing \n in panic(). approved perry.
|
| 1.66 | 22-Sep-2002 |
jdolecek | don't need <sys/conf.h> here
|
| 1.65 | 16-Jun-2002 |
perseant | For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.64 | 17-May-2002 |
perseant | branches: 1.64.2;
use macros from <sys/queue.h>
|
| 1.63 | 14-May-2002 |
perseant | branches: 1.63.2;
Phase one of my three-phase plan to make LFS play nice with UBC, and bug-fixes
I found while making sure there weren't any new ones.
* Make the write clusters keep track of the buffers whose blocks they contain.
This should make it possible to (1) write clusters using a page mapping
instead of malloc, if desired, and (2) schedule blocks for rewriting
(somewhere else) if a write error occurs. Code is present to use
pagemove() to construct the clusters but that is untested and will go away
anyway in favor of page mapping.
* DEBUG now keeps a log of Ifile writes, so that any lingering instances of
the "dirty bufs" problem can be properly debugged.
* Keep track of whether the Ifile has been dirtied by various routines that
can be called by lfs_segwrite, and loop on that until it is clean, for
a checkpoint. Checkpoints need to be squeaky clean.
* Warn the user (once) if the Ifile grows larger than is reasonable for their
buffer cache. Both lfs_mountfs and lfs_unmount check since the Ifile can
grow.
* If an inode is not found in a disk block, try rereading the block, under
the assumption that the block was copied to a cluster and then freed.
* Protect WRITEINPROG() with splbio() to fix a hang in lfs_update.
|
| 1.62 | 27-Apr-2002 |
perseant | Make exported LFSes not panic on the first file create.
|
| 1.61 | 11-Feb-2002 |
perseant | Include the space taken by inodes in the count made by lfs_check();
make VOP_SETATTR call lfs_check. This prevents large numbers of inode
changes (say, at the end of tar(1)) from filling the buffer cache.
|
| 1.60 | 18-Dec-2001 |
chs | use the new compatibility routines to allow mmap() to work
(in the same non-coherent fashion that it worked pre-UBC)
until someone has time to do it the right way.
|
| 1.59 | 23-Nov-2001 |
chs | add spaces for KNF. confirmed to produce identical objects.
|
| 1.58 | 08-Nov-2001 |
lukem | add RCSID
|
| 1.57 | 26-Oct-2001 |
lukem | remove #include <ufs/ufs/quota.h> where it was just to appease
<ufs/ufs/inode.h>, since the latter now includes the former. leave the former
in source that obviously uses specific bits of it (for completeness.)
|
| 1.56 | 22-Sep-2001 |
sommerfeld | branches: 1.56.2;
Add fifo_putpages() placebo so that the vnode's uobj is unlocked.
|
| 1.55 | 15-Sep-2001 |
chs | a whole bunch of changes to improve performance and robustness under load:
- remove special treatment of pager_map mappings in pmaps. this is
required now, since I've removed the globals that expose the address range.
pager_map now uses pmap_kenter_pa() instead of pmap_enter(), so there's
no longer any need to special-case it.
- eliminate struct uvm_vnode by moving its fields into struct vnode.
- rewrite the pageout path. the pager is now responsible for handling the
high-level requests instead of only getting control after a bunch of work
has already been done on its behalf. this will allow us to UBCify LFS,
which needs tighter control over its pages than other filesystems do.
writing a page to disk no longer requires making it read-only, which
allows us to write wired pages without causing all kinds of havoc.
- use a new PG_PAGEOUT flag to indicate that a page should be freed
on behalf of the pagedaemon when it's unlocked. this flag is very similar
to PG_RELEASED, but unlike PG_RELEASED, PG_PAGEOUT can be cleared if the
pageout fails due to eg. an indirect-block buffer being locked.
this allows us to remove the "version" field from struct vm_page,
and together with shrinking "loan_count" from 32 bits to 16,
struct vm_page is now 4 bytes smaller.
- no longer use PG_RELEASED for swap-backed pages. if the page is busy
because it's being paged out, we can't release the swap slot to be
reallocated until that write is complete, but unlike with vnodes we
don't keep a count of in-progress writes so there's no good way to
know when the write is done. instead, when we need to free a busy
swap-backed page, just sleep until we can get it busy ourselves.
- implement a fast-path for extending writes which allows us to avoid
zeroing new pages. this substantially reduces cpu usage.
- encapsulate the data used by the genfs code in a struct genfs_node,
which must be the first element of the filesystem-specific vnode data
for filesystems which use genfs_{get,put}pages().
- eliminate many of the UVM pagerops, since they aren't needed anymore
now that the pager "put" operation is a higher-level operation.
- enhance the genfs code to allow NFS to use the genfs_{get,put}pages
instead of a modified copy.
- clean up struct vnode by removing all the fields that used to be used by
the vfs_cluster.c code (which we don't use anymore with UBC).
- remove kmem_object and mb_object since they were useless.
instead of allocating pages to these objects, we now just allocate
pages with no object. such pages are mapped in the kernel until they
are freed, so we can use the mapping to find the page to free it.
this allows us to remove splvm() protection in several places.
The sum of all these changes improves write throughput on my
decstation 5000/200 to within 1% of the rate of NetBSD 1.5
and reduces the elapsed time for "make release" of a NetBSD 1.5
source tree on my 128MB pc to 10% less than a 1.5 kernel took.
|
| 1.54 | 24-Aug-2001 |
chs | branches: 1.54.2;
disable mmap() for LFS until it is fixed.
|
| 1.53 | 17-Aug-2001 |
chs | add getpages/putpages entries for spec vnodes.
|
| 1.52 | 24-Jul-2001 |
assar | change vop_symlink and vop_mknod to return vpp (the created node)
refed, so that the caller can actually use it. update callers and
file systems that implement these vnode operations
|
| 1.51 | 13-Jul-2001 |
perseant | Merge the short-lived perseant-lfsv2 branch into the trunk.
Kernels and tools understand both v1 and v2 filesystems; newfs_lfs
generates v2 by default. Changes for the v2 layout include:
- Segments of non-PO2 size and arbitrary block offset, so these can be
matched to convenient physical characteristics of the partition (e.g.,
stripe or track size and offset).
- Address by fragment instead of by disk sector, paving the way for
non-512-byte-sector devices. In theory fragments can be as large
as you like, though in reality they must be smaller than MAXBSIZE in size.
- Use serial number and filesystem identifier to ensure that roll-forward
doesn't get old data and think it's new. Roll-forward is enabled for
v2 filesystems, though not for v1 filesystems by default.
- The inode free list is now a tailq, paving the way for undelete (undelete
is not yet implemented, but can be without further non-backwards-compatible
changes to disk structures).
- Inode atime information is kept in the Ifile, instead of on the inode;
that is, the inode is never written *just* because atime was changed.
Because of this the inodes remain near the file data on the disk, rather
than wandering all over as the disk is read repeatedly. This speeds up
repeated reads by a small but noticeable amount.
Other changes of note include:
- The ifile written by newfs_lfs can now be of arbitrary length, it is no
longer restricted to a single indirect block.
- Fixed an old bug where ctime was changed every time a vnode was created.
I need to look more closely to make sure that the times are only updated
during write(2) and friends, not after-the-fact during a segment write,
and certainly not by the cleaner.
|
| 1.50 | 22-Jan-2001 |
jdolecek | branches: 1.50.2; 1.50.4; 1.50.6;
make filesystem vnodeop, specop, fifoop and vnodeopv_* arrays const
|
| 1.49 | 18-Nov-2000 |
toshii | Make buildable again.
The previous commit was a backout of rev. 1.45, which must be an accident.
|
| 1.48 | 17-Nov-2000 |
perseant | Correct accounting of lfs_avail, locked_queue_count, and locked_queue_bytes.
(PR #11468). In the case of fragment allocation, check to see if enough
space is available before extending a fragment already scheduled for writing.
The locked_queue_* variables indicate the number of buffer headers and bytes,
respectively, that are unavailable to getnewbuf() because they are locked up
waiting for LFS to flush them; make sure that that is actually what we're
counting, i.e., never count malloced buffers, and always use b_bufsize instead
of b_bcount.
If DEBUG is defined, the periodic calls to lfs_countlocked will now complain
if either counter is incorrect. (In the future lfs_countlocked will not need
to be called at all if DEBUG is not defined.)
|
| 1.47 | 12-Nov-2000 |
perseant | Do not needlessly dirty segment table blocks during lfs_segwrite,
preventing needless disk activity when the filesystem is idle. (PR #10979.)
|
| 1.46 | 14-Oct-2000 |
perseant | In lfs_truncate, don't overcount the real blocks removed from the inode,
when deallocating a fragment that has not made it to disk yet.
Also, during dirops, give the directory vnode an extra reference in
SET_DIROP, to ensure its continued existence during SET_ENDOP, preventing
a possible NULL-dereference there.
These two changes should close PR #11064.
|
| 1.45 | 19-Sep-2000 |
fvdl | Adapt for VOP_FSYNC parameter change.
|
| 1.44 | 09-Sep-2000 |
perseant | Various bug-fixes to LFS, to wit:
Kernel:
* Add runtime quantity lfs_ravail, the number of disk-blocks reserved
for writing. Writes to the filesystem first reserve a maximum amount
of blocks before their write is allowed to proceed; after the blocks
are allocated the reserved total is reduced by a corresponding amount.
If the lfs_reserve function cannot immediately reserve the requested
number of blocks, the inode is unlocked, and the thread sleeps until
the cleaner has made enough space available for the blocks to be
reserved. In this way large files can be written to the filesystem
(or, smaller files can be written to a nearly-full but thoroughly
clean filesystem) and the cleaner can still function properly.
* Remove explicit switching on dlfs_minfreeseg from the kernel code; it
is now merely a fs-creation parameter used to compute dlfs_avail and
dlfs_bfree (and used by fsck_lfs(8) to check their accuracy). Its
former role is better assumed by a properly computed dlfs_avail.
* Bounds-check inode numbers submitted through lfs_bmapv and lfs_markv.
This prevents a panic, but, if the cleaner is feeding the filesystem
the wrong data, you are still in a world of hurt.
* Cleanup: remove explicit references of DEV_BSIZE in favor of
btodb()/dbtob().
lfs_cleanerd:
* Make -n mean "send N segments' blocks through a single call to
lfs_markv". Previously it had meant "clean N segments though N calls
to lfs_markv, before looking again to see if more need to be cleaned".
The new behavior gives better packing of direct data on disk with as
little metadata as possible, largely alleviating the problem that the
cleaner can consume more disk through inefficient use of metadata than
it frees by moving dirty data away from clean "holes" to produce
entirely clean segments.
* Make -b mean "read as many segments as necessary to write N segments
of dirty data back to disk", rather than its former meaning of "read
as many segments as necessary to free N segments worth of space". The
new meaning, combined with the new -n behavior described above,
further aids in cleaning storage efficiency as entire segments can be
written at once, using as few blocks as possible for segment summaries
and inode blocks.
* Make the cleaner take note of segments which could not be cleaned due
to error, and not attempt to clean them until they are entirely free
of dirty blocks. This prevents the case in which a cleanerd running
with -n 1 and without -b (formerly the default) would spin trying
repeatedly to clean a corrupt segment, while the remaining space
filled and deadlocked the filesystem.
* Update the lfs_cleanerd manual page to describe all the options,
including the changes mentioned here (in particular, the -b and -n
flags were previously undocumented).
fsck_lfs:
* Check, and optionally fix, lfs_avail (to an exact figure) and
lfs_bfree (within a margin of error) in pass 5.
newfs_lfs:
* Reduce the default dlfs_minfreeseg to 1/20 of the total segments.
* Add a warning if the sgs disklabel field is 16 (the default for FFS'
cpg, but not usually desirable for LFS' sgs: 5--8 is a better range).
* Change the calculation of lfs_avail and lfs_bfree, corresponding to
the kernel changes mentioned above.
mount_lfs:
* Add -N and -b options to pass corresponding -n and -b options to
lfs_cleanerd.
* Default to calling lfs_cleanerd with "-b -n 4".
[All of these changes were largely tested in the 1.5 branch, with the
idea that they (along with previous un-pulled-up work) could be applied
to the branch while it was still in ALPHA2; however my test system has
experienced corruption on another filesystem (/dev/console has gone
missing :^), and, while I believe this unrelated to the LFS changes, I
cannot with good conscience request that the changes be pulled up.]
|
| 1.43 | 05-Jul-2000 |
perseant | Clean up accounting of lfs_uinodes (dirty but unwritten inodes).
Make lfs_uinodes a signed quantity for debugging purposes, and set it to
zero as fs mount time.
Enclose setting/clearing of the dirty flags (IN_MODIFIED, IN_ACCESSED,
IN_CLEANING) in macros, and use those macros everywhere. Make
LFS_ITIMES use these macros; updated the ITIMES macro in inode.h to know
about this. Make ufs_getattr use ITIMES instead of FFS_ITIMES.
|
| 1.42 | 01-Jul-2000 |
perseant | Move SET_ENDOP after vrele to avoid deactivating vnode twice, if
SET_ENDOP triggers a write.
|
| 1.41 | 28-Jun-2000 |
mrg | remove include of <vm/vm.h> and <uvm/uvm_extern.h>
|
| 1.40 | 27-Jun-2000 |
perseant | Fixes associated with filling an LFS:
Change the space computation to appear to change the size of the *disk*
rather than the *bytes used* when more segment summaries and inode
blocks are written. Try to estimate the amount of space that these will
take up when more files are written, so the disk size doesn't change too
much.
Regularize error returns from lfs_valloc, lfs_balloc, lfs_truncate: they
now fail entirely, rather than succeeding half-way and leaving the fs in
an inconsistent state.
Rewrite lfs_truncate, mostly stealing from ffs_truncate. The old
lfs_truncate had difficulty truncating a large file to a non-zero size
(indirect blocks were not handled appropriately).
Unmark VDIROP on fvp after ufs_remove, ufs_rmdir, so these can be
reclaimed immediately: this vnode would not be written to disk again
anyway if the removal succeeded, and if it failed, no directory
operation occurred.
ufs_makeinode and ufs_mkdir now remove IN_ADIROP on error.
|
| 1.39 | 22-Jun-2000 |
perseant | Update lfs_vunref for the fact that now a vnode can be locked with no
references (locked for VOP_INACTIVE at the end of vrele) and it's okay.
Check the return value of lfs_vref where appropriate.
Fixes PR #s 10285 and 10352.
|
| 1.38 | 31-May-2000 |
perseant | branches: 1.38.2;
update for IN_ACCESSED changes
|
| 1.37 | 27-May-2000 |
perseant | branches: 1.37.2;
Prevent dirops from getting around lfs_check and wedging the buffer cache.
All the dirop vnops now mark the inodes with a new flag, IN_ADIROP, which
is removed as soon as the dirop is done (as opposed to VDIROP which stays
until the file is written). To address one issue raised in PR#9357.
|
| 1.36 | 13-May-2000 |
perseant | Change the sementics of the last parameter from a boolean ("waitfor") to
a set of flags ("flags"). Two flags are defined, UPDATE_WAIT and
UPDATE_DIROP.
Under the old semantics, VOP_UPDATE would block if waitfor were set,
under the assumption that directory operations should be done
synchronously. At least LFS and FFS+softdep do not make this
assumption; FFS+softdep got around the problem by enclosing all relevant
calls to VOP_UPDATE in a "if(!DOINGSOFTDEP(vp))", while LFS simply
ignored waitfor, one of the reasons why NFS-serving an LFS filesystem
did not work properly.
Under the new semantics, the UPDATE_DIROP flag is a hint to the
fs-specific update routine that the call comes from a dirop routine, and
should be wait for, or not, accordingly.
Closes PR#8996.
|
| 1.35 | 30-Mar-2000 |
augustss | Remove register declarations.
|
| 1.34 | 15-Dec-1999 |
perseant | Fix error returns on lfs vnops so that locks and reference counts are
preserved. Handle dirop accounting in lfs_vfree for this case as well.
May address PR#8823.
|
| 1.33 | 03-Dec-1999 |
perseant | Handle the case of a vnode flush while dirops are active correctly in
lfs_segwrite. Also, make sure a flush is called in SET_DIROP before sleeping
on its results. Addresses PR #8863.
|
| 1.32 | 15-Nov-1999 |
fvdl | Add Kirk McKusick's soft updates code to the trunk. Not enabled by
default, as the copyright on the main file (ffs_softdep.c) is such
that is has been put into gnusrc. options SOFTDEP will pull this
in. This code also contains the trickle syncer.
Bump version number to 1.4O
|
| 1.31 | 06-Nov-1999 |
perseant | branches: 1.31.2;
Address ufs_hashlock/ufs_ihashins protocol bug, discovered while doing a
post-mortem of a production machine. Also, take the active dirop
count off of the fs and make it global (since it is measuring a global
resource) and tie the threshold value LFS_MAXDIROP to desiredvnodes.
|
| 1.30 | 05-Nov-1999 |
perseant | Better fix for PR# 8577: before setting dirops, check for cross-device
rename and error out. This avoids possible problems with attempting
rename between two LFSs.
|
| 1.29 | 01-Nov-1999 |
perseant | Check that the destination vnode is on an LFS before trying to twiddle
its superblock. Fixes PR#8577.
|
| 1.28 | 03-Sep-1999 |
perseant | branches: 1.28.2; 1.28.4; 1.28.6;
Make changes that will allow an LFS filesystem to be used as the root
filesystem. In particular,
- Fix mknod deadlock, described in PR 8172.
- Enable lfs_mountroot.
- Make lfs_writevnodes treat filesystems mounted on lfs device nodes properly,
by flushing that device rather than trying to add blocks to the device inode.
This, in combination with lfs boot blocks, will allow operation of an all-lfs
system.
|
| 1.27 | 03-Aug-1999 |
wrstuden | Add support for fcntl(2) to generate VOP_FCNTL calls. Any fcntl
call with F_FSCTL set and F_SETFL calls generate calls to a new
fileop fo_fcntl. Add genfs_fcntl() and soo_fcntl() which return 0
for F_SETFL and EOPNOTSUPP otherwise. Have all leaf filesystems
use genfs_fcntl().
Reviewed by: thorpej
Tested by: wrstuden
|
| 1.26 | 12-Apr-1999 |
perseant | Disallow threshold-initiated cache flush when dirops are active. Also, make
SET_ENDOP use lfs_check instead of inlining most of it.
|
| 1.25 | 29-Mar-1999 |
perseant | branches: 1.25.2;
Fix the other missing dirop wakeup
|
| 1.24 | 25-Mar-1999 |
perseant | Since dirop vnodes can't be flushed, they hold a reference until their
dirop is completely written to disk. This means that ordinary calls to
ufs vnops which would ordinarily call VOP_INACTIVE through vrele/vput,
don't. This patch detects that condition after such vnops have been
run, and calls VOP_INACTIVE if it would ordinarily have been called by
the ufs call.
|
| 1.23 | 25-Mar-1999 |
perseant | clean up unused/required #ifdefs
|
| 1.22 | 10-Mar-1999 |
perseant | New sources should leave the LFS in a more-or-less working state. Changes
include:
- DIROP segregation is enabled, and greater care is taken
to make sure that a checkpoint completes. Fsck is not
needed to remount the filesystem.
- Several checks to make sure that the LFS subsystem does not
overuse various resources (memory, in particular).
- The cleaner routines, lfs_markv in particular, are completely
rewritten. A buffer overflow is removed. Greater care is taken
to ensure that inodes come from where lfs_cleanerd say they come
from (so we know nothing has changed since lfs_bmapv was called).
- Fragment allocation is fixed, so that writes beyond end-of-file
do the right thing.
|
| 1.21 | 05-Mar-1999 |
mycroft | Pass null pointers to VOP_UPDATE rather than having all the callers fetch the
current time themselves.
|
| 1.20 | 06-Nov-1998 |
cgd | argument to dbtob needs to be cast to u_quad_t here to avoid shift lossage
|
| 1.19 | 01-Sep-1998 |
thorpej | Use the pool allocator and the "nointr" pool page allocator for LFS inodes.
|
| 1.18 | 24-Jun-1998 |
sommerfe | Always include fifos; "not an option any more".
|
| 1.17 | 22-Jun-1998 |
sommerfe | defopt for options FIFO
|
| 1.16 | 05-Jun-1998 |
kleink | Convert fsync vnode operator implementations and usage from the old `waitfor'
argument and MNT_WAIT/MNT_NOWAIT to `flags' and FSYNC_WAIT.
|
| 1.15 | 01-Mar-1998 |
fvdl | Merge with Lite2 + local changes
|
| 1.14 | 11-Jun-1997 |
bouyer | Add support for ext2fs, this needed a few modifications to ufs/ufs/inode.h:
- added an "union inode_ext" to struct inode, for the per-fs extentions.
For now only ext2fs uses it.
- i_din is now an union:
union {
struct dinode ffs_din; /* 128 bytes of the on-disk dinode. */
struct ext2fs_dinode e2fs_din; /* 128 bytes of the on-disk dinode. */
} i_din
Added a lot of #define i_ffs_* and i_e2fs_* to access the fields.
- Added two macros: FFS_ITIMES and EXT2FS_ITIMES. ITIMES calls the rigth
macro, depending on the time of the inode. ITIMES is used where necessary,
FFS_ITIMES and EXT2FS_ITIMES in other places.
|
| 1.13 | 07-Sep-1996 |
mycroft | Implement poll(2).
|
| 1.12 | 01-Sep-1996 |
mycroft | Add a set of generic file system operations that most file systems use.
Also, fix some time stamp bogosities.
|
| 1.11 | 11-May-1996 |
mycroft | Change VOP_UPDATE() semantics:
* Make 2nd and 3rd args timespecs, not timevals.
* Consistently pass a Boolean as the 4th arg (except in LFS).
Also, fix ffs_update() and lfs_update() to actually change the nsec fields.
|
| 1.10 | 09-Feb-1996 |
christos | lfs prototypes
|
| 1.9 | 09-Feb-1996 |
mycroft | Fix vop_link, vop_symlink, and vop_remove semantics in several ways:
* Change the argument names to vop_link so they actually make sense.
* Implement vop_link and vop_symlink for all file systems, so they do proper
cleanup.
* Require the file system to decide whether or not linking and unlinking of
directories is allowed, and disable it for all current file systems.
|
| 1.8 | 01-Feb-1996 |
jtc | Rename struct timespec fields to conform to POSIX.1b
|
| 1.7 | 15-Jun-1995 |
cgd | compensate for timeval/timespec/stat structure changes.
|
| 1.6 | 14-Dec-1994 |
mycroft | Sync with CSRG.
|
| 1.5 | 13-Dec-1994 |
mycroft | Not ready for part of the previous change yet...
|
| 1.4 | 13-Dec-1994 |
mycroft | Turn lease_check() into a vnode op, per CSRG.
|
| 1.3 | 20-Oct-1994 |
cgd | update for new syscall args description mechanism, and deal safely
with wider types.
|
| 1.2 | 29-Jun-1994 |
cgd | New RCS ID's, take two. they're more aesthecially pleasant, and use 'NetBSD'
|
| 1.1 | 08-Jun-1994 |
mycroft | branches: 1.1.1;
Update to 4.4-Lite fs code, with local changes.
|
| 1.1.1.2 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite2
|
| 1.1.1.1 | 01-Mar-1998 |
fvdl | Import 4.4BSD-Lite for reference
|
| 1.25.2.6 | 15-Jan-2000 |
he | Pull up revision 1.34 (requested by perseant):
Fix error returns on lfs vnops so that locks and reference counts
are preserved. Handle dirop accounting in lfs_vfree for this
case as well. Addresses PR#8823.
|
| 1.25.2.5 | 15-Jan-2000 |
he | Pull up revision 1.28 (requested by perseant):
Address problems related to using an LFS filesystem as the root
filesystem, including mknod hangs. Fixes PR#8172 and PR#9072.
|
| 1.25.2.4 | 18-Dec-1999 |
he | Pull up revision 1.33 (requested by perseant):
Handle the case of a vnode flush while dirops are active correctly
in lfs_segwrite. Also, make sure a flush is called in SET_DIROP
before sleeping on its results. Addresses PR#8863.
|
| 1.25.2.3 | 17-Dec-1999 |
he | Pull up revision 1.31 (requested by perseant):
Address locking protocol error for inode hash, and make the
maximum number of active dirops a global quantity.
|
| 1.25.2.2 | 08-Nov-1999 |
cgd | pull up revs 1.29-1.30 from trunk (requested by perseant):
Check for cross-device rename before setting up dirop markers in
lfs_rename. Addresses PR#8577.
|
| 1.25.2.1 | 13-Apr-1999 |
perseant | branches: 1.25.2.1.2;
Pull-up of changes made to the trunk on Sunday [1.25->1.26], to wit:
Take out the `#ifdef USE_UFSHASH'; use ufs_hashlock to lock the inode free
list instead of free_lock.
Fix inode reporting in lfs_statfs (the meaning of f_files and f_ffree was
reversed).
Fix "lfs_ifind: dinode xxx not found" panic. When inodes were freed, then
immediately reloaded, their dinodes were located in an inode block which
was not on disk at the advertized location, nor in the cache (although it
would be flushed to disk next segment write). Fix this by using getblk()
instead of lfs_newbuf() for inode blocks.
Better checking for held inode locks in lfs_fastvget, for a number of
error conditions. Also change the default setting of lfs_clean_vnhead to
0, which seems to make the locking problems go away (although this is
difficult to test as I can't reliably reproduce them).
Make sure that the wakeup occurs for vnodes that lfs_update might be
sleeping on (nodes which are not marked IN_MODIFIED/IN_CLEANING, but which
have dirty buffers), by marking them with the appropriate flag if
dirtybuffers were added while the write was in progress.
Fix block counting during file truncation, if not truncating to zero.
Disallow threshold-initiated cache flush when dirops are active. Also,
make SET_ENDOP use lfs_check instead of inlining most of it.
Improve the debugging printfs in the cleaner syscalls (in particular, make
it obvious that they're coming from lfs).
Check the superblock version field, and refuse to mount the filesystem if
the version number is higher than we know about. This allows, e.g.,
changes in the format of the ifile, segment size restrictions and
boundaries, etc., which would not affect existing fields in the
superblock, but which would drastically affect the filesystem, to be
smoothly integrated at a later date.
|
| 1.25.2.1.2.5 | 31-Aug-1999 |
perseant | Rudimentary support for LFS under UBC:
- LFS-specific VOP_BALLOC and VOP_PUTPAGES vnode ops.
- getblk VREG panic #ifdef'd out (can be reinstated when Ifile is
internalized and Ifile can be made another type from VREG)
- interface to VOP_PUTPAGES changed to pass all pager flags, not
just sync. FS putpages routines must know about the pager flags.
- new LFS magic disk address, -2 ("unwritten"), meaning accounted for
but not assigned to a fixed disk location (since LFS does these two
things separately, and the previous accounting method using buffer
headers no longer will work). Changed references to (foo == (daddr_t)-1)
to (foo < 0). Since disk drivers reject all addresses < 0, this should
not present a problem for other FSs.
|
| 1.25.2.1.2.4 | 11-Jul-1999 |
chs | add placeholders for getpages/putpages.
|
| 1.25.2.1.2.3 | 02-Jul-1999 |
thorpej | Take two at making a non-converted LFS work in a UBC kernel.
|
| 1.25.2.1.2.2 | 21-Jun-1999 |
thorpej | Pull in ffs_extern.h to get ffs_balloc_range() prototype for
ufs_readwrite.c
|
| 1.25.2.1.2.1 | 21-Jun-1999 |
thorpej | Sync w/ -current.
|
| 1.28.6.2 | 27-Dec-1999 |
wrstuden | Pull up to last week's -current.
|
| 1.28.6.1 | 21-Dec-1999 |
wrstuden | Initial commit of recent changes to make DEV_BSIZE go away.
Runs on i386, needs work on other arch's. Main kernel routines should be
fine, but a number of the stand programs need help.
cd, fd, ccd, wd, and sd have been updated. sd has been tested with non-512
byte block devices. vnd, raidframe, and lfs need work.
Non 2**n block support is automatic for LKM's and conditional for kernels
on "options NON_PO2_BLOCKS".
|
| 1.28.4.2 | 15-Nov-1999 |
fvdl | Sync with -current
|
| 1.28.4.1 | 19-Oct-1999 |
fvdl | Bring in Kirk McKusick's FFS softdep code on a branch.
|
| 1.28.2.3 | 11-Feb-2001 |
bouyer | Sync with HEAD.
|
| 1.28.2.2 | 22-Nov-2000 |
bouyer | Sync with HEAD.
|
| 1.28.2.1 | 20-Nov-2000 |
bouyer | Update thorpej_scsipi to -current as of a month ago
|
| 1.31.2.2 | 06-Nov-1999 |
perseant | Address ufs_hashlock/ufs_ihashins protocol bug, discovered while doing a
post-mortem of a production machine. Also, take the active dirop
count off of the fs and make it global (since it is measuring a global
resource) and tie the threshold value LFS_MAXDIROP to desiredvnodes.
|
| 1.31.2.1 | 06-Nov-1999 |
perseant | file lfs_vnops.c was added on branch comdex-fall-1999 on 1999-11-06 20:33:07 +0000
|
| 1.37.2.1 | 22-Jun-2000 |
minoura | Sync w/ netbsd-1-5-base.
|
| 1.38.2.6 | 03-Feb-2001 |
he | Pull up revisions 1.47-1.49 (requested by perseant):
o Don't write anything if the filesystem is idle (PR#10979).
o Close up accounting holes in LFS' accounting of immediately-
available-space, number of clean segments, and amount of dirty
space taken up by metadata (PR#11468, PR#11470, PR#11534).
|
| 1.38.2.5 | 14-Dec-2000 |
he | Pull up revision 1.45 (requested by fvdl):
Improve NFS performance, possibly with as much as 100% in
throughput. Please note: this implies a kernel interface change,
VOP_FSYNC gains two arguments.
|
| 1.38.2.4 | 01-Nov-2000 |
tv | Fix pullup of 1.46 [perseant, toshii]:
In lfs_truncate, don't overcount the real blocks removed from the inode,
when deallocating a fragment that has not made it to disk yet.
Also, during dirops, give the directory vnode an extra reference in
SET_DIROP, to ensure its continued existence during SET_ENDOP, preventing
a possible NULL-dereference there.
These two changes should close PR #11064.
|
| 1.38.2.3 | 01-Nov-2000 |
tv | Pullup 1.46 [perseant, toshii]:
In lfs_truncate, don't overcount the real blocks removed from the inode,
when deallocating a fragment that has not made it to disk yet.
Also, during dirops, give the directory vnode an extra reference in
SET_DIROP, to ensure its continued existence during SET_ENDOP, preventing
a possible NULL-dereference there.
These two changes should close PR #11064.
|
| 1.38.2.2 | 14-Sep-2000 |
perseant | Pull up recent LFS kernel changes (approved by thorpej):
ufs/ufs/inode.h, 1.20--1.22 (add i_lfs_effnblks extension ;
make ITIMES aware of LFS_ITIMES;
_LKM protection so userland progs
compile)
ufs/ufs/ufs_vnops.c, 1.69, 1.71 (remove IN_ADIROP;
use ITIMES instead of FFS_ITIMES)
ufs/ufs/ufs_readwrite.c, 1.27 (use lfs_reserve in lfs_write)
ufs/lfs/lfs.h, 1.26--1.32 (define LFS_EST_* macros ;
change MIN_FREE_SEGS to lfs_minfreesegs ;
add avail and bfree to CLEANERINFO ;
change lfs_uinodes to signed ;
change lfs_dmeta to signed ;
add whitespace to line up structure
members ;
explicit cast to int32_t in LFS_EST_*
macros)
ufs/lfs/lfs_alloc.c, back out 1.34.2.3 (pullups of 1.39, 1.40);
then pull up 1.38 (clean up on error)
1.39--1.43 (restore fvdl's ufs_hashlock fix ;
restore fvdl's ufs_hashlock fix ;
set i_lfs_effnblks ;
use UINO macros ;
add comments and fix long lines)
ufs/lfs/lfs_balloc.c, 1.19 (don't succeed halfway)
1.21--1.25 (use i_lfs_effnblks ;
fix i_lfs_effnblks computation and
quieten ;
fix i_ffs_blocks in unwritten fragment ;
remove useless debugging check ;
add comments and (c) 2000)
ufs/lfs/lfs_bio.c, 1.24--1.30 (cleanup and make lfs_flush_fs take
"struct lfs *" instead of "struct
mount *" ;
use lfs_minfreeseg instead of
MIN_FREE_SEGS ;
use UINO macros, and copy bfree/avail
to CLEANERINFO ;
add lfs_reserve function ;
1.28--1.30 fix printf formatting)
ufs/lfs/lfs_cksum.c, 1.13 (add (c) 2000)
ufs/lfs/lfs_debug.c, 1.11 (use btodb instead of DEV_BSIZE)
ufs/lfs/lfs_extern.h, 1.18, 1.20--1.21 (function prototype changes)
ufs/lfs/lfs_inode.c, 1.38 (rewrite lfs_truncate from
ffs_truncate)
1.40--1.44 (count written and unwritten blocks
seperately ;
use disk block units instead of bytes ;
remove unnecessary "mod" variable ;
correct B_DELWRI to avoid bawrite panic ;
use lfs_reserve)
ufs/lfs/lfs_segment.c, 1.52-1.59 (use lfs_dmeta to note used summaries ;
check for UNWRITTEN in indirect blocks ;
more debugging stuff inside #ifdef
DEBUG_LFS ;
use LK_CANRECURSE ;
don't drop dirty indirect blocks ;
use UINO macros ;
don't hose the free list ;
use btodb() instead of DEV_BSIZE ;
make it compile again (oops))
ufs/lfs/lfs_subr.c, 1.16--1.17 (check for locked inodes before
changing ;
use btodb() instead of DEV_BSIZE, (c)
2000)
ufs/lfs/lfs_syscalls.c, back out 1.41.4.2 (fvdl's ufs_hashlock fix);
then pull up 1.43 (use lfs_dmeta)
1.44--1.45 (restore fvdl's ufs_hashlock fix)
1.46--1.47 (fix lfs_avail leakage from sblock
segments ;
use UINO macros)
1.49 (bounds-check inode numbers in
lfs_markv)
ufs/lfs/lfs_vfsops.c, 1.53 (use LFS_EST_* macros in lfs_statfs)
1.56--1.58 (initialize lfs_minfreeseg, lfs_effnblk ;
initialize lfs_uinodes ;
initialize lfs_ravail)
ufs/lfs/lfs_vnops.c, 1.40 (remove VDIROP from removed files)
1.42--1.44 (move SET_ENDOP below the removal of
VDIROP ;
use UINO macros and add lfs_itimes
function ;
use lfs_reserve in dirops)
|
| 1.38.2.1 | 22-Jun-2000 |
perseant | Pull up lfs_vunref fix from the trunk.
|
| 1.50.6.9 | 10-Oct-2002 |
jdolecek | sync kqueue with -current; this includes merge of gehenna-devsw branch,
merge of i386 MP branch, and part of autoconf rototil work
|
| 1.50.6.8 | 26-Sep-2002 |
jdolecek | hook in genfs_kqfilter(), kevents seem to work fine
|
| 1.50.6.7 | 23-Sep-2002 |
jdolecek | add spec kqfilter vnode op
|
| 1.50.6.6 | 22-Sep-2002 |
jdolecek | add fifo_kqfilter() to fifo ops, to switch on support for kevents
|
| 1.50.6.5 | 23-Jun-2002 |
jdolecek | catch up with -current on kqueue branch
|
| 1.50.6.4 | 16-Mar-2002 |
jdolecek | Catch up with -current.
|
| 1.50.6.3 | 10-Jan-2002 |
thorpej | Sync kqueue branch with -current.
|
| 1.50.6.2 | 25-Aug-2001 |
thorpej | Merge Aug 24 -current into the kqueue branch.
|
| 1.50.6.1 | 03-Aug-2001 |
lukem | update to -current
|
| 1.50.4.2 | 02-Jul-2001 |
perseant | Change disk addressing unit to be the fragment, instead of the disk sector.
All quantities in the superblock, inodes, indirect blocks, etc. refer now
to this abstract unit (called "fsb" as it is in FFS) instead of disk sectors;
as a consequence segment summary blocks have to be multiples of a fragment in
size. In v1 filesystems, compatibility code ensures that 1 fsb == 1 sector,
regardless of fragment size.
Fragments can now range in size between 512 and 32k; in the event that
LFS_LABELPAD (8k) is smaller than the disk address unit size, an extra
proto-superblock is kept at 8k from the beginning of the disk, to be used
*only* to locate the real superblocks. (Not all of the userland knows about
this yet.)
Almost all of this was done not by me, but by joff.
|
| 1.50.4.1 | 29-Jun-2001 |
perseant | Get rid of __P(), protoizing where it had not already been done
|
| 1.50.2.15 | 08-Jan-2003 |
thorpej | Sync with HEAD.
|
| 1.50.2.14 | 08-Jan-2003 |
thorpej | Sync with HEAD.
|
| 1.50.2.13 | 29-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.50.2.12 | 11-Dec-2002 |
thorpej | Sync with HEAD.
|
| 1.50.2.11 | 11-Nov-2002 |
nathanw | Catch up to -current
|
| 1.50.2.10 | 18-Oct-2002 |
nathanw | Catch up to -current.
|
| 1.50.2.9 | 24-Jun-2002 |
nathanw | Curproc->curlwp renaming.
Change uses of "curproc->l_proc" back to "curproc", which is more like the
original use. Bare uses of "curproc" are now "curlwp".
"curproc" is now #defined in proc.h as ((curlwp) ? (curlwp)->l_proc) : NULL)
so that it is always safe to reference curproc (*de*referencing curproc
is another story, but that's always been true).
|
| 1.50.2.8 | 20-Jun-2002 |
nathanw | Catch up to -current.
|
| 1.50.2.7 | 28-Feb-2002 |
nathanw | Catch up to -current.
|
| 1.50.2.6 | 08-Jan-2002 |
nathanw | Catch up to -current.
|
| 1.50.2.5 | 14-Nov-2001 |
nathanw | Catch up to -current.
|
| 1.50.2.4 | 26-Sep-2001 |
nathanw | Catch up to -current.
Again.
|
| 1.50.2.3 | 21-Sep-2001 |
nathanw | Catch up to -current.
|
| 1.50.2.2 | 24-Aug-2001 |
nathanw | Catch up with -current.
|
| 1.50.2.1 | 05-Mar-2001 |
nathanw | Initial commit of scheduler activations and lightweight process support.
|
| 1.54.2.1 | 01-Oct-2001 |
fvdl | Catch up with -current.
|
| 1.56.2.1 | 12-Nov-2001 |
thorpej | Sync the thorpej-mips-cache branch with -current.
|
| 1.63.2.2 | 20-Jun-2002 |
gehenna | catch up with -current.
|
| 1.63.2.1 | 30-May-2002 |
gehenna | Catch up with -current.
|
| 1.64.2.1 | 20-Jun-2002 |
lukem | Pull up revision 1.65 (requested by perseant in ticket #325):
For synchronous writes, keep separate i/o counters for each write, so
processes don't have to wait for one another to finish (e.g., nfsd seems
to be a little happier now, though I haven't measured the difference).
Synchronous checkpoints, however, must always wait for all i/o to finish.
Take the contents of the callback functions and have them run in thread
context instead (aiodoned thread). lfs_iocount no longer has to be
protected in splbio(), and quite a bit less of the segment construction
loop needs to be in splbio() as well.
If lfs_markv is handed a block that is not the correct size according to
the inode, refuse to process it. (Formerly it was extended to the "correct"
size.) This is possibly more prone to deadlock, but less prone to corruption.
lfs_segclean now outright refuses to clean segments that appear to have live
bytes in them. Again this may be more prone to deadlock but avoids
corruption.
Replace ufsspec_close and ufsfifo_close with LFS equivalents; this means
that no UFS functions need to know about LFS_ITIMES any more. Remove
the reference from ufs/inode.h.
Tested on i386, test-compiled on alpha.
|
| 1.109.2.11 | 10-Nov-2005 |
skrll | Sync with HEAD. Here we go again...
|
| 1.109.2.10 | 01-Apr-2005 |
skrll | Sync with HEAD.
|
| 1.109.2.9 | 08-Mar-2005 |
skrll | Sync with HEAD.
|
| 1.109.2.8 | 04-Mar-2005 |
skrll | Sync with HEAD.
Hi Perry!
|
| 1.109.2.7 | 04-Feb-2005 |
skrll | Sync with HEAD.
|
| 1.109.2.6 | 30-Oct-2004 |
skrll | Reduced diff to HEAD by restoring the struct proc * argument to lfs_bmapv
|
| 1.109.2.5 | 21-Sep-2004 |
skrll | Fix the sync with head I botched.
|
| 1.109.2.4 | 18-Sep-2004 |
skrll | Sync with HEAD.
|
| 1.109.2.3 | 24-Aug-2004 |
skrll | Undo part of the ktrace/lwp changes. In particular:
* Remove the "lwp *" argument that was added to vget(). Turns out
that nothing actually used it!
* Remove the "lwp *" arguments that were added to VFS_ROOT(), VFS_VGET(),
and VFS_FHTOVP(); all they did was pass it to vget() (which, as noted
above, didn't use it).
* Remove all of the "lwp *" arguments to internal functions that were added
just to appease the above.
|
| 1.109.2.2 | 03-Aug-2004 |
skrll | Sync with HEAD
|
| 1.109.2.1 | 02-Jul-2003 |
darrenr | Apply the aborted ktrace-lwp changes to a specific branch. This is just for
others to review, I'm concerned that patch fuziness may have resulted in some
errant code being generated but I'll look at that later by comparing the diff
from the base to the branch with the file I attempt to apply to it. This will,
at the very least, put the changes in a better context for others to review
them and attempt to tinker with removing passing of 'struct lwp' through
the kernel.
|
| 1.129.4.1 | 10-May-2005 |
riz | Pull up the following revisions (requested by perseant in ticket #1281):
1.8 sys/ufs/lfs/TODO
1.75 sys/ufs/lfs/lfs.h (via patch)
1.74 sys/ufs/lfs/lfs_alloc.c (via patch)
1.49, 1.51 sys/ufs/lfs/lfs_balloc.c (1.51 via patch)
1.78 sys/ufs/lfs/lfs_bio.c
1.62 sys/ufs/lfs/lfs_extern.h (via patch)
1.156 sys/ufs/lfs/lfs_segment.c (via patch)
1.48 sys/ufs/lfs/lfs_subr.c
1.101 sys/ufs/lfs/lfs_syscalls.c
1.163 sys/ufs/lfs/lfs_vfsops.c (via patch)
1.134 sys/ufs/lfs/lfs_vnops.c (via patch)
1.61 sys/ufs/ufs/ufs_readwrite.c (via patch)
1.20 libexec/lfs_cleanerd/clean.h (via patch)
1.52 libexec/lfs_cleanerd/cleanerd.c (via patch)
1.41 libexec/lfs_cleanerd/library.c (via patch)
1.4 regress/sys/fs/lfs/newfs_fsck/Makefile
1.2 regress/sys/fs/lfs/newfs_fsck/mkfs_mount
1.2 regress/sys/fs/lfs/newfs_fsck/smallfiles
1.3 sbin/fsck_lfs/bufcache.c
1.3 sbin/fsck_lfs/bufcache.h
1.3 sbin/fsck_lfs/lfs.h
1.8 sbin/fsck_lfs/lfs.c (via patch)
1.8 sbin/fsck_lfs/pass3.c (via patch)
1.18 sbin/fsck_lfs/pass0.c (via patch)
1.18 sbin/fsck_lfs/utilities.c (via patch)
1.7 sbin/fsck_lfs/segwrite.c
1.19 sbin/fsck_lfs/setup.c (via patch)
1.3 sbin/newfs_lfs/Makefile
0 sbin/newfs_lfs/lfs.c (yes, remove it)
1.1 sbin/newfs_lfs/make_lfs.c
1.15 sbin/newfs_lfs/newfs.c (via patch)
Various minor LFS improvements.
Kernel:
* Note when lfs_putpages(9) thinks it is not going to be writing any
pages before calling genfs_putpages(9). This prevents a situation in
which blocks can be queued for writing without a segment header.
* Correct computation of NRESERVE(), though it is still a gross
overestimate in most cases. Note that if NRESERVE() is too high, it
may be impossible to create files on the filesystem. We catch this
case on filesystem mount and refuse to mount r/w.
* Allow filesystems to be mounted whose block size is == MAXBSIZE.
* Somewhere along the line, ufs_bmaparray(9) started mangling UNWRITTEN
entries in indirect blocks again, triggering a failed assertion "daddr
<= LFS_MAX_DADDR". Explicitly convert to and from int32_t to correct
this. Should fix PR #29045.
* Add a high-water mark for the number of dirty pages any given LFS can
hold before triggering a flush. This is settable by sysctl, but off
(zero) by default.
* Be more careful about the MAX_BYTES and MAX_BUFS computations so we
shouldn't see "please increase to at least zero" messages.
* Note that VBLK and VCHR vnodes can have nonzero values in di_db[0]
even though their v_size == 0. Don't panic when we see this.
Fixes PR #26680.
* Change lfs_bfree to a signed quantity. The manner in which it is
processed before being passed to the cleaner means that sometimes it
may drop below zero, and the cleaner must be aware of this.
* Never report bfree < 0 (or higher than lfs_dsize) through
lfs_statfs(9). This prevents df(1) from ever telling us that our full
filesystems have 16TB free.
* Account space allocated through lfs_balloc(9) that does not have
associated buffer headers, so that the pagedaemon doesn't run us out
of segments.
* Return ENOSPC from lfs_balloc(9) when bfree drops to zero.
* Address a deadlock in lfs_bmapv/lfs_markv when the filesystem is being
unmounted. Because vfs_busy() is a shared lock, and
lfs_bmapv/lfs_markv mark the filesystem vfs_busy(), the cleaner can be
holding the lock that umount() is blocking on, then try to vfs_busy()
again in getnewvnode().
cleaner:
* Adapt lfs_cleanerd to use the fcntl call to get the Ifile filehandle,
so it need not be in the namespace.
* Make lfs_cleanerd be more careful when there are very few available
segments.
* Make lfs_cleanerd less verbose when the filesystem is unmounted.
newfs_lfs, fsck_lfs, and regression:
* Extend the lfs library from fsck_lfs(8) so that it can be used with a
not-yet-existent LFS. Make newfs_lfs(8) use this library, so it can
create LFSs whose Ifile is larger than one segment. Addresses PR #11110.
* Make newfs_lfs(8) use strsuftoi64() for its arguments, a la newfs(8).
* Make fsck_lfs(8) respect the "file system is clean" flag.
* Don't let fsck_lfs(8) think it has dirty blocks when invoked with the
-n flag.
* Remove the Ifile from the filesystem namespace. The cleaner now uses
a fcntl call on the root inode to find the Ifile filehandle. (As a
side-effect, addresses PR #29144.)
|
| 1.132.6.3 | 26-Mar-2005 |
yamt | sync with head.
|
| 1.132.6.2 | 19-Mar-2005 |
yamt | sync with head. xen and whitespace. xen part is not finished.
|
| 1.132.6.1 | 12-Feb-2005 |
yamt | sync with head.
|
| 1.132.4.1 | 29-Apr-2005 |
kent | sync with -current
|
| 1.137.2.25 | 10-Aug-2006 |
tron | Apply patch (requested by fair in perseant #1457):
Bring LFS up to current, including a patch (1.95 lfs_alloc.c) that
should prevent the inode free list errors seen on the STABLE branch
subsequent to pullup ticket #1327.
|
| 1.137.2.24 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.177
Don't be quite so eager to error out from lfs_putpages() when pages are
busy; if we've sensed a possible 3-way deadlock and are not the pagedaemon,
relock and try again.
|
| 1.137.2.23 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.93
sys/ufs/lfs/lfs.h: revision 1.106
sys/ufs/lfs/lfs_vfsops.c: revision 1.209
sys/ufs/lfs/lfs_vnops.c: revision 1.175
sys/ufs/lfs/lfs_segment.c: revision 1.178
Fixes to address the "vinvalbuf: dirty blocks" panic that can occur when
many inodes are cleaned at once. Make sure that we write all the pages
on vnodes that are being flushed, even if we don't think there's room;
drain v_numoutput before lfs_vflush() completes.
Also, don't allow a vnode that is in the process of being cleaned to be
chosen by getnewvnode(); this avoids a segment accounting panic in the case
that a large number of inodes are fed to lfs_markv() all at once.
|
| 1.137.2.22 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_alloc.c: revision 1.92
sys/ufs/lfs/lfs.h: revision 1.105
sys/ufs/lfs/lfs_vfsops.c: revision 1.207
sys/ufs/lfs/lfs_subr.c: revision 1.59
sys/ufs/lfs/lfs_vnops.c: revision 1.173
sys/ufs/lfs/lfs_bio.c: revision 1.92
Introduce another per-filesystem parameter, lfs_resvseg, to separate the
notion of "how many segments are reserved for the cleaner" from that of
"how many segments are not counted in lfs_bfree". The default value
used for existing filesystems is the same as the previous implicit value
of (lfs_minfreeseg / 2 + 1), modulo some sanity checking.
Count pending dirops on a per-filesystem basis, since once we start
writing them we can't stop until we're done. This seems to help stave off
the "no clean segments" panic in the case of filling the filesystem with
directories and small files (e.g. simultaneously unpacking more copies of
pkgsrc than will fit).
|
| 1.137.2.21 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.172
Fix a "locking against myself": lfs_flush_dirops() doesn't need to lock the
vnodes to write their blocks, since it holds the segment lock.
|
| 1.137.2.20 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.171
sys/ufs/lfs/lfs_extern.h: revision 1.81
sys/ufs/lfs/lfs_segment.c: revision 1.177
Don't ever partially write dirops, even if we need the cleaner to run.
This increases the chances of the "no clean segments" panic slightly,
but allows us to run the ckckp regression test successfully to completion.
|
| 1.137.2.19 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.104
sys/ufs/lfs/lfs_vfsops.c: revision 1.206
sys/ufs/lfs/lfs_vnops.c: revision 1.170
sys/ufs/lfs/lfs_extern.h: revision 1.80
sys/ufs/lfs/lfs_segment.c: revision 1.176
sys/ufs/lfs/lfs_inode.c: revision 1.103 via patch
sys/ufs/lfs/lfs_alloc.c: revision 1.90
Postpone the segment accounting changes coming from truncation until the
inode that makes those changes valid is either written to disk by
lfs_writeinode() or discarded by lfs_vfree().
A couple of locking fixes are also included as well.
|
| 1.137.2.18 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.169
Yet another MP locking issue.
|
| 1.137.2.17 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.103
sys/ufs/lfs/lfs_segment.c: revision 1.174
sys/ufs/lfs/lfs_vnops.c: revision 1.168
Introduce two fcntl calls that freeze the filesystem right at the point
where segment 0 is being considered for writing. This allows for automated
checkpoint vailidity scanning, and could be used (in conjunction with the
existing LFCNREWIND) for e.g. snapshot dumps as well.
Include a regression test that does such scanning.
When writing the Ifile, loop through the dirty block list three times to
make sure that the checkpoint is always consistent (the first and second
times the Ifile blocks can cross a segment boundary; not so the third time
unless the segments are very small). Discovered by using the aforementioned
regression test.
|
| 1.137.2.16 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs.h: revision 1.102
sys/ufs/lfs/lfs_segment.c: revision 1.173
sys/ufs/lfs/lfs_vnops.c: revision 1.167 via patch
sys/ufs/lfs/lfs_bio.c: revision 1.91
Make lfs_vref/lfs_vunref not need to know about VXLOCK and VFREEING
explicitly (especially since we didn't know about VFREEING at all before),
but notice the EBUSY return from vget() instead.
Fix some more MP locking protocol issues, most of which were pointed out by
Christian Ehrhardt this morning on tech-kern.
|
| 1.137.2.15 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.166
Another MP locking fix.
|
| 1.137.2.14 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.165
Don't leak vnode references if we fail to lock a vnode in lfs_flush_pchain().
Also fix another (probably only academic) simple_lock protocol error.
|
| 1.137.2.13 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.200
sys/ufs/lfs/lfs_vnops.c: revision 1.164
sys/ufs/lfs/lfs_inode.c: revision 1.101
sys/ufs/lfs/lfs_extern.h: revision 1.78
sys/ufs/lfs/lfs.h: revision 1.100
Implement a somewhat finer-grained mechanism for paging LFS-backed pages.
The writer daemon, if it does not need to flush the whole filesystem,
now only writes the vnodes for which the pagedaemon has requested pageouts
(although it does not pay attention to the page ranges the pagedaemon
supplies).
|
| 1.137.2.12 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_balloc.c: revision 1.60
sys/ufs/lfs/lfs_syscalls.c: revision 1.111
sys/ufs/lfs/lfs_segment.c: revision 1.172
sys/ufs/lfs/lfs_vnops.c: revision 1.163
Several minor bug fixes:
* Correct (weak) segment lock assertions in lfs_fragextend and lfs_putpages.
* Keep IN_MODIFIED set if we run out of avail in lfs_putpages.
* Don't try to (re)write buffers on a VBLK vnode; fixes a panic I found
while running with an LFS root.
* Raise priority of LFCNSEGWAIT to PVFS; PUSER is way too low for
something the pagedaemon is relying on.
|
| 1.137.2.11 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.162
Make sure we unlock to zero when avoiding 3-way deadlock; otherwise we
simply have a different form of deadlock.
|
| 1.137.2.10 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vfsops.c: revision 1.198
sys/ufs/lfs/lfs_vnops.c: revision 1.161
Handle the "filesystem is clean" flag correctly when upgrading from
read-only to read-write mount. This makes "root on lfs" work for me,
although it looks like a different traceback from PR#32667.
|
| 1.137.2.9 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.159
Don't let the pagedaemon wait for pages, since that is just asking for
a deadlock.
|
| 1.137.2.8 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.158
sys/ufs/lfs/lfs_subr.c: revision 1.57
sys/ufs/lfs/lfs_segment.c: revision 1.171
sys/ufs/lfs/lfs.h: revision 1.97
sys/ufs/lfs/lfs_vfsops.c: revision 1.195
sys/ufs/lfs/lfs_extern.h: revision 1.76
Improvements to LFS's paging mechanism, to wit:
* Acknowledge that sometimes there are more dirty pages to be written to
disk than clean segments. When we reach the danger line,
lfs_gop_write() now returns EAGAIN. The caller of VOP_PUTPAGES(), if
it holds the segment lock, drops it and waits for the cleaner to make
room before continuing.
* Note and avoid a three-way deadlock in lfs_putpages (a writer holding
a page busy blocks on the cleaner while the cleaner blocks on the
segment lock while lfs_putpages blocks on the page).
|
| 1.137.2.7 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.153
sys/ufs/lfs/lfs_debug.c: revision 1.32
sys/ufs/lfs/lfs_alloc.c: revision 1.84
sys/ufs/lfs/lfs_vfsops.c: revision 1.185
sys/ufs/lfs/lfs_segment.c: revision 1.165
64 bit inode changes.
|
| 1.137.2.6 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.152
sys/ufs/lfs/lfs_debug.c: revision 1.31
sys/ufs/lfs/lfs_subr.c: revision 1.53
sys/ufs/lfs/lfs_extern.h: revision 1.68
sys/ufs/lfs/lfs_inode.c: revision 1.96
sys/ufs/lfs/lfs_bio.c: revision 1.86
sys/ufs/lfs/lfs_alloc.c: revision 1.83
sys/ufs/lfs/lfs_vfsops.c: revision 1.181
sys/ufs/lfs/lfs.h: revision 1.88
sys/ufs/lfs/lfs_segment.c: revision 1.164
- sprinkle const
- avoid shadow variables.
|
| 1.137.2.5 | 20-May-2006 |
riz | Pull up following revision(s) (requested by perseant in ticket #1327):
sys/ufs/lfs/lfs_vnops.c: revision 1.151
VOP_LOCK drops the interlock; pick it up again to avoid an "already unlocked"
panic in lfs_putpages.
|
| 1.137.2.4 | 07-May-2005 |
tron | Apply patch (requested by perseant in ticket #242):
* fsck_lfs buffer cache fixes, including PR #29151
* Change fsck_lfs phase 0 message to reflect reality
* fsck_lfs: check phase 5 (cleanerinfo accounting) even on
roll-forward
* Keep better track of the free list during roll-forward, avoiding
a core dump
* Improve hash table use for fsck_lfs buffer and vnode cache
* Document fsck_lfs flag -f, and implement -q
* Add resize_lfs, including kernel support
* Add LFS to mountd's list of exportable filesystem types
* Make the LFS lkm work again [christos@]
* Add MP locking to the LFS kernel subsystem
* Fix pager_map deadlock in lfs_putpages()
* Avoid incomplete file extension that looks like "partial
truncation" to fsck
* Use lfs_malloc for cleaner malloc, since the cleaner often runs
in low-memory conditions.
* Use splay trees, not hash table, to track page allocation for
write.
* Fix mkdir panic on full fs
* Fix page accounting leak by counting differently.
* Use rightly named structure for lfs_getattr [skrll@]
* Cosmetic changes for readability.
|
| 1.137.2.3 | 30-Mar-2005 |
tron | Pull up revision 1.140 (requested by perseant in ticket #74):
Don't sleep while holding the vnode interlock. Should take care of the
first panic case in PR #26043.
|
| 1.137.2.2 | 30-Mar-2005 |
tron | Pull up revision 1.139 (requested by perseant in ticket #74):
avoid the need for recursive locking lfs_flush_dirops() by unlocking
the vnode around the call to this in the caller.
|
| 1.137.2.1 | 30-Mar-2005 |
tron | Pull up revision 1.138 (requested by perseant in ticket #74):
Make LFS dirops get their vnode first, before incrementing the dirop
count, to prevent a deadlock trying to call VOP_PUTPAGES() on a VDIROP
vnode. This can happen when a stacked filesystem is mounted on top of an
LFS: an LFS dirop needs to get a vnode, which is available from the upper
layer. The corresponding lower layer vnode, however, is VDIROP, so the
upper layer can't be cleaned out since its VOP_PUTPAGES() is passed
through to the lower layer, which waits for dirops to drain before it can
proceed. Deadlock.
Tweak ufs_makeinode() and ufs_mkdir() to pass the a_vpp argument through
to VOP_VALLOC().
Partially addresses PR # 26043, though it probably does not completely fix
the problem described there.
|
| 1.152.2.8 | 04-Feb-2008 |
yamt | sync with head.
|
| 1.152.2.7 | 21-Jan-2008 |
yamt | sync with head
|
| 1.152.2.6 | 07-Dec-2007 |
yamt | sync with head
|
| 1.152.2.5 | 27-Oct-2007 |
yamt | sync with head.
|
| 1.152.2.4 | 03-Sep-2007 |
yamt | sync with head.
|
| 1.152.2.3 | 26-Feb-2007 |
yamt | sync with head.
|
| 1.152.2.2 | 30-Dec-2006 |
yamt | sync with head.
|
| 1.152.2.1 | 21-Jun-2006 |
yamt | sync with head.
|
| 1.155.2.2 | 29-Oct-2005 |
yamt | use lfs_* directly rather than via ufs_ops.
suggested by Chuck Silvers.
|
| 1.155.2.1 | 20-Oct-2005 |
yamt | adapt ufs.
|
| 1.157.12.3 | 24-May-2006 |
tron | Merge 2006-05-24 NetBSD-current into the "peter-altq" branch.
|
| 1.157.12.2 | 31-Mar-2006 |
tron | Merge 2006-03-31 NetBSD-current into the "peter-altq" branch.
|
| 1.157.12.1 | 28-Mar-2006 |
tron | Merge 2006-03-28 NetBSD-current into the "peter-altq" branch.
|
| 1.157.10.5 | 11-May-2006 |
elad | sync with head
|
| 1.157.10.4 | 06-May-2006 |
christos | - Move kauth_cred_t declaration to <sys/types.h>
- Cleanup struct ucred; forward declarations that are unused.
- Don't include <sys/kauth.h> in any header, but include it in the c files
that need it.
Approved by core.
|
| 1.157.10.3 | 19-Apr-2006 |
elad | sync with head.
|
| 1.157.10.2 | 10-Mar-2006 |
elad | generic_authorize() -> kauth_authorize_generic().
|
| 1.157.10.1 | 08-Mar-2006 |
elad | Adapt to kernel authorization KPI.
|
| 1.157.8.6 | 03-Sep-2006 |
yamt | sync with head.
|
| 1.157.8.5 | 11-Aug-2006 |
yamt | sync with head
|
| 1.157.8.4 | 26-Jun-2006 |
yamt | sync with head.
|
| 1.157.8.3 | 24-May-2006 |
yamt | sync with head.
|
| 1.157.8.2 | 11-Apr-2006 |
yamt | sync with head
|
| 1.157.8.1 | 01-Apr-2006 |
yamt | sync with head.
|
| 1.157.6.2 | 01-Jun-2006 |
kardel | Sync with head.
|
| 1.157.6.1 | 22-Apr-2006 |
simonb | Sync with head.
|
| 1.157.4.1 | 09-Sep-2006 |
rpaulo | sync with head
|
| 1.178.4.1 | 13-Jul-2006 |
gdamore | Merge from HEAD.
|
| 1.188.2.4 | 01-Feb-2007 |
ad | Sync with head.
|
| 1.188.2.3 | 30-Jan-2007 |
ad | Remove support for SA. Ok core@.
|
| 1.188.2.2 | 12-Jan-2007 |
ad | Sync with head.
|
| 1.188.2.1 | 18-Nov-2006 |
ad | Sync with head.
|
| 1.189.2.2 | 10-Dec-2006 |
yamt | sync with head.
|
| 1.189.2.1 | 22-Oct-2006 |
yamt | sync with head
|
| 1.193.2.3 | 25-Nov-2007 |
xtraeme | Pull up following revision(s) (requested by christos in ticket #994):
sys/ufs/lfs/lfs_vnops.c: revision 1.208 (patch)
Move the "vp = NULL" assignment after the code that requires vp != NULL.
Reported by Chris Ross on current-users.
|
| 1.193.2.2 | 05-Jun-2007 |
bouyer | Pull up following revision(s) (requested by perseant in ticket #703):
sys/miscfs/genfs/genfs.h 1.21
sys/miscfs/genfs/genfs_vnops.c 1.151
sys/ufs/lfs/lfs.h 1.119, 1.120
sys/ufs/lfs/lfs_bio.c 1.99-101
sys/ufs/lfs/lfs_extern.h 1.89
sys/ufs/lfs/lfs_inode.c 1.108, 1.109
sys/ufs/lfs/lfs_segment.c 1.197, 1.199, 1.200
sys/ufs/lfs/lfs_subr.c 1.69, 1.70
sys/ufs/lfs/lfs_syscalls.c 1.119
sys/ufs/lfs/lfs_vfsops.c 1.234, 1.235
sys/ufs/lfs/lfs_vnops.c 1.195, 1.196, 1.200, 1.202-206
Reduce busy waiting in lfs_putpages(), and other LFS improvements.
|
| 1.193.2.1 | 17-Feb-2007 |
tron | branches: 1.193.2.1.2;
Apply patch (requested by chs in ticket #422):
- Fix various deadlock problems with nullfs and unionfs.
- Speed up path lookups by upto 25%.
|
| 1.193.2.1.2.2 | 06-Jan-2008 |
wrstuden | Catch up to netbsd-4.0 release.
|
| 1.193.2.1.2.1 | 03-Sep-2007 |
wrstuden | Sync w/ NetBSD-4-RC_1
|
| 1.198.2.4 | 07-May-2007 |
yamt | sync with head.
|
| 1.198.2.3 | 15-Apr-2007 |
yamt | sync with head.
|
| 1.198.2.2 | 12-Mar-2007 |
rmind | Sync with HEAD.
|
| 1.198.2.1 | 27-Feb-2007 |
yamt | - sync with head.
- move sched_changepri back to kern_synch.c as it doesn't know PPQ anymore.
|
| 1.201.4.1 | 11-Jul-2007 |
mjf | Sync with head.
|
| 1.201.2.12 | 16-Sep-2007 |
ad | - Checkpoint work in progress on the vnode lifecycle and reference counting
stuff. This makes it work properly without kernel_lock and fixes a few
quite old bugs. See vfs_subr.c 1.283.2.17 for details.
- Fix some problems with softdep. Unfortunately our softdep code appears
to have some longstanding bugs that cause it fail under stress test.
|
| 1.201.2.11 | 20-Aug-2007 |
ad | Sync with HEAD.
|
| 1.201.2.10 | 19-Aug-2007 |
ad | - Back out the biodone() changes.
- Eliminate B_ERROR (from HEAD).
|
| 1.201.2.9 | 15-Jul-2007 |
ad | Sync with head.
|
| 1.201.2.8 | 23-Jun-2007 |
ad | - Lock v_cleanblkhd, v_dirtyblkhd, v_numoutput with the vnode's interlock.
Get rid of global_v_numoutput_lock. Partially incomplete as the buffer
cache locking doesn't work very well and needs an overhaul.
- Some changes to try and make softdep MP safe. Untested.
|
| 1.201.2.7 | 17-Jun-2007 |
ad | - Increase the number of thread priorities from 128 to 256. How the space
is set up is to be revisited.
- Implement soft interrupts as kernel threads. A generic implementation
is provided, with hooks for fast-path MD code that can run the interrupt
threads over the top of other threads executing in the kernel.
- Split vnode::v_flag into three fields, depending on how the flag is
locked (by the interlock, by the vnode lock, by the file system).
- Miscellaneous locking fixes and improvements.
|
| 1.201.2.6 | 08-Jun-2007 |
ad | Sync with head.
|
| 1.201.2.5 | 13-May-2007 |
ad | - Pass the error number and residual count to biodone(), and let it handle
setting error indicators. Prepare to eliminate B_ERROR.
- Add a flag argument to brelse() to be set into the buf's flags, instead
of doing it directly. Typically used to set B_INVAL.
- Add a "struct cpu_info *" argument to kthread_create(), to be used to
create bound threads. Change "bool mpsafe" to "int flags".
- Allow exit of LWPs in the IDL state when (l != curlwp).
- More locking fixes & conversion to the new API.
|
| 1.201.2.4 | 10-Apr-2007 |
ad | Sync with head.
|
| 1.201.2.3 | 09-Apr-2007 |
ad | - Add two new arguments to kthread_create1: pri_t pri, bool mpsafe.
- Fork kthreads off proc0 as new LWPs, not new processes.
|
| 1.201.2.2 | 21-Mar-2007 |
ad | - Replace more simple_locks, and fix up in a few places.
- Use condition variables.
- LOCK_ASSERT -> KASSERT.
|
| 1.201.2.1 | 13-Mar-2007 |
ad | Pull in the initial set of changes for the vmlocking branch.
|
| 1.208.2.1 | 15-Aug-2007 |
skrll | Sync with HEAD.
|
| 1.210.10.2 | 29-Jul-2007 |
ad | It's not a good idea for device drivers to modify b_flags, as they don't
need to understand the locking around that field. Instead of setting
B_ERROR, set b_error instead. b_error is 'owned' by whoever completes
the I/O request.
|
| 1.210.10.1 | 29-Jul-2007 |
ad | file lfs_vnops.c was added on branch matt-mips64 on 2007-07-29 13:31:16 +0000
|
| 1.210.8.1 | 14-Oct-2007 |
yamt | sync with head.
|
| 1.210.6.3 | 23-Mar-2008 |
matt | sync with HEAD
|
| 1.210.6.2 | 09-Jan-2008 |
matt | sync with HEAD
|
| 1.210.6.1 | 06-Nov-2007 |
matt | sync with HEAD
|
| 1.210.4.2 | 27-Nov-2007 |
joerg | Sync with HEAD. amd64 Xen support needs testing.
|
| 1.210.4.1 | 26-Oct-2007 |
joerg | Sync with HEAD.
Follow the merge of pmap.c on i386 and amd64 and move
pmap_init_tmp_pgtbl into arch/x86/x86/pmap.c. Modify the ACPI wakeup
code to restore CR4 before jumping back into kernel space as the large
page option might cover that.
|
| 1.212.4.2 | 18-Feb-2008 |
mjf | Sync with HEAD.
|
| 1.212.4.1 | 08-Dec-2007 |
mjf | Sync with HEAD.
|
| 1.213.6.1 | 02-Jan-2008 |
bouyer | Sync with HEAD
|
| 1.213.2.4 | 19-Dec-2007 |
ad | Use a global lfs_lock.
|
| 1.213.2.3 | 19-Dec-2007 |
ad | Fix some more problems w/lfs on this branch.
|
| 1.213.2.2 | 19-Dec-2007 |
ad | Get lfs mostly working.
|
| 1.213.2.1 | 04-Dec-2007 |
ad | Pull the vmlocking changes into a new branch.
|
| 1.215.10.5 | 11-Aug-2010 |
yamt | sync with head.
|
| 1.215.10.4 | 11-Mar-2010 |
yamt | sync with head
|
| 1.215.10.3 | 16-May-2009 |
yamt | sync with head
|
| 1.215.10.2 | 04-May-2009 |
yamt | sync with head.
|
| 1.215.10.1 | 16-May-2008 |
yamt | sync with head.
|
| 1.215.8.2 | 17-Jun-2008 |
yamt | sync with head.
|
| 1.215.8.1 | 18-May-2008 |
yamt | sync with head.
|
| 1.215.6.4 | 17-Jan-2009 |
mjf | Sync with HEAD.
|
| 1.215.6.3 | 29-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.215.6.2 | 05-Jun-2008 |
mjf | Sync with HEAD.
Also fix build.
|
| 1.215.6.1 | 02-Jun-2008 |
mjf | Sync with HEAD.
|
| 1.216.2.2 | 18-Sep-2008 |
wrstuden | Sync with wrstuden-revivesa-base-2.
|
| 1.216.2.1 | 23-Jun-2008 |
wrstuden | Sync w/ -current. 34 merge conflicts to follow.
|
| 1.217.2.1 | 27-Jun-2008 |
simonb | Sync with head.
|
| 1.218.12.1 | 29-Feb-2012 |
matt | Deal with UVM_PAGE_OWN changes.
|
| 1.218.6.1 | 19-May-2012 |
riz | Apply patch (requested by buhrow in ticket #1759):
sys/ufs/lfs/lfs_vnops.c patch
sys/ufs/ufs/inode.h patch
sys/ufs/ufs/ufs_extern.h patch
sys/ufs/ufs/ufs_lookup.c patch
sys/ufs/ufs/ufs_vnops.c patch
sys/ufs/ufs/ufs_wapbl.c patch
Port dholland's ufs_rename locking changes to netbsd-5.
[buhrow, ticket #1759]
Hello. More testing has revealed a minor misunderstanding between the
vnode API in -current and 5.x. The below patch, against NetBSD-5.1
sources, rolls all the accumulated patches into one patch set. With this
patch, I believe you can now run with WAPBL, softdep or traditional ufs
semantics with heavy file loads and avoid panics due to resource exhaustion
and/or tstile deadlocks. Testing has been done on I386, both uniprocessor
and multiprocessor, and on Sparc machines in uniprocessor mode, though I
think multiprocessor Sparc would be fine as well. Since these changes are
machine independent, I don't anticipate any issues on any platform. It is
my hope that modulo any final issues that come up in the final round of
testing I'm currently performing, these patches will be ready to be pulled
up into the NetBSD-5 branch.
Finally, I'd like to thank mouse@ and hannken@ for their help and
patience in helping me track down and test the final versions of these
patches. With their assistance, I'm confident these patches make NetBSD-5
a much more stable and robust operating environment in a variety of
setings.
|
| 1.218.4.2 | 03-Mar-2009 |
skrll | Sync with HEAD.
|
| 1.218.4.1 | 19-Jan-2009 |
skrll | Sync with HEAD.
|
| 1.219.2.1 | 13-May-2009 |
jym | Sync with HEAD.
Commit is split, to avoid a "too many arguments" protocol error.
|
| 1.226.4.6 | 21-May-2011 |
rmind | Fix the build.
|
| 1.226.4.5 | 19-May-2011 |
rmind | Implement sharing of vnode_t::v_interlock amongst vnodes:
- Lock is shared amongst UVM objects using uvm_obj_setlock() or getnewvnode().
- Adjust vnode cache to handle unsharing, add VI_LOCKSHARE flag for that.
- Use sharing in tmpfs and layerfs for underlying object.
- Simplify locking in ubc_fault().
- Sprinkle some asserts.
Discussed with ad@.
|
| 1.226.4.4 | 05-Mar-2011 |
rmind | sync with head
|
| 1.226.4.3 | 03-Jul-2010 |
rmind | sync with head
|
| 1.226.4.2 | 30-May-2010 |
rmind | sync with head
|
| 1.226.4.1 | 16-Mar-2010 |
rmind | Change struct uvm_object::vmobjlock to be dynamically allocated with
mutex_obj_alloc(). It allows us to share the locks among UVM objects.
|
| 1.226.2.2 | 17-Aug-2010 |
uebayasi | Sync with HEAD.
|
| 1.226.2.1 | 30-Apr-2010 |
uebayasi | Sync with HEAD.
|
| 1.234.6.1 | 23-Jun-2011 |
cherry | Catchup with rmind-uvmplock merge.
|
| 1.238.6.3 | 02-Jun-2012 |
mrg | sync to latest -current.
|
| 1.238.6.2 | 05-Apr-2012 |
mrg | sync to latest -current.
|
| 1.238.6.1 | 18-Feb-2012 |
mrg | merge to -current.
|
| 1.238.2.6 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.238.2.5 | 23-Jan-2013 |
yamt | sync with head
|
| 1.238.2.4 | 23-May-2012 |
yamt | sync with head.
|
| 1.238.2.3 | 17-Apr-2012 |
yamt | sync with head
|
| 1.238.2.2 | 06-Nov-2011 |
yamt | remove pg->listq and uobj->memq
|
| 1.238.2.1 | 02-Nov-2011 |
yamt | page cache related changes
- maintain object pages in radix tree rather than rb tree.
- reduce unnecessary page scan in putpages. esp. when an object has a ton of
pages cached but only a few of them are dirty.
- reduce the number of pmap operations by tracking page dirtiness more
precisely in uvm layer.
- fix nfs commit range tracking.
- fix nfs write clustering. XXX hack
|
| 1.239.2.2 | 27-Aug-2016 |
bouyer | Pull up following revision(s) (requested by dholland in ticket #1389):
sys/ufs/lfs/lfs_vnops.c: revision 1.304
Fix a deadlock
ok dholland@
|
| 1.239.2.1 | 17-Mar-2012 |
bouyer | Pull up following revision(s) (requested by perseant in ticket #116):
sys/ufs/lfs/lfs_alloc.c: revision 1.112
tests/fs/vfs/t_rmdirrace.c: revision 1.9
tests/fs/vfs/t_renamerace.c: revision 1.25
sys/ufs/lfs/lfs_vnops.c: revision 1.240
sys/ufs/lfs/lfs_segment.c: revision 1.224
sys/ufs/lfs/lfs_bio.c: revision 1.122
sys/ufs/lfs/lfs_vfsops.c: revision 1.294
sbin/newfs_lfs/make_lfs.c: revision 1.19
sys/ufs/lfs/lfs.h: revision 1.136
Pass t_renamerace and t_rmdirrace tests.
Adapt dholland@'s fix to ufs_rename to fix PR kern/43582. Address several
other MP locking issues discovered during the course of investigating the
same problem.
Removed extraneous vn_lock() calls on the Ifile, since the Ifile writes
are controlled by the segment lock.
Fix PR kern/45982 by deemphasizing the estimate of how much metadata
will fill the empty space on disk when the disk is nearly empty
(t_renamerace crates a lot of inode blocks on a tiny empty disk).
|
| 1.242.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.242.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.242.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.242.2.1 | 25-Feb-2013 |
tls | resync with head
|
| 1.248.4.1 | 23-Jul-2013 |
riastradh | sync with HEAD
|
| 1.248.2.2 | 18-May-2014 |
rmind | sync with head
|
| 1.248.2.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.262.2.1 | 10-Aug-2014 |
tls | Rebase.
|
| 1.269.4.6 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.269.4.5 | 05-Oct-2016 |
skrll | Sync with HEAD
|
| 1.269.4.4 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.269.4.3 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.269.4.2 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.269.4.1 | 06-Apr-2015 |
skrll | Sync with HEAD
|
| 1.269.2.2 | 14-Jul-2016 |
martin | Pull up following revision(s) (requested by dholland in ticket #1205):
sys/ufs/lfs/lfs_vnops.c: revision 1.304
Fix a deadlock
ok dholland@
|
| 1.269.2.1 | 06-Aug-2015 |
snj | Apply patch (requested by dholland in ticket #935):
Comment out some KASSERTs.
|
| 1.304.4.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.304.2.2 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.304.2.1 | 20-Mar-2017 |
pgoyette | Sync with HEAD
|
| 1.315.2.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.321.4.4 | 21-Apr-2020 |
martin | Sync with HEAD
|
| 1.321.4.3 | 13-Apr-2020 |
martin | Mostly merge changes from HEAD upto 20200411
|
| 1.321.4.2 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.321.4.1 | 10-Jun-2019 |
christos | Sync with HEAD
|
| 1.321.2.2 | 18-Jan-2019 |
pgoyette | Synch with HEAD
|
| 1.321.2.1 | 06-Sep-2018 |
pgoyette | Sync with HEAD
Resolve a couple of conflicts (result of the uimin/uimax changes)
|
| 1.324.2.1 | 17-Aug-2020 |
martin | Pull up following revision(s) (requested by riastradh in ticket #1050):
sys/ufs/lfs/lfs_subr.c: revision 1.101
sys/ufs/lfs/lfs_subr.c: revision 1.102
sys/ufs/lfs/lfs_inode.c: revision 1.158
sys/ufs/lfs/lfs_inode.h: revision 1.25
sys/ufs/lfs/lfs_balloc.c: revision 1.95
sys/ufs/lfs/lfs_pages.c: revision 1.21
sys/ufs/lfs/lfs_vnops.c: revision 1.330
sys/ufs/lfs/lfs_alloc.c: revision 1.140 (patch)
sys/ufs/lfs/lfs_alloc.c: revision 1.141 (patch)
lib/libp2k/p2k.c: revision 1.72
sys/ufs/lfs/lfs.h: revision 1.205
sys/ufs/lfs/lfs.h: revision 1.206
sys/ufs/lfs/lfs_segment.c: revision 1.284
sys/ufs/lfs/lfs.h: revision 1.207
sys/ufs/lfs/lfs_segment.c: revision 1.285
sys/ufs/lfs/lfs_debug.c: revision 1.55
sys/ufs/lfs/lfs_rename.c: revision 1.23
usr.sbin/dumplfs/dumplfs.c: revision 1.65
sys/ufs/lfs/lfs_vfsops.c: revision 1.371
sys/arch/i386/stand/efiboot/bootx64/Makefile: revision 1.3
sys/ufs/lfs/lfs_vfsops.c: revision 1.372
sys/ufs/lfs/lfs_vfsops.c: revision 1.373
sbin/fsck_lfs/pass1.c: revision 1.46
sys/ufs/lfs/lfs_vnops.c: revision 1.326
sys/ufs/lfs/lfs_vnops.c: revision 1.327
sys/ufs/lfs/lfs_vfsops.c: revision 1.375 (patch)
sys/ufs/lfs/lfs_vnops.c: revision 1.328
sys/ufs/lfs/lfs_subr.c: revision 1.98
sys/ufs/lfs/lfs_extern.h: revision 1.116
sys/ufs/lfs/lfs_vnops.c: revision 1.329
sys/ufs/lfs/lfs_subr.c: revision 1.99
sys/ufs/lfs/lfs_extern.h: revision 1.117
sys/ufs/lfs/lfs_accessors.h: revision 1.49
sys/ufs/lfs/lfs_extern.h: revision 1.118
sys/rump/fs/lib/liblfs/Makefile: revision 1.15
sys/ufs/lfs/lfs_bio.c: revision 1.146 (patch)
sys/ufs/lfs/lfs_bio.c: revision 1.147
sys/ufs/lfs/lfs_subr.c: revision 1.100
Fix kassert in lfs by initializing vp first.
Use a marker node to iterate lfs_dchainhd / i_lfs_dchain.
I believe elements can be removed while the lock is dropped,
including the next node we're hanging on to.
Just use VOP_BWRITE for lfs_bwrite_log.
Hope this doesn't cause trouble with vfs_suspend.
Teach lfs to transition ro<->rw.
Prevent new dirops while we issue lfs_flush_dirops.
lfs_flush_dirops assumes (by KASSERT((ip->i_state & IN_ADIROP) == 0))
that vnodes on the dchain will not become involved in active dirops
even while holding no other locks (lfs_lock, v_interlock), so we must
set lfs_writer here. All other callers already set lfs_writer.
We set fs->lfs_writer++ without explicitly doing lfs_writer_enter
because
(a) we already waited for the dirops to drain, and
(b) we hold lfs_lock and cannot drop it before setting lfs_writer.
Assert lfs_writer where I think we can now prove it.
Serialize access to the splay tree with lfs_lock.
Change some cheap KDASSERT into KASSERT.
Take a reference and fix assertions in lfs_flush_dirops.
Fixes panic:
KASSERT((ip->i_state & IN_ADIROP) == 0) at lfs_vnops.c:1670
lfs_flush_dirops
lfs_check
lfs_setattr
VOP_SETATTR
change_mode
sys_fchmod
syscall
This assertion -- and the assertion that vp->v_uflag has VU_DIROP set
-- is valid only until we release lfs_lock, because we may race with
lfs_unmark_dirop which will remove the nodes and change the flags.
Further, vp itself is valid only as long as it is referenced, which it
is as long as it's on the dchain, but lfs_unmark_dirop drops the
dchain's reference.
Don't lfs_writer_enter while holding v_interlock.
There's no need to lfs_writer_enter at all here, as far as I can see.
lfs_flush_fs will do it for us.
Break deadlock in PR kern/52301.
The lock order is lfs_writer -> lfs_seglock. The problem in 52301 is
that lfs_segwrite violates this lock order by sometimes doing
lfs_seglock -> lfs_writer, either (a) when doing a checkpoint or (b),
opportunistically, when there are no dirops pending. Both cases can
deadlock, because dirops sometimes take the seglock (lfs_truncate,
lfs_valloc, lfs_vfree):
(a) There may be dirops pending, and they may be waiting for the
seglock, so we can't wait for them to complete while holding the
seglock.
(b) The test for fs->lfs_dirops == 0 happens unlocked, and the state
may change by the time lfs_writer_enter acquires lfs_lock.
To resolve this in each case:
(a) Do lfs_writer_enter before lfs_seglock, since we will need it
unconditionally anyway. The worst performance impact of this should
be that some dirops get delayed a little bit.
(b) Create a new lfs_writer_tryenter to use at this point so that the
test for fs->lfs_dirops == 0 and the acquisition of lfs_writer happen
atomically under lfs_lock.
Initialize/destroy lfs_allclean_wakeup in modcmd, not lfs_mountfs.
Fixes reloading lfs.kmod.
In lfs_update, hold lfs_writer around lfs_vflush.
Otherwise, we might do
lfs_vflush
-> lfs_seglock
-> lfs_segwait(SEGM_CKP)
-> lfs_writer_enter
which is the reverse of the lfs_writer -> lfs_seglock ordering.
Call lfs_orphan in lfs_rename while we're still in the dirop.
lfs_writer_enter can't fail; keep it simple and don't pretend it can.
Assert that mtsleep can't fail either -- it doesn't catch signals and
there's no timeout.
Teach LFS_ORPHAN_NEXTFREE about lfs64.
Dust off the orphan detection code and try to make it work.
Fix !DIAGNOSTIC compile
Fix userland references to LFS_ORPHAN_NEXTFREE.
Forgot to grep for these or do a full distribution build, oops!
Fix missing <sys/evcnt.h> by removing the evcnts instead.
Just wanted to confirm that a race might happen, and indeed it did.
These serve little diagnostic value otherwise.
OR into bp->b_cflags; don't overwrite.
CTASSERT lfs on-disk structure sizes.
Avoid misaligned access to lfs64 on-disk records in memory.
lfs64 directory entries are only 32-bit aligned in order to conserve
space in directory blocks, and we had a hack to stuff a 64-bit inode
in them. This replaces the hack by __aligned(4) __packed, and goes
further:
1. It's not clear that all the other lfs64 data structures are 64-bit
aligned on disk to begin with. We can go through these later and
upgrade them from
struct foo64 {
...
} __aligned(4) __packed;
union foo {
struct foo64 f64;
...
};
to
struct foo64 {
...
};
union foo {
struct foo64 f64 __aligned(8);
...
} __aligned(4) __packed;
if we really want to take advantage of 64-bit memory accesses.
However, the __aligned(4) __packed must remain on the union
because:
2. We access even the lfs32 data structures via a union that has
lfs64 members, and it turns out that compilers will assume access
through a union with 64-bit aligned members implies the whole
union has 64-bit alignment, even if we're only accessing a 32-bit
aligned member.
Fix clang build after packed lfs64 accessor change.
Suppress spurious address-of-packed error in rump lfs too.
|
| 1.325.2.1 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.331.4.1 | 20-Apr-2020 |
bouyer | Sync with HEAD
|
| 1.336.6.1 | 01-Aug-2021 |
thorpej | Sync with HEAD.
|
| 1.9 | 30-Mar-2017 |
hannken | Remove now redundant calls to fstrans_start()/fstrans_done().
Add fstrans_start()/fstrans_done() to lfs_putpages().
|
| 1.8 | 13-Mar-2017 |
riastradh | #if DIAGNOSTIC panic ---> KASSERT
Replace some #if DEBUG by this too. DEBUG is only for expensive
assertions; these are not.
|
| 1.7 | 01-Sep-2015 |
dholland | branches: 1.7.2; 1.7.4;
Use the lfs dinode accessors in place of the ufs-derived ones.
(Mostly.)
The ufs-derived ones are fake structure member macros, which are gross
and not very safe. Also, it seems that a lot of places in the lfs code
were using the ffsv1 branch of them unconditionally, and this way it's
guaranteed all those places have been updated.
Found while doing this: for non-devices, have getattr produce NODEV
in the rdev field instead of leaking the address of the first direct
block.
|
| 1.6 | 02-Aug-2015 |
dholland | Pass the fs object to LFS_MAX_DADDR so it can check lfs_is64.
Remove some hackish intentional 64->32 truncations next to the checks
using LFS_MAX_DADDR, and tackle the problem they handled in bmap
instead.
The problem: the magic block pointer value UNWRITTEN has magic value
-2, and if it's not handled specifically, uint32 -> uint64 promotion
turns it into 4294967294, which then causes consternation and
monkeyhouse downstream.
What's here is still kind of a hack, but it's a step forward.
|
| 1.5 | 28-Jul-2013 |
dholland | branches: 1.5.4; 1.5.8;
Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.4 | 06-Jun-2013 |
dholland | branches: 1.4.2; 1.4.4;
Remove stray references to ext2fs, chfs, ffs, and mfs.
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.4.4.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.4.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.4.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.4.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.4.2.1 | 06-Jun-2013 |
tls | file ulfs_bmap.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.5.8.2 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.5.8.1 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.5.4.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.5.4.1 | 28-Jul-2013 |
yamt | file ulfs_bmap.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.7.4.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.7.2.2 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.7.2.1 | 20-Mar-2017 |
pgoyette | Sync with HEAD
|
| 1.9 | 19-Apr-2018 |
christos | s/static inline/static __inline/g for consistency.
|
| 1.8 | 20-Jun-2016 |
dholland | branches: 1.8.16;
u_int{8,16,32,64}_t -> uint{8,16,32,64}_t in remaining lfs headers.
|
| 1.7 | 19-Jun-2016 |
dholland | we already have changes here comparable to ufs_bswap.h -r1.20 and -r1.21.
|
| 1.6 | 18-Oct-2013 |
christos | branches: 1.6.4; 1.6.8;
use __USE() in the right place, instead of (void)var.
|
| 1.5 | 17-Oct-2013 |
christos | - remove unused variables
- add debug ifdefs for debugging variables
- __USE() where appropriate.
|
| 1.4 | 28-Jul-2013 |
dholland | Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.3 | 06-Jun-2013 |
dholland | branches: 1.3.2; 1.3.4;
Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs step 3: rearrange config stuff.
Add new options:
LFS_EI
LFS_DIRHASH
LFS_EXTATTR
LFS_EXTATTR_AUTOSTART
LFS_QUOTA
LFS_QUOTA2
and update code referring to the corresponding FFS and UFS config
symbols to use the LFS versions. Disable the one extant reference
to APPLE_UFS in the ulfs files. Use opt_lfs.h only, not opt_ffs.h.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.3.4.2 | 18-May-2014 |
rmind | sync with head
|
| 1.3.4.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.3.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.3.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.3.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.3.2.1 | 06-Jun-2013 |
tls | file ulfs_bswap.h was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.6.8.1 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.6.4.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.6.4.1 | 18-Oct-2013 |
yamt | file ulfs_bswap.h was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.8.16.1 | 22-Apr-2018 |
pgoyette | Sync with HEAD
|
| 1.13 | 20-Jun-2016 |
dholland | Massedit u_int{8,16,32,64}_t to uint{8,16,32,64}_t. This effectively
merges ufs/dinode.h 1.25.
|
| 1.12 | 19-Jun-2016 |
dholland | we are actually synced with ufs/dinode.h 1.24 and ufs/dir.h 1.25.
|
| 1.11 | 08-Jun-2013 |
dholland | branches: 1.11.2; 1.11.10; 1.11.14;
Move a comment to lfs.h that belongs better there.
|
| 1.10 | 08-Jun-2013 |
dholland | Move more symbols to lfs.h:
LFS_DIRBLKSIZ
LFS_DIRECTSIZ
LFS_DIRSIZ
LFS_OLDDIRFMT
LFS_NEWDIRFMT
LFS_IFTODT
LFS_DTTOIF
ULFS{,1,2}_MAXSYMLINKLEN
|
| 1.9 | 08-Jun-2013 |
dholland | Move stuff to lfs.h that's needed by userland:
LFS_DT_*
ULFS_ROOTINO
ULFS_WINO
struct lfs_direct
struct lfs_dirtemplate
struct lfs_odirtemplate
struct ulfs_args
Also fix FFS_MAXNAMLEN -> LFS_MAXNAMLEN in several places.
|
| 1.8 | 08-Jun-2013 |
dholland | struct direct -> struct lfs_direct
struct dirtemplate -> struct lfs_dirtemplate
struct odirtemplate -> struct lfs_odirtemplate
DT_* -> LFS_DT_*
|
| 1.7 | 08-Jun-2013 |
dholland | Now move LFS_IFMT and friends from ulfs_dinode.h to lfs.h.
|
| 1.6 | 08-Jun-2013 |
dholland | Stick LFS_ in front of IFMT, IFIFO, IFREG, etc. so as not to conflict
with the UFS copies of these symbols. (Which themselves ought to have
UFS_ stuck on.)
|
| 1.5 | 08-Jun-2013 |
dholland | Move the dinode (on-disk inode) structures to lfs.h, since they are
and will be obviously required by userland tools that need to read
the on-disk structures.
Also, DINODE{1,2}_SIZE -> LFS_DINODE{1,2}_SIZE.
|
| 1.4 | 06-Jun-2013 |
dholland | Remove references to Apple UFS.
|
| 1.3 | 06-Jun-2013 |
dholland | Cleanups to reduce symbol and header exposure:
- move struct ufid from ulfs_inode.h to lfs.h
- lfs.h needs sys/mount.h and sys/pool.h
- ulfs_quota2_subr.c needs lfs_inode.h
- remove ulfs_inode.h from lfs.h in favor of ulfs_dinode.h
- move ULFS_NDADDR, ULFS_NIADDR, ULFS_NXADDR from ulfs_dinode.h to lfs.h
- remove ulfs_dinode.h from lfs.h
- add lfs.h to ulfs_dinode.h
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.11.14.1 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.11.10.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.11.10.1 | 08-Jun-2013 |
yamt | file ulfs_dinode.h was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.11.2.3 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.11.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.11.2.1 | 08-Jun-2013 |
tls | file ulfs_dinode.h was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.9 | 08-Jun-2013 |
dholland | ulfs_dir.h has been emptied; remove it.
|
| 1.8 | 08-Jun-2013 |
dholland | Move more symbols to lfs.h:
LFS_DIRBLKSIZ
LFS_DIRECTSIZ
LFS_DIRSIZ
LFS_OLDDIRFMT
LFS_NEWDIRFMT
LFS_IFTODT
LFS_DTTOIF
ULFS{,1,2}_MAXSYMLINKLEN
|
| 1.7 | 08-Jun-2013 |
dholland | DIRBLKSIZ -> LFS_DIRBLKSIZ
DIRECTSIZ -> LFS_DIRECTSIZ
DIRSIZ -> LFS_DIRSIZ
OLDDIRFMT -> LFS_OLDDIRFMT
NEWDIRFMT -> LFS_NEWDIRFMT
IFTODT -> LFS_IFTODT
DTTOIF -> LFS_DTTOIF
|
| 1.6 | 08-Jun-2013 |
dholland | Move stuff to lfs.h that's needed by userland:
LFS_DT_*
ULFS_ROOTINO
ULFS_WINO
struct lfs_direct
struct lfs_dirtemplate
struct lfs_odirtemplate
struct ulfs_args
Also fix FFS_MAXNAMLEN -> LFS_MAXNAMLEN in several places.
|
| 1.5 | 08-Jun-2013 |
dholland | struct direct -> struct lfs_direct
struct dirtemplate -> struct lfs_dirtemplate
struct odirtemplate -> struct lfs_odirtemplate
DT_* -> LFS_DT_*
|
| 1.4 | 08-Jun-2013 |
dholland | Split the definitions suitable for userland out of ulfs_inode.h into
lfs_inode.h. Since fsck_lfs, newfs_lfs, and lfs_cleanerd want to reuse
the inode structure for their own internal use, and some of them share
parts of the kernel code as well, the best way forward is to provide a
relatively sanitized header that doesn't bring in stray material.
Shuffle a few other definitions around so that lfs_inode.h depends
only on lfs.h.
Install lfs_inode.h into /usr/include.
|
| 1.3 | 06-Jun-2013 |
dholland | Remove references to Apple UFS.
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.19 | 07-Aug-2022 |
simonb | If UFS or LFS dirhash is enabled in the kernel, set the dirhash cache
size dependant on memory size. If less than 128MB of memory, default
to no cache. With 128MB of memory or more, use a maximum cache size of
1/64th of memory; cap maximum default cache size to 32MB (for systems
with 2GB of memory or more).
The dirhash cache sizes are still explicityly setable by sysctl(8) or
by adding relevant entry(s) to sysctl.conf(5).
|
| 1.18 | 14-Mar-2020 |
ad | - Hide the details of SPCF_SHOULDYIELD and related behind a couple of small
functions: preempt_point() and preempt_needed().
- preempt(): if the LWP has exceeded its timeslice in kernel, strip it of
any priority boost gained earlier from blocking.
|
| 1.17 | 20-Jun-2016 |
dholland | branches: 1.17.18;
Merge -r1.37 of ufs_dirhash.c:
clear i_dirhash sooner, but what lock protects it?
|
| 1.16 | 20-Jun-2016 |
dholland | More already-merged or equivalent changes:
ufs_dirhash.c 1.36 corresponds to ulfs_dirhash.c 1.8
ufs_extattr.c 1.43 corresponds to ulfs_extattr.c 1.7
ufs_lookup.c 1.126 does not apply to lfs
ufs_lookup.c 1.127 we already have
ufs_lookup.c 1.128 does not apply to lfs
ufs_lookup.c 1.129 corresponds to ulfs_lookup.c 1.19
ufs_quota1.c 1.19 corresponds to ulfs_quota1.c 1.7
ufs_quota1.c 1.20 corresponds to ulfs_quota1.c 1.8
ufs_quota2.c 1.36 we have equivalent changes for
ufs_rename.c 1.9 corresponds to lfs_rename.c 1.5
ufs_rename.c 1.10 corresponds to lfs_rename.c 1.6
ufs_vnops.c 1.219 corresponds to lfs_vnops.c 1.260 and ulfs_vnops.c 1.19
ufs_vnops.c 1.220 corresponds to lfs_vnops.c 1.261 and ulfs_vnops.c 1.20
ufs_vnops.c 1.221 was superseded by later changes
ufs_vnops.c 1.222 got fixed independently in lfs
|
| 1.15 | 19-Jun-2016 |
dholland | Mark ufs file versions we're already synced with.
|
| 1.14 | 21-Sep-2015 |
dholland | Add 64-bit directory entry structures, and adjust accessors accordingly.
The LFS64 directory entry has a 64-bit inode number. This is stored as
two 32-bit values to avoid inducing 64-bit alignment requirements.
The exposed type for manipulating directory entries is now
LFS_DIRHEADER, following the same convention as e.g. IFILE and SEGUSE.
(But with LFS_ on it, because.)
|
| 1.13 | 21-Sep-2015 |
dholland | Oops; LFS_DIRECTSIZ() is going to need the fs as an argument.
Also, it turns out that dirhash needs a compile-time-constant version
of LFS_DIRECTSIZ(LFS_MAXNAMLEN+1), independent of 64-vs-32, so create
LFS_MAXDIRENTRYSIZE for this. Sigh.
|
| 1.12 | 15-Sep-2015 |
dholland | Pass around struct lfs_dirheader instead of struct lfs_direct.
|
| 1.11 | 15-Sep-2015 |
dholland | Add an accessor function for directory names.
|
| 1.10 | 15-Sep-2015 |
dholland | Add and use accessor functions for more of the directory entry fields.
|
| 1.9 | 01-Sep-2015 |
dholland | Add new accessors for the d_type and d_namlen fields of struct lfs_direct.
Napalm the old byteswap access logic for these.
|
| 1.8 | 25-Feb-2014 |
pooka | branches: 1.8.4; 1.8.8;
Ensure that the top level sysctl nodes (kern, vfs, net, ...) exist before
the sysctl link sets are processed, and remove redundancy.
Shaves >13kB off of an amd64 GENERIC, not to mention >1k duplicate
lines of code.
|
| 1.7 | 28-Jul-2013 |
dholland | Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.6 | 08-Jun-2013 |
dholland | branches: 1.6.2; 1.6.4;
ulfs_dir.h has been emptied; remove it.
|
| 1.5 | 08-Jun-2013 |
dholland | DIRBLKSIZ -> LFS_DIRBLKSIZ
DIRECTSIZ -> LFS_DIRECTSIZ
DIRSIZ -> LFS_DIRSIZ
OLDDIRFMT -> LFS_OLDDIRFMT
NEWDIRFMT -> LFS_NEWDIRFMT
IFTODT -> LFS_IFTODT
DTTOIF -> LFS_DTTOIF
|
| 1.4 | 08-Jun-2013 |
dholland | struct direct -> struct lfs_direct
struct dirtemplate -> struct lfs_dirtemplate
struct odirtemplate -> struct lfs_odirtemplate
DT_* -> LFS_DT_*
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.6.4.2 | 18-May-2014 |
rmind | sync with head
|
| 1.6.4.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.6.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.6.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.6.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.6.2.1 | 08-Jun-2013 |
tls | file ulfs_dirhash.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.8.8.2 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.8.8.1 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.8.4.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.8.4.1 | 25-Feb-2014 |
yamt | file ulfs_dirhash.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.17.18.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.12 | 19-Aug-2021 |
andvar | s/memry/memory+s/softare/software/+s/grapics/graphics+s/ouput/output
|
| 1.11 | 27-Dec-2019 |
msaitoh | s/inital/initial/
|
| 1.10 | 20-Jun-2016 |
dholland | branches: 1.10.18;
u_int{8,16,32,64}_t -> uint{8,16,32,64}_t in remaining lfs headers.
|
| 1.9 | 19-Jun-2016 |
dholland | Mark ufs file versions we're already synced with.
|
| 1.8 | 21-Sep-2015 |
dholland | Add 64-bit directory entry structures, and adjust accessors accordingly.
The LFS64 directory entry has a 64-bit inode number. This is stored as
two 32-bit values to avoid inducing 64-bit alignment requirements.
The exposed type for manipulating directory entries is now
LFS_DIRHEADER, following the same convention as e.g. IFILE and SEGUSE.
(But with LFS_ on it, because.)
|
| 1.7 | 21-Sep-2015 |
dholland | Oops; LFS_DIRECTSIZ() is going to need the fs as an argument.
Also, it turns out that dirhash needs a compile-time-constant version
of LFS_DIRECTSIZ(LFS_MAXNAMLEN+1), independent of 64-vs-32, so create
LFS_MAXDIRENTRYSIZE for this. Sigh.
|
| 1.6 | 15-Sep-2015 |
dholland | Pass around struct lfs_dirheader instead of struct lfs_direct.
|
| 1.5 | 08-Jun-2013 |
dholland | branches: 1.5.2; 1.5.10; 1.5.14;
DIRBLKSIZ -> LFS_DIRBLKSIZ
DIRECTSIZ -> LFS_DIRECTSIZ
DIRSIZ -> LFS_DIRSIZ
OLDDIRFMT -> LFS_OLDDIRFMT
NEWDIRFMT -> LFS_NEWDIRFMT
IFTODT -> LFS_IFTODT
DTTOIF -> LFS_DTTOIF
|
| 1.4 | 08-Jun-2013 |
dholland | Move stuff to lfs.h that's needed by userland:
LFS_DT_*
ULFS_ROOTINO
ULFS_WINO
struct lfs_direct
struct lfs_dirtemplate
struct lfs_odirtemplate
struct ulfs_args
Also fix FFS_MAXNAMLEN -> LFS_MAXNAMLEN in several places.
|
| 1.3 | 08-Jun-2013 |
dholland | struct direct -> struct lfs_direct
struct dirtemplate -> struct lfs_dirtemplate
struct odirtemplate -> struct lfs_odirtemplate
DT_* -> LFS_DT_*
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.5.14.2 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.5.14.1 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.5.10.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.5.10.1 | 08-Jun-2013 |
yamt | file ulfs_dirhash.h was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.5.2.3 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.5.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.5.2.1 | 08-Jun-2013 |
tls | file ulfs_dirhash.h was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.10.18.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.18 | 10-Feb-2024 |
andvar | Fix various typos in comments, log messages and documentation.
|
| 1.17 | 29-Jun-2021 |
dholland | Add containment for the cloning devices hack in vn_open.
Cloning devices (and also things like /dev/stderr) work by allocating
a struct file, stuffing it in the file table (which is a layer
violation), stuffing the file descriptor number for it in a magic
field of struct lwp (which is gross), and then "failing" with one of
two magic errnos, EDUPFD or EMOVEFD.
Before this commit, all callers of vn_open in the kernel (there are
quite a few) were expected to check for these errors and handle the
situation. Needless to say, none of them except for open() itself did,
resulting in internal negative errnos being returned to userspace.
This hack is fairly deeply rooted and cannot be eliminated all at
once. This commit adds logic to handle the magic errnos inside
vn_open; now on success vn_open returns either a vnode or an integer
file descriptor, along with a flag that says whether the underlying
code requested EDUPFD or EMOVEFD. Callers not prepared to cope with
file descriptors can pass NULL for the extra return values, in which
case if a file descriptor would be produced vn_open fails with
EOPNOTSUPP.
Since I'm rearranging vn_open's signature anyway, stop exposing struct
nameidata. Instead, take three arguments: an optional vnode to use as
the starting point (like openat()), the path, and additional namei
flags to use, restricted to NOCHROOT and TRYEMULROOT. (Other namei
behavior, e.g. NOFOLLOW, can be requested via the open flags.)
This change requires a kernel bump. Ride the one an hour ago.
(That was supposed to be coordinated; did not intend to let an hour
slip by. My fault.)
|
| 1.16 | 16-May-2020 |
christos | branches: 1.16.6;
Add ACL support for FFS. From FreeBSD.
|
| 1.15 | 17-Jan-2020 |
ad | VFS_VGET(), VFS_ROOT(), VFS_FHTOVP(): give them a "int lktype" argument, to
allow us to get shared locks (or no lock) on the returned vnode. Matches
FreeBSD.
|
| 1.14 | 09-Nov-2016 |
dholland | branches: 1.14.16; 1.14.22;
Apply ufs_extattr.c 1.48:
Explain why the lock in here needs to be recursive. Related to PR 46997.
ufs_extattr 1.47 was also committed directly here, so this file is still
fully synced with it.
|
| 1.13 | 07-Jul-2016 |
msaitoh | branches: 1.13.2;
KNF. Remove extra spaces. No functional change.
|
| 1.12 | 20-Jun-2016 |
dholland | Merge -r1.46 of ufs_extattr.c: Fix uninitialized mutex usage
|
| 1.11 | 20-Jun-2016 |
dholland | Merge -r1.45 of ufs_extattr.c:
Fix UFS1 extended attribute backend autocreation deadlock
|
| 1.10 | 20-Jun-2016 |
dholland | Merge -r1.44 of ufs_extattr.c and related change -r1.302 of ffs_vfops.c:
fix use-after-free on failed unmount with extended attributes enabled.
|
| 1.9 | 20-Jun-2016 |
dholland | More already-merged or equivalent changes:
ufs_dirhash.c 1.36 corresponds to ulfs_dirhash.c 1.8
ufs_extattr.c 1.43 corresponds to ulfs_extattr.c 1.7
ufs_lookup.c 1.126 does not apply to lfs
ufs_lookup.c 1.127 we already have
ufs_lookup.c 1.128 does not apply to lfs
ufs_lookup.c 1.129 corresponds to ulfs_lookup.c 1.19
ufs_quota1.c 1.19 corresponds to ulfs_quota1.c 1.7
ufs_quota1.c 1.20 corresponds to ulfs_quota1.c 1.8
ufs_quota2.c 1.36 we have equivalent changes for
ufs_rename.c 1.9 corresponds to lfs_rename.c 1.5
ufs_rename.c 1.10 corresponds to lfs_rename.c 1.6
ufs_vnops.c 1.219 corresponds to lfs_vnops.c 1.260 and ulfs_vnops.c 1.19
ufs_vnops.c 1.220 corresponds to lfs_vnops.c 1.261 and ulfs_vnops.c 1.20
ufs_vnops.c 1.221 was superseded by later changes
ufs_vnops.c 1.222 got fixed independently in lfs
|
| 1.8 | 19-Jun-2016 |
dholland | Mark ufs file versions we're already synced with.
|
| 1.7 | 07-Feb-2014 |
hannken | branches: 1.7.4; 1.7.8;
Change vnode operation lookup to return the resulting vnode *vpp unlocked.
Change cache_lookup() to return an unlocked vnode.
Discussed on tech-kern@
Welcome to 6.99.31
|
| 1.6 | 08-Jun-2013 |
dholland | branches: 1.6.2; 1.6.4;
ulfs_dir.h has been emptied; remove it.
|
| 1.5 | 08-Jun-2013 |
dholland | DIRBLKSIZ -> LFS_DIRBLKSIZ
DIRECTSIZ -> LFS_DIRECTSIZ
DIRSIZ -> LFS_DIRSIZ
OLDDIRFMT -> LFS_OLDDIRFMT
NEWDIRFMT -> LFS_NEWDIRFMT
IFTODT -> LFS_IFTODT
DTTOIF -> LFS_DTTOIF
|
| 1.4 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 3: rearrange config stuff.
Add new options:
LFS_EI
LFS_DIRHASH
LFS_EXTATTR
LFS_EXTATTR_AUTOSTART
LFS_QUOTA
LFS_QUOTA2
and update code referring to the corresponding FFS and UFS config
symbols to use the LFS versions. Disable the one extant reference
to APPLE_UFS in the ulfs files. Use opt_lfs.h only, not opt_ffs.h.
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.6.4.1 | 18-May-2014 |
rmind | sync with head
|
| 1.6.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.6.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.6.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.6.2.1 | 08-Jun-2013 |
tls | file ulfs_extattr.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.7.8.2 | 05-Dec-2016 |
skrll | Sync with HEAD
|
| 1.7.8.1 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.7.4.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.7.4.1 | 07-Feb-2014 |
yamt | file ulfs_extattr.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.13.2.1 | 07-Jan-2017 |
pgoyette | Sync with HEAD. (Note that most of these changes are simply $NetBSD$
tag issues.)
|
| 1.14.22.1 | 17-Jan-2020 |
ad | Sync with head.
|
| 1.14.16.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.16.6.1 | 01-Aug-2021 |
thorpej | Sync with HEAD.
|
| 1.3 | 20-Jun-2016 |
dholland | Merge -r1.11 of extattr.h:
Bump UFS1 extended attribute max name length to 256
|
| 1.2 | 06-Jun-2013 |
dholland | branches: 1.2.2; 1.2.10; 1.2.14;
Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.2.14.1 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.2.10.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.2.10.1 | 06-Jun-2013 |
yamt | file ulfs_extattr.h was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.2.2.3 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.2.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.2.2.1 | 06-Jun-2013 |
tls | file ulfs_extattr.h was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.26 | 18-Jul-2021 |
dholland | Abolish all the silly indirection macros for initializing vnode ops tables.
These are things of the form #define foofs_op genfs_op, or #define
foofs_op genfs_eopnotsupp, or similar. They serve no purpose besides
obfuscation, and have gotten cutpasted all over everywhere.
|
| 1.25 | 17-Jan-2020 |
ad | branches: 1.25.10;
VFS_VGET(), VFS_ROOT(), VFS_FHTOVP(): give them a "int lktype" argument, to
allow us to get shared locks (or no lock) on the returned vnode. Matches
FreeBSD.
|
| 1.24 | 20-Jun-2016 |
dholland | branches: 1.24.18; 1.24.24;
One more batch of already-synced ufs changes:
ufs_extern.h 1.79 is equivalent to ulfs_extern.h 1.14
ufsmount.h 1.43 is (roughly) equivalent to lfs_extern.h 1.102
ufs_inode.c 1.94 does not apply to lfs
ufs_inode.c 1.95 does not apply to lfs either
ufs_readwrite.c 1.108 is equivalent to ulfs_readwrite.c 1.8
ufs_readwrite.c 1.109 is equivalent to ulfs_readwrite.c 1.9
ufs_readwrite.c 1.110 is equivalent to ulfs_readwrite.c 1.10
ufs_readwrite.c 1.111 does not apply to lfs
ufs_readwrite.c 1.112 is equivalent to ulfs_readwrite.c 1.11
ufs_readwrite.c 1.113 is equivalent to ulfs_readwrite.c 1.13
ufs_readwrite.c 1.114 is equivalent to ulfs_readwrite.c 1.14
ufs_readwrite.c 1.115 is equivalent to ulfs_readwrite.c 1.15
ufs_readwrite.c 1.116-1.118 does not apply to lfs
ufs_readwrite.c 1.119-1.120 are equivalent to ulfs_readwrite.c 1.16
ufs_rename.c 1.12 is equivalent to lfs_rename.c 1.8
ufs_vnops.c 1.226 is equivalent to ulfs_vnops.c 1.22 and lfs_vnops.c 1.270
ufs_vnops.c 1.227 is equivalent to ulfs_vnops.c 1.23
ufs_vnops.c 1.228-1.229 are equivalent to ulfs_vnops.c 1.24
ufs_vnops.c 1.230 is equivalent to ulfs_vnops.c 1.25 and lfs_vnops.c 1.271
ufs_vnops.c 1.231 originated in lfs
ufs_vnops.c 1.232 does not apply to lfs
|
| 1.23 | 20-Jun-2016 |
dholland | Merge (effectively) -r1.78 of ufs_extern.h: shift ulfs_makeinode to
lfs_vnops.c and make it file-static there, as that's the only place
it's used.
|
| 1.22 | 20-Jun-2016 |
dholland | Note more already-merged versions:
inode.h 1.68 is subsumed by ulfs_inode.h 1.19
inode.h 1.69-1.72 do not apply to lfs
ufs_extern.h 1.74 was covered when lfs was moved to the new vnode cache
ufs_extern.h 1.75 is equivalent to ulfs_extern.h 1.13
ufs_extern.h 1.76-1.77 do not apply to lfs
ufsmount.h 1.42 does not apply to lfs
ufs_inode.c 1.90 is subsumed by ulfs_inode.c 1.10
ufs_inode.c 1.91-1.92 do not apply to lfs
ufs_lookup.c 1.130 is subsumed by ulfs_lookup.c 1.24
ufs_lookup.c 1.131 is equivalent to ulfs_lookup.c 1.20
ufs_lookup.c 1.132 is equivalent to ulfs_lookup.c 1.21
ufs_lookup.c 1.133 is equivalent to ulfs_lookup.c 1.22
ufs_lookup.c 1.134 is equivalent to ulfs_lookup.c 1.23
ufs_lookup.c 1.135 is equivalent to ulfs_lookup.c 1.25
ufs_quota2.c 1.38 is equivalent to ulfs_quota2.c 1.17
ufs_quota2.c 1.39 is equivalent to ulfs_quota2.c 1.16
ufs_quota2.c 1.40 is equivalent to ulfs_quota2.c 1.18
ufs_vfsops.c 1.53 is subsumed by lfs_vfsops.c 1.324
ufs_vfsops.c 1.54 is subsumed by lfs_vfsops.c 1.324
ufs_vnops.c 1.223-1.224 do not apply to lfs
|
| 1.21 | 19-Jun-2016 |
dholland | Update the ufs versions these files are synced with by 1: the
201306016 commit by hannken@ that removed references to ffs_snapgone
in ufs doesn't need to be synced into lfs.
|
| 1.20 | 21-Sep-2015 |
dholland | Add 64-bit directory entry structures, and adjust accessors accordingly.
The LFS64 directory entry has a 64-bit inode number. This is stored as
two 32-bit values to avoid inducing 64-bit alignment requirements.
The exposed type for manipulating directory entries is now
LFS_DIRHEADER, following the same convention as e.g. IFILE and SEGUSE.
(But with LFS_ on it, because.)
|
| 1.19 | 15-Sep-2015 |
dholland | Pass around struct lfs_dirheader instead of struct lfs_direct.
|
| 1.18 | 15-Sep-2015 |
dholland | Kill off the ulfs_direct_cache pool.
We no longer allocate temporary struct directs, so we don't need a
pool for them.
|
| 1.17 | 15-Sep-2015 |
dholland | Kill off ulfs_makedirentry; just pass the data to ulfs_direnter instead.
For now, move one copy of the code that allocates and fills in a
temporary struct lfs_direct to the top of ulfs_direnter; but it should
go away shortly.
|
| 1.16 | 01-Sep-2015 |
dholland | Add new accessors for the d_type and d_namlen fields of struct lfs_direct.
Napalm the old byteswap access logic for these.
|
| 1.15 | 31-May-2015 |
hannken | Change lfs from hash table to vcache.
- Change lfs_valloc() to return an inode number and version instead of
a vnode and move lfs_ialloc() and lfs_vcreate() to new lfs_init_vnode().
- Add lfs_valloc_fixed() to allocate a known inode, used by kernel
roll forward.
- Remove lfs_*ref(), these functions cannot coexist with vcache and
their commented behaviour is far away from their implementation.
- Add the cleaner lwp and blockinfo to struct ulfsmount so lfs_loadvnode()
may use hints from the cleaner.
- Remove vnode locks from ulfs_lookup() like we did with ufs_lookup().
|
| 1.14 | 27-Mar-2015 |
riastradh | Disentangle buffer-cached I/O from page-cached I/O in UFS.
Page-cached I/O is used for regular files, and is initiated by VFS
users such as userland and NFS.
Buffer-cached I/O is used for directories and symlinks, and is issued
only internally by UFS.
New UFS routine ufs_bufio replaces vn_rdwr for internal use.
ufs_bufio is implemented by new UFS operations uo_bufrd/uo_bufwr,
which sit in ufs_readwrite.c alongside the VOP_READ/VOP_WRITE
implementations.
I preserved the code as much as possible and will leave further
simplification for future commits. I kept the ulfs_readwrite.c
copypasta close to ufs_readwrite.c in case we ever want to merge them
back; likewise ext2fs_readwrite.c.
No externally visible semantic change. All atf fs tests still pass.
|
| 1.13 | 25-May-2014 |
hannken | branches: 1.13.4;
Remove ulfs_checkpath() and ulfs_readdotdot(). These are relics
from the pre-genfs_rename era.
|
| 1.12 | 17-May-2014 |
dholland | branches: 1.12.2;
Merge ulfs_mkdir into lfs_mkdir.
|
| 1.11 | 17-May-2014 |
dholland | Merge ulfs_symlink into lfs_symlink.
|
| 1.10 | 28-Jul-2013 |
dholland | branches: 1.10.2;
Bring in a copy of ffs_quota2_mount() for reference.
Add stuff to struct lfs that it needs to initialize.
Clear these fields in mount as there's no on-disk support for quota2;
but this increases the chances of being able to add it (or something
like it) in the future.
|
| 1.9 | 28-Jul-2013 |
dholland | Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.8 | 20-Jul-2013 |
dholland | G/C unused pieces.
|
| 1.7 | 20-Jul-2013 |
dholland | Collect the pieces of lfs rename into lfs_rename.c, and sprinkle static.
|
| 1.6 | 08-Jun-2013 |
dholland | branches: 1.6.2; 1.6.4; 1.6.6;
struct direct -> struct lfs_direct
struct dirtemplate -> struct lfs_dirtemplate
struct odirtemplate -> struct lfs_odirtemplate
DT_* -> LFS_DT_*
|
| 1.5 | 06-Jun-2013 |
dholland | Fix some exposed symbols:
LOSTFOUNDINO -> LFS_LOSTFOUNDINO
struct ufid -> struct ulfs_ufid
|
| 1.4 | 06-Jun-2013 |
dholland | Apparently we also need to cut and paste ffs_snapgone() in order to be
able to link the ufs code.
Instead of actually cutting and pasting it (as it depends on ffs-only
things) implement it as panic. Probably we'll be able to demonstrate
later that it's unreachable.
XXX: Someone should add snapgone to struct ufs_ops in ufs/ufsmount.h,
XXX: and fix ufs/ufs_lookup.c to not hardwire ffs.
|
| 1.3 | 06-Jun-2013 |
dholland | Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.6.6.1 | 23-Jul-2013 |
riastradh | sync with HEAD
|
| 1.6.4.2 | 18-May-2014 |
rmind | sync with head
|
| 1.6.4.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.6.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.6.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.6.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.6.2.1 | 08-Jun-2013 |
tls | file ulfs_extern.h was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.10.2.1 | 10-Aug-2014 |
tls | Rebase.
|
| 1.12.2.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.12.2.1 | 17-May-2014 |
yamt | file ulfs_extern.h was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.13.4.4 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.13.4.3 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.13.4.2 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.13.4.1 | 06-Apr-2015 |
skrll | Sync with HEAD
|
| 1.24.24.1 | 17-Jan-2020 |
ad | Sync with head.
|
| 1.24.18.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.25.10.1 | 01-Aug-2021 |
thorpej | Sync with HEAD.
|
| 1.6 | 31-May-2015 |
hannken | Change lfs from hash table to vcache.
- Change lfs_valloc() to return an inode number and version instead of
a vnode and move lfs_ialloc() and lfs_vcreate() to new lfs_init_vnode().
- Add lfs_valloc_fixed() to allocate a known inode, used by kernel
roll forward.
- Remove lfs_*ref(), these functions cannot coexist with vcache and
their commented behaviour is far away from their implementation.
- Add the cleaner lwp and blockinfo to struct ulfsmount so lfs_loadvnode()
may use hints from the cleaner.
- Remove vnode locks from ulfs_lookup() like we did with ufs_lookup().
|
| 1.5 | 20-Apr-2015 |
riastradh | Make vget always return vnode unlocked.
Convert callers who want locks to use vn_lock afterward.
Add extra argument so the compiler will report stragglers.
|
| 1.4 | 27-Feb-2014 |
hannken | branches: 1.4.4; 1.4.8;
The current implementation of vn_lock() is racy. Modification of
the vnode operations vector for active vnodes is unsafe because it
is not known whether deadfs or the original file system will be
called.
- Pass down LK_RETRY to the lock operation (hint for deadfs only).
- Change deadfs lock operation to return ENOENT if LK_RETRY is unset.
- Change all other lock operations to check for dead vnode once
the vnode is locked and unlock and return ENOENT in this case.
With these changes in place vnode lock operations will never succeed
after vclean() has marked the vnode as VI_XLOCK and before vclean()
has changed the operations vector.
Adresses PR kern/37706 (Forced unmount of file systems is unsafe)
Discussed on tech-kern.
Welcome to 6.99.33
|
| 1.3 | 06-Jun-2013 |
dholland | branches: 1.3.2; 1.3.4;
Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.3.4.1 | 18-May-2014 |
rmind | sync with head
|
| 1.3.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.3.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.3.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.3.2.1 | 06-Jun-2013 |
tls | file ulfs_ihash.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.4.8.1 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.4.4.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.4.4.1 | 27-Feb-2014 |
yamt | file ulfs_ihash.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.26 | 05-Sep-2020 |
riastradh | Round of uvm.h cleanup.
The poorly named uvm.h is generally supposed to be for uvm-internal
users only.
- Narrow it to files that actually need it -- mostly files that need
to query whether curlwp is the pagedaemon, which should maybe be
exposed by an external header.
- Use uvm_extern.h where feasible and uvm_*.h for things not exposed
by it. We should split up uvm_extern.h but this will serve for now
to reduce the uvm.h dependencies.
- Use uvm_stat.h and #ifdef UVMHIST uvm.h for files that use
UVMHIST(ubchist), since ubchist is declared in uvm.h but the
reference evaporates if UVMHIST is not defined, so we reduce header
file dependencies.
- Make uvm_device.h and uvm_swap.h independently includable while
here.
ok chs@
|
| 1.25 | 23-Feb-2020 |
ad | UVM locking changes, proposed on tech-kern:
- Change the lock on uvm_object, vm_amap and vm_anon to be a RW lock.
- Break v_interlock and vmobjlock apart. v_interlock remains a mutex.
- Do partial PV list locking in the x86 pmap. Others to follow later.
|
| 1.24 | 15-Jan-2020 |
ad | Merge from yamt-pagecache (after much testing):
- Reduce unnecessary page scan in putpages esp. when an object has a ton of
pages cached but only a few of them are dirty.
- Reduce the number of pmap operations by tracking page dirtiness more
precisely in uvm layer.
|
| 1.23 | 31-Dec-2019 |
ad | branches: 1.23.2;
- Add and use wrapper functions that take and acquire page interlocks, and pairs
of page interlocks. Require that the page interlock be held over calls to
uvm_pageactivate(), uvm_pagewire() and similar.
- Solve the concurrency problem with page replacement state. Rather than
updating the global state synchronously, set an intended state on
individual pages (active, inactive, enqueued, dequeued) while holding the
page interlock. After the interlock is released put the pages on a 128
entry per-CPU queue for their state changes to be made real in batch.
This results in in a ~400 fold decrease in contention on my test system.
Proposed on tech-kern but modified to use the page interlock rather than
atomics to synchronise as it's much easier to maintain that way, and
cheaper.
|
| 1.22 | 13-Dec-2019 |
ad | Break the global uvm_pageqlock into a per-page identity lock and a private
lock for use of the pagedaemon policy code. Discussed on tech-kern.
PR kern/54209: NetBSD 8 large memory performance extremely low
PR kern/54210: NetBSD-8 processes presumably not exiting
PR kern/54727: writing a large file causes unreasonable system behaviour
|
| 1.21 | 28-Oct-2017 |
pgoyette | branches: 1.21.4;
Update the kernhist(9) kernel history code to address issues identified
in PR kern/52639, as well as some general cleaning-up...
(As proposed on tech-kern@ with additional changes and enhancements.)
Details of changes:
* All history arguments are now stored as uintmax_t values[1], both in
the kernel and in the structures used for exporting the history data
to userland via sysctl(9). This avoids problems on some architectures
where passing a 64-bit (or larger) value to printf(3) can cause it to
process the value as multiple arguments. (This can be particularly
problematic when printf()'s format string is not a literal, since in
that case the compiler cannot know how large each argument should be.)
* Update the data structures used for exporting kernel history data to
include a version number as well as the length of history arguments.
* All [2] existing users of kernhist(9) have had their format strings
updated. Each format specifier now includes an explicit length
modifier 'j' to refer to numeric values of the size of uintmax_t.
* All [2] existing users of kernhist(9) have had their format strings
updated to replace uses of "%p" with "%#jx", and the pointer
arguments are now cast to (uintptr_t) before being subsequently cast
to (uintmax_t). This is needed to avoid compiler warnings about
casting "pointer to integer of a different size."
* All [2] existing users of kernhist(9) have had instances of "%s" or
"%c" format strings replaced with numeric formats; several instances
of mis-match between format string and argument list have been fixed.
* vmstat(1) has been modified to handle the new size of arguments in the
history data as exported by sysctl(9).
* vmstat(1) now provides a warning message if the history requested with
the -u option does not exist (previously, this condition was silently
ignored, with only a single blank line being printed).
* vmstat(1) now checks the version and argument length included in the
data exported via sysctl(9) and exits if they do not match the values
with which vmstat was built.
* The kernhist(9) man-page has been updated to note the additional
requirements imposed on the format strings, along with several other
minor changes and enhancements.
[1] It would have been possible to use an explicit length (for example,
uint64_t) for the history arguments. But that would require another
"rototill" of all the users in the future when we add support for an
architecture that supports a larger size. Also, the printf(3) format
specifiers for explicitly-sized values, such as "%"PRIu64, are much
more verbose (and less aesthetically appealing, IMHO) than simply
using "%ju".
[2] I've tried very hard to find "all [the] existing users of kernhist(9)"
but it is possible that I've missed some of them. I would be glad to
update any stragglers that anyone identifies.
|
| 1.20 | 10-Jun-2017 |
maya | Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.19 | 26-May-2017 |
riastradh | branches: 1.19.2;
Eliminate crusty debugging sludge.
We have a mostly sane vnode lifecycle now. If this needs debugging,
it should be done once at the call site of VOP_RECLAIM.
|
| 1.18 | 11-Apr-2017 |
riastradh | Make VOP_INACTIVE preserve vnode lock on return.
Discussed on tech-kern:
https://mail-index.netbsd.org/tech-kern/2017/04/01/msg021751.html
Ride 7.99.68, a bumpy bus of incremental vfs improvements!
|
| 1.17 | 30-Mar-2017 |
hannken | Remove now redundant calls to fstrans_start()/fstrans_done().
Add fstrans_start()/fstrans_done() to lfs_putpages().
|
| 1.16 | 20-Aug-2016 |
hannken | branches: 1.16.2;
Remove now obsolete operation vcache_remove().
Welcome to 7.99.36
|
| 1.15 | 20-Jun-2016 |
dholland | branches: 1.15.2;
One more batch of already-synced ufs changes:
ufs_extern.h 1.79 is equivalent to ulfs_extern.h 1.14
ufsmount.h 1.43 is (roughly) equivalent to lfs_extern.h 1.102
ufs_inode.c 1.94 does not apply to lfs
ufs_inode.c 1.95 does not apply to lfs either
ufs_readwrite.c 1.108 is equivalent to ulfs_readwrite.c 1.8
ufs_readwrite.c 1.109 is equivalent to ulfs_readwrite.c 1.9
ufs_readwrite.c 1.110 is equivalent to ulfs_readwrite.c 1.10
ufs_readwrite.c 1.111 does not apply to lfs
ufs_readwrite.c 1.112 is equivalent to ulfs_readwrite.c 1.11
ufs_readwrite.c 1.113 is equivalent to ulfs_readwrite.c 1.13
ufs_readwrite.c 1.114 is equivalent to ulfs_readwrite.c 1.14
ufs_readwrite.c 1.115 is equivalent to ulfs_readwrite.c 1.15
ufs_readwrite.c 1.116-1.118 does not apply to lfs
ufs_readwrite.c 1.119-1.120 are equivalent to ulfs_readwrite.c 1.16
ufs_rename.c 1.12 is equivalent to lfs_rename.c 1.8
ufs_vnops.c 1.226 is equivalent to ulfs_vnops.c 1.22 and lfs_vnops.c 1.270
ufs_vnops.c 1.227 is equivalent to ulfs_vnops.c 1.23
ufs_vnops.c 1.228-1.229 are equivalent to ulfs_vnops.c 1.24
ufs_vnops.c 1.230 is equivalent to ulfs_vnops.c 1.25 and lfs_vnops.c 1.271
ufs_vnops.c 1.231 originated in lfs
ufs_vnops.c 1.232 does not apply to lfs
|
| 1.14 | 20-Jun-2016 |
dholland | Merge ufs_inode.c 1.93: missing unlock on error path.
|
| 1.13 | 20-Jun-2016 |
dholland | Note more already-merged versions:
inode.h 1.68 is subsumed by ulfs_inode.h 1.19
inode.h 1.69-1.72 do not apply to lfs
ufs_extern.h 1.74 was covered when lfs was moved to the new vnode cache
ufs_extern.h 1.75 is equivalent to ulfs_extern.h 1.13
ufs_extern.h 1.76-1.77 do not apply to lfs
ufsmount.h 1.42 does not apply to lfs
ufs_inode.c 1.90 is subsumed by ulfs_inode.c 1.10
ufs_inode.c 1.91-1.92 do not apply to lfs
ufs_lookup.c 1.130 is subsumed by ulfs_lookup.c 1.24
ufs_lookup.c 1.131 is equivalent to ulfs_lookup.c 1.20
ufs_lookup.c 1.132 is equivalent to ulfs_lookup.c 1.21
ufs_lookup.c 1.133 is equivalent to ulfs_lookup.c 1.22
ufs_lookup.c 1.134 is equivalent to ulfs_lookup.c 1.23
ufs_lookup.c 1.135 is equivalent to ulfs_lookup.c 1.25
ufs_quota2.c 1.38 is equivalent to ulfs_quota2.c 1.17
ufs_quota2.c 1.39 is equivalent to ulfs_quota2.c 1.16
ufs_quota2.c 1.40 is equivalent to ulfs_quota2.c 1.18
ufs_vfsops.c 1.53 is subsumed by lfs_vfsops.c 1.324
ufs_vfsops.c 1.54 is subsumed by lfs_vfsops.c 1.324
ufs_vnops.c 1.223-1.224 do not apply to lfs
|
| 1.12 | 14-Nov-2015 |
pgoyette | Remove historic references to wapbl.
|
| 1.11 | 01-Sep-2015 |
dholland | Add new accessors for the d_type and d_namlen fields of struct lfs_direct.
Napalm the old byteswap access logic for these.
|
| 1.10 | 31-May-2015 |
hannken | Change lfs from hash table to vcache.
- Change lfs_valloc() to return an inode number and version instead of
a vnode and move lfs_ialloc() and lfs_vcreate() to new lfs_init_vnode().
- Add lfs_valloc_fixed() to allocate a known inode, used by kernel
roll forward.
- Remove lfs_*ref(), these functions cannot coexist with vcache and
their commented behaviour is far away from their implementation.
- Add the cleaner lwp and blockinfo to struct ulfsmount so lfs_loadvnode()
may use hints from the cleaner.
- Remove vnode locks from ulfs_lookup() like we did with ufs_lookup().
|
| 1.9 | 28-Jul-2013 |
dholland | branches: 1.9.4; 1.9.6; 1.9.8;
Remove the now-pointless ulfs ops macros.
|
| 1.8 | 28-Jul-2013 |
dholland | Get rid of the ulfs_ops table as we only have one fs in here now.
|
| 1.7 | 08-Jun-2013 |
dholland | branches: 1.7.2; 1.7.4;
There is no WAPBL in LFS.
|
| 1.6 | 08-Jun-2013 |
dholland | mp->mnt_wapbl and mp->mnt_wapbl_replay are always NULL in here.
|
| 1.5 | 06-Jun-2013 |
dholland | Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.4 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 3: rearrange config stuff.
Add new options:
LFS_EI
LFS_DIRHASH
LFS_EXTATTR
LFS_EXTATTR_AUTOSTART
LFS_QUOTA
LFS_QUOTA2
and update code referring to the corresponding FFS and UFS config
symbols to use the LFS versions. Disable the one extant reference
to APPLE_UFS in the ulfs files. Use opt_lfs.h only, not opt_ffs.h.
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.7.4.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.7.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.7.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.7.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.7.2.1 | 08-Jun-2013 |
tls | file ulfs_inode.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.9.8.6 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.9.8.5 | 05-Oct-2016 |
skrll | Sync with HEAD
|
| 1.9.8.4 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.9.8.3 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.9.8.2 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.9.8.1 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.9.6.1 | 10-Jul-2016 |
martin | Pull up following revision(s) (requested by dholland in ticket #1188):
sys/ufs/lfs/ulfs_inode.c: revision 1.14
Merge ufs_inode.c 1.93: missing unlock on error path.
|
| 1.9.4.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.9.4.1 | 28-Jul-2013 |
yamt | file ulfs_inode.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.15.2.1 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.16.2.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.19.2.2 | 02-Nov-2017 |
snj | Pull up following revision(s) (requested by pgoyette in ticket #335):
share/man/man9/kernhist.9: 1.5-1.8
sys/arch/acorn26/acorn26/pmap.c: 1.39
sys/arch/arm/arm32/fault.c: 1.105 via patch
sys/arch/arm/arm32/pmap.c: 1.350, 1.359
sys/arch/arm/broadcom/bcm2835_bsc.c: 1.7
sys/arch/arm/omap/if_cpsw.c: 1.20
sys/arch/arm/omap/tiotg.c: 1.7
sys/arch/evbarm/conf/RPI2_INSTALL: 1.3
sys/dev/ic/sl811hs.c: 1.98
sys/dev/usb/ehci.c: 1.256
sys/dev/usb/if_axe.c: 1.83
sys/dev/usb/motg.c: 1.18
sys/dev/usb/ohci.c: 1.274
sys/dev/usb/ucom.c: 1.119
sys/dev/usb/uhci.c: 1.277
sys/dev/usb/uhub.c: 1.137
sys/dev/usb/umass.c: 1.160-1.162
sys/dev/usb/umass_quirks.c: 1.100
sys/dev/usb/umass_scsipi.c: 1.55
sys/dev/usb/usb.c: 1.168
sys/dev/usb/usb_mem.c: 1.70
sys/dev/usb/usb_subr.c: 1.221
sys/dev/usb/usbdi.c: 1.175
sys/dev/usb/usbdi_util.c: 1.67-1.70
sys/dev/usb/usbroothub.c: 1.3
sys/dev/usb/xhci.c: 1.75
sys/external/bsd/drm2/dist/drm/i915/i915_gem.c: 1.34
sys/kern/kern_history.c: 1.15
sys/kern/kern_xxx.c: 1.74
sys/kern/vfs_bio.c: 1.275-1.276
sys/miscfs/genfs/genfs_io.c: 1.71
sys/sys/kernhist.h: 1.21
sys/ufs/ffs/ffs_balloc.c: 1.63
sys/ufs/lfs/lfs_vfsops.c: 1.361
sys/ufs/lfs/ulfs_inode.c: 1.21
sys/ufs/lfs/ulfs_vnops.c: 1.52
sys/ufs/ufs/ufs_inode.c: 1.102
sys/ufs/ufs/ufs_vnops.c: 1.239
sys/uvm/pmap/pmap.c: 1.37-1.39
sys/uvm/pmap/pmap_tlb.c: 1.22
sys/uvm/uvm_amap.c: 1.108
sys/uvm/uvm_anon.c: 1.64
sys/uvm/uvm_aobj.c: 1.126
sys/uvm/uvm_bio.c: 1.91
sys/uvm/uvm_device.c: 1.66
sys/uvm/uvm_fault.c: 1.201
sys/uvm/uvm_km.c: 1.144
sys/uvm/uvm_loan.c: 1.85
sys/uvm/uvm_map.c: 1.353
sys/uvm/uvm_page.c: 1.194
sys/uvm/uvm_pager.c: 1.111
sys/uvm/uvm_pdaemon.c: 1.109
sys/uvm/uvm_swap.c: 1.175
sys/uvm/uvm_vnode.c: 1.103
usr.bin/vmstat/vmstat.c: 1.219
Reorder to test for null before null deref in debug code
--
Reorder to test for null before null deref in debug code
--
KNF
--
No need for '\n' in UVMHIST_LOG
--
normalise a BIOHIST log message
--
Update the kernhist(9) kernel history code to address issues identified
in PR kern/52639, as well as some general cleaning-up...
(As proposed on tech-kern@ with additional changes and enhancements.)
Details of changes:
* All history arguments are now stored as uintmax_t values[1], both in
the kernel and in the structures used for exporting the history data
to userland via sysctl(9). This avoids problems on some architectures
where passing a 64-bit (or larger) value to printf(3) can cause it to
process the value as multiple arguments. (This can be particularly
problematic when printf()'s format string is not a literal, since in
that case the compiler cannot know how large each argument should be.)
* Update the data structures used for exporting kernel history data to
include a version number as well as the length of history arguments.
* All [2] existing users of kernhist(9) have had their format strings
updated. Each format specifier now includes an explicit length
modifier 'j' to refer to numeric values of the size of uintmax_t.
* All [2] existing users of kernhist(9) have had their format strings
updated to replace uses of "%p" with "%#jx", and the pointer
arguments are now cast to (uintptr_t) before being subsequently cast
to (uintmax_t). This is needed to avoid compiler warnings about
casting "pointer to integer of a different size."
* All [2] existing users of kernhist(9) have had instances of "%s" or
"%c" format strings replaced with numeric formats; several instances
of mis-match between format string and argument list have been fixed.
* vmstat(1) has been modified to handle the new size of arguments in the
history data as exported by sysctl(9).
* vmstat(1) now provides a warning message if the history requested with
the -u option does not exist (previously, this condition was silently
ignored, with only a single blank line being printed).
* vmstat(1) now checks the version and argument length included in the
data exported via sysctl(9) and exits if they do not match the values
with which vmstat was built.
* The kernhist(9) man-page has been updated to note the additional
requirements imposed on the format strings, along with several other
minor changes and enhancements.
[1] It would have been possible to use an explicit length (for example,
uint64_t) for the history arguments. But that would require another
"rototill" of all the users in the future when we add support for an
architecture that supports a larger size. Also, the printf(3)
format
specifiers for explicitly-sized values, such as "%"PRIu64, are much
more verbose (and less aesthetically appealing, IMHO) than simply
using "%ju".
[2] I've tried very hard to find "all [the] existing users of
kernhist(9)"
but it is possible that I've missed some of them. I would be glad
to
update any stragglers that anyone identifies.
--
For some reason this single kernel seems to have outgrown its declared
size as a result of the kernhist(9) changes. Bump the size.
XXX The amount of increase may be excessive - anyone with more detailed
XXX knowledge please feel free to further adjust the value
appropriately.
--
Misssed one cast of pointer --> uintptr_t in previous kernhist(9) commit
--
And yet another one. :(
--
Use correct mark-up for NetBSD version.
--
More improvements in grammar and readability.
--
Remove a stray '"' (obvious typo) and add a couple of casts that are
probably needed.
--
And replace an instance of "%p" conversion with "%#jx"
--
Whitespace fix. Give Bl tag table a width. Fix Xr.
|
| 1.19.2.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.21.4.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.23.2.2 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.23.2.1 | 17-Jan-2020 |
ad | Sync with head.
|
| 1.25 | 17-Feb-2024 |
mlelstv | Whitespace.
|
| 1.24 | 10-Jun-2017 |
maya | Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.23 | 08-Jun-2017 |
chs | move some buffer cache internals declarations from buf.h to vfs_bio.c.
this is needed to avoid name conflicts with ZFS and also
makes it clearer that other code shouldn't be messing with these.
remove the LFS debug code that poked around in bufqueues and
remove the BQ_EMPTY bufqueue since nothing uses it anymore.
provide a function to let LFS and wapbl read the value of nbuf for now.
|
| 1.22 | 21-Jun-2016 |
dholland | branches: 1.22.10;
Revert version 1.19 (make ufid_ino in struct ulfs_ufid 64-bit) -- via
a twisty maze of marginal if not illegal type punning it breaks the
cleaner.
This will need to be done over, but it requires substantially more
mechanism and compat ioctls. Booo.
|
| 1.21 | 20-Jun-2016 |
dholland | u_int{8,16,32,64}_t -> uint{8,16,32,64}_t in remaining lfs headers.
|
| 1.20 | 20-Jun-2016 |
dholland | Note more already-merged versions:
inode.h 1.68 is subsumed by ulfs_inode.h 1.19
inode.h 1.69-1.72 do not apply to lfs
ufs_extern.h 1.74 was covered when lfs was moved to the new vnode cache
ufs_extern.h 1.75 is equivalent to ulfs_extern.h 1.13
ufs_extern.h 1.76-1.77 do not apply to lfs
ufsmount.h 1.42 does not apply to lfs
ufs_inode.c 1.90 is subsumed by ulfs_inode.c 1.10
ufs_inode.c 1.91-1.92 do not apply to lfs
ufs_lookup.c 1.130 is subsumed by ulfs_lookup.c 1.24
ufs_lookup.c 1.131 is equivalent to ulfs_lookup.c 1.20
ufs_lookup.c 1.132 is equivalent to ulfs_lookup.c 1.21
ufs_lookup.c 1.133 is equivalent to ulfs_lookup.c 1.22
ufs_lookup.c 1.134 is equivalent to ulfs_lookup.c 1.23
ufs_lookup.c 1.135 is equivalent to ulfs_lookup.c 1.25
ufs_quota2.c 1.38 is equivalent to ulfs_quota2.c 1.17
ufs_quota2.c 1.39 is equivalent to ulfs_quota2.c 1.16
ufs_quota2.c 1.40 is equivalent to ulfs_quota2.c 1.18
ufs_vfsops.c 1.53 is subsumed by lfs_vfsops.c 1.324
ufs_vfsops.c 1.54 is subsumed by lfs_vfsops.c 1.324
ufs_vnops.c 1.223-1.224 do not apply to lfs
|
| 1.19 | 20-Jun-2016 |
dholland | Merge -r1.67 of ufs/inode.h: make the inode field of a filehandle
64-bit instead of truncating to 32 bits. Note that if you're serving
nfs off lfs (but I don't think you are as I think there are known
fatal problems doing so) you'll need to reboot your clients after this
change.
I've used a 64-bit value explicitly instead of ino_t (as in the ufs
structure) because this is a structure whose size ought to be well
defined. I remember some discussion of this when the ufs change was
committed, but not the conclusion (if any) -- if anyone hates this it
can be changed to ino_t easily enough.
|
| 1.18 | 20-Jun-2016 |
dholland | Merge ufs/inode.h 1.66: remove i_hash from struct inode. This is the
hash table entry link from the old per-fs vnode cache and we don't
need it any more.
|
| 1.17 | 19-Jun-2016 |
dholland | Mark ufs file versions we're already synced with.
|
| 1.16 | 01-Sep-2015 |
dholland | Use the lfs dinode accessors in place of the ufs-derived ones.
(Mostly.)
The ufs-derived ones are fake structure member macros, which are gross
and not very safe. Also, it seems that a lot of places in the lfs code
were using the ffsv1 branch of them unconditionally, and this way it's
guaranteed all those places have been updated.
Found while doing this: for non-devices, have getattr produce NODEV
in the rdev field instead of leaking the address of the first direct
block.
|
| 1.15 | 12-Aug-2015 |
dholland | Hack up dinode usage to be 64 vs. 32 as needed. Part 1.
(This part changes the native lfs code; the ufs-derived code already
has 64 vs. 32 logic, but as aspects of it are unsafe, and don't
entirely interoperate cleanly with the lfs 64/32 stuff, pass 2 will be
rehashing that.)
|
| 1.14 | 24-Jul-2015 |
dholland | More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)
The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.
|
| 1.13 | 24-Jul-2015 |
dholland | Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)
It also gets rid of cpp abuse in the form of fake structure member
macros.
Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)
XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.
|
| 1.12 | 17-May-2014 |
dholland | branches: 1.12.2; 1.12.6;
Remove the DIROP macros. They are evil, especially the CREATE ones.
This results in some duplicate logic in the creation vnops (symlink,
mknod, create, mkdir) but we will probably be able to factor it out in
a more sensible way later.
Now the creation vnops call getnewvnode explicitly instead of under
multiple layers of obscure gunk. Then we explicitly do lfs_set_dirop,
and afterwards lfs_unset_dirop.
|
| 1.11 | 18-Mar-2014 |
riastradh | branches: 1.11.2;
Merge riastradh-drm2 to HEAD.
|
| 1.10 | 20-Jul-2013 |
dholland | Collect the pieces of lfs rename into lfs_rename.c, and sprinkle static.
|
| 1.9 | 18-Jun-2013 |
christos | branches: 1.9.2; 1.9.4; 1.9.6;
Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.
|
| 1.8 | 18-Jun-2013 |
dholland | Tuck away a bunch of symbols that don't need to be public.
|
| 1.7 | 08-Jun-2013 |
dholland | ulfs_dir.h has been emptied; remove it.
|
| 1.6 | 08-Jun-2013 |
dholland | Split the definitions suitable for userland out of ulfs_inode.h into
lfs_inode.h. Since fsck_lfs, newfs_lfs, and lfs_cleanerd want to reuse
the inode structure for their own internal use, and some of them share
parts of the kernel code as well, the best way forward is to provide a
relatively sanitized header that doesn't bring in stray material.
Shuffle a few other definitions around so that lfs_inode.h depends
only on lfs.h.
Install lfs_inode.h into /usr/include.
|
| 1.5 | 06-Jun-2013 |
dholland | Cleanups to reduce symbol and header exposure:
- move struct ufid from ulfs_inode.h to lfs.h
- lfs.h needs sys/mount.h and sys/pool.h
- ulfs_quota2_subr.c needs lfs_inode.h
- remove ulfs_inode.h from lfs.h in favor of ulfs_dinode.h
- move ULFS_NDADDR, ULFS_NIADDR, ULFS_NXADDR from ulfs_dinode.h to lfs.h
- remove ulfs_dinode.h from lfs.h
- add lfs.h to ulfs_dinode.h
|
| 1.4 | 06-Jun-2013 |
dholland | Remove stray references to ext2fs, chfs, ffs, and mfs.
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.9.6.1 | 23-Jul-2013 |
riastradh | sync with HEAD
|
| 1.9.4.2 | 18-May-2014 |
rmind | sync with head
|
| 1.9.4.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.9.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.9.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.9.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.9.2.1 | 18-Jun-2013 |
tls | file ulfs_inode.h was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.11.2.1 | 10-Aug-2014 |
tls | Rebase.
|
| 1.12.6.3 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.12.6.2 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.12.6.1 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.12.2.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.12.2.1 | 17-May-2014 |
yamt | file ulfs_inode.h was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.22.10.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.48 | 08-Sep-2024 |
rillig | fix a/an grammar in obvious cases
|
| 1.47 | 06-Aug-2022 |
andvar | branches: 1.47.10;
s/blity/bility/ in various words, mainly in comments.
|
| 1.46 | 05-Sep-2020 |
riastradh | Revert "ufs: Prevent mkdir from choking on deleted directories."
This change made no sense and should not have been committed.
|
| 1.45 | 05-Sep-2020 |
riastradh | ufs: Prevent mkdir from choking on deleted directories.
Fix some missing uvm_vnp_setsize in screw cases while here.
|
| 1.44 | 16-May-2020 |
christos | Add ACL support for FFS. From FreeBSD.
|
| 1.43 | 04-Apr-2020 |
ad | Merge the remaining changes from the ad-namecache branch, affecting namei()
and getcwd():
- push vnode locking back as far as possible.
- do most lookups directly in the namecache, avoiding vnode locks & refs.
- don't block new refs to vnodes across VOP_INACTIVE().
- get shared locks for VOP_LOOKUP() if the file system supports it.
- correct lock types for VOP_ACCESS() / VOP_GETATTR() in a few places.
Possible future enhancements:
- make the lookups lockless.
- support dotdot lookups by being lockless and inferring absence of chroot.
- maybe make it work for layered file systems.
- avoid vnode references at the root & cwd.
|
| 1.42 | 14-Mar-2020 |
ad | - Hide the details of SPCF_SHOULDYIELD and related behind a couple of small
functions: preempt_point() and preempt_needed().
- preempt(): if the LWP has exceeded its timeslice in kernel, strip it of
any priority boost gained earlier from blocking.
|
| 1.41 | 10-Jun-2017 |
maya | branches: 1.41.6; 1.41.12;
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.40 | 30-Mar-2017 |
hannken | branches: 1.40.6;
Remove now redundant calls to fstrans_start()/fstrans_done().
Add fstrans_start()/fstrans_done() to lfs_putpages().
|
| 1.39 | 20-Jun-2016 |
dholland | branches: 1.39.2; 1.39.4;
Note more already-merged versions:
inode.h 1.68 is subsumed by ulfs_inode.h 1.19
inode.h 1.69-1.72 do not apply to lfs
ufs_extern.h 1.74 was covered when lfs was moved to the new vnode cache
ufs_extern.h 1.75 is equivalent to ulfs_extern.h 1.13
ufs_extern.h 1.76-1.77 do not apply to lfs
ufsmount.h 1.42 does not apply to lfs
ufs_inode.c 1.90 is subsumed by ulfs_inode.c 1.10
ufs_inode.c 1.91-1.92 do not apply to lfs
ufs_lookup.c 1.130 is subsumed by ulfs_lookup.c 1.24
ufs_lookup.c 1.131 is equivalent to ulfs_lookup.c 1.20
ufs_lookup.c 1.132 is equivalent to ulfs_lookup.c 1.21
ufs_lookup.c 1.133 is equivalent to ulfs_lookup.c 1.22
ufs_lookup.c 1.134 is equivalent to ulfs_lookup.c 1.23
ufs_lookup.c 1.135 is equivalent to ulfs_lookup.c 1.25
ufs_quota2.c 1.38 is equivalent to ulfs_quota2.c 1.17
ufs_quota2.c 1.39 is equivalent to ulfs_quota2.c 1.16
ufs_quota2.c 1.40 is equivalent to ulfs_quota2.c 1.18
ufs_vfsops.c 1.53 is subsumed by lfs_vfsops.c 1.324
ufs_vfsops.c 1.54 is subsumed by lfs_vfsops.c 1.324
ufs_vnops.c 1.223-1.224 do not apply to lfs
|
| 1.38 | 20-Jun-2016 |
dholland | More already-merged or equivalent changes:
ufs_dirhash.c 1.36 corresponds to ulfs_dirhash.c 1.8
ufs_extattr.c 1.43 corresponds to ulfs_extattr.c 1.7
ufs_lookup.c 1.126 does not apply to lfs
ufs_lookup.c 1.127 we already have
ufs_lookup.c 1.128 does not apply to lfs
ufs_lookup.c 1.129 corresponds to ulfs_lookup.c 1.19
ufs_quota1.c 1.19 corresponds to ulfs_quota1.c 1.7
ufs_quota1.c 1.20 corresponds to ulfs_quota1.c 1.8
ufs_quota2.c 1.36 we have equivalent changes for
ufs_rename.c 1.9 corresponds to lfs_rename.c 1.5
ufs_rename.c 1.10 corresponds to lfs_rename.c 1.6
ufs_vnops.c 1.219 corresponds to lfs_vnops.c 1.260 and ulfs_vnops.c 1.19
ufs_vnops.c 1.220 corresponds to lfs_vnops.c 1.261 and ulfs_vnops.c 1.20
ufs_vnops.c 1.221 was superseded by later changes
ufs_vnops.c 1.222 got fixed independently in lfs
|
| 1.37 | 19-Jun-2016 |
dholland | we already have ufs_lookup.c 1.125 and ufs_vnops.c 1.218.
|
| 1.36 | 19-Jun-2016 |
dholland | Update the ufs versions these files are synced with by 1: the
201306016 commit by hannken@ that removed references to ffs_snapgone
in ufs doesn't need to be synced into lfs.
|
| 1.35 | 14-Nov-2015 |
pgoyette | Remove historic references to wapbl.
|
| 1.34 | 21-Sep-2015 |
dholland | Add 64-bit directory entry structures, and adjust accessors accordingly.
The LFS64 directory entry has a 64-bit inode number. This is stored as
two 32-bit values to avoid inducing 64-bit alignment requirements.
The exposed type for manipulating directory entries is now
LFS_DIRHEADER, following the same convention as e.g. IFILE and SEGUSE.
(But with LFS_ on it, because.)
|
| 1.33 | 21-Sep-2015 |
dholland | Oops; LFS_DIRECTSIZ() is going to need the fs as an argument.
Also, it turns out that dirhash needs a compile-time-constant version
of LFS_DIRECTSIZ(LFS_MAXNAMLEN+1), independent of 64-vs-32, so create
LFS_MAXDIRENTRYSIZE for this. Sigh.
|
| 1.32 | 15-Sep-2015 |
dholland | Pass around struct lfs_dirheader instead of struct lfs_direct.
|
| 1.31 | 15-Sep-2015 |
dholland | Add an accessor function for directory names.
|
| 1.30 | 15-Sep-2015 |
dholland | Tidyups/fixes preparatory to making d_name[] in struct lfs_direct size
0 instead of size LFS_MAXNAMLEN+1, and preparatory to having accessor
functions for d_name. In particular, don't create prototype entries
and copy them, and access the name field only for directory structures
that are in buffers with space for the name to exist.
|
| 1.29 | 15-Sep-2015 |
dholland | Tidy up ulfs_direnter: don't malloc a temporary struct lfs_direct
and double-copy it. Just write to the destination buffer.
|
| 1.28 | 15-Sep-2015 |
dholland | Kill off ulfs_makedirentry; just pass the data to ulfs_direnter instead.
For now, move one copy of the code that allocates and fills in a
temporary struct lfs_direct to the top of ulfs_direnter; but it should
go away shortly.
|
| 1.27 | 15-Sep-2015 |
dholland | Add and use accessor functions for more of the directory entry fields.
|
| 1.26 | 01-Sep-2015 |
dholland | Add new accessors for the d_type and d_namlen fields of struct lfs_direct.
Napalm the old byteswap access logic for these.
|
| 1.25 | 11-Jul-2015 |
mlelstv | mp->mnt_stat.f_flag is never set. Use the mnt_flag directly.
This will now actually prevent the 'bad dir' panic if the filesystem
is read-only.
|
| 1.24 | 31-May-2015 |
hannken | Change lfs from hash table to vcache.
- Change lfs_valloc() to return an inode number and version instead of
a vnode and move lfs_ialloc() and lfs_vcreate() to new lfs_init_vnode().
- Add lfs_valloc_fixed() to allocate a known inode, used by kernel
roll forward.
- Remove lfs_*ref(), these functions cannot coexist with vcache and
their commented behaviour is far away from their implementation.
- Add the cleaner lwp and blockinfo to struct ulfsmount so lfs_loadvnode()
may use hints from the cleaner.
- Remove vnode locks from ulfs_lookup() like we did with ufs_lookup().
|
| 1.23 | 28-Mar-2015 |
maxv | Remove the 'cred' argument from breadn(), and update the man page
accordingly.
ok hannken@
|
| 1.22 | 27-Mar-2015 |
riastradh | Disentangle buffer-cached I/O from page-cached I/O in UFS.
Page-cached I/O is used for regular files, and is initiated by VFS
users such as userland and NFS.
Buffer-cached I/O is used for directories and symlinks, and is issued
only internally by UFS.
New UFS routine ufs_bufio replaces vn_rdwr for internal use.
ufs_bufio is implemented by new UFS operations uo_bufrd/uo_bufwr,
which sit in ufs_readwrite.c alongside the VOP_READ/VOP_WRITE
implementations.
I preserved the code as much as possible and will leave further
simplification for future commits. I kept the ulfs_readwrite.c
copypasta close to ufs_readwrite.c in case we ever want to merge them
back; likewise ext2fs_readwrite.c.
No externally visible semantic change. All atf fs tests still pass.
|
| 1.21 | 03-Jun-2014 |
joerg | branches: 1.21.4;
Introduce two helper functions to centralise the namecache statistics
in vfs_cache.c. Use consistent locking around the per-cpu data.
|
| 1.20 | 25-May-2014 |
hannken | Remove ulfs_checkpath() and ulfs_readdotdot(). These are relics
from the pre-genfs_rename era.
|
| 1.19 | 07-Feb-2014 |
hannken | branches: 1.19.2; 1.19.4;
Change vnode operation lookup to return the resulting vnode *vpp unlocked.
Change cache_lookup() to return an unlocked vnode.
Discussed on tech-kern@
Welcome to 6.99.31
|
| 1.18 | 28-Jan-2014 |
martin | Bogus gcc 4.8 maybe-used-uninitialized warning
|
| 1.17 | 25-Oct-2013 |
martin | Mark a diagnostic-only variable
|
| 1.16 | 17-Oct-2013 |
christos | - remove unused variables
- add debug ifdefs for debugging variables
- __USE() where appropriate.
|
| 1.15 | 28-Jul-2013 |
dholland | Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.14 | 28-Jul-2013 |
dholland | Remove the now-pointless ulfs ops macros.
|
| 1.13 | 28-Jul-2013 |
dholland | Get rid of the ulfs_ops table as we only have one fs in here now.
|
| 1.12 | 18-Jun-2013 |
christos | branches: 1.12.2; 1.12.4;
Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.
|
| 1.11 | 08-Jun-2013 |
dholland | ulfs_dir.h has been emptied; remove it.
|
| 1.10 | 08-Jun-2013 |
dholland | There is no WAPBL in LFS.
|
| 1.9 | 08-Jun-2013 |
dholland | DIRBLKSIZ -> LFS_DIRBLKSIZ
DIRECTSIZ -> LFS_DIRECTSIZ
DIRSIZ -> LFS_DIRSIZ
OLDDIRFMT -> LFS_OLDDIRFMT
NEWDIRFMT -> LFS_NEWDIRFMT
IFTODT -> LFS_IFTODT
DTTOIF -> LFS_DTTOIF
|
| 1.8 | 08-Jun-2013 |
dholland | Move stuff to lfs.h that's needed by userland:
LFS_DT_*
ULFS_ROOTINO
ULFS_WINO
struct lfs_direct
struct lfs_dirtemplate
struct lfs_odirtemplate
struct ulfs_args
Also fix FFS_MAXNAMLEN -> LFS_MAXNAMLEN in several places.
|
| 1.7 | 08-Jun-2013 |
dholland | struct direct -> struct lfs_direct
struct dirtemplate -> struct lfs_dirtemplate
struct odirtemplate -> struct lfs_odirtemplate
DT_* -> LFS_DT_*
|
| 1.6 | 06-Jun-2013 |
dholland | Apparently we also need to cut and paste ffs_snapgone() in order to be
able to link the ufs code.
Instead of actually cutting and pasting it (as it depends on ffs-only
things) implement it as panic. Probably we'll be able to demonstrate
later that it's unreachable.
XXX: Someone should add snapgone to struct ufs_ops in ufs/ufsmount.h,
XXX: and fix ufs/ufs_lookup.c to not hardwire ffs.
|
| 1.5 | 06-Jun-2013 |
dholland | Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.4 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 3: rearrange config stuff.
Add new options:
LFS_EI
LFS_DIRHASH
LFS_EXTATTR
LFS_EXTATTR_AUTOSTART
LFS_QUOTA
LFS_QUOTA2
and update code referring to the corresponding FFS and UFS config
symbols to use the LFS versions. Disable the one extant reference
to APPLE_UFS in the ulfs files. Use opt_lfs.h only, not opt_ffs.h.
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.12.4.2 | 18-May-2014 |
rmind | sync with head
|
| 1.12.4.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.12.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.12.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.12.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.12.2.1 | 18-Jun-2013 |
tls | file ulfs_lookup.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.19.4.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.19.4.1 | 07-Feb-2014 |
yamt | file ulfs_lookup.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.19.2.1 | 10-Aug-2014 |
tls | Rebase.
|
| 1.21.4.5 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.21.4.4 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.21.4.3 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.21.4.2 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.21.4.1 | 06-Apr-2015 |
skrll | Sync with HEAD
|
| 1.39.4.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.39.2.1 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.40.6.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.41.12.1 | 19-Jan-2020 |
ad | Set IMNT_SHRLOOKUP and use it for the in-cache case. Need to check what
more can be done with tmpfs though, it can probably do the whole lookup.
|
| 1.41.6.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.47.10.1 | 02-Aug-2025 |
perseant | Sync with HEAD
|
| 1.13 | 19-Jun-2016 |
dholland | Mark ufs file versions we're already synced with.
|
| 1.12 | 28-Jun-2014 |
dholland | branches: 1.12.4;
Revert the following changes:
src/sys/sys/quotactl.h 1.37
src/sys/compat/netbsd32/netbsd32.h 1.101
src/sys/compat/netbsd32/netbsd32_netbsd.c 1.188, 1.189
src/sys/kern/vfs_quotactl.c 1.39
src/sys/kern/vfs_syscalls.c 1.483
src/sys/ufs/lfs/ulfs_quota.c 1.11
src/sys/ufs/ufs/ufs_quota.c 1.116
src/lib/libquota/quota_kernel.c 1.5
and do them correctly.
If you're going to change the name of something, you need to change
the name of *all* the things with the same name, not just a handful,
and you should change it to something similar so it still matches the
rest of the system rather than just picking an arbitrarily different
name.
Hi, Joerg.
To wit, rename the quotactl "delete" operation to "del", because
"delete" is a reserved word in C++ and for some reason Joerg wants to
run internal interfaces used only by C code through his C++ compiler.
Do not rename it to "remove" instead, because this doesn't match
libquota or the rest of the usage throughout the system; and rename
all the related identifiers, not just the ones that blew the mind of
Joerg's C++ compiler.
Because this is not a user-facing API (the only userland consumer
sys/quotactl.h is libquota) it is sort of ok to make arbitrary
source-incompatible changes; however, by the same token it's completely
unnecessary. If it *were* a user-facing API that someone might have a
semi-rational reason to want to run a C++ compiler on, it would be
incorrect to change it at this point.
|
| 1.11 | 12-Jun-2014 |
joerg | Don't t use a C++ keyword as field name.
|
| 1.10 | 22-Nov-2013 |
dholland | branches: 1.10.2; 1.10.4;
fix typo; hi christos
|
| 1.9 | 16-Nov-2013 |
dholland | This is now equivalent to ufs_quota.c -r1.115.
(it isn't quite the same textually in a few places but this doesn't
really matter)
|
| 1.8 | 18-Oct-2013 |
christos | fix unused variable warnings
|
| 1.7 | 28-Jul-2013 |
dholland | Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.6 | 06-Jun-2013 |
dholland | branches: 1.6.2; 1.6.4;
Cleanups and hacks to make lfs userland stuff build:
- lfs_cksum.c doesn't actually need ulfs_inode.h any more.
- neither does lfs_itimes.c.
- add hacks to fsck_lfs to make it compile.
- add hacks to newfs_lfs to make it compile.
- fix warning in ulfs_quota.c when quotas are fully disabled
(as I guess is happening with the rumpity version)
XXX: This commit adds -I${NETBSDSRCDIR}/sys to the Makefiles for
XXX: fsck_lfs, newfs_lfs, and lfs_cleanerd. This needs to be cleaned
XXX: up ASAP; but I consider this less problematic in the short term
XXX: than spewing ulfs_*.h into /usr/include.
|
| 1.5 | 06-Jun-2013 |
dholland | Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.4 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 3: rearrange config stuff.
Add new options:
LFS_EI
LFS_DIRHASH
LFS_EXTATTR
LFS_EXTATTR_AUTOSTART
LFS_QUOTA
LFS_QUOTA2
and update code referring to the corresponding FFS and UFS config
symbols to use the LFS versions. Disable the one extant reference
to APPLE_UFS in the ulfs files. Use opt_lfs.h only, not opt_ffs.h.
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.6.4.2 | 18-May-2014 |
rmind | sync with head
|
| 1.6.4.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.6.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.6.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.6.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.6.2.1 | 06-Jun-2013 |
tls | file ulfs_quota.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.10.4.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.10.4.1 | 22-Nov-2013 |
yamt | file ulfs_quota.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.10.2.1 | 10-Aug-2014 |
tls | Rebase.
|
| 1.12.4.1 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.7 | 20-Jun-2016 |
dholland | u_int{8,16,32,64}_t -> uint{8,16,32,64}_t in remaining lfs headers.
|
| 1.6 | 19-Jun-2016 |
dholland | Mark ufs file versions we're already synced with.
|
| 1.5 | 28-Jun-2014 |
dholland | branches: 1.5.4;
Revert the following changes:
src/sys/sys/quotactl.h 1.37
src/sys/compat/netbsd32/netbsd32.h 1.101
src/sys/compat/netbsd32/netbsd32_netbsd.c 1.188, 1.189
src/sys/kern/vfs_quotactl.c 1.39
src/sys/kern/vfs_syscalls.c 1.483
src/sys/ufs/lfs/ulfs_quota.c 1.11
src/sys/ufs/ufs/ufs_quota.c 1.116
src/lib/libquota/quota_kernel.c 1.5
and do them correctly.
If you're going to change the name of something, you need to change
the name of *all* the things with the same name, not just a handful,
and you should change it to something similar so it still matches the
rest of the system rather than just picking an arbitrarily different
name.
Hi, Joerg.
To wit, rename the quotactl "delete" operation to "del", because
"delete" is a reserved word in C++ and for some reason Joerg wants to
run internal interfaces used only by C code through his C++ compiler.
Do not rename it to "remove" instead, because this doesn't match
libquota or the rest of the usage throughout the system; and rename
all the related identifiers, not just the ones that blew the mind of
Joerg's C++ compiler.
Because this is not a user-facing API (the only userland consumer
sys/quotactl.h is libquota) it is sort of ok to make arbitrary
source-incompatible changes; however, by the same token it's completely
unnecessary. If it *were* a user-facing API that someone might have a
semi-rational reason to want to run a C++ compiler on, it would be
incorrect to change it at this point.
|
| 1.4 | 06-Jun-2013 |
dholland | branches: 1.4.2; 1.4.8; 1.4.10;
Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.4.10.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.4.10.1 | 06-Jun-2013 |
yamt | file ulfs_quota.h was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.4.8.1 | 10-Aug-2014 |
tls | Rebase.
|
| 1.4.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.4.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.4.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.4.2.1 | 06-Jun-2013 |
tls | file ulfs_quota.h was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.5.4.1 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.12 | 29-Jun-2021 |
dholland | Add containment for the cloning devices hack in vn_open.
Cloning devices (and also things like /dev/stderr) work by allocating
a struct file, stuffing it in the file table (which is a layer
violation), stuffing the file descriptor number for it in a magic
field of struct lwp (which is gross), and then "failing" with one of
two magic errnos, EDUPFD or EMOVEFD.
Before this commit, all callers of vn_open in the kernel (there are
quite a few) were expected to check for these errors and handle the
situation. Needless to say, none of them except for open() itself did,
resulting in internal negative errnos being returned to userspace.
This hack is fairly deeply rooted and cannot be eliminated all at
once. This commit adds logic to handle the magic errnos inside
vn_open; now on success vn_open returns either a vnode or an integer
file descriptor, along with a flag that says whether the underlying
code requested EDUPFD or EMOVEFD. Callers not prepared to cope with
file descriptors can pass NULL for the extra return values, in which
case if a file descriptor would be produced vn_open fails with
EOPNOTSUPP.
Since I'm rearranging vn_open's signature anyway, stop exposing struct
nameidata. Instead, take three arguments: an optional vnode to use as
the starting point (like openat()), the path, and additional namei
flags to use, restricted to NOCHROOT and TRYEMULROOT. (Other namei
behavior, e.g. NOFOLLOW, can be requested via the open flags.)
This change requires a kernel bump. Ride the one an hour ago.
(That was supposed to be coordinated; did not intend to let an hour
slip by. My fault.)
|
| 1.11 | 20-Jun-2016 |
dholland | branches: 1.11.34;
Merge -r1.20 and -r1.21 of ufs_quota1.c: widen before multiplying.
|
| 1.10 | 20-Jun-2016 |
dholland | More already-merged or equivalent changes:
ufs_dirhash.c 1.36 corresponds to ulfs_dirhash.c 1.8
ufs_extattr.c 1.43 corresponds to ulfs_extattr.c 1.7
ufs_lookup.c 1.126 does not apply to lfs
ufs_lookup.c 1.127 we already have
ufs_lookup.c 1.128 does not apply to lfs
ufs_lookup.c 1.129 corresponds to ulfs_lookup.c 1.19
ufs_quota1.c 1.19 corresponds to ulfs_quota1.c 1.7
ufs_quota1.c 1.20 corresponds to ulfs_quota1.c 1.8
ufs_quota2.c 1.36 we have equivalent changes for
ufs_rename.c 1.9 corresponds to lfs_rename.c 1.5
ufs_rename.c 1.10 corresponds to lfs_rename.c 1.6
ufs_vnops.c 1.219 corresponds to lfs_vnops.c 1.260 and ulfs_vnops.c 1.19
ufs_vnops.c 1.220 corresponds to lfs_vnops.c 1.261 and ulfs_vnops.c 1.20
ufs_vnops.c 1.221 was superseded by later changes
ufs_vnops.c 1.222 got fixed independently in lfs
|
| 1.9 | 26-Jul-2015 |
hannken | Remove bogus "mutex_enter(&mntvnode_lock)".
|
| 1.8 | 24-May-2014 |
christos | branches: 1.8.4;
Introduce a selector function to the vfs vnode iterator so that we don't
need to vget() vnodes that we are not interested at, and optimize locking
a bit. Iterator changes reviewed by Hannken (thanks), the rest of the bugs
are mine.
|
| 1.7 | 17-Mar-2014 |
hannken | branches: 1.7.2; 1.7.4;
Change lfsquota1_handle_cmd_quotaon() and lfs_q1sync()
to use vfs_vnode_iterator.
|
| 1.6 | 28-Jul-2013 |
dholland | Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.5 | 08-Jun-2013 |
dholland | branches: 1.5.2; 1.5.4;
mp->mnt_wapbl and mp->mnt_wapbl_replay are always NULL in here.
|
| 1.4 | 06-Jun-2013 |
dholland | Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.5.4.2 | 18-May-2014 |
rmind | sync with head
|
| 1.5.4.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.5.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.5.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.5.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.5.2.1 | 08-Jun-2013 |
tls | file ulfs_quota1.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.7.4.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.7.4.1 | 17-Mar-2014 |
yamt | file ulfs_quota1.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.7.2.1 | 10-Aug-2014 |
tls | Rebase.
|
| 1.8.4.2 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.8.4.1 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.11.34.1 | 01-Aug-2021 |
thorpej | Sync with HEAD.
|
| 1.5 | 20-Jun-2016 |
dholland | u_int{8,16,32,64}_t -> uint{8,16,32,64}_t in remaining lfs headers.
|
| 1.4 | 06-Jun-2013 |
dholland | branches: 1.4.2; 1.4.10; 1.4.14;
Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.4.14.1 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.4.10.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.4.10.1 | 06-Jun-2013 |
yamt | file ulfs_quota1.h was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.4.2.3 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.4.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.4.2.1 | 06-Jun-2013 |
tls | file ulfs_quota1.h was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.4 | 25-Jul-2021 |
skrll | #include <sys/param.h> for COHERENCY_UNIT (and KNF)
|
| 1.3 | 06-Jun-2013 |
dholland | branches: 1.3.2; 1.3.10; 1.3.54;
Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.3.54.1 | 01-Aug-2021 |
thorpej | Sync with HEAD.
|
| 1.3.10.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.3.10.1 | 06-Jun-2013 |
yamt | file ulfs_quota1_subr.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.3.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.3.2.1 | 06-Jun-2013 |
tls | file ulfs_quota1_subr.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.35 | 28-May-2022 |
andvar | s/grabing/grabbing/ in comments.
|
| 1.34 | 15-Oct-2021 |
andvar | fix typos in comments.
|
| 1.33 | 05-Dec-2020 |
thorpej | Remove unnecessary inclusion of <sys/timevar.h>.
|
| 1.32 | 17-Jan-2020 |
ad | branches: 1.32.6;
VFS_VGET(), VFS_ROOT(), VFS_FHTOVP(): give them a "int lktype" argument, to
allow us to get shared locks (or no lock) on the returned vnode. Matches
FreeBSD.
|
| 1.31 | 10-Jun-2017 |
maya | branches: 1.31.6; 1.31.12;
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.30 | 30-Mar-2017 |
hannken | branches: 1.30.6;
Remove now redundant calls to fstrans_start()/fstrans_done().
Add fstrans_start()/fstrans_done() to lfs_putpages().
|
| 1.29 | 20-Nov-2016 |
riastradh | branches: 1.29.2;
KASSERT(mutex_owner(...)) ---> KASSERT(mutex_owned(...))
|
| 1.28 | 07-Jul-2016 |
msaitoh | branches: 1.28.2;
KNF. Remove extra spaces. No functional change.
|
| 1.27 | 20-Jun-2016 |
dholland | Note more already-merged versions:
inode.h 1.68 is subsumed by ulfs_inode.h 1.19
inode.h 1.69-1.72 do not apply to lfs
ufs_extern.h 1.74 was covered when lfs was moved to the new vnode cache
ufs_extern.h 1.75 is equivalent to ulfs_extern.h 1.13
ufs_extern.h 1.76-1.77 do not apply to lfs
ufsmount.h 1.42 does not apply to lfs
ufs_inode.c 1.90 is subsumed by ulfs_inode.c 1.10
ufs_inode.c 1.91-1.92 do not apply to lfs
ufs_lookup.c 1.130 is subsumed by ulfs_lookup.c 1.24
ufs_lookup.c 1.131 is equivalent to ulfs_lookup.c 1.20
ufs_lookup.c 1.132 is equivalent to ulfs_lookup.c 1.21
ufs_lookup.c 1.133 is equivalent to ulfs_lookup.c 1.22
ufs_lookup.c 1.134 is equivalent to ulfs_lookup.c 1.23
ufs_lookup.c 1.135 is equivalent to ulfs_lookup.c 1.25
ufs_quota2.c 1.38 is equivalent to ulfs_quota2.c 1.17
ufs_quota2.c 1.39 is equivalent to ulfs_quota2.c 1.16
ufs_quota2.c 1.40 is equivalent to ulfs_quota2.c 1.18
ufs_vfsops.c 1.53 is subsumed by lfs_vfsops.c 1.324
ufs_vfsops.c 1.54 is subsumed by lfs_vfsops.c 1.324
ufs_vnops.c 1.223-1.224 do not apply to lfs
|
| 1.26 | 20-Jun-2016 |
dholland | Merge some cosmetic changes from ffs_quota2.c 1.5. I didn't merge the
whitespace changes.
|
| 1.25 | 20-Jun-2016 |
dholland | Remove stray 'n' in file. silly control key...
|
| 1.24 | 20-Jun-2016 |
dholland | Merge ufs_quota2.c 1.37: set grace time if lowering the limit causes
the user/group to now be over quota. From Edgar Fu�.
|
| 1.23 | 20-Jun-2016 |
dholland | More already-merged or equivalent changes:
ufs_dirhash.c 1.36 corresponds to ulfs_dirhash.c 1.8
ufs_extattr.c 1.43 corresponds to ulfs_extattr.c 1.7
ufs_lookup.c 1.126 does not apply to lfs
ufs_lookup.c 1.127 we already have
ufs_lookup.c 1.128 does not apply to lfs
ufs_lookup.c 1.129 corresponds to ulfs_lookup.c 1.19
ufs_quota1.c 1.19 corresponds to ulfs_quota1.c 1.7
ufs_quota1.c 1.20 corresponds to ulfs_quota1.c 1.8
ufs_quota2.c 1.36 we have equivalent changes for
ufs_rename.c 1.9 corresponds to lfs_rename.c 1.5
ufs_rename.c 1.10 corresponds to lfs_rename.c 1.6
ufs_vnops.c 1.219 corresponds to lfs_vnops.c 1.260 and ulfs_vnops.c 1.19
ufs_vnops.c 1.220 corresponds to lfs_vnops.c 1.261 and ulfs_vnops.c 1.20
ufs_vnops.c 1.221 was superseded by later changes
ufs_vnops.c 1.222 got fixed independently in lfs
|
| 1.22 | 14-Nov-2015 |
pgoyette | Remove historic references to wapbl.
|
| 1.21 | 28-Jul-2015 |
dholland | Add a new lfs header file: lfs_accessors.h.
This contains all the accessor functions and macros out of lfs.h.
Add an include of lfs_accessors.h after all uses of lfs.h... except
for code that wants to define its own struct lfs-alike that the
accessors are supposed to play along with. For these, set STRUCT_LFS
and include lfs_accessors.h after the necessary structure has been
defined, so that lfs_accessors.h can emit functions in terms of it.
|
| 1.20 | 24-Jul-2015 |
dholland | More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)
The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.
|
| 1.19 | 24-Jul-2015 |
dholland | Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)
It also gets rid of cpp abuse in the form of fake structure member
macros.
Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)
XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.
|
| 1.18 | 28-Mar-2015 |
maxv | Remove the 'cred' argument from bread(). Remove a now unused var in
ffs_snapshot.c. Update the man page accordingly.
ok hannken@
|
| 1.17 | 08-Dec-2014 |
justin | Avoid uninitialized variable error in some cases with gcc
|
| 1.16 | 28-Jun-2014 |
dholland | branches: 1.16.4;
Revert the following changes:
src/sys/sys/quotactl.h 1.37
src/sys/compat/netbsd32/netbsd32.h 1.101
src/sys/compat/netbsd32/netbsd32_netbsd.c 1.188, 1.189
src/sys/kern/vfs_quotactl.c 1.39
src/sys/kern/vfs_syscalls.c 1.483
src/sys/ufs/lfs/ulfs_quota.c 1.11
src/sys/ufs/ufs/ufs_quota.c 1.116
src/lib/libquota/quota_kernel.c 1.5
and do them correctly.
If you're going to change the name of something, you need to change
the name of *all* the things with the same name, not just a handful,
and you should change it to something similar so it still matches the
rest of the system rather than just picking an arbitrarily different
name.
Hi, Joerg.
To wit, rename the quotactl "delete" operation to "del", because
"delete" is a reserved word in C++ and for some reason Joerg wants to
run internal interfaces used only by C code through his C++ compiler.
Do not rename it to "remove" instead, because this doesn't match
libquota or the rest of the usage throughout the system; and rename
all the related identifiers, not just the ones that blew the mind of
Joerg's C++ compiler.
Because this is not a user-facing API (the only userland consumer
sys/quotactl.h is libquota) it is sort of ok to make arbitrary
source-incompatible changes; however, by the same token it's completely
unnecessary. If it *were* a user-facing API that someone might have a
semi-rational reason to want to run a C++ compiler on, it would be
incorrect to change it at this point.
|
| 1.15 | 18-Oct-2013 |
christos | branches: 1.15.2; 1.15.4;
fix unused variable warnings
|
| 1.14 | 18-Oct-2013 |
christos | use __USE() in the right place, instead of (void)var.
|
| 1.13 | 29-Jul-2013 |
dholland | Fix build both with and without options LFS_EI.
|
| 1.12 | 29-Jul-2013 |
dholland | Revert previous; it is wrong.
|
| 1.11 | 28-Jul-2013 |
pgoyette | Remove more unused variables to unbreak the build.
|
| 1.10 | 28-Jul-2013 |
dholland | Bring in a copy of ffs_quota2_mount() for reference.
Add stuff to struct lfs that it needs to initialize.
Clear these fields in mount as there's no on-disk support for quota2;
but this increases the chances of being able to add it (or something
like it) in the future.
|
| 1.9 | 28-Jul-2013 |
dholland | Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.8 | 28-Jul-2013 |
dholland | Remove the now-pointless ulfs ops macros.
|
| 1.7 | 28-Jul-2013 |
dholland | Get rid of the ulfs_ops table as we only have one fs in here now.
|
| 1.6 | 08-Jun-2013 |
dholland | branches: 1.6.2; 1.6.4;
There is no WAPBL in LFS.
|
| 1.5 | 06-Jun-2013 |
dholland | Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.4 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 3: rearrange config stuff.
Add new options:
LFS_EI
LFS_DIRHASH
LFS_EXTATTR
LFS_EXTATTR_AUTOSTART
LFS_QUOTA
LFS_QUOTA2
and update code referring to the corresponding FFS and UFS config
symbols to use the LFS versions. Disable the one extant reference
to APPLE_UFS in the ulfs files. Use opt_lfs.h only, not opt_ffs.h.
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.6.4.2 | 18-May-2014 |
rmind | sync with head
|
| 1.6.4.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.6.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.6.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.6.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.6.2.1 | 08-Jun-2013 |
tls | file ulfs_quota2.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.15.4.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.15.4.1 | 18-Oct-2013 |
yamt | file ulfs_quota2.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.15.2.1 | 10-Aug-2014 |
tls | Rebase.
|
| 1.16.4.6 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.16.4.5 | 05-Dec-2016 |
skrll | Sync with HEAD
|
| 1.16.4.4 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.16.4.3 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.16.4.2 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.16.4.1 | 06-Apr-2015 |
skrll | Sync with HEAD
|
| 1.28.2.2 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.28.2.1 | 07-Jan-2017 |
pgoyette | Sync with HEAD. (Note that most of these changes are simply $NetBSD$
tag issues.)
|
| 1.29.2.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.30.6.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.31.12.1 | 17-Jan-2020 |
ad | Sync with head.
|
| 1.31.6.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.32.6.1 | 14-Dec-2020 |
thorpej | Sync w/ HEAD.
|
| 1.4 | 06-Jun-2013 |
dholland | branches: 1.4.2; 1.4.10;
Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.4.10.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.4.10.1 | 06-Jun-2013 |
yamt | file ulfs_quota2.h was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.4.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.4.2.1 | 06-Jun-2013 |
tls | file ulfs_quota2.h was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.7 | 24-Aug-2023 |
andvar | s/defaut/default/ in comments.
|
| 1.6 | 06-Jun-2013 |
dholland | branches: 1.6.2; 1.6.10;
Cleanups to reduce symbol and header exposure:
- move struct ufid from ulfs_inode.h to lfs.h
- lfs.h needs sys/mount.h and sys/pool.h
- ulfs_quota2_subr.c needs lfs_inode.h
- remove ulfs_inode.h from lfs.h in favor of ulfs_dinode.h
- move ULFS_NDADDR, ULFS_NIADDR, ULFS_NXADDR from ulfs_dinode.h to lfs.h
- remove ulfs_dinode.h from lfs.h
- add lfs.h to ulfs_dinode.h
|
| 1.5 | 06-Jun-2013 |
dholland | Remove stray references to ext2fs, chfs, ffs, and mfs.
|
| 1.4 | 06-Jun-2013 |
dholland | Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.6.10.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.6.10.1 | 06-Jun-2013 |
yamt | file ulfs_quota2_subr.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.6.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.6.2.1 | 06-Jun-2013 |
tls | file ulfs_quota2_subr.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.4 | 08-Jun-2013 |
dholland | branches: 1.4.2; 1.4.10;
Split the definitions suitable for userland out of ulfs_inode.h into
lfs_inode.h. Since fsck_lfs, newfs_lfs, and lfs_cleanerd want to reuse
the inode structure for their own internal use, and some of them share
parts of the kernel code as well, the best way forward is to provide a
relatively sanitized header that doesn't bring in stray material.
Shuffle a few other definitions around so that lfs_inode.h depends
only on lfs.h.
Install lfs_inode.h into /usr/include.
|
| 1.3 | 06-Jun-2013 |
dholland | Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.4.10.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.4.10.1 | 08-Jun-2013 |
yamt | file ulfs_quotacommon.h was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.4.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.4.2.1 | 08-Jun-2013 |
tls | file ulfs_quotacommon.h was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.28 | 20-Oct-2021 |
thorpej | Overhaul of the EVFILT_VNODE kevent(2) filter:
- Centralize vnode kevent handling in the VOP_*() wrappers, rather than
forcing each individual file system to deal with it (except VOP_RENAME(),
because VOP_RENAME() is a mess and we currently have 2 different ways
of handling it; at least it's reasonably well-centralized in the "new"
way).
- Add support for NOTE_OPEN, NOTE_CLOSE, NOTE_CLOSE_WRITE, and NOTE_READ,
compatible with the same events in FreeBSD.
- Track which kevent notifications clients are interested in receiving
to avoid doing work for events no one cares about (avoiding, e.g.
taking locks and traversing the klist to send a NOTE_WRITE when
someone is merely watching for a file to be deleted, for example).
In support of the above:
- Add support in vnode_if.sh for specifying PRE- and POST-op handlers,
to be invoked before and after vop_pre() and vop_post(), respectively.
Basic idea from FreeBSD, but implemented differently.
- Add support in vnode_if.sh for specifying CONTEXT fields in the
vop_*_args structures. These context fields are used to convey information
between the file system VOP function and the VOP wrapper, but do not
occupy an argument slot in the VOP_*() call itself. These context fields
are initialized and subsequently interpreted by PRE- and POST-op handlers.
- Version VOP_REMOVE(), uses the a context field for the file system to report
back the resulting link count of the target vnode. Return this in tmpfs,
udf, nfs, chfs, ext2fs, lfs, and ufs.
NetBSD 9.99.92.
|
| 1.27 | 23-Apr-2020 |
ad | PR kern/54759 (vm.ubc_direct deadlock when read()/write() into mapping of itself)
- Add new flag UBC_ISMAPPED which tells ubc_uiomove() the object is mmap()ed
somewhere. Use it to decide whether to do direct-mapped copy, rather than
poking around directly in the vnode in ubc_uiomove(), which is ugly and
doesn't work for tmpfs. It would be nicer to contain all this in UVM but
the filesystem provides the needed locking here (VV_MAPPED) and to
reinvent that would suck more.
- Rename UBC_UNMAP_FLAG() to UBC_VNODE_FLAGS(). Pass in UBC_ISMAPPED where
appropriate.
|
| 1.26 | 23-Feb-2020 |
ad | branches: 1.26.4;
UVM locking changes, proposed on tech-kern:
- Change the lock on uvm_object, vm_amap and vm_anon to be a RW lock.
- Break v_interlock and vmobjlock apart. v_interlock remains a mutex.
- Do partial PV list locking in the x86 pmap. Others to follow later.
|
| 1.25 | 20-Jun-2019 |
christos | branches: 1.25.4;
unifdef -DLFS_READWRITE ulfs_readwrite.c
|
| 1.24 | 10-Jun-2017 |
maya | branches: 1.24.6;
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.23 | 30-Mar-2017 |
hannken | branches: 1.23.6;
Remove now redundant calls to fstrans_start()/fstrans_done().
Add fstrans_start()/fstrans_done() to lfs_putpages().
|
| 1.22 | 20-Jun-2016 |
dholland | branches: 1.22.2; 1.22.4;
One more batch of already-synced ufs changes:
ufs_extern.h 1.79 is equivalent to ulfs_extern.h 1.14
ufsmount.h 1.43 is (roughly) equivalent to lfs_extern.h 1.102
ufs_inode.c 1.94 does not apply to lfs
ufs_inode.c 1.95 does not apply to lfs either
ufs_readwrite.c 1.108 is equivalent to ulfs_readwrite.c 1.8
ufs_readwrite.c 1.109 is equivalent to ulfs_readwrite.c 1.9
ufs_readwrite.c 1.110 is equivalent to ulfs_readwrite.c 1.10
ufs_readwrite.c 1.111 does not apply to lfs
ufs_readwrite.c 1.112 is equivalent to ulfs_readwrite.c 1.11
ufs_readwrite.c 1.113 is equivalent to ulfs_readwrite.c 1.13
ufs_readwrite.c 1.114 is equivalent to ulfs_readwrite.c 1.14
ufs_readwrite.c 1.115 is equivalent to ulfs_readwrite.c 1.15
ufs_readwrite.c 1.116-1.118 does not apply to lfs
ufs_readwrite.c 1.119-1.120 are equivalent to ulfs_readwrite.c 1.16
ufs_rename.c 1.12 is equivalent to lfs_rename.c 1.8
ufs_vnops.c 1.226 is equivalent to ulfs_vnops.c 1.22 and lfs_vnops.c 1.270
ufs_vnops.c 1.227 is equivalent to ulfs_vnops.c 1.23
ufs_vnops.c 1.228-1.229 are equivalent to ulfs_vnops.c 1.24
ufs_vnops.c 1.230 is equivalent to ulfs_vnops.c 1.25 and lfs_vnops.c 1.271
ufs_vnops.c 1.231 originated in lfs
ufs_vnops.c 1.232 does not apply to lfs
|
| 1.21 | 19-Jun-2016 |
dholland | Mark ufs file versions we're already synced with.
|
| 1.20 | 23-Nov-2015 |
mlelstv | fix assertion checking that bufrd function is used only for large symlinks
that aren't embedded in the inode.
|
| 1.19 | 24-Jul-2015 |
dholland | More lfs superblock accessors.
(This changes the rest of the code over; all the accessors were
already added.)
The difference between this commit and the previous one is arbitrary,
but the previous one passed the regression tests on its own so I'm
keeping it separate to help with any bisections that might be needed
in the future.
|
| 1.18 | 24-Jul-2015 |
dholland | Switch to accessor functions for elements of the LFS on-disk
superblock. This will allow switching between 32/64 bit forms on the
fly; it will also allow handling LFS_EI reasonably tidily. (That
currently doesn't work on the superblock.)
It also gets rid of cpp abuse in the form of fake structure member
macros.
Also, instead of doing sleep/wakeup on &lfs_avail and &lfs_nextseg
inside the on-disk superblock, add extra elements to the in-memory
struct lfs for this. (XXX: these should be changed to condvars, but
not right now)
XXX: this migrates a structure needed by the lfs code in libsa (struct
salfs) into lfs.h, where it doesn't belong, but for the time being
this is necessary in order to allow the accessors (and the various
lfs macros and other goop that relies on them) to compile.
|
| 1.17 | 12-Apr-2015 |
riastradh | Strip IO_JOURNALLOCKED, PGO_JOURNALLOCKED out of ulfs_readwrite.c.
These are vestigial from ufs_readwrite.c with wapbl -- lfs does not
have a journal but only the explicit wapbl calls, not these flags,
got ripped out in the transition to ulfs_readwrite.c.
|
| 1.16 | 12-Apr-2015 |
riastradh | Same putpages->kassert in ulfs_readwrite.c
|
| 1.15 | 28-Mar-2015 |
maxv | Remove the 'cred' argument from bread(). Remove a now unused var in
ffs_snapshot.c. Update the man page accordingly.
ok hannken@
|
| 1.14 | 28-Mar-2015 |
riastradh | Let I/O errors override inode update errors in UFS.
Fixes tests/fs/vfs/t_io:read_fault for UFS.
|
| 1.13 | 28-Mar-2015 |
maxv | Remove the 'cred' argument from breadn(), and update the man page
accordingly.
ok hannken@
|
| 1.12 | 28-Mar-2015 |
riastradh | Make some comments match better in ulfs_readwrite.
|
| 1.11 | 28-Mar-2015 |
riastradh | Factor out post-read/write inode updates in UFS.
|
| 1.10 | 28-Mar-2015 |
riastradh | Turn some `#if DIAGNOSTIC' into KASSERT.
|
| 1.9 | 27-Mar-2015 |
riastradh | Tighten some kasserts in ufs_bufio code paths.
|
| 1.8 | 27-Mar-2015 |
riastradh | Disentangle buffer-cached I/O from page-cached I/O in UFS.
Page-cached I/O is used for regular files, and is initiated by VFS
users such as userland and NFS.
Buffer-cached I/O is used for directories and symlinks, and is issued
only internally by UFS.
New UFS routine ufs_bufio replaces vn_rdwr for internal use.
ufs_bufio is implemented by new UFS operations uo_bufrd/uo_bufwr,
which sit in ufs_readwrite.c alongside the VOP_READ/VOP_WRITE
implementations.
I preserved the code as much as possible and will leave further
simplification for future commits. I kept the ulfs_readwrite.c
copypasta close to ufs_readwrite.c in case we ever want to merge them
back; likewise ext2fs_readwrite.c.
No externally visible semantic change. All atf fs tests still pass.
|
| 1.7 | 17-Oct-2013 |
christos | branches: 1.7.4; 1.7.8;
- remove unused variables
- add debug ifdefs for debugging variables
- __USE() where appropriate.
|
| 1.6 | 28-Jul-2013 |
dholland | Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.5 | 28-Jul-2013 |
dholland | Remove the now-pointless ulfs ops macros.
|
| 1.4 | 18-Jun-2013 |
christos | branches: 1.4.2; 1.4.4;
Prefix most of the cpp macros with lfs_ and LFS_ to avoid conflicts with ffs.
This was done so that boot blocks that want to compile both FFS and LFS in
the same file work.
|
| 1.3 | 08-Jun-2013 |
dholland | There is no WAPBL in LFS.
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.4.4.2 | 18-May-2014 |
rmind | sync with head
|
| 1.4.4.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.4.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.4.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.4.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.4.2.1 | 18-Jun-2013 |
tls | file ulfs_readwrite.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.7.8.5 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.7.8.4 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.7.8.3 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.7.8.2 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.7.8.1 | 06-Apr-2015 |
skrll | Sync with HEAD
|
| 1.7.4.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.7.4.1 | 17-Oct-2013 |
yamt | file ulfs_readwrite.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.22.4.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.22.2.1 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.23.6.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.24.6.2 | 13-Apr-2020 |
martin | Mostly merge changes from HEAD upto 20200411
|
| 1.24.6.1 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.25.4.1 | 29-Feb-2020 |
ad | Sync with head.
|
| 1.26.4.1 | 25-Apr-2020 |
bouyer | Sync with bouyer-xenpvh-base2 (HEAD)
|
| 1.10 | 20-Jul-2013 |
dholland | Collect the pieces of lfs rename into lfs_rename.c, and sprinkle static.
|
| 1.9 | 19-Jun-2013 |
dholland | branches: 1.9.2; 1.9.4; 1.9.6;
Rename ambiguous macros:
MAXDIRSIZE -> UFS_MAXDIRSIZE or LFS_MAXDIRSIZE
NINDIR -> FFS_NINDIR, EXT2_NINDIR, LFS_NINDIR, or MFS_NINDIR
INOPB -> FFS_INOPB, LFS_INOPB
INOPF -> FFS_INOPF, LFS_INOPF
blksize -> ffs_blksize, ext2_blksize, or lfs_blksize
sblksize -> ffs_blksize
These are not the only ambiguously defined filesystem macros, of
course, there's a pile more. I may not have found all the ambiguous
definitions of blksize(), too, as there are a lot of other things
called 'blksize' in the system.
|
| 1.8 | 08-Jun-2013 |
dholland | ulfs_dir.h has been emptied; remove it.
|
| 1.7 | 08-Jun-2013 |
dholland | There is no WAPBL in LFS.
|
| 1.6 | 08-Jun-2013 |
dholland | DIRBLKSIZ -> LFS_DIRBLKSIZ
DIRECTSIZ -> LFS_DIRECTSIZ
DIRSIZ -> LFS_DIRSIZ
OLDDIRFMT -> LFS_OLDDIRFMT
NEWDIRFMT -> LFS_NEWDIRFMT
IFTODT -> LFS_IFTODT
DTTOIF -> LFS_DTTOIF
|
| 1.5 | 08-Jun-2013 |
dholland | struct direct -> struct lfs_direct
struct dirtemplate -> struct lfs_dirtemplate
struct odirtemplate -> struct lfs_odirtemplate
DT_* -> LFS_DT_*
|
| 1.4 | 08-Jun-2013 |
dholland | Stick LFS_ in front of IFMT, IFIFO, IFREG, etc. so as not to conflict
with the UFS copies of these symbols. (Which themselves ought to have
UFS_ stuck on.)
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.9.6.1 | 23-Jul-2013 |
riastradh | sync with HEAD
|
| 1.9.4.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.9.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.9.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.9.2.1 | 19-Jun-2013 |
tls | file ulfs_rename.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.4 | 05-Sep-2020 |
riastradh | Round of uvm.h cleanup.
The poorly named uvm.h is generally supposed to be for uvm-internal
users only.
- Narrow it to files that actually need it -- mostly files that need
to query whether curlwp is the pagedaemon, which should maybe be
exposed by an external header.
- Use uvm_extern.h where feasible and uvm_*.h for things not exposed
by it. We should split up uvm_extern.h but this will serve for now
to reduce the uvm.h dependencies.
- Use uvm_stat.h and #ifdef UVMHIST uvm.h for files that use
UVMHIST(ubchist), since ubchist is declared in uvm.h but the
reference evaporates if UVMHIST is not defined, so we reduce header
file dependencies.
- Make uvm_device.h and uvm_swap.h independently includable while
here.
ok chs@
|
| 1.3 | 14-Nov-2015 |
pgoyette | Remove historic references to wapbl.
|
| 1.2 | 08-Jun-2013 |
dholland | branches: 1.2.2; 1.2.10; 1.2.14;
There is no WAPBL in LFS.
|
| 1.1 | 06-Jun-2013 |
dholland | Apparently we also need to cut and paste ffs_snapgone() in order to be
able to link the ufs code.
Instead of actually cutting and pasting it (as it depends on ffs-only
things) implement it as panic. Probably we'll be able to demonstrate
later that it's unreachable.
XXX: Someone should add snapgone to struct ufs_ops in ufs/ufsmount.h,
XXX: and fix ufs/ufs_lookup.c to not hardwire ffs.
|
| 1.2.14.1 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.2.10.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.2.10.1 | 08-Jun-2013 |
yamt | file ulfs_snapshot.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.2.2.3 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.2.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.2.2.1 | 08-Jun-2013 |
tls | file ulfs_snapshot.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.16 | 17-Jan-2020 |
ad | VFS_VGET(), VFS_ROOT(), VFS_FHTOVP(): give them a "int lktype" argument, to
allow us to get shared locks (or no lock) on the returned vnode. Matches
FreeBSD.
|
| 1.15 | 22-Dec-2019 |
ad | branches: 1.15.2;
Make mntvnode_lock per-mount, and address false sharing of struct mount.
|
| 1.14 | 10-Dec-2018 |
maxv | Remove unused mbuf.h includes.
|
| 1.13 | 17-Apr-2017 |
hannken | branches: 1.13.10; 1.13.12;
Remove unused argument "nextp" from vfs_busy() and vfs_unbusy().
Remove argument "keepref" from vfs_unbusy() and add vfs_ref() where needed.
|
| 1.12 | 20-Jun-2016 |
dholland | branches: 1.12.2; 1.12.4;
Note more already-merged versions:
inode.h 1.68 is subsumed by ulfs_inode.h 1.19
inode.h 1.69-1.72 do not apply to lfs
ufs_extern.h 1.74 was covered when lfs was moved to the new vnode cache
ufs_extern.h 1.75 is equivalent to ulfs_extern.h 1.13
ufs_extern.h 1.76-1.77 do not apply to lfs
ufsmount.h 1.42 does not apply to lfs
ufs_inode.c 1.90 is subsumed by ulfs_inode.c 1.10
ufs_inode.c 1.91-1.92 do not apply to lfs
ufs_lookup.c 1.130 is subsumed by ulfs_lookup.c 1.24
ufs_lookup.c 1.131 is equivalent to ulfs_lookup.c 1.20
ufs_lookup.c 1.132 is equivalent to ulfs_lookup.c 1.21
ufs_lookup.c 1.133 is equivalent to ulfs_lookup.c 1.22
ufs_lookup.c 1.134 is equivalent to ulfs_lookup.c 1.23
ufs_lookup.c 1.135 is equivalent to ulfs_lookup.c 1.25
ufs_quota2.c 1.38 is equivalent to ulfs_quota2.c 1.17
ufs_quota2.c 1.39 is equivalent to ulfs_quota2.c 1.16
ufs_quota2.c 1.40 is equivalent to ulfs_quota2.c 1.18
ufs_vfsops.c 1.53 is subsumed by lfs_vfsops.c 1.324
ufs_vfsops.c 1.54 is subsumed by lfs_vfsops.c 1.324
ufs_vnops.c 1.223-1.224 do not apply to lfs
|
| 1.11 | 15-Sep-2015 |
dholland | Kill off the ulfs_direct_cache pool.
We no longer allocate temporary struct directs, so we don't need a
pool for them.
|
| 1.10 | 01-Sep-2015 |
dholland | Add new accessors for the d_type and d_namlen fields of struct lfs_direct.
Napalm the old byteswap access logic for these.
|
| 1.9 | 31-May-2015 |
hannken | Change lfs from hash table to vcache.
- Change lfs_valloc() to return an inode number and version instead of
a vnode and move lfs_ialloc() and lfs_vcreate() to new lfs_init_vnode().
- Add lfs_valloc_fixed() to allocate a known inode, used by kernel
roll forward.
- Remove lfs_*ref(), these functions cannot coexist with vcache and
their commented behaviour is far away from their implementation.
- Add the cleaner lwp and blockinfo to struct ulfsmount so lfs_loadvnode()
may use hints from the cleaner.
- Remove vnode locks from ulfs_lookup() like we did with ufs_lookup().
|
| 1.8 | 08-Jun-2013 |
dholland | branches: 1.8.2; 1.8.10; 1.8.14;
struct direct -> struct lfs_direct
struct dirtemplate -> struct lfs_dirtemplate
struct odirtemplate -> struct lfs_odirtemplate
DT_* -> LFS_DT_*
|
| 1.7 | 06-Jun-2013 |
dholland | Fix some exposed symbols:
LOSTFOUNDINO -> LFS_LOSTFOUNDINO
struct ufid -> struct ulfs_ufid
|
| 1.6 | 06-Jun-2013 |
dholland | Cleanups to reduce symbol and header exposure:
- move struct ufid from ulfs_inode.h to lfs.h
- lfs.h needs sys/mount.h and sys/pool.h
- ulfs_quota2_subr.c needs lfs_inode.h
- remove ulfs_inode.h from lfs.h in favor of ulfs_dinode.h
- move ULFS_NDADDR, ULFS_NIADDR, ULFS_NXADDR from ulfs_dinode.h to lfs.h
- remove ulfs_dinode.h from lfs.h
- add lfs.h to ulfs_dinode.h
|
| 1.5 | 06-Jun-2013 |
dholland | Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.4 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 3: rearrange config stuff.
Add new options:
LFS_EI
LFS_DIRHASH
LFS_EXTATTR
LFS_EXTATTR_AUTOSTART
LFS_QUOTA
LFS_QUOTA2
and update code referring to the corresponding FFS and UFS config
symbols to use the LFS versions. Disable the one extant reference
to APPLE_UFS in the ulfs files. Use opt_lfs.h only, not opt_ffs.h.
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.8.14.3 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.8.14.2 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.8.14.1 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.8.10.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.8.10.1 | 08-Jun-2013 |
yamt | file ulfs_vfsops.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.8.2.3 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.8.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.8.2.1 | 08-Jun-2013 |
tls | file ulfs_vfsops.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.12.4.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.12.2.1 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.13.12.2 | 08-Apr-2020 |
martin | Merge changes from current as of 20200406
|
| 1.13.12.1 | 10-Jun-2019 |
christos | Sync with HEAD
|
| 1.13.10.1 | 26-Dec-2018 |
pgoyette | Sync with HEAD, resolve a few conflicts
|
| 1.15.2.1 | 17-Jan-2020 |
ad | Sync with head.
|
| 1.56 | 27-Mar-2022 |
christos | add a kauth vnode check for creating links
|
| 1.55 | 20-Oct-2021 |
thorpej | Overhaul of the EVFILT_VNODE kevent(2) filter:
- Centralize vnode kevent handling in the VOP_*() wrappers, rather than
forcing each individual file system to deal with it (except VOP_RENAME(),
because VOP_RENAME() is a mess and we currently have 2 different ways
of handling it; at least it's reasonably well-centralized in the "new"
way).
- Add support for NOTE_OPEN, NOTE_CLOSE, NOTE_CLOSE_WRITE, and NOTE_READ,
compatible with the same events in FreeBSD.
- Track which kevent notifications clients are interested in receiving
to avoid doing work for events no one cares about (avoiding, e.g.
taking locks and traversing the klist to send a NOTE_WRITE when
someone is merely watching for a file to be deleted, for example).
In support of the above:
- Add support in vnode_if.sh for specifying PRE- and POST-op handlers,
to be invoked before and after vop_pre() and vop_post(), respectively.
Basic idea from FreeBSD, but implemented differently.
- Add support in vnode_if.sh for specifying CONTEXT fields in the
vop_*_args structures. These context fields are used to convey information
between the file system VOP function and the VOP wrapper, but do not
occupy an argument slot in the VOP_*() call itself. These context fields
are initialized and subsequently interpreted by PRE- and POST-op handlers.
- Version VOP_REMOVE(), uses the a context field for the file system to report
back the resulting link count of the target vnode. Return this in tmpfs,
udf, nfs, chfs, ext2fs, lfs, and ufs.
NetBSD 9.99.92.
|
| 1.54 | 05-Sep-2020 |
riastradh | Round of uvm.h cleanup.
The poorly named uvm.h is generally supposed to be for uvm-internal
users only.
- Narrow it to files that actually need it -- mostly files that need
to query whether curlwp is the pagedaemon, which should maybe be
exposed by an external header.
- Use uvm_extern.h where feasible and uvm_*.h for things not exposed
by it. We should split up uvm_extern.h but this will serve for now
to reduce the uvm.h dependencies.
- Use uvm_stat.h and #ifdef UVMHIST uvm.h for files that use
UVMHIST(ubchist), since ubchist is declared in uvm.h but the
reference evaporates if UVMHIST is not defined, so we reduce header
file dependencies.
- Make uvm_device.h and uvm_swap.h independently includable while
here.
ok chs@
|
| 1.53 | 16-May-2020 |
christos | Add ACL support for FFS. From FreeBSD.
|
| 1.52 | 28-Oct-2017 |
pgoyette | Update the kernhist(9) kernel history code to address issues identified
in PR kern/52639, as well as some general cleaning-up...
(As proposed on tech-kern@ with additional changes and enhancements.)
Details of changes:
* All history arguments are now stored as uintmax_t values[1], both in
the kernel and in the structures used for exporting the history data
to userland via sysctl(9). This avoids problems on some architectures
where passing a 64-bit (or larger) value to printf(3) can cause it to
process the value as multiple arguments. (This can be particularly
problematic when printf()'s format string is not a literal, since in
that case the compiler cannot know how large each argument should be.)
* Update the data structures used for exporting kernel history data to
include a version number as well as the length of history arguments.
* All [2] existing users of kernhist(9) have had their format strings
updated. Each format specifier now includes an explicit length
modifier 'j' to refer to numeric values of the size of uintmax_t.
* All [2] existing users of kernhist(9) have had their format strings
updated to replace uses of "%p" with "%#jx", and the pointer
arguments are now cast to (uintptr_t) before being subsequently cast
to (uintmax_t). This is needed to avoid compiler warnings about
casting "pointer to integer of a different size."
* All [2] existing users of kernhist(9) have had instances of "%s" or
"%c" format strings replaced with numeric formats; several instances
of mis-match between format string and argument list have been fixed.
* vmstat(1) has been modified to handle the new size of arguments in the
history data as exported by sysctl(9).
* vmstat(1) now provides a warning message if the history requested with
the -u option does not exist (previously, this condition was silently
ignored, with only a single blank line being printed).
* vmstat(1) now checks the version and argument length included in the
data exported via sysctl(9) and exits if they do not match the values
with which vmstat was built.
* The kernhist(9) man-page has been updated to note the additional
requirements imposed on the format strings, along with several other
minor changes and enhancements.
[1] It would have been possible to use an explicit length (for example,
uint64_t) for the history arguments. But that would require another
"rototill" of all the users in the future when we add support for an
architecture that supports a larger size. Also, the printf(3) format
specifiers for explicitly-sized values, such as "%"PRIu64, are much
more verbose (and less aesthetically appealing, IMHO) than simply
using "%ju".
[2] I've tried very hard to find "all [the] existing users of kernhist(9)"
but it is possible that I've missed some of them. I would be glad to
update any stragglers that anyone identifies.
|
| 1.51 | 07-Aug-2017 |
dholland | Tidy up ufs_readdir. First step only; there's plenty more that could be
done to improve this code.
|
| 1.50 | 04-Aug-2017 |
maya | fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
|
| 1.49 | 10-Jun-2017 |
maya | Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
|
| 1.48 | 26-Apr-2017 |
riastradh | branches: 1.48.4;
Change VOP_REMOVE and VOP_RMDIR to preserve lock/ref on dvp.
No change to vp -- the plan is to replace the node by the
componentname in the vop parameters, and let all directory vops do
lookups internally.
Proposed on tech-kern with no objections:
https://mail-index.netbsd.org/tech-kern/2017/04/17/msg021825.html
|
| 1.47 | 11-Apr-2017 |
riastradh | Sprinkle lock ownership assertions.
|
| 1.46 | 30-Mar-2017 |
hannken | Remove now redundant calls to fstrans_start()/fstrans_done().
Add fstrans_start()/fstrans_done() to lfs_putpages().
|
| 1.45 | 13-Mar-2017 |
riastradh | #if DIAGNOSTIC panic ---> KASSERT
Replace some #if DEBUG by this too. DEBUG is only for expensive
assertions; these are not.
|
| 1.44 | 20-Jun-2016 |
dholland | branches: 1.44.2; 1.44.4;
One more batch of already-synced ufs changes:
ufs_extern.h 1.79 is equivalent to ulfs_extern.h 1.14
ufsmount.h 1.43 is (roughly) equivalent to lfs_extern.h 1.102
ufs_inode.c 1.94 does not apply to lfs
ufs_inode.c 1.95 does not apply to lfs either
ufs_readwrite.c 1.108 is equivalent to ulfs_readwrite.c 1.8
ufs_readwrite.c 1.109 is equivalent to ulfs_readwrite.c 1.9
ufs_readwrite.c 1.110 is equivalent to ulfs_readwrite.c 1.10
ufs_readwrite.c 1.111 does not apply to lfs
ufs_readwrite.c 1.112 is equivalent to ulfs_readwrite.c 1.11
ufs_readwrite.c 1.113 is equivalent to ulfs_readwrite.c 1.13
ufs_readwrite.c 1.114 is equivalent to ulfs_readwrite.c 1.14
ufs_readwrite.c 1.115 is equivalent to ulfs_readwrite.c 1.15
ufs_readwrite.c 1.116-1.118 does not apply to lfs
ufs_readwrite.c 1.119-1.120 are equivalent to ulfs_readwrite.c 1.16
ufs_rename.c 1.12 is equivalent to lfs_rename.c 1.8
ufs_vnops.c 1.226 is equivalent to ulfs_vnops.c 1.22 and lfs_vnops.c 1.270
ufs_vnops.c 1.227 is equivalent to ulfs_vnops.c 1.23
ufs_vnops.c 1.228-1.229 are equivalent to ulfs_vnops.c 1.24
ufs_vnops.c 1.230 is equivalent to ulfs_vnops.c 1.25 and lfs_vnops.c 1.271
ufs_vnops.c 1.231 originated in lfs
ufs_vnops.c 1.232 does not apply to lfs
|
| 1.43 | 20-Jun-2016 |
dholland | With the previous we seem to have the changes from -r1.225 of ufs_vnops.c.
(as that was stuff from moving ffs to the new vcache and lfs has also been
moved, this is not surprising)
|
| 1.42 | 20-Jun-2016 |
dholland | Merge (effectively) -r1.78 of ufs_extern.h: shift ulfs_makeinode to
lfs_vnops.c and make it file-static there, as that's the only place
it's used.
|
| 1.41 | 20-Jun-2016 |
dholland | Note more already-merged versions:
inode.h 1.68 is subsumed by ulfs_inode.h 1.19
inode.h 1.69-1.72 do not apply to lfs
ufs_extern.h 1.74 was covered when lfs was moved to the new vnode cache
ufs_extern.h 1.75 is equivalent to ulfs_extern.h 1.13
ufs_extern.h 1.76-1.77 do not apply to lfs
ufsmount.h 1.42 does not apply to lfs
ufs_inode.c 1.90 is subsumed by ulfs_inode.c 1.10
ufs_inode.c 1.91-1.92 do not apply to lfs
ufs_lookup.c 1.130 is subsumed by ulfs_lookup.c 1.24
ufs_lookup.c 1.131 is equivalent to ulfs_lookup.c 1.20
ufs_lookup.c 1.132 is equivalent to ulfs_lookup.c 1.21
ufs_lookup.c 1.133 is equivalent to ulfs_lookup.c 1.22
ufs_lookup.c 1.134 is equivalent to ulfs_lookup.c 1.23
ufs_lookup.c 1.135 is equivalent to ulfs_lookup.c 1.25
ufs_quota2.c 1.38 is equivalent to ulfs_quota2.c 1.17
ufs_quota2.c 1.39 is equivalent to ulfs_quota2.c 1.16
ufs_quota2.c 1.40 is equivalent to ulfs_quota2.c 1.18
ufs_vfsops.c 1.53 is subsumed by lfs_vfsops.c 1.324
ufs_vfsops.c 1.54 is subsumed by lfs_vfsops.c 1.324
ufs_vnops.c 1.223-1.224 do not apply to lfs
|
| 1.40 | 20-Jun-2016 |
dholland | More already-merged or equivalent changes:
ufs_dirhash.c 1.36 corresponds to ulfs_dirhash.c 1.8
ufs_extattr.c 1.43 corresponds to ulfs_extattr.c 1.7
ufs_lookup.c 1.126 does not apply to lfs
ufs_lookup.c 1.127 we already have
ufs_lookup.c 1.128 does not apply to lfs
ufs_lookup.c 1.129 corresponds to ulfs_lookup.c 1.19
ufs_quota1.c 1.19 corresponds to ulfs_quota1.c 1.7
ufs_quota1.c 1.20 corresponds to ulfs_quota1.c 1.8
ufs_quota2.c 1.36 we have equivalent changes for
ufs_rename.c 1.9 corresponds to lfs_rename.c 1.5
ufs_rename.c 1.10 corresponds to lfs_rename.c 1.6
ufs_vnops.c 1.219 corresponds to lfs_vnops.c 1.260 and ulfs_vnops.c 1.19
ufs_vnops.c 1.220 corresponds to lfs_vnops.c 1.261 and ulfs_vnops.c 1.20
ufs_vnops.c 1.221 was superseded by later changes
ufs_vnops.c 1.222 got fixed independently in lfs
|
| 1.39 | 19-Jun-2016 |
dholland | we already have ufs_lookup.c 1.125 and ufs_vnops.c 1.218.
|
| 1.38 | 19-Jun-2016 |
dholland | note that we're synced with ufs_vnops.c -r1.217 and ufsmount.h -r1.41
(those changes removed lfs hooks from ufs so shouldn't be merged across)
|
| 1.37 | 19-Jun-2016 |
dholland | Merge -r1.216 of ufs_vnops.c: comments about maxsymlinklen handling
|
| 1.36 | 19-Jun-2016 |
dholland | Merge -r1.215 of ufs_vnops.c: the speed limit is 80
(-r1.214 was ffs-only)
|
| 1.35 | 14-Nov-2015 |
pgoyette | Remove historic references to wapbl.
|
| 1.34 | 21-Sep-2015 |
dholland | Add 64-bit directory entry structures, and adjust accessors accordingly.
The LFS64 directory entry has a 64-bit inode number. This is stored as
two 32-bit values to avoid inducing 64-bit alignment requirements.
The exposed type for manipulating directory entries is now
LFS_DIRHEADER, following the same convention as e.g. IFILE and SEGUSE.
(But with LFS_ on it, because.)
|
| 1.33 | 21-Sep-2015 |
dholland | Oops; LFS_DIRECTSIZ() is going to need the fs as an argument.
Also, it turns out that dirhash needs a compile-time-constant version
of LFS_DIRECTSIZ(LFS_MAXNAMLEN+1), independent of 64-vs-32, so create
LFS_MAXDIRENTRYSIZE for this. Sigh.
|
| 1.32 | 15-Sep-2015 |
dholland | Pass around struct lfs_dirheader instead of struct lfs_direct.
|
| 1.31 | 15-Sep-2015 |
dholland | Add an accessor function for directory names.
|
| 1.30 | 15-Sep-2015 |
dholland | Kill off ulfs_makedirentry; just pass the data to ulfs_direnter instead.
For now, move one copy of the code that allocates and fills in a
temporary struct lfs_direct to the top of ulfs_direnter; but it should
go away shortly.
|
| 1.29 | 15-Sep-2015 |
dholland | Add and use accessor functions for more of the directory entry fields.
|
| 1.28 | 01-Sep-2015 |
dholland | Add new accessors for the d_type and d_namlen fields of struct lfs_direct.
Napalm the old byteswap access logic for these.
|
| 1.27 | 01-Sep-2015 |
dholland | Use the lfs dinode accessors in place of the ufs-derived ones.
(Mostly.)
The ufs-derived ones are fake structure member macros, which are gross
and not very safe. Also, it seems that a lot of places in the lfs code
were using the ffsv1 branch of them unconditionally, and this way it's
guaranteed all those places have been updated.
Found while doing this: for non-devices, have getattr produce NODEV
in the rdev field instead of leaking the address of the first direct
block.
|
| 1.26 | 31-May-2015 |
hannken | Change lfs from hash table to vcache.
- Change lfs_valloc() to return an inode number and version instead of
a vnode and move lfs_ialloc() and lfs_vcreate() to new lfs_init_vnode().
- Add lfs_valloc_fixed() to allocate a known inode, used by kernel
roll forward.
- Remove lfs_*ref(), these functions cannot coexist with vcache and
their commented behaviour is far away from their implementation.
- Add the cleaner lwp and blockinfo to struct ulfsmount so lfs_loadvnode()
may use hints from the cleaner.
- Remove vnode locks from ulfs_lookup() like we did with ufs_lookup().
|
| 1.25 | 20-Apr-2015 |
riastradh | Make VOP_LINK return directory still locked and referenced.
Ride 7.99.10 bump.
|
| 1.24 | 20-Apr-2015 |
riastradh | Fix more dvp->v_mount after vput(dvp).
|
| 1.23 | 27-Mar-2015 |
riastradh | Tighten some kasserts in ufs_bufio code paths.
|
| 1.22 | 27-Mar-2015 |
riastradh | Disentangle buffer-cached I/O from page-cached I/O in UFS.
Page-cached I/O is used for regular files, and is initiated by VFS
users such as userland and NFS.
Buffer-cached I/O is used for directories and symlinks, and is issued
only internally by UFS.
New UFS routine ufs_bufio replaces vn_rdwr for internal use.
ufs_bufio is implemented by new UFS operations uo_bufrd/uo_bufwr,
which sit in ufs_readwrite.c alongside the VOP_READ/VOP_WRITE
implementations.
I preserved the code as much as possible and will leave further
simplification for future commits. I kept the ulfs_readwrite.c
copypasta close to ufs_readwrite.c in case we ever want to merge them
back; likewise ext2fs_readwrite.c.
No externally visible semantic change. All atf fs tests still pass.
|
| 1.21 | 17-May-2014 |
dholland | branches: 1.21.2; 1.21.6;
Move the ulfs-level (copy of ufs) vnops for symlink, create, and mkdir
into lfs_vnops.c preparatory to folding them into the lfs entry points.
(lfs_vnops.c now has four licenses. sigh.)
|
| 1.20 | 23-Jan-2014 |
hannken | branches: 1.20.2;
Change vnode operations create, mknod, mkdir and symlink to return
the resulting vnode *vpp unlocked.
Discussed on tech-kern@
Welcome to 6.99.30
|
| 1.19 | 17-Jan-2014 |
hannken | Change vnode operations create, mknod, mkdir and symlink to keep the
directory node dvp locked on return.
Discussed on tech-kern@
Welcome to 6.99.29
|
| 1.18 | 28-Jul-2013 |
dholland | Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.17 | 28-Jul-2013 |
dholland | Remove the now-pointless ulfs ops macros.
|
| 1.16 | 28-Jul-2013 |
dholland | Remove ulfsspec_close and ulfsfifo_close as they're not used.
|
| 1.15 | 21-Jul-2013 |
dholland | Merge logic from ulfs_close(), ulfs_getattr(), and ulfs_strategy()
into the preexisting lfs_*() versions of these functions, and delete
the unused ulfs copies.
|
| 1.14 | 20-Jul-2013 |
dholland | Remove ulfs_mknod, which is not used.
|
| 1.13 | 08-Jun-2013 |
dholland | branches: 1.13.2; 1.13.4; 1.13.6;
ulfs_dir.h has been emptied; remove it.
|
| 1.12 | 08-Jun-2013 |
dholland | There is no WAPBL in LFS.
|
| 1.11 | 08-Jun-2013 |
dholland | mp->mnt_wapbl and mp->mnt_wapbl_replay are always NULL in here.
|
| 1.10 | 08-Jun-2013 |
dholland | Merge -r1.213 of ufs_vnops.c:
Committed By: kardel
Date: Sat Jun 8 05:47:02 UTC 2013
fix clearing of system-flags (schg, sappnd). clearing system flags is
possible again at securelevel < 1.
reviewed by christos@
|
| 1.9 | 08-Jun-2013 |
dholland | DIRBLKSIZ -> LFS_DIRBLKSIZ
DIRECTSIZ -> LFS_DIRECTSIZ
DIRSIZ -> LFS_DIRSIZ
OLDDIRFMT -> LFS_OLDDIRFMT
NEWDIRFMT -> LFS_NEWDIRFMT
IFTODT -> LFS_IFTODT
DTTOIF -> LFS_DTTOIF
|
| 1.8 | 08-Jun-2013 |
dholland | struct direct -> struct lfs_direct
struct dirtemplate -> struct lfs_dirtemplate
struct odirtemplate -> struct lfs_odirtemplate
DT_* -> LFS_DT_*
|
| 1.7 | 08-Jun-2013 |
dholland | Stick LFS_ in front of IFMT, IFIFO, IFREG, etc. so as not to conflict
with the UFS copies of these symbols. (Which themselves ought to have
UFS_ stuck on.)
|
| 1.6 | 06-Jun-2013 |
dholland | Remove stray references to ext2fs, chfs, ffs, and mfs.
|
| 1.5 | 06-Jun-2013 |
dholland | Add lfs_ or ulfs_ in front of extern symbols lacking them, mostly
quota-related (and particularly quota2-related) stuff.
|
| 1.4 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 3: rearrange config stuff.
Add new options:
LFS_EI
LFS_DIRHASH
LFS_EXTATTR
LFS_EXTATTR_AUTOSTART
LFS_QUOTA
LFS_QUOTA2
and update code referring to the corresponding FFS and UFS config
symbols to use the LFS versions. Disable the one extant reference
to APPLE_UFS in the ulfs files. Use opt_lfs.h only, not opt_ffs.h.
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.13.6.1 | 23-Jul-2013 |
riastradh | sync with HEAD
|
| 1.13.4.2 | 18-May-2014 |
rmind | sync with head
|
| 1.13.4.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.13.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.13.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.13.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.13.2.1 | 08-Jun-2013 |
tls | file ulfs_vnops.c was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.20.2.1 | 10-Aug-2014 |
tls | Rebase.
|
| 1.21.6.6 | 28-Aug-2017 |
skrll | Sync with HEAD
|
| 1.21.6.5 | 09-Jul-2016 |
skrll | Sync with HEAD
|
| 1.21.6.4 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.21.6.3 | 22-Sep-2015 |
skrll | Sync with HEAD
|
| 1.21.6.2 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.21.6.1 | 06-Apr-2015 |
skrll | Sync with HEAD
|
| 1.21.2.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.21.2.1 | 17-May-2014 |
yamt | file ulfs_vnops.c was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
| 1.44.4.1 | 21-Apr-2017 |
bouyer | Sync with HEAD
|
| 1.44.2.2 | 26-Apr-2017 |
pgoyette | Sync with HEAD
|
| 1.44.2.1 | 20-Mar-2017 |
pgoyette | Sync with HEAD
|
| 1.48.4.2 | 02-Nov-2017 |
snj | Pull up following revision(s) (requested by pgoyette in ticket #335):
share/man/man9/kernhist.9: 1.5-1.8
sys/arch/acorn26/acorn26/pmap.c: 1.39
sys/arch/arm/arm32/fault.c: 1.105 via patch
sys/arch/arm/arm32/pmap.c: 1.350, 1.359
sys/arch/arm/broadcom/bcm2835_bsc.c: 1.7
sys/arch/arm/omap/if_cpsw.c: 1.20
sys/arch/arm/omap/tiotg.c: 1.7
sys/arch/evbarm/conf/RPI2_INSTALL: 1.3
sys/dev/ic/sl811hs.c: 1.98
sys/dev/usb/ehci.c: 1.256
sys/dev/usb/if_axe.c: 1.83
sys/dev/usb/motg.c: 1.18
sys/dev/usb/ohci.c: 1.274
sys/dev/usb/ucom.c: 1.119
sys/dev/usb/uhci.c: 1.277
sys/dev/usb/uhub.c: 1.137
sys/dev/usb/umass.c: 1.160-1.162
sys/dev/usb/umass_quirks.c: 1.100
sys/dev/usb/umass_scsipi.c: 1.55
sys/dev/usb/usb.c: 1.168
sys/dev/usb/usb_mem.c: 1.70
sys/dev/usb/usb_subr.c: 1.221
sys/dev/usb/usbdi.c: 1.175
sys/dev/usb/usbdi_util.c: 1.67-1.70
sys/dev/usb/usbroothub.c: 1.3
sys/dev/usb/xhci.c: 1.75
sys/external/bsd/drm2/dist/drm/i915/i915_gem.c: 1.34
sys/kern/kern_history.c: 1.15
sys/kern/kern_xxx.c: 1.74
sys/kern/vfs_bio.c: 1.275-1.276
sys/miscfs/genfs/genfs_io.c: 1.71
sys/sys/kernhist.h: 1.21
sys/ufs/ffs/ffs_balloc.c: 1.63
sys/ufs/lfs/lfs_vfsops.c: 1.361
sys/ufs/lfs/ulfs_inode.c: 1.21
sys/ufs/lfs/ulfs_vnops.c: 1.52
sys/ufs/ufs/ufs_inode.c: 1.102
sys/ufs/ufs/ufs_vnops.c: 1.239
sys/uvm/pmap/pmap.c: 1.37-1.39
sys/uvm/pmap/pmap_tlb.c: 1.22
sys/uvm/uvm_amap.c: 1.108
sys/uvm/uvm_anon.c: 1.64
sys/uvm/uvm_aobj.c: 1.126
sys/uvm/uvm_bio.c: 1.91
sys/uvm/uvm_device.c: 1.66
sys/uvm/uvm_fault.c: 1.201
sys/uvm/uvm_km.c: 1.144
sys/uvm/uvm_loan.c: 1.85
sys/uvm/uvm_map.c: 1.353
sys/uvm/uvm_page.c: 1.194
sys/uvm/uvm_pager.c: 1.111
sys/uvm/uvm_pdaemon.c: 1.109
sys/uvm/uvm_swap.c: 1.175
sys/uvm/uvm_vnode.c: 1.103
usr.bin/vmstat/vmstat.c: 1.219
Reorder to test for null before null deref in debug code
--
Reorder to test for null before null deref in debug code
--
KNF
--
No need for '\n' in UVMHIST_LOG
--
normalise a BIOHIST log message
--
Update the kernhist(9) kernel history code to address issues identified
in PR kern/52639, as well as some general cleaning-up...
(As proposed on tech-kern@ with additional changes and enhancements.)
Details of changes:
* All history arguments are now stored as uintmax_t values[1], both in
the kernel and in the structures used for exporting the history data
to userland via sysctl(9). This avoids problems on some architectures
where passing a 64-bit (or larger) value to printf(3) can cause it to
process the value as multiple arguments. (This can be particularly
problematic when printf()'s format string is not a literal, since in
that case the compiler cannot know how large each argument should be.)
* Update the data structures used for exporting kernel history data to
include a version number as well as the length of history arguments.
* All [2] existing users of kernhist(9) have had their format strings
updated. Each format specifier now includes an explicit length
modifier 'j' to refer to numeric values of the size of uintmax_t.
* All [2] existing users of kernhist(9) have had their format strings
updated to replace uses of "%p" with "%#jx", and the pointer
arguments are now cast to (uintptr_t) before being subsequently cast
to (uintmax_t). This is needed to avoid compiler warnings about
casting "pointer to integer of a different size."
* All [2] existing users of kernhist(9) have had instances of "%s" or
"%c" format strings replaced with numeric formats; several instances
of mis-match between format string and argument list have been fixed.
* vmstat(1) has been modified to handle the new size of arguments in the
history data as exported by sysctl(9).
* vmstat(1) now provides a warning message if the history requested with
the -u option does not exist (previously, this condition was silently
ignored, with only a single blank line being printed).
* vmstat(1) now checks the version and argument length included in the
data exported via sysctl(9) and exits if they do not match the values
with which vmstat was built.
* The kernhist(9) man-page has been updated to note the additional
requirements imposed on the format strings, along with several other
minor changes and enhancements.
[1] It would have been possible to use an explicit length (for example,
uint64_t) for the history arguments. But that would require another
"rototill" of all the users in the future when we add support for an
architecture that supports a larger size. Also, the printf(3)
format
specifiers for explicitly-sized values, such as "%"PRIu64, are much
more verbose (and less aesthetically appealing, IMHO) than simply
using "%ju".
[2] I've tried very hard to find "all [the] existing users of
kernhist(9)"
but it is possible that I've missed some of them. I would be glad
to
update any stragglers that anyone identifies.
--
For some reason this single kernel seems to have outgrown its declared
size as a result of the kernhist(9) changes. Bump the size.
XXX The amount of increase may be excessive - anyone with more detailed
XXX knowledge please feel free to further adjust the value
appropriately.
--
Misssed one cast of pointer --> uintptr_t in previous kernhist(9) commit
--
And yet another one. :(
--
Use correct mark-up for NetBSD version.
--
More improvements in grammar and readability.
--
Remove a stray '"' (obvious typo) and add a couple of casts that are
probably needed.
--
And replace an instance of "%p" conversion with "%#jx"
--
Whitespace fix. Give Bl tag table a width. Fix Xr.
|
| 1.48.4.1 | 30-Oct-2017 |
snj | Pull up following revision(s) (requested by maya in ticket #330):
sbin/fsck_lfs/inode.c: 1.69
sbin/fsck_lfs/lfs.c: 1.73
sbin/fsck_lfs/pass6.c: 1.50
sbin/fsck_lfs/segwrite.c: 1.46
sys/ufs/lfs/lfs.h: 1.202-1.203
sys/ufs/lfs/lfs_accessors.h: 1.48
sys/ufs/lfs/lfs_alloc.c: 1.136-1.137
sys/ufs/lfs/lfs_balloc.c: 1.94
sys/ufs/lfs/lfs_bio.c: 1.141
sys/ufs/lfs/lfs_extern.h: 1.113
sys/ufs/lfs/lfs_inode.c: 1.156-1.157
sys/ufs/lfs/lfs_inode.h: 1.20, 1.21, 1.23
sys/ufs/lfs/lfs_itimes.c: 1.20
sys/ufs/lfs/lfs_pages.c: 1.13-1.15
sys/ufs/lfs/lfs_rename.c: 1.22
sys/ufs/lfs/lfs_segment.c: 1.270-1.275
sys/ufs/lfs/lfs_subr.c: 1.94-1.97
sys/ufs/lfs/lfs_syscalls.c: 1.175
sys/ufs/lfs/lfs_vfsops.c: 1.360
sys/ufs/lfs/lfs_vnops.c: 1.316-1.321
sys/ufs/lfs/ulfs_inode.c: 1.20
sys/ufs/lfs/ulfs_inode.h: 1.24
sys/ufs/lfs/ulfs_lookup.c: 1.41
sys/ufs/lfs/ulfs_quota2.c: 1.31
sys/ufs/lfs/ulfs_readwrite.c: 1.24
sys/ufs/lfs/ulfs_vnops.c: 1.49-1.50
Update inode member i_flag --> i_state to keep up with kernel changes
Move definition of IN_ALLMOD near the flag it's a mask for.
Now we can see that it doesn't match all the flags, but changing that will
require more careful thought.
Correct confusion between i_flag and i_flags
These will have to be renamed.
Spotted by Riastradh, thanks!
Add an XXX about the missing flags so it's not buried in a commit
message.
now the XXX count for LFS is 260
Rename i_flag to i_state.
The similarity to i_flags has previously caused errors.
Use continue to denote the no-op loop to match netbsd style
newline for extra clarity.
It isn't safe to drain dirops with seglock held, it'll deadlock if there
are any dirops. drain before grabbing seglock.
lfs_dirops == 0 is always true (as we already drained dirops), so omit
that part of the comparison.
Fixes a lot of LFS deadlocks. PR kern/52301
Many thanks to dholland for help analyzing coredumps
Ifdef out KDASSERT which fires on my machine.
Deduplicate sanity check that seglock is held on segunlock
Revert r1.272 fix to PR kern/52301, the performance hit is making things
unusable.
change lfs_nextsegsleep and lfs_allclean_wakeup to use condvar
XXX had to use lfs_lock in lfs_segwait, removed kernel_lock, is this
appropriate?
fix buffer overflow/KASSERT when cookies are supplied
lfs no longer uses the ffs-style struct direct, use the correct minimum
size
from dholland
XXX more wrong
Consistently use {,UN}MARK_VNODE macros rather than function calls.
Not much point doing anything after a panic call
Ask some question about the code in a XXX comment
XXX question our double-flushing of dirops
Fix typo in comment
|
| 1.6 | 08-Jun-2013 |
dholland | G/C
|
| 1.5 | 08-Jun-2013 |
dholland | There is no WAPBL in LFS.
|
| 1.4 | 06-Jun-2013 |
dholland | Remove stray references to ext2fs, chfs, ffs, and mfs.
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.5 | 08-Jun-2013 |
dholland | G/C
|
| 1.4 | 08-Jun-2013 |
dholland | There is no WAPBL in LFS.
|
| 1.3 | 08-Jun-2013 |
dholland | mp->mnt_wapbl and mp->mnt_wapbl_replay are always NULL in here.
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.18 | 20-Jun-2016 |
dholland | One more batch of already-synced ufs changes:
ufs_extern.h 1.79 is equivalent to ulfs_extern.h 1.14
ufsmount.h 1.43 is (roughly) equivalent to lfs_extern.h 1.102
ufs_inode.c 1.94 does not apply to lfs
ufs_inode.c 1.95 does not apply to lfs either
ufs_readwrite.c 1.108 is equivalent to ulfs_readwrite.c 1.8
ufs_readwrite.c 1.109 is equivalent to ulfs_readwrite.c 1.9
ufs_readwrite.c 1.110 is equivalent to ulfs_readwrite.c 1.10
ufs_readwrite.c 1.111 does not apply to lfs
ufs_readwrite.c 1.112 is equivalent to ulfs_readwrite.c 1.11
ufs_readwrite.c 1.113 is equivalent to ulfs_readwrite.c 1.13
ufs_readwrite.c 1.114 is equivalent to ulfs_readwrite.c 1.14
ufs_readwrite.c 1.115 is equivalent to ulfs_readwrite.c 1.15
ufs_readwrite.c 1.116-1.118 does not apply to lfs
ufs_readwrite.c 1.119-1.120 are equivalent to ulfs_readwrite.c 1.16
ufs_rename.c 1.12 is equivalent to lfs_rename.c 1.8
ufs_vnops.c 1.226 is equivalent to ulfs_vnops.c 1.22 and lfs_vnops.c 1.270
ufs_vnops.c 1.227 is equivalent to ulfs_vnops.c 1.23
ufs_vnops.c 1.228-1.229 are equivalent to ulfs_vnops.c 1.24
ufs_vnops.c 1.230 is equivalent to ulfs_vnops.c 1.25 and lfs_vnops.c 1.271
ufs_vnops.c 1.231 originated in lfs
ufs_vnops.c 1.232 does not apply to lfs
|
| 1.17 | 20-Jun-2016 |
dholland | Note more already-merged versions:
inode.h 1.68 is subsumed by ulfs_inode.h 1.19
inode.h 1.69-1.72 do not apply to lfs
ufs_extern.h 1.74 was covered when lfs was moved to the new vnode cache
ufs_extern.h 1.75 is equivalent to ulfs_extern.h 1.13
ufs_extern.h 1.76-1.77 do not apply to lfs
ufsmount.h 1.42 does not apply to lfs
ufs_inode.c 1.90 is subsumed by ulfs_inode.c 1.10
ufs_inode.c 1.91-1.92 do not apply to lfs
ufs_lookup.c 1.130 is subsumed by ulfs_lookup.c 1.24
ufs_lookup.c 1.131 is equivalent to ulfs_lookup.c 1.20
ufs_lookup.c 1.132 is equivalent to ulfs_lookup.c 1.21
ufs_lookup.c 1.133 is equivalent to ulfs_lookup.c 1.22
ufs_lookup.c 1.134 is equivalent to ulfs_lookup.c 1.23
ufs_lookup.c 1.135 is equivalent to ulfs_lookup.c 1.25
ufs_quota2.c 1.38 is equivalent to ulfs_quota2.c 1.17
ufs_quota2.c 1.39 is equivalent to ulfs_quota2.c 1.16
ufs_quota2.c 1.40 is equivalent to ulfs_quota2.c 1.18
ufs_vfsops.c 1.53 is subsumed by lfs_vfsops.c 1.324
ufs_vfsops.c 1.54 is subsumed by lfs_vfsops.c 1.324
ufs_vnops.c 1.223-1.224 do not apply to lfs
|
| 1.16 | 19-Jun-2016 |
dholland | note that we're synced with ufs_vnops.c -r1.217 and ufsmount.h -r1.41
(those changes removed lfs hooks from ufs so shouldn't be merged across)
|
| 1.15 | 19-Jun-2016 |
dholland | Update the ufs versions these files are synced with by 1: the
201306016 commit by hannken@ that removed references to ffs_snapgone
in ufs doesn't need to be synced into lfs.
|
| 1.14 | 15-Oct-2015 |
dholland | Move stuff from struct ulfsmount to struct lfs.
|
| 1.13 | 31-May-2015 |
hannken | Change lfs from hash table to vcache.
- Change lfs_valloc() to return an inode number and version instead of
a vnode and move lfs_ialloc() and lfs_vcreate() to new lfs_init_vnode().
- Add lfs_valloc_fixed() to allocate a known inode, used by kernel
roll forward.
- Remove lfs_*ref(), these functions cannot coexist with vcache and
their commented behaviour is far away from their implementation.
- Add the cleaner lwp and blockinfo to struct ulfsmount so lfs_loadvnode()
may use hints from the cleaner.
- Remove vnode locks from ulfs_lookup() like we did with ufs_lookup().
|
| 1.12 | 28-Jul-2013 |
dholland | branches: 1.12.4; 1.12.8;
Migrate the miscellaneous ulfs-level info from struct ulfsmount to
struct lfs.
Put them inside #ifdef _KERNEL there. They are not the only such
members, gross as that is. Unfortunately, moving struct lfs to
lfs_kernel.h does not work.
|
| 1.11 | 28-Jul-2013 |
dholland | Remove the now-pointless ulfs ops macros.
|
| 1.10 | 28-Jul-2013 |
dholland | Get rid of the ulfs_ops table as we only have one fs in here now.
|
| 1.9 | 28-Jul-2013 |
dholland | Improve comments in struct ulfsmount.
Also rearrange it to group related items together.
|
| 1.8 | 28-Jul-2013 |
dholland | Prune unused stuff from struct ulfsmount.
|
| 1.7 | 08-Jun-2013 |
dholland | branches: 1.7.2; 1.7.4;
Move stuff to lfs.h that's needed by userland:
LFS_DT_*
ULFS_ROOTINO
ULFS_WINO
struct lfs_direct
struct lfs_dirtemplate
struct lfs_odirtemplate
struct ulfs_args
Also fix FFS_MAXNAMLEN -> LFS_MAXNAMLEN in several places.
|
| 1.6 | 06-Jun-2013 |
dholland | Remove references to Apple UFS.
|
| 1.5 | 06-Jun-2013 |
dholland | Remove stray references to ext2fs, chfs, ffs, and mfs.
|
| 1.4 | 06-Jun-2013 |
dholland | Split lfs from ufs step 4:
Massedit all ufs symbols to be "ulfs" instead, to make sure there are
no conflicts with ufs. Confirmed with grep.
(This required changing a few comments that maybe should have been
left alone to say "ulfs", but we'll survive that.)
|
| 1.3 | 06-Jun-2013 |
dholland | Split lfs from ufs step 3: rearrange config stuff.
Add new options:
LFS_EI
LFS_DIRHASH
LFS_EXTATTR
LFS_EXTATTR_AUTOSTART
LFS_QUOTA
LFS_QUOTA2
and update code referring to the corresponding FFS and UFS config
symbols to use the LFS versions. Disable the one extant reference
to APPLE_UFS in the ulfs files. Use opt_lfs.h only, not opt_ffs.h.
|
| 1.2 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 2:
Change all <ufs/ufs/foo.h> includes to <ufs/lfs/ulfs_foo.h>.
|
| 1.1 | 06-Jun-2013 |
dholland | Split lfs from ufs, part 1: cut and paste 15000 lines of ufs as "ulfs".
These are verbatim copies except that I've preserved the ufs rcsids
for reference. Also,
ufs/quota.h -> ulfs_quotacommon.h
ufs/ufs_quota.h -> ulfs_quota.h
Splitting lfs from ufs was ok'd by core some years ago. This is not
from my original tree, which became unmergeable after the several sets
of quota changes; I've done the work over again over the last couple
days.
|
| 1.7.4.1 | 28-Aug-2013 |
rmind | sync with head
|
| 1.7.2.4 | 03-Dec-2017 |
jdolecek | update from HEAD
|
| 1.7.2.3 | 20-Aug-2014 |
tls | Rebase to HEAD as of a few days ago.
|
| 1.7.2.2 | 23-Jun-2013 |
tls | resync from head
|
| 1.7.2.1 | 08-Jun-2013 |
tls | file ulfsmount.h was added on branch tls-maxphys on 2013-06-23 06:18:39 +0000
|
| 1.12.8.2 | 27-Dec-2015 |
skrll | Sync with HEAD (as of 26th Dec)
|
| 1.12.8.1 | 06-Jun-2015 |
skrll | Sync with HEAD
|
| 1.12.4.2 | 22-May-2014 |
yamt | sync with head.
for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.
this commit was splitted into small chunks to avoid
a limitation of cvs. ("Protocol error: too many arguments")
|
| 1.12.4.1 | 28-Jul-2013 |
yamt | file ulfsmount.h was added on branch yamt-pagecache on 2014-05-22 11:41:19 +0000
|
