Define _KERNTYPES for things that need it.

If a library needs a symbol from another library, pull that library in
explicitly, even if the DT_NEEDED closure would normally already ensure
the presence.

branches: 1.6.2;
These directories default to WARNS=2

branches: 1.5.6;
uncommit accidental changes

Use range locking (fnctl(F_SETLK)) instead of file locking (flock()).
This allows to mount multiple (non-overlapping) partitions from the
same disk image.

Add what is essentially disklabel support. All disk-based file
systems should be mounted with ukfs_mount_disk() from now on. The
partition argument specifies which label is being mounted (or the
entire image). E.g. partition 4 should be label 'e'.

branches: 1.2.2;
LIBDPLIBS on librumpvfs

branches: 1.1.2;
Install rump libraries and utilities to the base system and remove the
private non-installed build infrastructure from sys/rump.

breakdown of commit:
* install relevant headers into /usr/include/rump
* build sys/rump/librump/rumpuser and sys/rump/librump/rumpkern
from src/lib and install as librumpuser and librump, respectively
+ this retains the ability to test a librump build with just the
kernel sources at hand
* move sys/rump/fs/lib/libukfs and sys/rump/fs/lib/libp2k to src/lib
for general consumption, they are not kernel-space dwellers anyway
* build and install sys/rump/fs/lib/lib$fs as librumpfs_$fs
* add chapter 3 manual pages for rump, rumpuser, ukfs and p2k
* build and install userspace kernel file system daemons if MKPUFFS=yes
is spexified
* retire fsconsole for now, it will make a comeback with an actually
implemented version shortly

branches: 1.1.2;
file Version.map was initially added on branch agc-symver.

merge christos-time_t

branches: 1.1.2; 1.1.8;
Install rump libraries and utilities to the base system and remove the
private non-installed build infrastructure from sys/rump.

breakdown of commit:
* install relevant headers into /usr/include/rump
* build sys/rump/librump/rumpuser and sys/rump/librump/rumpkern
from src/lib and install as librumpuser and librump, respectively
+ this retains the ability to test a librump build with just the
kernel sources at hand
* move sys/rump/fs/lib/libukfs and sys/rump/fs/lib/libp2k to src/lib
for general consumption, they are not kernel-space dwellers anyway
* build and install sys/rump/fs/lib/lib$fs as librumpfs_$fs
* add chapter 3 manual pages for rump, rumpuser, ukfs and p2k
* build and install userspace kernel file system daemons if MKPUFFS=yes
is spexified
* retire fsconsole for now, it will make a comeback with an actually
implemented version shortly

s/it desired/is desired/ in documentation.

Fixes PR lib/59409. From Zhaoming Luo

branches: 1.18.2; 1.18.4;
remove double t from targeted, add missing r to arbitrary
And fix few more typos along the way in comments and man pages.

fix typos in word "successful".

Remove exgtraneous comma

branches: 1.15.2;
Typos.

New sentence, new line.

some minor updates and reality-checks

Use Mt for email addresses.

branches: 1.11.6; 1.11.10;
Add missing Fa macro argument.

branches: 1.10.6;
Update. Especially, describe ukfs_mount_disk(), ukfs_release() flags
and return value, and remove obsolete info in BUGS.

fix markup

Use .Bl -ohang instead of physical markup.

MNT_SOFTDEP -> MNT_LOG in example. pointed out by mrg.

branches: 1.6.2; 1.6.4;
Fix two mdoclint warnings.

Add BUGS section which notes that dynamic linker features in dealing
with link sets prevent from including more than one file system on
the link line and that the rest must be loaded dynamically at runtime.

branches: 1.4.2;
fix type for a few functions

Introduce ukfs_vfstypes() which returns the available file system
types loaded into rump. Now it is possible to iterate over all
file system types supported by a rump program in an attempt to
access an unknown file system image instead of having to know
beforehand which type of file system is on the image.

Support ukfs_modload(), which dlopens and vfs_attaches rump file
system modules for use. Sneakily this solves the problem with the
dynamic linker not wanting to handle the modules link set for
binaries where more than one file system library is included during
the link phase and therefore only one of the file systems getting
vfs_attach()ed in rump "boot". But more importantly, this is really
TRTTD, since now applications can be built, linked and shipped
completely independently of the file systems they support.

tested by Arnaud Ysmal

Install rump libraries and utilities to the base system and remove the
private non-installed build infrastructure from sys/rump.

breakdown of commit:
* install relevant headers into /usr/include/rump
* build sys/rump/librump/rumpuser and sys/rump/librump/rumpkern
from src/lib and install as librumpuser and librump, respectively
+ this retains the ability to test a librump build with just the
kernel sources at hand
* move sys/rump/fs/lib/libukfs and sys/rump/fs/lib/libp2k to src/lib
for general consumption, they are not kernel-space dwellers anyway
* build and install sys/rump/fs/lib/lib$fs as librumpfs_$fs
* add chapter 3 manual pages for rump, rumpuser, ukfs and p2k
* build and install userspace kernel file system daemons if MKPUFFS=yes
is spexified
* retire fsconsole for now, it will make a comeback with an actually
implemented version shortly

ukfs(3): Don't assume that <sys/endian.h> is transcluded via <sys/types.h>

lib: remove CONSTCOND comment

Since 2021-01-31, lint doesn't need it anymore for the common pattern of
'do ... while (0)'.

Don't include <rump/rumpvnode_if.h> from rump.h. It's not needed
unless you're doing something special, but requires register_t.
Adjust the few places which actually need rumpvnode_if.h.

fix error messages

Add support for a byteswapped disklabel so that I can mount
NetBSD/sparc anita images on my i386.

branches: 1.56.2;
rump_lwproc_newproc -> rump_lwproc_rfork

rumpuser should not be included outside of the rump kernel (and
rumpuser itself)

Migrate from rump private interfaces to syscalls.

update to new rump proc/lwp interfaces

convert to newstyle automagic rump compat syscalls

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.

Some _t purification. no functional change.

mfs is a bit off in the sense that mount(2) doesn't return since
mfs uses the mounting process for the backing store memory. I
guess mfs could be fixed to just reference the process vmspace and
let it return, but that would probably cause wait() to return for
other worms. So it's easier to dance according to mfs's tune: if
mounting mfs, create a thread for extra execution context.

Update to use newstyle rump_module interfaces.

Refcount ukfs_part. Otherwise it's not possible to call ukfs_mount()
several times with only one ukfs_part_probe().

fix error branch

Use range locking (fnctl(F_SETLK)) instead of file locking (flock()).
This allows to mount multiple (non-overlapping) partitions from the
same disk image.

Init ukfs__part_{na,none} statically instead of at runtime to avoid
init-order lossage from p2k/rump_smbfs, as noted by Tron.

Improve the ukfs magicpath interface a bit:
* rename PART to DISKLABEL
* allow to take an OFFSET range to make it possible to access an image
where there is no disklabel present
* make future extensions easier and less intrusive for the callers

Add RTLD_LAZY to dlopen call, to make it work on Linux hosts.

ok by pooka.

use newstyle rump lwp interfaces

Adjust names of external rump control interfaces to match the
new rump_pub namespace.

fix warning whine

Add what is essentially disklabel support. All disk-based file
systems should be mounted with ukfs_mount_disk() from now on. The
partition argument specifies which label is being mounted (or the
entire image). E.g. partition 4 should be label 'e'.

* allow callers to store a private data pointer behind the ukfs handle
* release reference on root vnode before unmounting
(and reaquire the root vnode if unmount fails)
* return correct error value if unmount fails

Give rump_sys_unmount() in ukfs_release() a curlwp context. This
prevents use-after-free and should fix crashy cases reported by
Arnaud Ysmal.

If opening the device fails, skip etfs registration only if it is
because of ENOENT. Otherwise, let the file system handle it.

fix errno reporting

* use new etfs interface
+ attempt to register only if we can stat() the devpath. this works
around e.g. nfs devpath, which is useless to register to etfs
XXX: the caller should decide

return proper errno from mount

delint

Since rump rootfs now supports creating directories, mount ukfs
mounts at the requested mountpoint instead of as the rump kernel
rootfs. While doing so, generally convert everything to use system
calls instead of handcrafter rump interfaces (e.g. rump_sys_mount()
instead of rump_mnt_mount()). This avoids a lot of unnecessary
fuss.

Change UKFS_DEFAULTMP from / to /ukfs and bump ukfs version.

Can't see any reason to stat() the .so before dlopen(), so get rid
of that call. uses normal ldlib paths now

use compat syscalls if necessary

Fix calls to rump syscalls now that the padding argument is no
longer advertised.

Since rump_module_load() doesn't actually load the module, change
the name to rump_module_init(). Also, adjust the signature to take
a direct pointer to modinfo and allow passing of props. Finally,
provide rump_module_fini().

Merge librumpfs_ufs into librumpfs_ffs. This reflects what happened
with the ffs kernel module and follows the trend of retiring ufs.
It also allows to get rid of a special case kludge in runtime module
loading, since ufs was not really a module. librumpfs_ufs is now
obsoleted and ffs consumers should be linked solely against
librumpfs_ffs.

Provide a stateful mode for readdir/read/write. Using them gives
a measurable boost to some fs-utils commands.

release root vnode at unmount

Do not flock() block devices. The kernel recently started to guard
against opening a block device twice and flock requires that we
keep the device open. Hence, the same mount would attempt to open
for locking and then VOP_OPEN for kernel access and run into
problems. This makes it possible to rump_foofs block devices again,
i.e. take the same command line arguments as regular mounts.
Noticed by apb.

branches: 1.21.2;
Adjust for explicit errno parameter removal from rump syscalls

Add #ifdef to disable pthread linkage requirement at compile-time.

Update per rump_sys call __RENAME().

merge christos-time_t

Nuke macros which haven't been used since the advent of RUMP_VOPs.

Closing file in ukfs_release

branches: 1.15.2;
Return ENODEV instead of ENOENT if the file system is not available.

rump_vfs_load -> rump_module_load

dlerror() resets the error message, so save it since we want to
access it twice.

Use RUMP_VOP instead of prehistoric rump_vp.

* use advisory file locks to avoid multiple mounts on the same device
(multiple rdonly mounts are still allowed)
* shuffle some assignments to avoid leaking resources in error branches

Embed library build version into the init call and refuse to init
if they don't match. In other words, this is a lightweight library
major number.

adapt to RUMPCRED_SUSER change

remove verbosity attack

branches: 1.7.2;
Fix rump_vfs_sync() credarg also.

Fix arg to RUMP_VOP_READDIR, NULL no longer works there.

Avoid pid 0
XXX: pid allocation should not be done here.

Introduce ukfs_vfstypes() which returns the available file system
types loaded into rump. Now it is possible to iterate over all
file system types supported by a rump program in an attempt to
access an unknown file system image instead of having to know
beforehand which type of file system is on the image.

Support ukfs_modload(), which dlopens and vfs_attaches rump file
system modules for use. Sneakily this solves the problem with the
dynamic linker not wanting to handle the modules link set for
binaries where more than one file system library is included during
the link phase and therefore only one of the file systems getting
vfs_attach()ed in rump "boot". But more importantly, this is really
TRTTD, since now applications can be built, linked and shipped
completely independently of the file systems they support.

tested by Arnaud Ysmal

fix pasto

Install rump libraries and utilities to the base system and remove the
private non-installed build infrastructure from sys/rump.

breakdown of commit:
* install relevant headers into /usr/include/rump
* build sys/rump/librump/rumpuser and sys/rump/librump/rumpkern
from src/lib and install as librumpuser and librump, respectively
+ this retains the ability to test a librump build with just the
kernel sources at hand
* move sys/rump/fs/lib/libukfs and sys/rump/fs/lib/libp2k to src/lib
for general consumption, they are not kernel-space dwellers anyway
* build and install sys/rump/fs/lib/lib$fs as librumpfs_$fs
* add chapter 3 manual pages for rump, rumpuser, ukfs and p2k
* build and install userspace kernel file system daemons if MKPUFFS=yes
is spexified
* retire fsconsole for now, it will make a comeback with an actually
implemented version shortly

branches: 1.1.4;
libukfs: Add expected symbols list.

PR lib/58838: shared libraries in base should all have expsym lists

lib: remove CONSTCOND comment

Since 2021-01-31, lint doesn't need it anymore for the common pattern of
'do ... while (0)'.

Always initialize part to avoid triggering uninitialized variable
warnings.

branches: 1.13.6;
Improve the ukfs magicpath interface a bit:
* rename PART to DISKLABEL
* allow to take an OFFSET range to make it possible to access an image
where there is no disklabel present
* make future extensions easier and less intrusive for the callers

default to UKFS_PARTITION_NONE in argvprobe

Add what is essentially disklabel support. All disk-based file
systems should be mounted with ukfs_mount_disk() from now on. The
partition argument specifies which label is being mounted (or the
entire image). E.g. partition 4 should be label 'e'.

* allow callers to store a private data pointer behind the ukfs handle
* release reference on root vnode before unmounting
(and reaquire the root vnode if unmount fails)
* return correct error value if unmount fails

Since rump rootfs now supports creating directories, mount ukfs
mounts at the requested mountpoint instead of as the rump kernel
rootfs. While doing so, generally convert everything to use system
calls instead of handcrafter rump interfaces (e.g. rump_sys_mount()
instead of rump_mnt_mount()). This avoids a lot of unnecessary
fuss.

Change UKFS_DEFAULTMP from / to /ukfs and bump ukfs version.

Provide a stateful mode for readdir/read/write. Using them gives
a measurable boost to some fs-utils commands.

branches: 1.7.4; 1.7.6;
Embed library build version into the init call and refuse to init
if they don't match. In other words, this is a lightweight library
major number.

branches: 1.6.2;
Introduce ukfs_vfstypes() which returns the available file system
types loaded into rump. Now it is possible to iterate over all
file system types supported by a rump program in an attempt to
access an unknown file system image instead of having to know
beforehand which type of file system is on the image.

__{BEGIN,END}_DECLS

Support ukfs_modload(), which dlopens and vfs_attaches rump file
system modules for use. Sneakily this solves the problem with the
dynamic linker not wanting to handle the modules link set for
binaries where more than one file system library is included during
the link phase and therefore only one of the file systems getting
vfs_attach()ed in rump "boot". But more importantly, this is really
TRTTD, since now applications can be built, linked and shipped
completely independently of the file systems they support.

tested by Arnaud Ysmal

No reason to forward def componentname anymore

Include <sys/types.h> for completeness, uid_t etc. in prototypes

noticed by Arnaud Ysmal

Install rump libraries and utilities to the base system and remove the
private non-installed build infrastructure from sys/rump.

breakdown of commit:
* install relevant headers into /usr/include/rump
* build sys/rump/librump/rumpuser and sys/rump/librump/rumpkern
from src/lib and install as librumpuser and librump, respectively
+ this retains the ability to test a librump build with just the
kernel sources at hand
* move sys/rump/fs/lib/libukfs and sys/rump/fs/lib/libp2k to src/lib
for general consumption, they are not kernel-space dwellers anyway
* build and install sys/rump/fs/lib/lib$fs as librumpfs_$fs
* add chapter 3 manual pages for rump, rumpuser, ukfs and p2k
* build and install userspace kernel file system daemons if MKPUFFS=yes
is spexified
* retire fsconsole for now, it will make a comeback with an actually
implemented version shortly

ukfs(3): Don't assume that <sys/endian.h> is transcluded via <sys/types.h>

Add support for a byteswapped disklabel so that I can mount
NetBSD/sparc anita images on my i386.

branches: 1.2.2;
Improve the ukfs magicpath interface a bit:
* rename PART to DISKLABEL
* allow to take an OFFSET range to make it possible to access an image
where there is no disklabel present
* make future extensions easier and less intrusive for the callers

Add what is essentially disklabel support. All disk-based file
systems should be mounted with ukfs_mount_disk() from now on. The
partition argument specifies which label is being mounted (or the
entire image). E.g. partition 4 should be label 'e'.

Add support for a byteswapped disklabel so that I can mount
NetBSD/sparc anita images on my i386.

branches: 1.2.2;
Improve the ukfs magicpath interface a bit:
* rename PART to DISKLABEL
* allow to take an OFFSET range to make it possible to access an image
where there is no disklabel present
* make future extensions easier and less intrusive for the callers

Add what is essentially disklabel support. All disk-based file
systems should be mounted with ukfs_mount_disk() from now on. The
partition argument specifies which label is being mounted (or the
entire image). E.g. partition 4 should be label 'e'.