branches: 1.1.4;
Install netbsd-dm.h file to include/dev/dm so it can be used by public.

Remove wrong TODO item, it violates abstraction between kernel-libdevmapper
and lvm tools.

branches: 1.5.2;
Update to reality.

branches: 1.4.2; 1.4.4; 1.4.6;
Add more tasks and add more details about them.

Remove done task and add new one.

Merge the haad-dm branch to -current. This branch adds LVM functionality to
the base NetBSD system. It uses Linux LVM2 tools and our BSD licensed
device-mapper driver.

The device-mapper driver can be used to create virtual block devices which
maps virtual blocks to real with target mapping called target. Currently
these targets are available a linear, zero, error and a snapshot (this is
work in progress and doesn't work yet).

The lvm2tools adds lvm and dmsetup binary to based system, where the lvm
tool is used to manage and administer whole LVM and the dmestup is used to
communicate iwith device-mapper kernel driver. With these tools also
a libdevmapper library is instaled to the base system.

Building of tools and driver is currently disable and can be enabled with
MKLVM=yes in mk.conf. I will add sets lists and rc.d script soon.

Oked by agc@ and cube@.

branches: 1.1.2;
Add dmgetdefaultdisklabel to get virtual disklabel for Logical Volume device. Add snapshot targets (snapshot, snapshot-origin), add dm_type to dm_dev structure to identify type of device. e.g. we can have mirrored disk device with snapshot on them and spare disk on them. When driver want to work with snapshot devices, it looks to upcalls list and finds all DM_SNAPSHOT devices.

Use xdevice_name for DIOCGWEDGEINFO answer to align with the registered
disk name. getdisksize() relies on it to find the disk structure.

Fixes PR 59674.
Pullup -11.

branches: 1.64.12;
For device modules that provide both auto-config and /dev/xxx
interfaces, make sure that initialization and destruction
follow the proper sequence. This is triggered by the recent
changes to the devsw stuff; per riastradh@ the required call
sequence is:

devsw_attach()
config_init_component() or config_cf*_attach()
...
config_fini_component() or config_cf*_detach()
devsw_detach()

While here, add a few missing calls to some of the detach
routines.

Testing of these changes has been limited to:
1. compile without build break
2. no related test failures from atf
3. modload/modunload work as well as
before.

No functional device testing done, since I don't have any
of these devices. Let me know of any damage I might cause
here!

XXX Some of the modules affected by this commit are already
XXX broken; see kern/56772. This commit does not break
any additional modules (as far as I know).

Media size is in bytes (off_t), not sectors.

Track the number of cdev and bdev opens and fail dm_detach()
on open devices unless detach is forced.

PR kern/54969 (Disk cache is no longer flushed on shutdown)

branches: 1.61.6;
Adapt to proplib API chanages.

dm(8) doesn't cause problems on suspend, don't block it.

Tested by Matthias Petermann, thanks!

branches: 1.59.2;
dm: Make numsec/secsize arguments in dm_table_disksize() optional

dm: Add missing "ioctl called" debug prints

dm: Refactor dmioctl()

More readable without dm_ioctl_switch() as a separate function.

dm: u_{int,long} -> unsigned {int,long}

dm: Rename dm specific atoi() to atoi64()

This is uint64_t version, not sys/lib/libsa/atoi.c.

dm: Style cleanups (no functional changes)

dm: Make targets' ->sync() optional

Apparently some targets have nothing to sync, so make it optional.

dm: "unsigned" -> "unsigned int" for consistency

Use either one, but not both.

dm: Move extern declaration of global variables to dm.h

dm: Remove void casts of function calls

Use of void casts in dm is not consistent, just get rid of them.

dm: Revert "Fix error handling in dmioctl()" for now

This change seems to break "deps" in dm ioctl(2) for linear target.
Revert the change, will revisit later.
https://mail-index.netbsd.org/current-users/2019/12/11/msg037179.html

dm: Fix error handling in dmioctl()

Looks like cleanup_exit label is in a wrong place.
Should skip prop_dictionary_copyout_ioctl() on error.

dm: Fix indentation

dm: Fix typos in comments/messages

taken-from: DragonFlyBSD

dm: Remove unneeded dm_get_version_ioctl()

"version" is implemented and handled in userspace,
hence dm ioctl doesn't need to support it.

taken-from: DragonFlyBSD

dm: Minor fixes/cleanups

dm: style + whitespace + indentation fixes

No functional changes.

dm: Doesn't need to expose struct cmd_function in a header

dm: Add a comment on race window on unload

There is a minor race window on unload vs device creation
that can cause panic.
https://github.com/DragonFlyBSD/DragonFlyBSD/commit/53a07f3ae7313aa58948a60f46428bfc2254dc3c

taken-from: DragonFlyBSD

branches: 1.40.4;
Add ioctls to query geometry.

branches: 1.39.2; 1.39.4;
Kill some more extern struct cfdriver declarations.

Down with externs in .c!

branches: 1.38.2; 1.38.8; 1.38.10;
KNF. No functional change.

include "ioconf.h" to get the 'void <driver>attach(int count);' prototype.

disk_blocksize and disk_set_info relay the same information
to the disk subsystem.

Make disk_set_info also set blocksize shift values.
Remove every call to disk_blocksize.

Keep disk_blocksize for ABI compatibility, make it also set dg_secsize.

branches: 1.35.2;
devmajor_t not int

Add d_discard to all struct cdevsw instances I could find.

All have been set to "nodiscard"; some should get a real implementation.

Add d_discard to all struct bdevsw instances I could find.

I've set them all to nodiscard. Some of them (wd, dk, vnd, ld,
raidframe, maybe cgd) should be implemented for real.

branches: 1.32.2;
Make dksubr.c into a module, and make the cgd and dm modules depend on
it.

Now that cgd is completely modularized, descend into modules/cgd to
actually create the module.

remove unused variables

branches: 1.30.2;
phase 1 of disk geometry cleanup:
- centralize the geometry -> plist code so that we don't have
n useless copies of it.

branches: 1.29.2;
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.

branches: 1.28.8; 1.28.12;
Now that we have allowed operator to access the control node, make sure
that he cannot cause damage, by only allowing the superuser to do ioctls
that can cause damage.

make dm aware of physical sector sizes.

For aggregates of multiple disks we use the largest sector size from
all disks. For standard power-of-2 sizes this is the same as the least
common multiple. We still require proper alignment of the targets in
the mapping table.

ok by haad@

I have forgot to fix dm_dev_counter here, too.

Add old file describing locking schema used in dm driver.

In rome do as romans do. If I will get EEXIST from devsw_attach don't exit and
continue. This unbreaks usage of libdm in RUMP.

Add support for DIOCCACHESYNC ioctl for dm devices. Add new sync function
pointer to dm_target_t because that is the only part of dm which know real
block device. disk_ioctl_switch parses whole device table and for every
entry it calls particular sync routine which propagates DIOCCACHESYNC
to real disk.

While I was here implement some KNF fixes and remove unneeded symbols from
dm.h.

Problem reported on port-xen@ by Hugo Silva.

Fix build of i386 ALL kernel with patch from Greg A. Woods in PR/43056.

Rework module/builtin code so it works in both cases.
(Tested recently in the module case, slightly less recently as builtin.)

<haad> [if it works] go for it

branches: 1.20.2;
Disable disk_ioctl_switch fo device-mapper control device.

Use RUN_ONCE(9) again. (Everyone happy now?)

Use correct prototype for dmattach().

Fix device-mapper to work in builtin case. This commit fixes

PR: 42799

REwork fix commited by jak@ to be more readable and remove unneeded ifdef _MODULE.

Be sure to config_cfattach_attach() in dmattach().
Loosely based on patch from Brian Brombacher,
with other inspriation from vnd(4).

This should fix PR kern/42799.

While here, a few prototype, style and namespace pollution-related changes.

branches: 1.15.2;
Fix this for non-module (which, i guess, is only i386/ALL).

XXX: autoconfusion vs. modules is an intolerable mess of stepping
on each others' toes.

Refactor dmioctl to by cleaner and to copyout dictionary in case when something
went wrong.

KNF police, remove unnecessary whitespaces and tabulators. Refactor dmioctl
to not use prop_dictionary_copy* functions if NetBSD_DM_IOCTL command was
not called on device. If disk_ioctl returms anything else then ENOTTY exit
from dmioctl.

Hook device-mapper to autoconf framework. Add dm_attach, dm_match and dm_detach
routines used by autoconf users. Change dm_dev_remove_ioctl to call dm_detach.

This should be primary used by kernel to disable devices during shutdown of
system with nested disk devices.

Requested by dyoung@.

Add private lock to dm_dev_t used for mutual exclusion for diks(9) api
routines. This change fixes PR kern/42532.

Add support for DM_QUERY_INACTIVE_TABLE_FLAG to dm_table_status and dm_table_deps
this flag was introduced to dm->lvm protocol in 4.16.0 version of it. Restore
vrsion check functionality and sent back actual kernel driver version.

Don't spam the console with debug messages.

Some fixes for dm(4).

- Limit accesses to the size of the volume.
This fixes a KASSERT in physio_biodone().
- Change dm_table_size() to return the size of the volume in
DEV_BSIZE units. It was reporting it in DEV_BSIZE^2 units.
- Remove a bit of trailing whitespace.

ok haad

Add support for DIOCGDISKINFO to disk like device drivers. Change
partutil.c::getdiskinfo to use it to get disk geometry info.
Use DIOCGWEDGEINFO ioctl to get information about partition size, if disk
driver doesn't support it use old DIOCGDINFO. This patch adds support for
wedge like devices(lvm logical volumes, ZFS zvol partitions) to newfs and
other tools.

No objections on tech-userlevel@.

branches: 1.6.2;
Use functions from disk(9) framework. Initialize disk/disklabel during
dm_device_create_ioctl, before calling dmgetdisklabel.
Use disk_busy/disk_unbusy in dmstrategy to display LVM LV's in iostat
output.

branches: 1.5.2;
Move back to printing the minor device as a 32-bit entity

branches: 1.4.2; 1.4.4;
Disable unloading of dm driver when there are any devices defined. This means
that we have to run lvm vgchange -a n before modunload in NetBSD.

We really need to disable module unloading only for mounted devices, I have
to look at vfs_hooks for mount/umount calls and find way how to mark devices
as mounted in dm driver.

Fix aprint_debug after christos-time_t merge.

Merge the haad-dm branch to -current. This branch adds LVM functionality to
the base NetBSD system. It uses Linux LVM2 tools and our BSD licensed
device-mapper driver.

The device-mapper driver can be used to create virtual block devices which
maps virtual blocks to real with target mapping called target. Currently
these targets are available a linear, zero, error and a snapshot (this is
work in progress and doesn't work yet).

The lvm2tools adds lvm and dmsetup binary to based system, where the lvm
tool is used to manage and administer whole LVM and the dmestup is used to
communicate iwith device-mapper kernel driver. With these tools also
a libdevmapper library is instaled to the base system.

Building of tools and driver is currently disable and can be enabled with
MKLVM=yes in mk.conf. I will add sets lists and rc.d script soon.

Oked by agc@ and cube@.

branches: 1.1.2;
Add dmgetdefaultdisklabel to get virtual disklabel for Logical Volume device. Add snapshot targets (snapshot, snapshot-origin), add dm_type to dm_dev structure to identify type of device. e.g. we can have mirrored disk device with snapshot on them and spare disk on them. When driver want to work with snapshot devices, it looks to upcalls list and finds all DM_SNAPSHOT devices.

fix typos in sys/dev/dm code comments and documentation.
also remove some trailing space in documentation.

dm: #if0 target's ->upcall() handler

This is part of NetBSD's dm design, but unimplemented
(all handlers return 0) and also unused.

branches: 1.54.2;
dm: Add dm-delay target

Ported from DragonFlyBSD, but this target had originally existed in
Linux kernel. See below for details.
https://www.kernel.org/doc/Documentation/device-mapper/delay.txt

Due to "tick" in hz(9) not working (which results in dmdlthread spinning
forever in _submit_queue() without dp extracted from delayed list
and queued into submit list), this hasn't been hooked to dm.kmod yet.

taken-from: DragonFlyBSD

dm: Add dm-flakey target

Ported from DragonFlyBSD, but this target had originally existed in
Linux kernel. See below for details.
https://www.kernel.org/doc/Documentation/device-mapper/dm-flakey.txt

Due to two technical issues, this hasn't been hooked to dm.kmod yet.
1) "tick" in hz(9) not working.
2) Unable to use ->b_private in nestiobuf callback when it's already
used for mbp (see HAS_BUF_PRIV2).

taken-from: DragonFlyBSD

dm: Make target's ->table() optional

Since ->info() (counter part of ->table() in the original dm design
in Linux kernel in .status where both INFO and TABLE are optional)
is an optional handler, make ->table() optional as well. Some
targets don't have anything to do in ->table() just as in ->info().

taken-from: DragonFlyBSD

dm: Fix dm-stripe's "status" output format

As mentioned in "dm: Don't try to implement "status" as subset of "table"",
dm-stripe in NetBSD doesn't have correct "status" output format.
Implement ->info() to sync with Linux kernel.

Note that num_error for stripe device isn't implemented yet.

taken-from: DragonFlyBSD

dm: Remove target's ->deps() by implementing deps in dm core

Retrieving device dependencies doesn't need to be target specific.
The reason it currently needs ->deps() is because dm core doesn't
have data structure that allows table to walk through target's
underlying devices. Add struct dm_mapping to be able to do this,
and remove ->deps()'s from targets which basically do the same thing.

=====(A) before this commit
table
| [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

=====(B) this commit
table---->mapping-->mapping-->mapping-->...
| | | |
| v v v [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

taken-from: DragonFlyBSD

dm: Fix "table" output format of dm-linear and dm-stripe

The existing "table" output showing device file path of pdev is
not compatible with dm in Linux kernel (and also DragonFlyBSD).
It should be showing "major:minor" instead.

taken-from: DragonFlyBSD

dm: Cleanup dm.h (remove unneeded comments, etc)

dm: Enable dm-error and dm-zero target

Add these two targets to dm.kmod.
These are generally available in Linux and DragonFlyBSD,
so enable them in NetBSD as well.

dm: Rename dm specific atoi() to atoi64()

This is uint64_t version, not sys/lib/libsa/atoi.c.

dm: Make targets' ->sync() optional

Apparently some targets have nothing to sync, so make it optional.

dm: Make targets' ->secsize() optional

and make a caller assume secsize 0 if ->secsize not present.
This allows a dummy function to be removed which was added in
"dm: Add dummy target ->sync()/->secsize() to prevent panic on modload(8)".

dm: "unsigned" -> "unsigned int" for consistency

Use either one, but not both.

dm: Rename targets' ->status() to ->table() given ->info() exists

Since now that dm targets in NetBSD have ->info() for "status",
->status() should be renamed to ->table() for "table",
given how dm target status was originally designed in Linux kernel.

taken-from: DragonFlyBSD

dm: Don't try to implement "status" as subset of "table"

The way dm_table_status_ioctl() implements "status" and "table" is
not compatible with Linux kernel. Some targets have different outputs
that "status" can't be implemented as subset of "table".

Add ->info() handler to sync with "status" behavior in Linux kernel.
Some targets which currently exist in NetBSD (I think striped)
as well as some minor targets that I plan to port to NetBSD
can/should implement ->info(), but will do that in a different commit.

taken-from: DragonFlyBSD

dm: Move extern declaration of global variables to dm.h

dm: Make target's ->init() take parsed argc and argv

This gets rid of the same parser code in each target using strsep(3).
taken-from: DragonFlyBSD

dm: Add dummy target ->sync()/->secsize() to prevent panic on modload(8)

dm_target_insert() has assertions to ensure targets implement all handlers.
Adding dummy ones at least prevents panic on modload(8).

dm: Move targets specific structs to .c files

These don't need to be defined and exposed in dm.h.

dm: Unbreak compilation of kernel modules

The dm kernel modules (MK_DM_TARGETS enabled in sys/modules/dm/Makefile)
have been broken. Unbreak the build.

dm: Refactor target's ->init() i/f

Take dm_table_entry_t* instead of void**.
Remove dm_dev_t* unneeded by target code.
No functional change, but for future changes.

taken-from: DragonFlyBSD

dm: Fix strange pointer declarations

Should be "type *name" or "type* name", but not "type * name".
taken-from: DragonFlyBSD

dm: Remove unneeded dm_get_version_ioctl()

"version" is implemented and handled in userspace,
hence dm ioctl doesn't need to support it.

taken-from: DragonFlyBSD

dm: Comment out unused dm_dev fields documented in sys/dev/dm/doc

dm: Remove unused macro MAX_TARGET_STRING_LEN

that was added but never used in both NetBSD and DragonFlyBSD.
taken-from: DragonFlyBSD

dm: Include <sys/vnode.h> in dm.h

dm.h already depends on vnode, so have dm.h include <sys/vnode.h>
instead of other .c files.

taken-from: DragonFlyBSD

dm: Doesn't need to expose struct cmd_function in a header

dm: Remove unused dm_dev::dev_type

Given OOP-like architecture of dm target device structure,
dm_dev doesn't need to have self contained target type field,
and in fact this is unused.

branches: 1.27.20;
dm_target_t reference count is modified with atomic_{inc,dec}_32 so should
be uint_32_t not int

Change dk_lookup() to return an anonymous vnode not associated with
any file system. Change all consumers of dk_lookup() to get the
device from "v_rdev" instead of VOP_GETATTR() as specfs does not
support VOP_GETATTR(). Devices obtained with dk_lookup() will no
longer disappear on forced unmounts.

Fix for PR kern/48849 (root mirror raid fails on shutdown)

Welcome to 6.99.44

branches: 1.25.2;
Fix typo ("then" instead of "than")

branches: 1.24.2; 1.24.4;
Fix -fno-common found by building i386/conf/ALL

branches: 1.23.2;
- add a function to get the inactive table's size
- some whitespace fix from emacs...

Now that we have allowed operator to access the control node, make sure
that he cannot cause damage, by only allowing the superuser to do ioctls
that can cause damage.

make dm aware of physical sector sizes.

For aggregates of multiple disks we use the largest sector size from
all disks. For standard power-of-2 sizes this is the same as the least
common multiple. We still require proper alignment of the targets in
the mapping table.

ok by haad@

Add comment about dm_dev minor, and change it to 32bit value we realy not
going to use so much device minor numbers.

Add old file describing locking schema used in dm driver.

Add multi device strip support written by Guillermo Amaral and reviewed by me.

branches: 1.17.2; 1.17.4;
Add private lock to dm_dev_t used for mutual exclusion for diks(9) api
routines. This change fixes PR kern/42532.

Make our driver version equal to linux driver.

Revert my commit which have added knowledge about dm targets to libdevmapper,
this breaks abstraction. Because only lvmtools/lvmlib and device-mapper can
have knowledge about target mapping and libdevmapper only passes requests
from lvmtools to kernel and back. Bump major library and driver version.

Requested by: yamt@

Add support for DIOCGDISKINFO to disk like device drivers. Change
partutil.c::getdiskinfo to use it to get disk geometry info.
Use DIOCGWEDGEINFO ioctl to get information about partition size, if disk
driver doesn't support it use old DIOCGDINFO. This patch adds support for
wedge like devices(lvm logical volumes, ZFS zvol partitions) to newfs and
other tools.

No objections on tech-userlevel@.

Parse dm param string in libdevmapper and not in a dm target init function.
Create proplib param dictionary entry in libdevmapper and pass it to dm in
dm_ioctl dict.
Param target is then passed to target init function, where is parse. I like
this aproach much better than passing char **argv and trusting to user input.

I have bumped minor lib/driver version.

XXX. Add more sanity checks in kernel.

branches: 1.12.2;
It is only by chance that this gets the prop_array_t definition that it
needs, so explicitly #include <prop/proplib.h>.

Fix lvm lvrename command. There was bug in dm_dev_lookup where
dm_dev_lookup_name was called with device uuid. Remove dm_dev_t:dk_label is
it not used anymore.

Add initial support for striping target, hardcode maximal number of stripes
to 2 for now. strategy routine must be written, for now only parsing and
support routines are available. This is work in progress code and should be
taken very carefully.

Add support for autoloading of device-mapper targets modules. Add
dm_target_autoload function which tries to load target module. Fix two
deadlocks in dm_table_load_ioctl error path(I forgot to call dm_dev_unbusy).

branches: 1.8.2; 1.8.4; 1.8.6;
Add struct disk to dm_dev so our LV will beshowed in iostat output.

Use 64bit long minor nuber in dm_dev structure.

Add stubs for mirror target, too. This target does nothing for now, but lvm
tools support mirror target and trying to create LV with mirrorred backend
caused panic in dm_table_load_ioctl.

Add stripe target functions stubs. Stripe target must be present in dm driver
because without it lvm2tools will not create LVsi and eventualy panic system.

Problem reported by agc@.

Add support for loading dm targets as separate modules. All targets except
linear can be loaded as module. Module is not loaded when there is target
with similar name already. Zero and error targets aresimple examples how
can be all future targets written to support dynamic loading. Target can't
be unloaded until there is at least one user.

Add infrastructure needed to load device-mapper targets as modules.
Targets wasn't converted yet and at least snapshot target will be converted
in a near future.

Merge the haad-dm branch to -current. This branch adds LVM functionality to
the base NetBSD system. It uses Linux LVM2 tools and our BSD licensed
device-mapper driver.

The device-mapper driver can be used to create virtual block devices which
maps virtual blocks to real with target mapping called target. Currently
these targets are available a linear, zero, error and a snapshot (this is
work in progress and doesn't work yet).

The lvm2tools adds lvm and dmsetup binary to based system, where the lvm
tool is used to manage and administer whole LVM and the dmestup is used to
communicate iwith device-mapper kernel driver. With these tools also
a libdevmapper library is instaled to the base system.

Building of tools and driver is currently disable and can be enabled with
MKLVM=yes in mk.conf. I will add sets lists and rc.d script soon.

Oked by agc@ and cube@.

branches: 1.1.2;
Add dmgetdefaultdisklabel to get virtual disklabel for Logical Volume device. Add snapshot targets (snapshot, snapshot-origin), add dm_type to dm_dev structure to identify type of device. e.g. we can have mirrored disk device with snapshot on them and spare disk on them. When driver want to work with snapshot devices, it looks to upcalls list and finds all DM_SNAPSHOT devices.

fix typos in sys/dev/dm code comments and documentation.
also remove some trailing space in documentation.

Adapt to proplib API chanages.

dm: Remove unnecessary inlining

These two don't really need to be inlined.

dm: Style cleanups (no functional changes)

dm: Remove void casts of function calls

Use of void casts in dm is not consistent, just get rid of them.

dm: Fix strange pointer declarations

Should be "type *name" or "type* name", but not "type * name".
taken-from: DragonFlyBSD

dm: style + whitespace + indentation fixes

No functional changes.

dm: Make global dm locks static

These are all per-file locks which protect dev/pdev/target structure.
taken-from: DragonFlyBSD

branches: 1.11.4;
use some size_t, add whitespace between functions, rcsids.

remove checks for failure after memory allocation calls that cannot fail:

kmem_alloc() with KM_SLEEP
kmem_zalloc() with KM_SLEEP
percpu_alloc()
pserialize_create()
psref_class_create()

all of these paths include an assertion that the allocation has not failed,
so callers should not assert that again.

branches: 1.9.2;
src is too big these days to tolerate superfluous apostrophes. It's
"its", people!

branches: 1.8.22;
Indent files remove unnecessary blank lines, white spaces and KNFize code.

Add private lock to dm_dev_t used for mutual exclusion for diks(9) api
routines. This change fixes PR kern/42532.

Fix bug in kmem_alloc/kmem_free of params string. Params string was
allocated with length DM_MAX_PARAMS_SIZE and released with strlen + 1 size.

Disable KM_NOSLEEP allocation because we do not need them here there is
nothing critical in ioctl part of dm driver.

Bug reported by jak@.

branches: 1.5.2;
Destroy locks in dm_dev_free, do not allocate struct disk twice.

Ansify function definitions w/o arguments. Generated with sed.

Fix lvm lvrename command. There was bug in dm_dev_lookup where
dm_dev_lookup_name was called with device uuid. Remove dm_dev_t:dk_label is
it not used anymore.

branches: 1.2.2; 1.2.4; 1.2.6;
Merge the haad-dm branch to -current. This branch adds LVM functionality to
the base NetBSD system. It uses Linux LVM2 tools and our BSD licensed
device-mapper driver.

The device-mapper driver can be used to create virtual block devices which
maps virtual blocks to real with target mapping called target. Currently
these targets are available a linear, zero, error and a snapshot (this is
work in progress and doesn't work yet).

The lvm2tools adds lvm and dmsetup binary to based system, where the lvm
tool is used to manage and administer whole LVM and the dmestup is used to
communicate iwith device-mapper kernel driver. With these tools also
a libdevmapper library is instaled to the base system.

Building of tools and driver is currently disable and can be enabled with
MKLVM=yes in mk.conf. I will add sets lists and rc.d script soon.

Oked by agc@ and cube@.

branches: 1.1.2;
Add dmgetdefaultdisklabel to get virtual disklabel for Logical Volume device. Add snapshot targets (snapshot, snapshot-origin), add dm_type to dm_dev structure to identify type of device. e.g. we can have mirrored disk device with snapshot on them and spare disk on them. When driver want to work with snapshot devices, it looks to upcalls list and finds all DM_SNAPSHOT devices.

Avoid leaving a configured device without resources.

s/exaple/example/

fix typos in sys/dev/dm code comments and documentation.
also remove some trailing space in documentation.

fix various typos in comments and log messages.

Fix all remaining typos, mainly in comments but also in few definitions and log messages, reported by me in PR kern/54889.
Also fixed some additional typos in comments, found on review of same files or typos.

fix proplib deprecation

Make sure the unit number of device-mapper devices matches their minor number.

branches: 1.50.6;
Adapt to proplib API chanages.

dm: Make target's ->table() optional

Since ->info() (counter part of ->table() in the original dm design
in Linux kernel in .status where both INFO and TABLE are optional)
is an optional handler, make ->table() optional as well. Some
targets don't have anything to do in ->table() just as in ->info().

taken-from: DragonFlyBSD

dm: Remove target's ->deps() by implementing deps in dm core

Retrieving device dependencies doesn't need to be target specific.
The reason it currently needs ->deps() is because dm core doesn't
have data structure that allows table to walk through target's
underlying devices. Add struct dm_mapping to be able to do this,
and remove ->deps()'s from targets which basically do the same thing.

=====(A) before this commit
table
| [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

=====(B) this commit
table---->mapping-->mapping-->mapping-->...
| | | |
| v v v [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

taken-from: DragonFlyBSD

dm: Minor dm_ioctl.c fixes (indentation/typo/type/etc)

dm: Style cleanups (no functional changes)

dm: Rename targets' ->status() to ->table() given ->info() exists

Since now that dm targets in NetBSD have ->info() for "status",
->status() should be renamed to ->table() for "table",
given how dm target status was originally designed in Linux kernel.

taken-from: DragonFlyBSD

dm: Don't try to implement "status" as subset of "table"

The way dm_table_status_ioctl() implements "status" and "table" is
not compatible with Linux kernel. Some targets have different outputs
that "status" can't be implemented as subset of "table".

Add ->info() handler to sync with "status" behavior in Linux kernel.
Some targets which currently exist in NetBSD (I think striped)
as well as some minor targets that I plan to port to NetBSD
can/should implement ->info(), but will do that in a different commit.

taken-from: DragonFlyBSD

dm: Make dm_dbg_print_flags() take uint32_t flag

which comes from prop_dictionary_get_uint32() result.
taken-from: DragonFlyBSD

dm: Move extern declaration of global variables to dm.h

dm: Make target's ->init() take parsed argc and argv

This gets rid of the same parser code in each target using strsep(3).
taken-from: DragonFlyBSD

dm: Refactor target's ->init() i/f

Take dm_table_entry_t* instead of void**.
Remove dm_dev_t* unneeded by target code.
No functional change, but for future changes.

taken-from: DragonFlyBSD

dm: Fix typos in comments/messages

taken-from: DragonFlyBSD

dm: Remove unneeded dm_get_version_ioctl()

"version" is implemented and handled in userspace,
hence dm ioctl doesn't need to support it.

taken-from: DragonFlyBSD

dm: style + whitespace + indentation fixes

No functional changes.

dm: Include <sys/vnode.h> in dm.h

dm.h already depends on vnode, so have dm.h include <sys/vnode.h>
instead of other .c files.

taken-from: DragonFlyBSD

dm: Remove unused dm_dev::dev_type

Given OOP-like architecture of dm target device structure,
dm_dev doesn't need to have self contained target type field,
and in fact this is unused.

dm: Always initialize target's status string

Explicitly clear status string to prevent dmsetup(8) from
printing binary junk to stdout. Similar change has been
applied to DragonFlyBSD since 2015.

https://github.com/DragonFlyBSD/DragonFlyBSD/commit/a6cf54187fd7c3e994c573215806ec03572c038e

taken-from: DragonFlyBSD

Only _if_ an inactive table has been loaded, make it active when resuming.
Fixes lvrename operation.

branches: 1.32.2; 1.32.4;
PR/52900: Tomohiro Kusumi: Fix kernel panic when target's ->init() failed
Reliably unbusy the target in the *all* the destroy routines, so ioctl
does not need to do it.

remove checks for failure after memory allocation calls that cannot fail:

kmem_alloc() with KM_SLEEP
kmem_zalloc() with KM_SLEEP
percpu_alloc()
pserialize_create()
psref_class_create()

all of these paths include an assertion that the allocation has not failed,
so callers should not assert that again.

branches: 1.30.8;
CID 976256: Help coverity understand that when the list is empty the head
is NULL.

branches: 1.29.2;
Call prop_dictionary_get_uint32 with uint32_t

release proplib iterator in failure paths.

remove unused variables

branches: 1.26.2; 1.26.12; 1.26.16;
- add a function to get the inactive table's size
- some whitespace fix from emacs...

Fix logic error

branches: 1.24.2;
We don't need 64 bit variable for dm_device coutner, too.

Add comment about dm_dev minor, and change it to 32bit value we realy not
going to use so much device minor numbers.

Fix typo in comment.

branches: 1.21.2;
Be sure to config_cfattach_attach() in dmattach().
Loosely based on patch from Brian Brombacher,
with other inspriation from vnd(4).

This should fix PR kern/42799.

While here, a few prototype, style and namespace pollution-related changes.

branches: 1.20.2;
Indent files remove unnecessary blank lines, white spaces and KNFize code.

Hook device-mapper to autoconf framework. Add dm_attach, dm_match and dm_detach
routines used by autoconf users. Change dm_dev_remove_ioctl to call dm_detach.

This should be primary used by kernel to disable devices during shutdown of
system with nested disk devices.

Requested by dyoung@.

Add private lock to dm_dev_t used for mutual exclusion for diks(9) api
routines. This change fixes PR kern/42532.

Add support for DM_QUERY_INACTIVE_TABLE_FLAG to dm_table_status and dm_table_deps
this flag was introduced to dm->lvm protocol in 4.16.0 version of it. Restore
vrsion check functionality and sent back actual kernel driver version.

Fix bug in dm_table_resume_ioctl where dmv->flags was sent back to libdevmapper
except flags variable. This fixes weird behaviour, when worng links to
devices in /dev/mapper were created after lvrename/lvresize.

Revert my commit which have added knowledge about dm targets to libdevmapper,
this breaks abstraction. Because only lvmtools/lvmlib and device-mapper can
have knowledge about target mapping and libdevmapper only passes requests
from lvmtools to kernel and back. Bump major library and driver version.

Requested by: yamt@

Fix bug in kmem_alloc/kmem_free of params string. Params string was
allocated with length DM_MAX_PARAMS_SIZE and released with strlen + 1 size.

Disable KM_NOSLEEP allocation because we do not need them here there is
nothing critical in ioctl part of dm driver.

Bug reported by jak@.

Add support for DIOCGDISKINFO to disk like device drivers. Change
partutil.c::getdiskinfo to use it to get disk geometry info.
Use DIOCGWEDGEINFO ioctl to get information about partition size, if disk
driver doesn't support it use old DIOCGDINFO. This patch adds support for
wedge like devices(lvm logical volumes, ZFS zvol partitions) to newfs and
other tools.

No objections on tech-userlevel@.

Parse dm param string in libdevmapper and not in a dm target init function.
Create proplib param dictionary entry in libdevmapper and pass it to dm in
dm_ioctl dict.
Param target is then passed to target init function, where is parse. I like
this aproach much better than passing char **argv and trusting to user input.

I have bumped minor lib/driver version.

XXX. Add more sanity checks in kernel.

branches: 1.11.2;
Destroy locks in dm_dev_free, do not allocate struct disk twice.

Use functions from disk(9) framework. Initialize disk/disklabel during
dm_device_create_ioctl, before calling dmgetdisklabel.
Use disk_busy/disk_unbusy in dmstrategy to display LVM LV's in iostat
output.

Fix tyop in previous.

Release tables before we drop reference on device, now creating device with
nonexistent target fails and not ends with deadlock.

Add support for autoloading of device-mapper targets modules. Add
dm_target_autoload function which tries to load target module. Fix two
deadlocks in dm_table_load_ioctl error path(I forgot to call dm_dev_unbusy).

branches: 1.6.2; 1.6.4; 1.6.6;
Disable unloading of dm driver when there are any devices defined. This means
that we have to run lvm vgchange -a n before modunload in NetBSD.

We really need to disable module unloading only for mounted devices, I have
to look at vfs_hooks for mount/umount calls and find way how to mark devices
as mounted in dm driver.

Use 64bit long minor nuber in dm_dev structure.

Fix error path in target_init call. When initializing target fails i have
to release table before I try to destroy it and I haveto unbusy device, too.

Add infrastructure needed to load device-mapper targets as modules.
Targets wasn't converted yet and at least snapshot target will be converted
in a near future.

Merge the haad-dm branch to -current. This branch adds LVM functionality to
the base NetBSD system. It uses Linux LVM2 tools and our BSD licensed
device-mapper driver.

The device-mapper driver can be used to create virtual block devices which
maps virtual blocks to real with target mapping called target. Currently
these targets are available a linear, zero, error and a snapshot (this is
work in progress and doesn't work yet).

The lvm2tools adds lvm and dmsetup binary to based system, where the lvm
tool is used to manage and administer whole LVM and the dmestup is used to
communicate iwith device-mapper kernel driver. With these tools also
a libdevmapper library is instaled to the base system.

Building of tools and driver is currently disable and can be enabled with
MKLVM=yes in mk.conf. I will add sets lists and rc.d script soon.

Oked by agc@ and cube@.

branches: 1.1.2;
Add dmgetdefaultdisklabel to get virtual disklabel for Logical Volume device. Add snapshot targets (snapshot, snapshot-origin), add dm_type to dm_dev structure to identify type of device. e.g. we can have mirrored disk device with snapshot on them and spare disk on them. When driver want to work with snapshot devices, it looks to upcalls list and finds all DM_SNAPSHOT devices.

fix typos in sys/dev/dm code comments and documentation.
also remove some trailing space in documentation.

dm: Fix "table" output format of dm-linear and dm-stripe

The existing "table" output showing device file path of pdev is
not compatible with dm in Linux kernel (and also DragonFlyBSD).
It should be showing "major:minor" instead.

taken-from: DragonFlyBSD

dm: Style cleanups (no functional changes)

dm: Fix memory leak in dm_pdev_rem()

Need to free dmp regardless of dm_pdev_rem() return value.
taken-from: DragonFlyBSD

Switch to vn_bdev_open* functions.

dm: Fix strange pointer declarations

Should be "type *name" or "type* name", but not "type * name".
taken-from: DragonFlyBSD

dm: Simplify list eviction code

taken-from: DragonFlyBSD

dm: Remove trailing whitespace

dm: Minor fixes/cleanups

dm: style + whitespace + indentation fixes

No functional changes.

dm: Make global dm locks static

These are all per-file locks which protect dev/pdev/target structure.
taken-from: DragonFlyBSD

dm: Include <sys/vnode.h> in dm.h

dm.h already depends on vnode, so have dm.h include <sys/vnode.h>
instead of other .c files.

taken-from: DragonFlyBSD

dm: Fix race on pdev create

List lookup and insert need to be atomic.
https://github.com/DragonFlyBSD/DragonFlyBSD/commit/ac816675c8321309b49d6a335f95c5388036803f

take-from: DragonFlyBSD

branches: 1.10.4;
use some size_t, add whitespace between functions, rcsids.

remove checks for failure after memory allocation calls that cannot fail:

kmem_alloc() with KM_SLEEP
kmem_zalloc() with KM_SLEEP
percpu_alloc()
pserialize_create()
psref_class_create()

all of these paths include an assertion that the allocation has not failed,
so callers should not assert that again.

branches: 1.8.18; 1.8.36;
make dm aware of physical sector sizes.

For aggregates of multiple disks we use the largest sector size from
all disks. For standard power-of-2 sizes this is the same as the least
common multiple. We still require proper alignment of the targets in
the mapping table.

ok by haad@

Introduce struct pathbuf. This is an abstraction to hold a pathname
and the metadata required to interpret it. Callers of namei must now
create a pathbuf and pass it to NDINIT (instead of a string and a
uio_seg), then destroy the pathbuf after the namei session is
complete.

Update all namei call sites accordingly. Add a pathbuf(9) man page and
update namei(9).

The pathbuf interface also now appears in a couple of related
additional places that were passing string/uio_seg pairs that were
later fed into NDINIT. Update other call sites accordingly.

branches: 1.6.4;
Indent files remove unnecessary blank lines, white spaces and KNFize code.

Replace aprint_Verbose with aprint_debug to shutup unnecessary logs.

Fix bug in kmem_alloc/kmem_free of params string. Params string was
allocated with length DM_MAX_PARAMS_SIZE and released with strlen + 1 size.

Disable KM_NOSLEEP allocation because we do not need them here there is
nothing critical in ioctl part of dm driver.

Bug reported by jak@.

branches: 1.3.2;
Ansify function definitions w/o arguments. Generated with sed.

branches: 1.2.2; 1.2.4; 1.2.6;
Merge the haad-dm branch to -current. This branch adds LVM functionality to
the base NetBSD system. It uses Linux LVM2 tools and our BSD licensed
device-mapper driver.

The device-mapper driver can be used to create virtual block devices which
maps virtual blocks to real with target mapping called target. Currently
these targets are available a linear, zero, error and a snapshot (this is
work in progress and doesn't work yet).

The lvm2tools adds lvm and dmsetup binary to based system, where the lvm
tool is used to manage and administer whole LVM and the dmestup is used to
communicate iwith device-mapper kernel driver. With these tools also
a libdevmapper library is instaled to the base system.

Building of tools and driver is currently disable and can be enabled with
MKLVM=yes in mk.conf. I will add sets lists and rc.d script soon.

Oked by agc@ and cube@.

branches: 1.1.2;
Add dmgetdefaultdisklabel to get virtual disklabel for Logical Volume device. Add snapshot targets (snapshot, snapshot-origin), add dm_type to dm_dev structure to identify type of device. e.g. we can have mirrored disk device with snapshot on them and spare disk on them. When driver want to work with snapshot devices, it looks to upcalls list and finds all DM_SNAPSHOT devices.

fix typos in sys/dev/dm code comments and documentation.
also remove some trailing space in documentation.

fix various typos in comments and log messages.

dm: Remove unnecessary inlining

These two don't really need to be inlined.

dm: Make numsec/secsize arguments in dm_table_disksize() optional

dm: Remove target's ->deps() by implementing deps in dm core

Retrieving device dependencies doesn't need to be target specific.
The reason it currently needs ->deps() is because dm core doesn't
have data structure that allows table to walk through target's
underlying devices. Add struct dm_mapping to be able to do this,
and remove ->deps()'s from targets which basically do the same thing.

=====(A) before this commit
table
| [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

=====(B) this commit
table---->mapping-->mapping-->mapping-->...
| | | |
| v v v [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

taken-from: DragonFlyBSD

dm: Style cleanups (no functional changes)

dm: Make targets' ->secsize() optional

and make a caller assume secsize 0 if ->secsize not present.
This allows a dummy function to be removed which was added in
"dm: Add dummy target ->sync()/->secsize() to prevent panic on modload(8)".

dm: "unsigned" -> "unsigned int" for consistency

Use either one, but not both.

dm: Remove void casts of function calls

Use of void casts in dm is not consistent, just get rid of them.

dm: Fix strange pointer declarations

Should be "type *name" or "type* name", but not "type * name".
taken-from: DragonFlyBSD

dm: Simplify list eviction code

taken-from: DragonFlyBSD

dm: Make dm core set config to NULL after destroy

Just let dm core do this instead of having a comment expecting each
target to do the right thing.

taken-from: DragonFlyBSD

dm: style + whitespace + indentation fixes

No functional changes.

branches: 1.8.4;
use some size_t, add whitespace between functions, rcsids.

add function to get the inactive table size (for during target init)

make dm aware of physical sector sizes.

For aggregates of multiple disks we use the largest sector size from
all disks. For standard power-of-2 sizes this is the same as the least
common multiple. We still require proper alignment of the targets in
the mapping table.

ok by haad@

branches: 1.5.4;
Indent files remove unnecessary blank lines, white spaces and KNFize code.

Reduce a noise from a debug printf to debug level

Some fixes for dm(4).

- Limit accesses to the size of the volume.
This fixes a KASSERT in physio_biodone().
- Change dm_table_size() to return the size of the volume in
DEV_BSIZE units. It was reporting it in DEV_BSIZE^2 units.
- Remove a bit of trailing whitespace.

ok haad

branches: 1.2.2; 1.2.4; 1.2.6; 1.2.8;
Merge the haad-dm branch to -current. This branch adds LVM functionality to
the base NetBSD system. It uses Linux LVM2 tools and our BSD licensed
device-mapper driver.

The device-mapper driver can be used to create virtual block devices which
maps virtual blocks to real with target mapping called target. Currently
these targets are available a linear, zero, error and a snapshot (this is
work in progress and doesn't work yet).

The lvm2tools adds lvm and dmsetup binary to based system, where the lvm
tool is used to manage and administer whole LVM and the dmestup is used to
communicate iwith device-mapper kernel driver. With these tools also
a libdevmapper library is instaled to the base system.

Building of tools and driver is currently disable and can be enabled with
MKLVM=yes in mk.conf. I will add sets lists and rc.d script soon.

Oked by agc@ and cube@.

branches: 1.1.2;
Add dmgetdefaultdisklabel to get virtual disklabel for Logical Volume device. Add snapshot targets (snapshot, snapshot-origin), add dm_type to dm_dev structure to identify type of device. e.g. we can have mirrored disk device with snapshot on them and spare disk on them. When driver want to work with snapshot devices, it looks to upcalls list and finds all DM_SNAPSHOT devices.

fix typos in sys/dev/dm code comments and documentation.
also remove some trailing space in documentation.

Adapt to proplib API chanages.

dm: #if0 target's ->upcall() handler

This is part of NetBSD's dm design, but unimplemented
(all handlers return 0) and also unused.

branches: 1.39.2;
dm: Add dm-delay target

Ported from DragonFlyBSD, but this target had originally existed in
Linux kernel. See below for details.
https://www.kernel.org/doc/Documentation/device-mapper/delay.txt

Due to "tick" in hz(9) not working (which results in dmdlthread spinning
forever in _submit_queue() without dp extracted from delayed list
and queued into submit list), this hasn't been hooked to dm.kmod yet.

taken-from: DragonFlyBSD

dm: Add dm-flakey target

Ported from DragonFlyBSD, but this target had originally existed in
Linux kernel. See below for details.
https://www.kernel.org/doc/Documentation/device-mapper/dm-flakey.txt

Due to two technical issues, this hasn't been hooked to dm.kmod yet.
1) "tick" in hz(9) not working.
2) Unable to use ->b_private in nestiobuf callback when it's already
used for mbp (see HAS_BUF_PRIV2).

taken-from: DragonFlyBSD

dm: Make target's ->table() optional

Since ->info() (counter part of ->table() in the original dm design
in Linux kernel in .status where both INFO and TABLE are optional)
is an optional handler, make ->table() optional as well. Some
targets don't have anything to do in ->table() just as in ->info().

taken-from: DragonFlyBSD

dm: Fix dm-stripe's "status" output format

As mentioned in "dm: Don't try to implement "status" as subset of "table"",
dm-stripe in NetBSD doesn't have correct "status" output format.
Implement ->info() to sync with Linux kernel.

Note that num_error for stripe device isn't implemented yet.

taken-from: DragonFlyBSD

dm: Remove target's ->deps() by implementing deps in dm core

Retrieving device dependencies doesn't need to be target specific.
The reason it currently needs ->deps() is because dm core doesn't
have data structure that allows table to walk through target's
underlying devices. Add struct dm_mapping to be able to do this,
and remove ->deps()'s from targets which basically do the same thing.

=====(A) before this commit
table
| [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

=====(B) this commit
table---->mapping-->mapping-->mapping-->...
| | | |
| v v v [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

taken-from: DragonFlyBSD

dm: u_{int,long} -> unsigned {int,long}

dm: Enable dm-error and dm-zero target

Add these two targets to dm.kmod.
These are generally available in Linux and DragonFlyBSD,
so enable them in NetBSD as well.

dm: Style cleanups (no functional changes)

dm: Make targets' ->sync() optional

Apparently some targets have nothing to sync, so make it optional.

dm: Make targets' ->secsize() optional

and make a caller assume secsize 0 if ->secsize not present.
This allows a dummy function to be removed which was added in
"dm: Add dummy target ->sync()/->secsize() to prevent panic on modload(8)".

dm: Rename targets' ->status() to ->table() given ->info() exists

Since now that dm targets in NetBSD have ->info() for "status",
->status() should be renamed to ->table() for "table",
given how dm target status was originally designed in Linux kernel.

taken-from: DragonFlyBSD

dm: Remove void casts of function calls

Use of void casts in dm is not consistent, just get rid of them.

dm: Don't KASSERT() target handlers

Having assertions here causes panic if target is missing anything
as shown in "dm: Add dummy target ->sync()/->secsize() to prevent panic on modload(8)".
Instead just return EINVAL if a handler(s) isn't implemented.

taken-from: DragonFlyBSD

dm: Unbreak compilation of kernel modules

The dm kernel modules (MK_DM_TARGETS enabled in sys/modules/dm/Makefile)
have been broken. Unbreak the build.

dm: Fix strange pointer declarations

Should be "type *name" or "type* name", but not "type * name".
taken-from: DragonFlyBSD

dm: Simplify list eviction code

taken-from: DragonFlyBSD

dm: Don't ignore dm_target_alloc() argument

dm_target_alloc() is supposed to be copying the name argument to its ->name.
taken-from: DragonFlyBSD

dm: style + whitespace + indentation fixes

No functional changes.

dm: Make global dm locks static

These are all per-file locks which protect dev/pdev/target structure.
taken-from: DragonFlyBSD

branches: 1.20.4;
use some size_t, add whitespace between functions, rcsids.

src is too big these days to tolerate superfluous apostrophes. It's
"its", people!

branches: 1.18.12;
readd assertions noew with the correct struct
im wondering why the module cflags didnt warn about this obvious typo

Revert last, assertions on undefined variables don't make sense.

Add a sanity check for missing functions in dmt

make dm aware of physical sector sizes.

For aggregates of multiple disks we use the largest sector size from
all disks. For standard power-of-2 sizes this is the same as the least
common multiple. We still require proper alignment of the targets in
the mapping table.

ok by haad@

Update the rest of the kernel to conform to the module subsystem's new
locking protocol.

Add support for DIOCCACHESYNC ioctl for dm devices. Add new sync function
pointer to dm_target_t because that is the only part of dm which know real
block device. disk_ioctl_switch parses whole device table and for every
entry it calls particular sync routine which propagates DIOCCACHESYNC
to real disk.

While I was here implement some KNF fixes and remove unneeded symbols from
dm.h.

Problem reported on port-xen@ by Hugo Silva.

branches: 1.12.2; 1.12.4;
Indent files remove unnecessary blank lines, white spaces and KNFize code.

Fix bug in kmem_alloc/kmem_free of params string. Params string was
allocated with length DM_MAX_PARAMS_SIZE and released with strlen + 1 size.

Disable KM_NOSLEEP allocation because we do not need them here there is
nothing critical in ioctl part of dm driver.

Bug reported by jak@.

fix a comment typo.

branches: 1.9.2;
Call dm_target_busy only on properly autoloaded module.

In NetBSD our target modules are called with prefix dm_target_. Add this
prefix to target module name when loading module, lvm2tools uses linux
target names e.g. zero not dm_target_zero.

Add support for autoloading of device-mapper targets modules. Add
dm_target_autoload function which tries to load target module. Fix two
deadlocks in dm_table_load_ioctl error path(I forgot to call dm_dev_unbusy).

branches: 1.6.2; 1.6.4; 1.6.6;
Add stubs for mirror target, too. This target does nothing for now, but lvm
tools support mirror target and trying to create LV with mirrorred backend
caused panic in dm_table_load_ioctl.

Add stripe target functions stubs. Stripe target must be present in dm driver
because without it lvm2tools will not create LVsi and eventualy panic system.

Problem reported by agc@.

Add support for loading dm targets as separate modules. All targets except
linear can be loaded as module. Module is not loaded when there is target
with similar name already. Zero and error targets aresimple examples how
can be all future targets written to support dynamic loading. Target can't
be unloaded until there is at least one user.

Add infrastructure needed to load device-mapper targets as modules.
Targets wasn't converted yet and at least snapshot target will be converted
in a near future.

Merge the haad-dm branch to -current. This branch adds LVM functionality to
the base NetBSD system. It uses Linux LVM2 tools and our BSD licensed
device-mapper driver.

The device-mapper driver can be used to create virtual block devices which
maps virtual blocks to real with target mapping called target. Currently
these targets are available a linear, zero, error and a snapshot (this is
work in progress and doesn't work yet).

The lvm2tools adds lvm and dmsetup binary to based system, where the lvm
tool is used to manage and administer whole LVM and the dmestup is used to
communicate iwith device-mapper kernel driver. With these tools also
a libdevmapper library is instaled to the base system.

Building of tools and driver is currently disable and can be enabled with
MKLVM=yes in mk.conf. I will add sets lists and rc.d script soon.

Oked by agc@ and cube@.

branches: 1.1.2;
Add dmgetdefaultdisklabel to get virtual disklabel for Logical Volume device. Add snapshot targets (snapshot, snapshot-origin), add dm_type to dm_dev structure to identify type of device. e.g. we can have mirrored disk device with snapshot on them and spare disk on them. When driver want to work with snapshot devices, it looks to upcalls list and finds all DM_SNAPSHOT devices.

branches: 1.2.4;
dm: #if0 target's ->upcall() handler

This is part of NetBSD's dm design, but unimplemented
(all handlers return 0) and also unused.

branches: 1.1.2;
dm: Add dm-delay target

Ported from DragonFlyBSD, but this target had originally existed in
Linux kernel. See below for details.
https://www.kernel.org/doc/Documentation/device-mapper/delay.txt

Due to "tick" in hz(9) not working (which results in dmdlthread spinning
forever in _submit_queue() without dp extracted from delayed list
and queued into submit list), this hasn't been hooked to dm.kmod yet.

taken-from: DragonFlyBSD

dm: #if0 target's ->upcall() handler

This is part of NetBSD's dm design, but unimplemented
(all handlers return 0) and also unused.

branches: 1.28.2;
dm: Make target's ->table() optional

Since ->info() (counter part of ->table() in the original dm design
in Linux kernel in .status where both INFO and TABLE are optional)
is an optional handler, make ->table() optional as well. Some
targets don't have anything to do in ->table() just as in ->info().

taken-from: DragonFlyBSD

dm: Remove target's ->deps() by implementing deps in dm core

Retrieving device dependencies doesn't need to be target specific.
The reason it currently needs ->deps() is because dm core doesn't
have data structure that allows table to walk through target's
underlying devices. Add struct dm_mapping to be able to do this,
and remove ->deps()'s from targets which basically do the same thing.

=====(A) before this commit
table
| [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

=====(B) this commit
table---->mapping-->mapping-->mapping-->...
| | | |
| v v v [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

taken-from: DragonFlyBSD

dm: Test # of args in target's ->init()

The # of args is part of target's spec.
Both Linux kernel and DragonFlyBSD test argc on ctr/init.

dm: Enable dm-error and dm-zero target

Add these two targets to dm.kmod.
These are generally available in Linux and DragonFlyBSD,
so enable them in NetBSD as well.

dm: Style cleanups (no functional changes)

dm: Make targets' ->sync() optional

Apparently some targets have nothing to sync, so make it optional.

dm: Make targets' ->secsize() optional

and make a caller assume secsize 0 if ->secsize not present.
This allows a dummy function to be removed which was added in
"dm: Add dummy target ->sync()/->secsize() to prevent panic on modload(8)".

dm: Rename targets' ->status() to ->table() given ->info() exists

Since now that dm targets in NetBSD have ->info() for "status",
->status() should be renamed to ->table() for "table",
given how dm target status was originally designed in Linux kernel.

taken-from: DragonFlyBSD

dm: Remove unconditional debug prints in targets' ->strategy()

Having debug prints in ->strategy() by default just to tell ->strategy()
is called is overkill.
taken-from: DragonFlyBSD

dm: Make target's ->init() take parsed argc and argv

This gets rid of the same parser code in each target using strsep(3).
taken-from: DragonFlyBSD

dm: Add dummy target ->sync()/->secsize() to prevent panic on modload(8)

dm_target_insert() has assertions to ensure targets implement all handlers.
Adding dummy ones at least prevents panic on modload(8).

dm: Refactor target's ->init() i/f

Take dm_table_entry_t* instead of void**.
Remove dm_dev_t* unneeded by target code.
No functional change, but for future changes.

taken-from: DragonFlyBSD

dm: Fix strange pointer declarations

Should be "type *name" or "type* name", but not "type * name".
taken-from: DragonFlyBSD

dm: Don't ignore dm_target_alloc() argument

dm_target_alloc() is supposed to be copying the name argument to its ->name.
taken-from: DragonFlyBSD

dm: Make dm core set config to NULL after destroy

Just let dm core do this instead of having a comment expecting each
target to do the right thing.

taken-from: DragonFlyBSD

dm: Remove unused dm_dev::dev_type

Given OOP-like architecture of dm target device structure,
dm_dev doesn't need to have self contained target type field,
and in fact this is unused.

branches: 1.12.4;
use some size_t, add whitespace between functions, rcsids.

Add support for DIOCCACHESYNC ioctl for dm devices. Add new sync function
pointer to dm_target_t because that is the only part of dm which know real
block device. disk_ioctl_switch parses whole device table and for every
entry it calls particular sync routine which propagates DIOCCACHESYNC
to real disk.

While I was here implement some KNF fixes and remove unneeded symbols from
dm.h.

Problem reported on port-xen@ by Hugo Silva.

branches: 1.10.2; 1.10.4;
Indent files remove unnecessary blank lines, white spaces and KNFize code.

Revert my commit which have added knowledge about dm targets to libdevmapper,
this breaks abstraction. Because only lvmtools/lvmlib and device-mapper can
have knowledge about target mapping and libdevmapper only passes requests
from lvmtools to kernel and back. Bump major library and driver version.

Requested by: yamt@

Parse dm param string in libdevmapper and not in a dm target init function.
Create proplib param dictionary entry in libdevmapper and pass it to dm in
dm_ioctl dict.
Param target is then passed to target init function, where is parse. I like
this aproach much better than passing char **argv and trusting to user input.

I have bumped minor lib/driver version.

XXX. Add more sanity checks in kernel.

branches: 1.7.2;
Add support for autoloading of device-mapper targets modules. Add
dm_target_autoload function which tries to load target module. Fix two
deadlocks in dm_table_load_ioctl error path(I forgot to call dm_dev_unbusy).

branches: 1.6.2; 1.6.4; 1.6.6;
Add dm module as dependency and remove module_hold/module_rele calls.

Properly register dm_target_* modules with module framework.

Add support for loading dm targets as separate modules. All targets except
linear can be loaded as module. Module is not loaded when there is target
with similar name already. Zero and error targets aresimple examples how
can be all future targets written to support dynamic loading. Target can't
be unloaded until there is at least one user.

Add infrastructure needed to load device-mapper targets as modules.
Targets wasn't converted yet and at least snapshot target will be converted
in a near future.

Merge the haad-dm branch to -current. This branch adds LVM functionality to
the base NetBSD system. It uses Linux LVM2 tools and our BSD licensed
device-mapper driver.

The device-mapper driver can be used to create virtual block devices which
maps virtual blocks to real with target mapping called target. Currently
these targets are available a linear, zero, error and a snapshot (this is
work in progress and doesn't work yet).

The lvm2tools adds lvm and dmsetup binary to based system, where the lvm
tool is used to manage and administer whole LVM and the dmestup is used to
communicate iwith device-mapper kernel driver. With these tools also
a libdevmapper library is instaled to the base system.

Building of tools and driver is currently disable and can be enabled with
MKLVM=yes in mk.conf. I will add sets lists and rc.d script soon.

Oked by agc@ and cube@.

branches: 1.1.2;
Add dmgetdefaultdisklabel to get virtual disklabel for Logical Volume device. Add snapshot targets (snapshot, snapshot-origin), add dm_type to dm_dev structure to identify type of device. e.g. we can have mirrored disk device with snapshot on them and spare disk on them. When driver want to work with snapshot devices, it looks to upcalls list and finds all DM_SNAPSHOT devices.

Fix all remaining typos, mainly in comments but also in few definitions and log messages, reported by me in PR kern/54889.
Also fixed some additional typos in comments, found on review of same files or typos.

branches: 1.3.4; 1.3.12;
dm: #if0 target's ->upcall() handler

This is part of NetBSD's dm design, but unimplemented
(all handlers return 0) and also unused.

branches: 1.2.2;
dm: Add "Copyright (c) 2015 The DragonFly Project." to dm-flakey

which I missed in my initial dm-flakey commit.

dm: Add dm-flakey target

Ported from DragonFlyBSD, but this target had originally existed in
Linux kernel. See below for details.
https://www.kernel.org/doc/Documentation/device-mapper/dm-flakey.txt

Due to two technical issues, this hasn't been hooked to dm.kmod yet.
1) "tick" in hz(9) not working.
2) Unable to use ->b_private in nestiobuf callback when it's already
used for mbp (see HAS_BUF_PRIV2).

taken-from: DragonFlyBSD

dm: #if0 target's ->upcall() handler

This is part of NetBSD's dm design, but unimplemented
(all handlers return 0) and also unused.

branches: 1.36.2;
dm: Fix typo in comment dklinear -> linear

dm: Remove target's ->deps() by implementing deps in dm core

Retrieving device dependencies doesn't need to be target specific.
The reason it currently needs ->deps() is because dm core doesn't
have data structure that allows table to walk through target's
underlying devices. Add struct dm_mapping to be able to do this,
and remove ->deps()'s from targets which basically do the same thing.

=====(A) before this commit
table
| [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

=====(B) this commit
table---->mapping-->mapping-->mapping-->...
| | | |
| v v v [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

taken-from: DragonFlyBSD

dm: Fix "table" output format of dm-linear and dm-stripe

The existing "table" output showing device file path of pdev is
not compatible with dm in Linux kernel (and also DragonFlyBSD).
It should be showing "major:minor" instead.

taken-from: DragonFlyBSD

dm: Test # of args in target's ->init()

The # of args is part of target's spec.
Both Linux kernel and DragonFlyBSD test argc on ctr/init.

dm: Rename dm specific atoi() to atoi64()

This is uint64_t version, not sys/lib/libsa/atoi.c.

dm: Style cleanups (no functional changes)

dm: "unsigned" -> "unsigned int" for consistency

Use either one, but not both.

dm: Rename targets' ->status() to ->table() given ->info() exists

Since now that dm targets in NetBSD have ->info() for "status",
->status() should be renamed to ->table() for "table",
given how dm target status was originally designed in Linux kernel.

taken-from: DragonFlyBSD

dm: Remove unconditional debug prints in targets' ->strategy()

Having debug prints in ->strategy() by default just to tell ->strategy()
is called is overkill.
taken-from: DragonFlyBSD

dm: Make target's ->init() take parsed argc and argv

This gets rid of the same parser code in each target using strsep(3).
taken-from: DragonFlyBSD

dm: Silence aprint_normal() in dm_target_linear_status()

dm_target_linear_status() shouldn't print this debug message by default
on dmsetup "table" ioctl/command.

dm: Refactor target's ->init() i/f

Take dm_table_entry_t* instead of void**.
Remove dm_dev_t* unneeded by target code.
No functional change, but for future changes.

taken-from: DragonFlyBSD

dm: Fix strange pointer declarations

Should be "type *name" or "type* name", but not "type * name".
taken-from: DragonFlyBSD

dm: Remove trailing whitespace

dm: Make dm core set config to NULL after destroy

Just let dm core do this instead of having a comment expecting each
target to do the right thing.

taken-from: DragonFlyBSD

dm: Include <sys/vnode.h> in dm.h

dm.h already depends on vnode, so have dm.h include <sys/vnode.h>
instead of other .c files.

taken-from: DragonFlyBSD

dm: Remove misleading comment on linear target arg

The offset arg is mandatory.
Remove code and comment that makes it sounds like it's optional.

taken-from: DragonFlyBSD

dm: Remove unused dm_dev::dev_type

Given OOP-like architecture of dm target device structure,
dm_dev doesn't need to have self contained target type field,
and in fact this is unused.

convert more KM_NOSLEEP to KM_SLEEP and remove code to handle failures.

branches: 1.17.4;
PR/52900: Tomohiro Kusumi: Fix kernel panic when target's ->init() failed
Reliably unbusy the target in the *all* the destroy routines, so ioctl
does not need to do it.

more dependiences -> dependencies

remove checks for failure after memory allocation calls that cannot fail:

kmem_alloc() with KM_SLEEP
kmem_zalloc() with KM_SLEEP
percpu_alloc()
pserialize_create()
psref_class_create()

all of these paths include an assertion that the allocation has not failed,
so callers should not assert that again.

branches: 1.14.4;
Change dk_lookup() to return an anonymous vnode not associated with
any file system. Change all consumers of dk_lookup() to get the
device from "v_rdev" instead of VOP_GETATTR() as specfs does not
support VOP_GETATTR(). Devices obtained with dk_lookup() will no
longer disappear on forced unmounts.

Fix for PR kern/48849 (root mirror raid fails on shutdown)

Welcome to 6.99.44

branches: 1.13.12; 1.13.26;
Change the vnode locking protocol of VOP_GETATTR() to request at least
a shared lock. Make all calls outside of file systems respect it.

The calls from file systems need review.

No objections from tech-kern.

make dm aware of physical sector sizes.

For aggregates of multiple disks we use the largest sector size from
all disks. For standard power-of-2 sizes this is the same as the least
common multiple. We still require proper alignment of the targets in
the mapping table.

ok by haad@

curlwp needs sys/lwp.h.

Add support for DIOCCACHESYNC ioctl for dm devices. Add new sync function
pointer to dm_target_t because that is the only part of dm which know real
block device. disk_ioctl_switch parses whole device table and for every
entry it calls particular sync routine which propagates DIOCCACHESYNC
to real disk.

While I was here implement some KNF fixes and remove unneeded symbols from
dm.h.

Problem reported on port-xen@ by Hugo Silva.

branches: 1.9.2; 1.9.4;
Indent files remove unnecessary blank lines, white spaces and KNFize code.

Revert my commit which have added knowledge about dm targets to libdevmapper,
this breaks abstraction. Because only lvmtools/lvmlib and device-mapper can
have knowledge about target mapping and libdevmapper only passes requests
from lvmtools to kernel and back. Bump major library and driver version.

Requested by: yamt@

Fix bug in kmem_alloc/kmem_free of params string. Params string was
allocated with length DM_MAX_PARAMS_SIZE and released with strlen + 1 size.

Disable KM_NOSLEEP allocation because we do not need them here there is
nothing critical in ioctl part of dm driver.

Bug reported by jak@.

fix a comment typo.

Parse dm param string in libdevmapper and not in a dm target init function.
Create proplib param dictionary entry in libdevmapper and pass it to dm in
dm_ioctl dict.
Param target is then passed to target init function, where is parse. I like
this aproach much better than passing char **argv and trusting to user input.

I have bumped minor lib/driver version.

XXX. Add more sanity checks in kernel.

branches: 1.4.2;
Linear target has only 2 arguments in a params string not 9.

branches: 1.3.2; 1.3.4; 1.3.6;
Add infrastructure needed to load device-mapper targets as modules.
Targets wasn't converted yet and at least snapshot target will be converted
in a near future.

Merge the haad-dm branch to -current. This branch adds LVM functionality to
the base NetBSD system. It uses Linux LVM2 tools and our BSD licensed
device-mapper driver.

The device-mapper driver can be used to create virtual block devices which
maps virtual blocks to real with target mapping called target. Currently
these targets are available a linear, zero, error and a snapshot (this is
work in progress and doesn't work yet).

The lvm2tools adds lvm and dmsetup binary to based system, where the lvm
tool is used to manage and administer whole LVM and the dmestup is used to
communicate iwith device-mapper kernel driver. With these tools also
a libdevmapper library is instaled to the base system.

Building of tools and driver is currently disable and can be enabled with
MKLVM=yes in mk.conf. I will add sets lists and rc.d script soon.

Oked by agc@ and cube@.

branches: 1.1.2;
Add dmgetdefaultdisklabel to get virtual disklabel for Logical Volume device. Add snapshot targets (snapshot, snapshot-origin), add dm_type to dm_dev structure to identify type of device. e.g. we can have mirrored disk device with snapshot on them and spare disk on them. When driver want to work with snapshot devices, it looks to upcalls list and finds all DM_SNAPSHOT devices.

dm: #if0 target's ->upcall() handler

This is part of NetBSD's dm design, but unimplemented
(all handlers return 0) and also unused.

branches: 1.23.2;
dm: Remove target's ->deps() by implementing deps in dm core

Retrieving device dependencies doesn't need to be target specific.
The reason it currently needs ->deps() is because dm core doesn't
have data structure that allows table to walk through target's
underlying devices. Add struct dm_mapping to be able to do this,
and remove ->deps()'s from targets which basically do the same thing.

=====(A) before this commit
table
| [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

=====(B) this commit
table---->mapping-->mapping-->mapping-->...
| | | |
| v v v [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

taken-from: DragonFlyBSD

dm: Style cleanups (no functional changes)

dm: Make targets' ->secsize() optional

and make a caller assume secsize 0 if ->secsize not present.
This allows a dummy function to be removed which was added in
"dm: Add dummy target ->sync()/->secsize() to prevent panic on modload(8)".

dm: Rename targets' ->status() to ->table() given ->info() exists

Since now that dm targets in NetBSD have ->info() for "status",
->status() should be renamed to ->table() for "table",
given how dm target status was originally designed in Linux kernel.

taken-from: DragonFlyBSD

dm: Remove unconditional debug prints in targets' ->strategy()

Having debug prints in ->strategy() by default just to tell ->strategy()
is called is overkill.
taken-from: DragonFlyBSD

dm: Make target's ->init() take parsed argc and argv

This gets rid of the same parser code in each target using strsep(3).
taken-from: DragonFlyBSD

dm: Add dummy target ->sync()/->secsize() to prevent panic on modload(8)

dm_target_insert() has assertions to ensure targets implement all handlers.
Adding dummy ones at least prevents panic on modload(8).

dm: Move targets specific structs to .c files

These don't need to be defined and exposed in dm.h.

dm: Refactor target's ->init() i/f

Take dm_table_entry_t* instead of void**.
Remove dm_dev_t* unneeded by target code.
No functional change, but for future changes.

taken-from: DragonFlyBSD

dm: Fix strange pointer declarations

Should be "type *name" or "type* name", but not "type * name".
taken-from: DragonFlyBSD

dm: Don't ignore dm_target_alloc() argument

dm_target_alloc() is supposed to be copying the name argument to its ->name.
taken-from: DragonFlyBSD

dm: Make dm core set config to NULL after destroy

Just let dm core do this instead of having a comment expecting each
target to do the right thing.

taken-from: DragonFlyBSD

dm: Remove unused dm_dev::dev_type

Given OOP-like architecture of dm target device structure,
dm_dev doesn't need to have self contained target type field,
and in fact this is unused.

branches: 1.10.4;
use some size_t, add whitespace between functions, rcsids.

Add support for DIOCCACHESYNC ioctl for dm devices. Add new sync function
pointer to dm_target_t because that is the only part of dm which know real
block device. disk_ioctl_switch parses whole device table and for every
entry it calls particular sync routine which propagates DIOCCACHESYNC
to real disk.

While I was here implement some KNF fixes and remove unneeded symbols from
dm.h.

Problem reported on port-xen@ by Hugo Silva.

branches: 1.8.2; 1.8.4;
Indent files remove unnecessary blank lines, white spaces and KNFize code.

Revert my commit which have added knowledge about dm targets to libdevmapper,
this breaks abstraction. Because only lvmtools/lvmlib and device-mapper can
have knowledge about target mapping and libdevmapper only passes requests
from lvmtools to kernel and back. Bump major library and driver version.

Requested by: yamt@

Parse dm param string in libdevmapper and not in a dm target init function.
Create proplib param dictionary entry in libdevmapper and pass it to dm in
dm_ioctl dict.
Param target is then passed to target init function, where is parse. I like
this aproach much better than passing char **argv and trusting to user input.

I have bumped minor lib/driver version.

XXX. Add more sanity checks in kernel.

branches: 1.5.2;
Change comment and add description of line used to configure mirror device.

Add support for autoloading of device-mapper targets modules. Add
dm_target_autoload function which tries to load target module. Fix two
deadlocks in dm_table_load_ioctl error path(I forgot to call dm_dev_unbusy).

branches: 1.3.2; 1.3.4; 1.3.6;
Add dm module as dependency and remove module_hold/module_rele calls.

Return ENOSYS during from dm_target_init routine these tow targets aren't
implemented yet.

Add stubs for mirror target, too. This target does nothing for now, but lvm
tools support mirror target and trying to create LV with mirrorred backend
caused panic in dm_table_load_ioctl.

dm: #if0 target's ->upcall() handler

This is part of NetBSD's dm design, but unimplemented
(all handlers return 0) and also unused.

branches: 1.37.2;
dm: Remove target's ->deps() by implementing deps in dm core

Retrieving device dependencies doesn't need to be target specific.
The reason it currently needs ->deps() is because dm core doesn't
have data structure that allows table to walk through target's
underlying devices. Add struct dm_mapping to be able to do this,
and remove ->deps()'s from targets which basically do the same thing.

=====(A) before this commit
table
| [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

=====(B) this commit
table---->mapping-->mapping-->mapping-->...
| | | |
| v v v [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

taken-from: DragonFlyBSD

dm: Rename dm specific atoi() to atoi64()

This is uint64_t version, not sys/lib/libsa/atoi.c.

dm: Style cleanups (no functional changes)

dm: Make targets' ->sync() optional

Apparently some targets have nothing to sync, so make it optional.

dm: Make targets' ->secsize() optional

and make a caller assume secsize 0 if ->secsize not present.
This allows a dummy function to be removed which was added in
"dm: Add dummy target ->sync()/->secsize() to prevent panic on modload(8)".

dm: Rename targets' ->status() to ->table() given ->info() exists

Since now that dm targets in NetBSD have ->info() for "status",
->status() should be renamed to ->table() for "table",
given how dm target status was originally designed in Linux kernel.

taken-from: DragonFlyBSD

dm: Remove unconditional debug prints in targets' ->strategy()

Having debug prints in ->strategy() by default just to tell ->strategy()
is called is overkill.
taken-from: DragonFlyBSD

dm: Make target's ->init() take parsed argc and argv

This gets rid of the same parser code in each target using strsep(3).
taken-from: DragonFlyBSD

dm: Add dummy target ->sync()/->secsize() to prevent panic on modload(8)

dm_target_insert() has assertions to ensure targets implement all handlers.
Adding dummy ones at least prevents panic on modload(8).

dm: Move targets specific structs to .c files

These don't need to be defined and exposed in dm.h.

dm: Unbreak compilation of kernel modules

The dm kernel modules (MK_DM_TARGETS enabled in sys/modules/dm/Makefile)
have been broken. Unbreak the build.

dm: Refactor target's ->init() i/f

Take dm_table_entry_t* instead of void**.
Remove dm_dev_t* unneeded by target code.
No functional change, but for future changes.

taken-from: DragonFlyBSD

dm: Fix strange pointer declarations

Should be "type *name" or "type* name", but not "type * name".
taken-from: DragonFlyBSD

dm: Don't ignore dm_target_alloc() argument

dm_target_alloc() is supposed to be copying the name argument to its ->name.
taken-from: DragonFlyBSD

dm: Make dm core set config to NULL after destroy

Just let dm core do this instead of having a comment expecting each
target to do the right thing.

taken-from: DragonFlyBSD

dm: Include <sys/vnode.h> in dm.h

dm.h already depends on vnode, so have dm.h include <sys/vnode.h>
instead of other .c files.

taken-from: DragonFlyBSD

dm: Remove unused dm_dev::dev_type

Given OOP-like architecture of dm target device structure,
dm_dev doesn't need to have self contained target type field,
and in fact this is unused.

convert more KM_NOSLEEP to KM_SLEEP and remove code to handle failures.

branches: 1.19.4;
PR/52900: Tomohiro Kusumi: Fix kernel panic when target's ->init() failed
Reliably unbusy the target in the *all* the destroy routines, so ioctl
does not need to do it.

more dependiences -> dependencies

branches: 1.17.2;
Avoid a memory leak - from maxv

branches: 1.16.2;
Change dk_lookup() to return an anonymous vnode not associated with
any file system. Change all consumers of dk_lookup() to get the
device from "v_rdev" instead of VOP_GETATTR() as specfs does not
support VOP_GETATTR(). Devices obtained with dk_lookup() will no
longer disappear on forced unmounts.

Fix for PR kern/48849 (root mirror raid fails on shutdown)

Welcome to 6.99.44

branches: 1.15.12; 1.15.26;
Change the vnode locking protocol of VOP_GETATTR() to request at least
a shared lock. Make all calls outside of file systems respect it.

The calls from file systems need review.

No objections from tech-kern.

make dm aware of physical sector sizes.

For aggregates of multiple disks we use the largest sector size from
all disks. For standard power-of-2 sizes this is the same as the least
common multiple. We still require proper alignment of the targets in
the mapping table.

ok by haad@

Add support for DIOCCACHESYNC ioctl for dm devices. Add new sync function
pointer to dm_target_t because that is the only part of dm which know real
block device. disk_ioctl_switch parses whole device table and for every
entry it calls particular sync routine which propagates DIOCCACHESYNC
to real disk.

While I was here implement some KNF fixes and remove unneeded symbols from
dm.h.

Problem reported on port-xen@ by Hugo Silva.

branches: 1.12.2; 1.12.4;
Indent files remove unnecessary blank lines, white spaces and KNFize code.

Revert my commit which have added knowledge about dm targets to libdevmapper,
this breaks abstraction. Because only lvmtools/lvmlib and device-mapper can
have knowledge about target mapping and libdevmapper only passes requests
from lvmtools to kernel and back. Bump major library and driver version.

Requested by: yamt@

buidfix: change dm_target_snapshot_orig_init() to match prototype.

Parse dm param string in libdevmapper and not in a dm target init function.
Create proplib param dictionary entry in libdevmapper and pass it to dm in
dm_ioctl dict.
Param target is then passed to target init function, where is parse. I like
this aproach much better than passing char **argv and trusting to user input.

I have bumped minor lib/driver version.

XXX. Add more sanity checks in kernel.

branches: 1.8.2;
Add support for autoloading of device-mapper targets modules. Add
dm_target_autoload function which tries to load target module. Fix two
deadlocks in dm_table_load_ioctl error path(I forgot to call dm_dev_unbusy).

branches: 1.7.2; 1.7.4; 1.7.6;
Add dm module as dependency and remove module_hold/module_rele calls.

Properly register dm_target_* modules with module framework.

Add stripe target functions stubs. Stripe target must be present in dm driver
because without it lvm2tools will not create LVsi and eventualy panic system.

Problem reported by agc@.

Add support for loading dm targets as separate modules. All targets except
linear can be loaded as module. Module is not loaded when there is target
with similar name already. Zero and error targets aresimple examples how
can be all future targets written to support dynamic loading. Target can't
be unloaded until there is at least one user.

Add infrastructure needed to load device-mapper targets as modules.
Targets wasn't converted yet and at least snapshot target will be converted
in a near future.

Merge the haad-dm branch to -current. This branch adds LVM functionality to
the base NetBSD system. It uses Linux LVM2 tools and our BSD licensed
device-mapper driver.

The device-mapper driver can be used to create virtual block devices which
maps virtual blocks to real with target mapping called target. Currently
these targets are available a linear, zero, error and a snapshot (this is
work in progress and doesn't work yet).

The lvm2tools adds lvm and dmsetup binary to based system, where the lvm
tool is used to manage and administer whole LVM and the dmestup is used to
communicate iwith device-mapper kernel driver. With these tools also
a libdevmapper library is instaled to the base system.

Building of tools and driver is currently disable and can be enabled with
MKLVM=yes in mk.conf. I will add sets lists and rc.d script soon.

Oked by agc@ and cube@.

branches: 1.1.2;
Add dmgetdefaultdisklabel to get virtual disklabel for Logical Volume device. Add snapshot targets (snapshot, snapshot-origin), add dm_type to dm_dev structure to identify type of device. e.g. we can have mirrored disk device with snapshot on them and spare disk on them. When driver want to work with snapshot devices, it looks to upcalls list and finds all DM_SNAPSHOT devices.

fix a/an grammar in obvious cases

branches: 1.44.30;
dm: #if0 target's ->upcall() handler

This is part of NetBSD's dm design, but unimplemented
(all handlers return 0) and also unused.

branches: 1.43.2;
dm: Fix dm-stripe's "status" output format

As mentioned in "dm: Don't try to implement "status" as subset of "table"",
dm-stripe in NetBSD doesn't have correct "status" output format.
Implement ->info() to sync with Linux kernel.

Note that num_error for stripe device isn't implemented yet.

taken-from: DragonFlyBSD

dm: Remove target's ->deps() by implementing deps in dm core

Retrieving device dependencies doesn't need to be target specific.
The reason it currently needs ->deps() is because dm core doesn't
have data structure that allows table to walk through target's
underlying devices. Add struct dm_mapping to be able to do this,
and remove ->deps()'s from targets which basically do the same thing.

=====(A) before this commit
table
| [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

=====(B) this commit
table---->mapping-->mapping-->mapping-->...
| | | |
| v v v [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

taken-from: DragonFlyBSD

dm: Fix "table" output format of dm-linear and dm-stripe

The existing "table" output showing device file path of pdev is
not compatible with dm in Linux kernel (and also DragonFlyBSD).
It should be showing "major:minor" instead.

taken-from: DragonFlyBSD

dm: Test # of args in target's ->init()

The # of args is part of target's spec.
Both Linux kernel and DragonFlyBSD test argc on ctr/init.

dm: Rename dm specific atoi() to atoi64()

This is uint64_t version, not sys/lib/libsa/atoi.c.

dm: Style cleanups (no functional changes)

dm: "unsigned" -> "unsigned int" for consistency

Use either one, but not both.

dm: Rename targets' ->status() to ->table() given ->info() exists

Since now that dm targets in NetBSD have ->info() for "status",
->status() should be renamed to ->table() for "table",
given how dm target status was originally designed in Linux kernel.

taken-from: DragonFlyBSD

dm: Remove unconditional debug prints in targets' ->strategy()

Having debug prints in ->strategy() by default just to tell ->strategy()
is called is overkill.
taken-from: DragonFlyBSD

dm: Make target's ->init() take parsed argc and argv

This gets rid of the same parser code in each target using strsep(3).
taken-from: DragonFlyBSD

dm: Correct target name "stripe" -> "striped"

Sync with Linux kernel.
The file name is "stripe", but the target name is "striped".
DragonFlyBSD also follows this.

dm: Move targets specific structs to .c files

These don't need to be defined and exposed in dm.h.

dm: Refactor target's ->init() i/f

Take dm_table_entry_t* instead of void**.
Remove dm_dev_t* unneeded by target code.
No functional change, but for future changes.

taken-from: DragonFlyBSD

dm: Fix strange pointer declarations

Should be "type *name" or "type* name", but not "type * name".
taken-from: DragonFlyBSD

dm: Fix typos in comments/messages

taken-from: DragonFlyBSD

dm: Don't ignore dm_target_alloc() argument

dm_target_alloc() is supposed to be copying the name argument to its ->name.
taken-from: DragonFlyBSD

dm: Make dm core set config to NULL after destroy

Just let dm core do this instead of having a comment expecting each
target to do the right thing.

taken-from: DragonFlyBSD

dm: Include <sys/vnode.h> in dm.h

dm.h already depends on vnode, so have dm.h include <sys/vnode.h>
instead of other .c files.

taken-from: DragonFlyBSD

dm: Remove unused dm_dev::dev_type

Given OOP-like architecture of dm target device structure,
dm_dev doesn't need to have self contained target type field,
and in fact this is unused.

convert more KM_NOSLEEP to KM_SLEEP and remove code to handle failures.

branches: 1.23.4;
use some size_t, add whitespace between functions, rcsids.

remove checks for failure after memory allocation calls that cannot fail:

kmem_alloc() with KM_SLEEP
kmem_zalloc() with KM_SLEEP
percpu_alloc()
pserialize_create()
psref_class_create()

all of these paths include an assertion that the allocation has not failed,
so callers should not assert that again.

branches: 1.21.2;
cleanup properly on error.

Avoid a memory leak - from maxv

branches: 1.19.2;
Change dk_lookup() to return an anonymous vnode not associated with
any file system. Change all consumers of dk_lookup() to get the
device from "v_rdev" instead of VOP_GETATTR() as specfs does not
support VOP_GETATTR(). Devices obtained with dk_lookup() will no
longer disappear on forced unmounts.

Fix for PR kern/48849 (root mirror raid fails on shutdown)

Welcome to 6.99.44

branches: 1.18.2; 1.18.12;
Fix problem reported by jym on current-user@ patch was provided by mhitch@.

branches: 1.17.2;
Fix wrong KASSERT, reported by mhitch@.

branches: 1.16.2; 1.16.6;
Change the vnode locking protocol of VOP_GETATTR() to request at least
a shared lock. Make all calls outside of file systems respect it.

The calls from file systems need review.

No objections from tech-kern.

be consistent and define secsize here as well (though this seems unused?)
some white' from emacs

Fix simple typo which made stripe target not working very at all. Issue
found by mhitch@.

branches: 1.13.2; 1.13.6;
make dm aware of physical sector sizes.

For aggregates of multiple disks we use the largest sector size from
all disks. For standard power-of-2 sizes this is the same as the least
common multiple. We still require proper alignment of the targets in
the mapping table.

ok by haad@

curlwp needs sys/lwp.h.

Add old file describing locking schema used in dm driver.

Add multi device strip support written by Guillermo Amaral and reviewed by me.

branches: 1.9.2; 1.9.4;
Indent files remove unnecessary blank lines, white spaces and KNFize code.

Revert my commit which have added knowledge about dm targets to libdevmapper,
this breaks abstraction. Because only lvmtools/lvmlib and device-mapper can
have knowledge about target mapping and libdevmapper only passes requests
from lvmtools to kernel and back. Bump major library and driver version.

Requested by: yamt@

Fix bug in kmem_alloc/kmem_free of params string. Params string was
allocated with length DM_MAX_PARAMS_SIZE and released with strlen + 1 size.

Disable KM_NOSLEEP allocation because we do not need them here there is
nothing critical in ioctl part of dm driver.

Bug reported by jak@.

Parse dm param string in libdevmapper and not in a dm target init function.
Create proplib param dictionary entry in libdevmapper and pass it to dm in
dm_ioctl dict.
Param target is then passed to target init function, where is parse. I like
this aproach much better than passing char **argv and trusting to user input.

I have bumped minor lib/driver version.

XXX. Add more sanity checks in kernel.

branches: 1.5.2;
Fix build on amd64. Patch sent by dieter roelants.

Actually lmplement striping as LVM target. OK'd by Adam.

Add initial support for striping target, hardcode maximal number of stripes
to 2 for now. strategy routine must be written, for now only parsing and
support routines are available. This is work in progress code and should be
taken very carefully.

branches: 1.2.2; 1.2.4; 1.2.6;
Return ENOSYS during from dm_target_init routine these tow targets aren't
implemented yet.

Add stripe target functions stubs. Stripe target must be present in dm driver
because without it lvm2tools will not create LVsi and eventualy panic system.

Problem reported by agc@.

dm: #if0 target's ->upcall() handler

This is part of NetBSD's dm design, but unimplemented
(all handlers return 0) and also unused.

branches: 1.31.2;
s/suport/support/

dm: Make target's ->table() optional

Since ->info() (counter part of ->table() in the original dm design
in Linux kernel in .status where both INFO and TABLE are optional)
is an optional handler, make ->table() optional as well. Some
targets don't have anything to do in ->table() just as in ->info().

taken-from: DragonFlyBSD

dm: Remove target's ->deps() by implementing deps in dm core

Retrieving device dependencies doesn't need to be target specific.
The reason it currently needs ->deps() is because dm core doesn't
have data structure that allows table to walk through target's
underlying devices. Add struct dm_mapping to be able to do this,
and remove ->deps()'s from targets which basically do the same thing.

=====(A) before this commit
table
| [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

=====(B) this commit
table---->mapping-->mapping-->mapping-->...
| | | |
| v v v [dm core]
-------------------------------------------------------
| pdev pdev pdev [dm targets]
v ^ ^ ^
target----/---------/---------/
(void*)

taken-from: DragonFlyBSD

dm: Test # of args in target's ->init()

The # of args is part of target's spec.
Both Linux kernel and DragonFlyBSD test argc on ctr/init.

dm: Enable dm-error and dm-zero target

Add these two targets to dm.kmod.
These are generally available in Linux and DragonFlyBSD,
so enable them in NetBSD as well.

dm: Style cleanups (no functional changes)

dm: Make targets' ->sync() optional

Apparently some targets have nothing to sync, so make it optional.

dm: Make targets' ->secsize() optional

and make a caller assume secsize 0 if ->secsize not present.
This allows a dummy function to be removed which was added in
"dm: Add dummy target ->sync()/->secsize() to prevent panic on modload(8)".

dm: Rename targets' ->status() to ->table() given ->info() exists

Since now that dm targets in NetBSD have ->info() for "status",
->status() should be renamed to ->table() for "table",
given how dm target status was originally designed in Linux kernel.

taken-from: DragonFlyBSD

dm: Remove unconditional debug prints in targets' ->strategy()

Having debug prints in ->strategy() by default just to tell ->strategy()
is called is overkill.
taken-from: DragonFlyBSD

dm: Make target's ->init() take parsed argc and argv

This gets rid of the same parser code in each target using strsep(3).
taken-from: DragonFlyBSD

dm: Add dummy target ->sync()/->secsize() to prevent panic on modload(8)

dm_target_insert() has assertions to ensure targets implement all handlers.
Adding dummy ones at least prevents panic on modload(8).

dm: Refactor target's ->init() i/f

Take dm_table_entry_t* instead of void**.
Remove dm_dev_t* unneeded by target code.
No functional change, but for future changes.

taken-from: DragonFlyBSD

dm: Fix strange pointer declarations

Should be "type *name" or "type* name", but not "type * name".
taken-from: DragonFlyBSD

dm: Remove trailing whitespace

dm: Don't ignore dm_target_alloc() argument

dm_target_alloc() is supposed to be copying the name argument to its ->name.
taken-from: DragonFlyBSD

dm: Make dm core set config to NULL after destroy

Just let dm core do this instead of having a comment expecting each
target to do the right thing.

taken-from: DragonFlyBSD

dm: Remove unused dm_dev::dev_type

Given OOP-like architecture of dm target device structure,
dm_dev doesn't need to have self contained target type field,
and in fact this is unused.

branches: 1.13.4;
use some size_t, add whitespace between functions, rcsids.

fixup comments - no functional change

branches: 1.11.8; 1.11.12;
Add support for DIOCCACHESYNC ioctl for dm devices. Add new sync function
pointer to dm_target_t because that is the only part of dm which know real
block device. disk_ioctl_switch parses whole device table and for every
entry it calls particular sync routine which propagates DIOCCACHESYNC
to real disk.

While I was here implement some KNF fixes and remove unneeded symbols from
dm.h.

Problem reported on port-xen@ by Hugo Silva.

branches: 1.10.2; 1.10.4;
Indent files remove unnecessary blank lines, white spaces and KNFize code.

Revert my commit which have added knowledge about dm targets to libdevmapper,
this breaks abstraction. Because only lvmtools/lvmlib and device-mapper can
have knowledge about target mapping and libdevmapper only passes requests
from lvmtools to kernel and back. Bump major library and driver version.

Requested by: yamt@

Parse dm param string in libdevmapper and not in a dm target init function.
Create proplib param dictionary entry in libdevmapper and pass it to dm in
dm_ioctl dict.
Param target is then passed to target init function, where is parse. I like
this aproach much better than passing char **argv and trusting to user input.

I have bumped minor lib/driver version.

XXX. Add more sanity checks in kernel.

branches: 1.7.2;
Add support for autoloading of device-mapper targets modules. Add
dm_target_autoload function which tries to load target module. Fix two
deadlocks in dm_table_load_ioctl error path(I forgot to call dm_dev_unbusy).

branches: 1.6.2; 1.6.4; 1.6.6;
Add dm module as dependency and remove module_hold/module_rele calls.

Properly register dm_target_* modules with module framework.

Add support for loading dm targets as separate modules. All targets except
linear can be loaded as module. Module is not loaded when there is target
with similar name already. Zero and error targets aresimple examples how
can be all future targets written to support dynamic loading. Target can't
be unloaded until there is at least one user.

Add infrastructure needed to load device-mapper targets as modules.
Targets wasn't converted yet and at least snapshot target will be converted
in a near future.

Merge the haad-dm branch to -current. This branch adds LVM functionality to
the base NetBSD system. It uses Linux LVM2 tools and our BSD licensed
device-mapper driver.

The device-mapper driver can be used to create virtual block devices which
maps virtual blocks to real with target mapping called target. Currently
these targets are available a linear, zero, error and a snapshot (this is
work in progress and doesn't work yet).

The lvm2tools adds lvm and dmsetup binary to based system, where the lvm
tool is used to manage and administer whole LVM and the dmestup is used to
communicate iwith device-mapper kernel driver. With these tools also
a libdevmapper library is instaled to the base system.

Building of tools and driver is currently disable and can be enabled with
MKLVM=yes in mk.conf. I will add sets lists and rc.d script soon.

Oked by agc@ and cube@.

branches: 1.1.2;
Add dmgetdefaultdisklabel to get virtual disklabel for Logical Volume device. Add snapshot targets (snapshot, snapshot-origin), add dm_type to dm_dev structure to identify type of device. e.g. we can have mirrored disk device with snapshot on them and spare disk on them. When driver want to work with snapshot devices, it looks to upcalls list and finds all DM_SNAPSHOT devices.

Add error and zero targets to build.

branches: 1.6.64;
Hook device-mapper to autoconf framework. Add dm_attach, dm_match and dm_detach
routines used by autoconf users. Change dm_dev_remove_ioctl to call dm_detach.

This should be primary used by kernel to disable devices during shutdown of
system with nested disk devices.

Requested by dyoung@.

Remove unneeded targets from build. Should fix brekage of ALL kernel build
found by cegger@.

branches: 1.4.2; 1.4.4; 1.4.6; 1.4.8;
Add stubs for mirror target, too. This target does nothing for now, but lvm
tools support mirror target and trying to create LV with mirrorred backend
caused panic in dm_table_load_ioctl.

Add stripe target functions stubs. Stripe target must be present in dm driver
because without it lvm2tools will not create LVsi and eventualy panic system.

Problem reported by agc@.

Merge the haad-dm branch to -current. This branch adds LVM functionality to
the base NetBSD system. It uses Linux LVM2 tools and our BSD licensed
device-mapper driver.

The device-mapper driver can be used to create virtual block devices which
maps virtual blocks to real with target mapping called target. Currently
these targets are available a linear, zero, error and a snapshot (this is
work in progress and doesn't work yet).

The lvm2tools adds lvm and dmsetup binary to based system, where the lvm
tool is used to manage and administer whole LVM and the dmestup is used to
communicate iwith device-mapper kernel driver. With these tools also
a libdevmapper library is instaled to the base system.

Building of tools and driver is currently disable and can be enabled with
MKLVM=yes in mk.conf. I will add sets lists and rc.d script soon.

Oked by agc@ and cube@.