sfrench/cifs-2.6.git
11 years agodm snapshot: use merge origin if snapshot invalid
Mikulas Patocka [Thu, 10 Dec 2009 23:52:36 +0000 (23:52 +0000)]
dm snapshot: use merge origin if snapshot invalid

If the snapshot we are merging became invalid (e.g. it ran out of
space) redirect all I/O directly to the origin device.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: report merge failure in status
Mike Snitzer [Thu, 10 Dec 2009 23:52:35 +0000 (23:52 +0000)]
dm snapshot: report merge failure in status

Set 'merge_failed' flag if a snapshot fails to merge.  Update
snapshot_status() to report "Merge failed" if 'merge_failed' is set.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: merge consecutive chunks together
Mike Snitzer [Thu, 10 Dec 2009 23:52:34 +0000 (23:52 +0000)]
dm snapshot: merge consecutive chunks together

s->store->type->prepare_merge returns the number of chunks that can be
copied linearly working backwards from the returned chunk number.

For example, if it returns 3 chunks with old_chunk == 10 and new_chunk
== 20, then chunk 20 can be copied to 10, chunk 19 to 9 and 18 to 8.

Until now kcopyd only copied one chunk at a time.  This patch now copies
the full set at once.

Consequently, snapshot_merge_process() needs to delay the merging of all
chunks if any have writes in progress, not just the first chunk in the
region that is to be merged.

snapshot-merge's performance is now comparable to the original
snapshot-origin target.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: trigger exceptions in remaining snapshots during merge
Mikulas Patocka [Thu, 10 Dec 2009 23:52:34 +0000 (23:52 +0000)]
dm snapshot: trigger exceptions in remaining snapshots during merge

When there is one merging snapshot and other non-merging snapshots,
snapshot_merge_process() must make exceptions in the non-merging
snapshots.

Use a sequence count to resolve the race between I/O to chunks that are
about to be merged.  The count increases each time an exception
reallocation finishes.  Use wait_event() to wait until the count
changes.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: delay merging a chunk until writes to it complete
Mikulas Patocka [Thu, 10 Dec 2009 23:52:33 +0000 (23:52 +0000)]
dm snapshot: delay merging a chunk until writes to it complete

Track writes to chunks that are currently being merged and delay merging
a chunk until all writes to that chunk finish.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: queue writes to chunks being merged
Mikulas Patocka [Thu, 10 Dec 2009 23:52:33 +0000 (23:52 +0000)]
dm snapshot: queue writes to chunks being merged

While a set of chunks is being merged, any overlapping writes need to be
queued.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: add merging
Mikulas Patocka [Thu, 10 Dec 2009 23:52:32 +0000 (23:52 +0000)]
dm snapshot: add merging

Merging is started when origin is resumed and it is stopped when
origin is suspended or when the merging snapshot is destroyed or
errors are detected.

Merging is not yet interlocked with writes: this will be handled in
subsequent patches.

The code relies on callbacks from a private kcopyd thread.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: permit only one merge at once
Mikulas Patocka [Thu, 10 Dec 2009 23:52:32 +0000 (23:52 +0000)]
dm snapshot: permit only one merge at once

Merging more than one snapshot is not supported, so prevent
this happening.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: support barriers in snapshot merge target
Mike Snitzer [Thu, 10 Dec 2009 23:52:31 +0000 (23:52 +0000)]
dm snapshot: support barriers in snapshot merge target

Sets num_flush_requests=2 to support flushing both the origin and cow
devices used by the snapshot-merge target.

Also, snapshot_ctr() now gets the origin device using FMODE_WRITE if the
target is snapshot-merge (which writes to the origin device).

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: avoid allocating exceptions in merge
Mikulas Patocka [Thu, 10 Dec 2009 23:52:31 +0000 (23:52 +0000)]
dm snapshot: avoid allocating exceptions in merge

The snapshot-merge target should not allocate new exceptions because the
intent is to merge all of its exceptions as quickly and safely as
possible.

This patch introduces the snapshot-merge mapping function and updates
__origin_write() so that it doesn't allocate exceptions on any snapshots
that are being merged.

If a write request to a merging snapshot device is to be dispatched
directly to the origin (because the chunk is not remapped or was already
merged), snapshot_merge_map() must make exceptions in other snapshots so
calls do_origin().

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: rework writing to origin
Mikulas Patocka [Thu, 10 Dec 2009 23:52:30 +0000 (23:52 +0000)]
dm snapshot: rework writing to origin

To track the completion of exceptions relating to the same location on
the device, the current code selects one exception as primary_pe, links
the other exceptions to it and uses reference counting to wait until all
the reallocations are complete.

It is considered too complicated to extend this code to handle the new
snapshot-merge target, where sets of non-overlapping chunks would also
need to become linked.

Instead, a simpler (but less efficient) approach is taken.  Bios are
linked to one exception.  When it completes, bios are simply retried,
and if other related exceptions are still outstanding, they'll get
queued again to wait for another one.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: add merge target
Mikulas Patocka [Thu, 10 Dec 2009 23:52:30 +0000 (23:52 +0000)]
dm snapshot: add merge target

The snapshot-merge target allows a snapshot to be merged back into the
snapshot's origin device.

One anticipated use of snapshot merging is the rollback of filesystems
to back out problematic system upgrades.

This patch adds snapshot-merge target management to both
dm_snapshot_init() and dm_snapshot_exit().  As an initial place-holder,
snapshot-merge is identical to the snapshot target.  Documentation is
provided.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm exception store: add merge specific methods
Mikulas Patocka [Thu, 10 Dec 2009 23:52:29 +0000 (23:52 +0000)]
dm exception store: add merge specific methods

Add functions that decide how many consecutive chunks of snapshot to
merge back into the origin next and to update the metadata afterwards.

prepare_merge provides a pointer to the most recent still-to-be-merged
chunk and returns how many previous ones are consecutive and can be
processed together.

commit_merge removes the nr_merged most-recent chunks permanently from
the exception store.  The number must not exceed that returned by
prepare_merge.

Introduce NUM_SNAPSHOT_HDR_CHUNKS to show where the snapshot header
chunk is accounted for.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: create function for chunk_is_tracked wait
Mike Snitzer [Thu, 10 Dec 2009 23:52:29 +0000 (23:52 +0000)]
dm snapshot: create function for chunk_is_tracked wait

Move the __chunk_is_tracked() loop into a separate function as we will
also need to call it from the write path in the rare case of conflicting
writes to the same chunk.

Originally introduced in commit a8d41b59f3f5a7ac19452ef442a7fc1b5fa17366
("dm snapshot: fix race during exception creation").

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: make bio optional in __origin_write
Mikulas Patocka [Thu, 10 Dec 2009 23:52:28 +0000 (23:52 +0000)]
dm snapshot: make bio optional in __origin_write

To support the merging of snapshots back into their origin we need
to trigger exceptions in other snapshots not being merged without
any incoming bio on the origin device.  The bio parameter to
__origin_write() becomes optional and the sector needs supplying
separately.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm mpath: reject messages when device is suspended
Kiyoshi Ueda [Thu, 10 Dec 2009 23:52:27 +0000 (23:52 +0000)]
dm mpath: reject messages when device is suspended

This patch rejects messages that can generate I/O while the device
itself is suspended.

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Cc: Mike Anderson <andmike@linux.vnet.ibm.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: export suspended state to targets
Kiyoshi Ueda [Thu, 10 Dec 2009 23:52:27 +0000 (23:52 +0000)]
dm: export suspended state to targets

This patch adds the exported dm_suspended() function so that targets
can check whether or not they are suspended.

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Cc: Mike Anderson <andmike@linux.vnet.ibm.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: rename dm_suspended to dm_suspended_md
Kiyoshi Ueda [Thu, 10 Dec 2009 23:52:26 +0000 (23:52 +0000)]
dm: rename dm_suspended to dm_suspended_md

This patch renames dm_suspended() to dm_suspended_md() and
keeps it internal to dm.
No functional change.

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Cc: Mike Anderson <andmike@linux.vnet.ibm.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: swap target postsuspend call and setting suspended flag
Kiyoshi Ueda [Thu, 10 Dec 2009 23:52:26 +0000 (23:52 +0000)]
dm: swap target postsuspend call and setting suspended flag

This patch moves DMF_SUSPENDED flag set before postsuspend.
No one should care about the ordering, because the flag set and
the postsuspend are protected by a single lock, md->suspend_lock,
and all strict flag-checkers take the lock.

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Cc: Mike Anderson <andmike@linux.vnet.ibm.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm crypt: add plain64 iv
Milan Broz [Thu, 10 Dec 2009 23:52:25 +0000 (23:52 +0000)]
dm crypt: add plain64 iv

The default plain IV is 32-bit only.

This plain64 IV provides a compatible mode for encrypted devices bigger
than 4TB.

Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: trace request based remapping
Jun'ichi Nomura [Thu, 10 Dec 2009 23:52:25 +0000 (23:52 +0000)]
dm: trace request based remapping

This patch adds a remapping trace to request-based dm.
BIO-based dm already has the equivalent tracepoint.

For example, under this dm stack (linear LV on multipath):
  # dmsetup ls --tree -o ascii
  vg-lv0 (253:1)
   `-mpath0 (253:0)
      |- (8:160)
      |- (66:80)
      |- (65:176)
      `- (65:160)

Trace of 'dd of=/dev/vg/lv0 bs=128k count=1 oflag=direct' looks like this:

without the patch:
  dd-6674  [000]   539.727384: block_bio_queue: 253,1 WS 0 + 256 [dd]
  dd-6674  [000]   539.727392: block_remap: 253,0 WS 384 + 256 <- (253,1) 0
  dd-6674  [000]   539.727394: block_bio_queue: 253,0 WS 384 + 256 [dd]
  dd-6674  [000]   539.727405: block_getrq: 253,0 WS 384 + 256 [dd]
  dd-6674  [000]   539.727409: block_plug: [dd]
  dd-6674  [000]   539.727410: block_rq_insert: 253,0 W 0 () 384 + 256 [dd]
  dd-6674  [000]   539.727416: block_rq_issue: 253,0 W 0 () 384 + 256 [dd]
  dd-6674  [000]   539.727426: block_rq_insert: 65,176 W 0 () 384 + 256 [dd]
  dd-6674  [000]   539.727427: block_rq_issue: 65,176 W 0 () 384 + 256 [dd]
  ...

and with the patch: (the line with '**' is the trace added by this patch)
  dd-6617  [002]   162.914301: block_bio_queue: 253,1 WS 0 + 256 [dd]
  dd-6617  [002]   162.914314: block_remap: 253,0 WS 384 + 256 <- (253,1) 0
  dd-6617  [002]   162.914316: block_bio_queue: 253,0 WS 384 + 256 [dd]
  dd-6617  [002]   162.914331: block_getrq: 253,0 WS 384 + 256 [dd]
  dd-6617  [002]   162.914335: block_plug: [dd]
  dd-6617  [002]   162.914337: block_rq_insert: 253,0 W 0 () 384 + 256 [dd]
  dd-6617  [002]   162.914347: block_rq_issue: 253,0 W 0 () 384 + 256 [dd]
**dd-6617  [002]   162.914356: block_rq_remap: 65,176 W 384 + 256 <- (253,0) 384
  dd-6617  [002]   162.914358: block_rq_insert: 65,176 W 0 () 384 + 256 [dd]
  dd-6617  [002]   162.914359: block_rq_issue: 65,176 W 0 () 384 + 256 [dd]
  ...

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: allow live exception store handover between tables
Mike Snitzer [Thu, 10 Dec 2009 23:52:24 +0000 (23:52 +0000)]
dm snapshot: allow live exception store handover between tables

Permit in-use snapshot exception data to be 'handed over' from one
snapshot instance to another.  This is a pre-requisite for patches
that allow the changes made in a snapshot device to be merged back into
its origin device and also allows device resizing.

The basic call sequence is:

  dmsetup load new_snapshot (referencing the existing in-use cow device)
     - the ctr code detects that the cow is already in use and allows the
       two snapshot target instances to be linked together
  dmsetup suspend original_snapshot
  dmsetup resume new_snapshot
     - the new_snapshot becomes live, and if anything now tries to access
       the original one it will receive -EIO
  dmsetup remove original_snapshot

(There can only be two snapshot targets referencing the same cow device
simultaneously.)

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: keep old table until after resume succeeded
Alasdair G Kergon [Thu, 10 Dec 2009 23:52:24 +0000 (23:52 +0000)]
dm: keep old table until after resume succeeded

When swapping a new table into place, retain the old table until
its replacement is in place.

An old check for an empty table is removed because this is enforced
in populate_table().

__unbind() becomes redundant when followed by __bind().

Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: bind new table before destroying old
Alasdair G Kergon [Thu, 10 Dec 2009 23:52:23 +0000 (23:52 +0000)]
dm: bind new table before destroying old

When replacing a mapped device's table during a 'resume', delay the
destruction of the old table until the new one is successfully in place.

This will make it easier for a later patch to transfer internal state
information from the old table to the new one (something we do not currently
support) while giving us more options for reversion if a later part
of the operation fails.

Devices are always in the suspended state during dm_swap_table().
This patch reinforces the requirement that all I/O must have been
flushed from the table targets while in this state (including any in
workqueues).  In the case of 'noflush' suspending, unprocessed
I/O should have been 'pushed back' to the dm core prior to this point,
for resubmission after the new table is in place.

Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm ioctl: retrieve status from inactive table
Mike Snitzer [Thu, 10 Dec 2009 23:52:22 +0000 (23:52 +0000)]
dm ioctl: retrieve status from inactive table

Add the flag DM_QUERY_INACTIVE_TABLE_FLAG to the ioctls to return
infomation about the loaded-but-not-yet-active table instead of the live
table.  Prior to this patch it was impossible to obtain this information
until the device had been 'resumed'.

Userspace dmsetup and libdevmapper support the flag as of version 1.02.40.
e.g. dmsetup info --inactive vg1-lv1

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm io: handle empty barriers
Mikulas Patocka [Thu, 10 Dec 2009 23:52:22 +0000 (23:52 +0000)]
dm io: handle empty barriers

Accept empty barriers in dm-io.

dm-io will process empty write barrier requests just like the other
read/write requests.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm mpath: prevent io from work queue while suspended
Mike Anderson [Thu, 10 Dec 2009 23:52:21 +0000 (23:52 +0000)]
dm mpath: prevent io from work queue while suspended

Reject messages that can generate I/O while the device itself
is suspended.

Signed-off-by: Mike Anderson <andmike@linux.vnet.ibm.com>
Acked-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm mpath: add mutex to synchronize adding and flushing work
Mike Anderson [Thu, 10 Dec 2009 23:52:21 +0000 (23:52 +0000)]
dm mpath: add mutex to synchronize adding and flushing work

Add a mutex to allow possible creators of new work to synchronize with
flushing work queues.

Signed-off-by: Mike Anderson <andmike@linux.vnet.ibm.com>
Acked-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm ioctl: forbid messages to devices being deleted
Mike Anderson [Thu, 10 Dec 2009 23:52:20 +0000 (23:52 +0000)]
dm ioctl: forbid messages to devices being deleted

Once we begin deleting a device, prevent any further messages being sent
to targets of its table (to avoid races).

Signed-off-by: Mike Anderson <andmike@linux.vnet.ibm.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: add dm_deleting_md function
Mike Anderson [Thu, 10 Dec 2009 23:52:20 +0000 (23:52 +0000)]
dm: add dm_deleting_md function

Add dm_deleting_md to check whether or not a given mapped
device is currently being deleted.

Signed-off-by: Mike Anderson <andmike@linux.vnet.ibm.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm mpath: flush workqueues before suspend completes
Kiyoshi Ueda [Thu, 10 Dec 2009 23:52:19 +0000 (23:52 +0000)]
dm mpath: flush workqueues before suspend completes

This patch stops the remaining dm-mpath activity during the suspend
sequence by flushing workqueues in postsuspend function.

The current dm-mpath target may not be quiet even after suspend completes
because some workqueues (e.g. device_handler's work, event handling)
are not flushed during the suspend sequence, even though suspended
devices/targets are supposed to be quiet in this state.

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: rename dm_get_table to dm_get_live_table
Alasdair G Kergon [Thu, 10 Dec 2009 23:52:19 +0000 (23:52 +0000)]
dm: rename dm_get_table to dm_get_live_table

Rename dm_get_table to dm_get_live_table.

Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: add request based barrier support
Kiyoshi Ueda [Thu, 10 Dec 2009 23:52:18 +0000 (23:52 +0000)]
dm: add request based barrier support

This patch adds barrier support for request-based dm.

CORE DESIGN

The design is basically same as bio-based dm, which emulates barrier
by mapping empty barrier bios before/after a barrier I/O.
But request-based dm has been using struct request_queue for I/O
queueing, so the block-layer's barrier mechanism can be used.

o Summary of the block-layer's behavior (which is depended by dm-core)
  Request-based dm uses QUEUE_ORDERED_DRAIN_FLUSH ordered mode for
  I/O barrier.  It means that when an I/O requiring barrier is found
  in the request_queue, the block-layer makes pre-flush request and
  post-flush request just before and just after the I/O respectively.

  After the ordered sequence starts, the block-layer waits for all
  in-flight I/Os to complete, then gives drivers the pre-flush request,
  the barrier I/O and the post-flush request one by one.
  It means that the request_queue is stopped automatically by
  the block-layer until drivers complete each sequence.

o dm-core
  For the barrier I/O, treats it as a normal I/O, so no additional
  code is needed.

  For the pre/post-flush request, flushes caches by the followings:
    1. Make the number of empty barrier requests required by target's
       num_flush_requests, and map them (dm_rq_barrier()).
    2. Waits for the mapped barriers to complete (dm_rq_barrier()).
       If error has occurred, save the error value to md->barrier_error
       (dm_end_request()).
       (*) Basically, the first reported error is taken.
           But -EOPNOTSUPP supersedes any error and DM_ENDIO_REQUEUE
           follows.
    3. Requeue the pre/post-flush request if the error value is
       DM_ENDIO_REQUEUE.  Otherwise, completes with the error value
       (dm_rq_barrier_work()).
  The pre/post-flush work above is done in the kernel thread (kdmflush)
  context, since memory allocation which might sleep is needed in
  dm_rq_barrier() but sleep is not allowed in dm_request_fn(), which is
  an irq-disabled context.
  Also, clones of the pre/post-flush request share an original, so
  such clones can't be completed using the softirq context.
  Instead, complete them in the context of underlying device drivers.
  It should be safe since there is no I/O dispatching during
  the completion of such clones.

  For suspend, the workqueue of kdmflush needs to be flushed after
  the request_queue has been stopped.  Otherwise, the next flush work
  can be kicked even after the suspend completes.

TARGET INTERFACE

No new interface is added.
Just use the existing num_flush_requests in struct target_type
as same as bio-based dm.

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: move dm_end_request
Kiyoshi Ueda [Thu, 10 Dec 2009 23:52:17 +0000 (23:52 +0000)]
dm: move dm_end_request

This patch moves dm_end_request() to make the next patch more readable.
No functional change.

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: refactor request based completion functions
Kiyoshi Ueda [Thu, 10 Dec 2009 23:52:17 +0000 (23:52 +0000)]
dm: refactor request based completion functions

This patch factors out the clone completion code, dm_done(),
from dm_softirq_done() in preparation for a subsequent patch.
No functional change.

dm_done() will be used in barrier completion, which can't use and
doesn't need softirq.  The softirq_done callback needs to get a clone
from an original request but it can't in the case of barrier, where
an original request is shared by multiple clones.  On the other hand,
the completion of barrier clones doesn't involve re-submitting requests,
which was the primary reason of the need for softirq.

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: use md pending for in flight IO counting
Kiyoshi Ueda [Thu, 10 Dec 2009 23:52:16 +0000 (23:52 +0000)]
dm: use md pending for in flight IO counting

This patch changes the counter for the number of in_flight I/Os
to md->pending from q->in_flight in preparation for a later patch.
No functional change.

Request-based dm used q->in_flight to count the number of in-flight
clones assuming the counter is always incremented for an in-flight
original request and original:clone is 1:1 relationship.
However, it this no longer true for barrier requests.
So use md->pending to count the number of in-flight clones.

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: simplify request based suspend
Kiyoshi Ueda [Thu, 10 Dec 2009 23:52:16 +0000 (23:52 +0000)]
dm: simplify request based suspend

The semantics of bio-based dm were changed recently in the case of
suspend with "--nolockfs" but without "--noflush".
Before 2.6.30, I/Os submitted before the suspend invocation were always
flushed.  From 2.6.30 onwards, I/Os submitted before the suspend
invocation might not be flushed.  (For details, see
http://marc.info/?t=123994433400003&r=1&w=2)

This patch brings the behaviour of request-based dm into line with
bio-based dm, simplifying the code and preparing for a subsequent patch
that will wait for all in_flight I/Os to complete without stopping
request_queue and use dm_wait_for_completion() for it.

This change in semantics simplifies the suspend code as follows:
  o Suspend is implemented as stopping request_queue
    in request-based dm, and all I/Os are queued in the request_queue
    even after suspend is invoked.
  o In the old semantics, we had to track whether I/Os were
    queued before or after the suspend invocation, so a special
    barrier-like request called 'suspend marker' was introduced.
  o With the new semantics, we don't need to flush any I/O
    so we can remove the marker and the code related to the marker
    handling and I/O flushing.

After removing this codes, the suspend sequence is now:
  1. Flush all I/Os by lock_fs() if needed.
  2. Stop dispatching any I/O by stopping the request_queue.
  3. Wait for all in-flight I/Os to be completed or requeued.

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: abstract clone_rq
Kiyoshi Ueda [Thu, 10 Dec 2009 23:52:15 +0000 (23:52 +0000)]
dm: abstract clone_rq

This patch factors out the request cloning code in dm_prep_fn()
as clone_rq().  No functional change.

This patch is a preparation for a later patch in this series which needs to
make clones from an original barrier request.

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: pass gfp_mask to alloc_rq_tio
Kiyoshi Ueda [Thu, 10 Dec 2009 23:52:15 +0000 (23:52 +0000)]
dm: pass gfp_mask to alloc_rq_tio

This patch adds the gfp_mask argument to alloc_rq_tio().
No functional change.

This patch is a preparation for a later patch in this series which needs to
allocate tio (for barrier I/O) with different allocation flag (GFP_NOIO) from
the one in the normal I/O code path.

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: use clone in map_request function
Kiyoshi Ueda [Thu, 10 Dec 2009 23:52:14 +0000 (23:52 +0000)]
dm: use clone in map_request function

This patch changes the argument of map_request() to clone request
from original request.  No functional change.

This patch is a preparation for PATCH 9, which needs to use
map_request() for clones sharing an original barrier request.

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: abstract dm_in_flight function
Kiyoshi Ueda [Thu, 10 Dec 2009 23:52:13 +0000 (23:52 +0000)]
dm: abstract dm_in_flight function

This patch adds md_in_flight() to get the number of in_flight I/Os.
No functional change.

This patch is a preparation for a later patch in this series, which
changes I/O counter to md->pending from q->in_flight in request-based dm.

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm kcopyd: accept zero size jobs
Mikulas Patocka [Thu, 10 Dec 2009 23:52:13 +0000 (23:52 +0000)]
dm kcopyd: accept zero size jobs

dm-kcopyd: accept zero-size jobs

This patch changes dm-kcopyd so that it accepts zero-size jobs and completes
them immediatelly via its completion thread.

It is needed for multisnapshots snapshot resizing. When we are writing to
a chunk beyond origin end, no copying is done. To simplify the code, we submit
an empty request to kcopyd and let kcopyd complete it. If we didn't submit
a request to kcopyd and called the completion routine immediatelly, it would
violate the principle that completion is called only from one thread and
it would need additional locking.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: track suspended state in target
Mike Snitzer [Thu, 10 Dec 2009 23:52:12 +0000 (23:52 +0000)]
dm snapshot: track suspended state in target

Keep track of whether or not the device is suspended within the snapshot
target module, the same as we do in dm-raid1.

We will use this later to enforce the correct sequence of ioctls to
transfer the in-core exceptions from a snapshot target instance in
one table to a replacement one capable of merging them back
into the origin.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: move cow ref from exception store to snap core
Mike Snitzer [Thu, 10 Dec 2009 23:52:12 +0000 (23:52 +0000)]
dm snapshot: move cow ref from exception store to snap core

Store the reference to the snapshot cow device in the core snapshot
code instead of each exception store.  It can be accessed through the
new function dm_snap_cow().  Exception stores should each now maintain a
reference to their parent snapshot struct.

This is cleaner and makes part of the forthcoming snapshot merge code simpler.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Reviewed-by: Jonathan Brassow <jbrassow@redhat.com>
Cc: Mikulas Patocka <mpatocka@redhat.com>
11 years agodm snapshot: add allocated metadata to snapshot status
Mike Snitzer [Thu, 10 Dec 2009 23:52:11 +0000 (23:52 +0000)]
dm snapshot: add allocated metadata to snapshot status

Add number of sectors used by metadata to the end of the snapshot's status
line.

Renamed dm_exception_store_type's 'fraction_full' to 'usage'.  Renamed
arguments to be clearer about what is being returned.  Also added
'metadata_sectors'.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: rename exception functions
Jon Brassow [Thu, 10 Dec 2009 23:52:11 +0000 (23:52 +0000)]
dm snapshot: rename exception functions

Rename exception functions.  Preparing to pull them out of
dm-snap.c for broader use.

Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
Reviewed-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: rename exception_table to dm_exception_table
Jon Brassow [Thu, 10 Dec 2009 23:52:10 +0000 (23:52 +0000)]
dm snapshot: rename exception_table to dm_exception_table

Rename exception_table for broader use outside dm-snap.c

Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
Reviewed-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: rename dm_snap_exception to dm_exception
Jon Brassow [Thu, 10 Dec 2009 23:52:10 +0000 (23:52 +0000)]
dm snapshot: rename dm_snap_exception to dm_exception

The exception structure is not necessarily just a snapshot
element (especially after we pull it out of dm-snap.c).

Renaming appropriately.

Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
Reviewed-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: consolidate insert exception functions
Jon Brassow [Thu, 10 Dec 2009 23:52:09 +0000 (23:52 +0000)]
dm snapshot: consolidate insert exception functions

Consolidate the insert_*exception functions.  'insert_completed_exception'
already contains all the logic to handle 'insert_exception' (via
check for a hash_shift of 0), so remove redundant function.

Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
Reviewed-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: abstract minimum_chunk_size fn
Mikulas Patocka [Thu, 10 Dec 2009 23:52:08 +0000 (23:52 +0000)]
dm snapshot: abstract minimum_chunk_size fn

The origin needs to find minimum chunksize of all snapshots.  This logic is
moved to a separate function because it will be used at another place in
the snapshot merge patches.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Mike Snitzer <snitzer@redhat.com>
Reviewed-by: Jonathan Brassow <jbrassow@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: simplify sector_to_chunk expression
Mikulas Patocka [Thu, 10 Dec 2009 23:52:08 +0000 (23:52 +0000)]
dm snapshot: simplify sector_to_chunk expression

Removed unnecessary 'and' masking: The right shift discards the lower
bits so there is no need to clear them.

(A later patch needs this change to support a 32-bit chunk_mask.)

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Mike Snitzer <snitzer@redhat.com>
Reviewed-by: Jonathan Brassow <jbrassow@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: avoid else clause in persistent_read_metadata
Jon Brassow [Thu, 10 Dec 2009 23:52:07 +0000 (23:52 +0000)]
dm snapshot: avoid else clause in persistent_read_metadata

Minor code touch-up.  We don't need the 'else'.

Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
Reviewed-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm ioctl: prefer strlcpy over strncpy
Roel Kluin [Thu, 10 Dec 2009 23:52:07 +0000 (23:52 +0000)]
dm ioctl: prefer strlcpy over strncpy

strlcpy() will always null terminate the string.

    The code should already guarantee this as the last bytes are already
    NULs and the string lengths were restricted before being stored in
    hc.  Removing the '-1' becomes necessary so strlcpy() doesn't
    lose the last character of a maximum-length string.
- agk

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm raid1: explicitly initialise bio_lists
Mikulas Patocka [Thu, 10 Dec 2009 23:52:06 +0000 (23:52 +0000)]
dm raid1: explicitly initialise bio_lists

Explicitly initialize bio lists instead of relying on kzalloc.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Takahiro Yasui <tyasui@redhat.com>
Tested-by: Takahiro Yasui <tyasui@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm raid1: hold all write bios when leg fails
Mikulas Patocka [Thu, 10 Dec 2009 23:52:06 +0000 (23:52 +0000)]
dm raid1: hold all write bios when leg fails

Hold all write bios when leg fails and errors are handled

When using a userspace daemon such as dmeventd to handle errors, we must
delay completing  bios until it has done its job.
This patch prevents the following race:
  - primary leg fails
  - write "1" fail, the write is held, secondary leg is set default
  - write "2" goes straight to the secondary leg

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Takahiro Yasui <tyasui@redhat.com>
Tested-by: Takahiro Yasui <tyasui@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm raid1: hold write bios when errors are handled
Mikulas Patocka [Thu, 10 Dec 2009 23:52:05 +0000 (23:52 +0000)]
dm raid1: hold write bios when errors are handled

Hold all write bios when errors are handled.

Previously the failures list was used only when handling errors with
a userspace daemon such as dmeventd.  Now, it is always used for all bios.
The regions where some writes failed must be marked as nosync. This can only
be done in process context (i.e. in raid1 workqueue), not in the
write_callback function.

Previously the write would succeed if writing to at least one leg
succeeded.  This is wrong because data from the failed leg may be
replicated to the correct leg.  Now, if using a userspace daemon, the
write with some failures will be held until the daemon has done its job
and reconfigured the array.  If not using a daemon, the write still
succeeds if at least one leg succeeds. This is bad, but it is consistent
with current behavior.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Takahiro Yasui <tyasui@redhat.com>
Tested-by: Takahiro Yasui <tyasui@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm raid1: remove bio_endio from dm_rh_mark_nosync
Mikulas Patocka [Thu, 10 Dec 2009 23:52:05 +0000 (23:52 +0000)]
dm raid1: remove bio_endio from dm_rh_mark_nosync

Move bio completion out of dm_rh_mark_nosync in preparation for the
next patch.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Takahiro Yasui <tyasui@redhat.com>
Tested-by: Takahiro Yasui <tyasui@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm raid1: abstract get_valid_mirror function
Mikulas Patocka [Thu, 10 Dec 2009 23:52:04 +0000 (23:52 +0000)]
dm raid1: abstract get_valid_mirror function

Move the logic to get a valid mirror leg into a function for re-use
in a later patch.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Takahiro Yasui <tyasui@redhat.com>
Tested-by: Takahiro Yasui <tyasui@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm raid1: use hold framework in do_failures
Mikulas Patocka [Thu, 10 Dec 2009 23:52:04 +0000 (23:52 +0000)]
dm raid1: use hold framework in do_failures

Use the hold framework in do_failures.

This patch doesn't change the bio processing logic, it just simplifies
failure handling and avoids periodically polling the failures list.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Takahiro Yasui <tyasui@redhat.com>
Tested-by: Takahiro Yasui <tyasui@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm raid1: add framework to hold bios during suspend
Mikulas Patocka [Thu, 10 Dec 2009 23:52:03 +0000 (23:52 +0000)]
dm raid1: add framework to hold bios during suspend

Add framework to delay bios until a suspend and then resubmit them with
either DM_ENDIO_REQUEUE (if the suspend was noflush) or complete them
with -EIO.  I/O barrier support will use this.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Takahiro Yasui <tyasui@redhat.com>
Tested-by: Takahiro Yasui <tyasui@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm raid1: report flush errors separately in status
Mikulas Patocka [Thu, 10 Dec 2009 23:52:02 +0000 (23:52 +0000)]
dm raid1: report flush errors separately in status

Report flush errors as 'F' instead of 'D' for log and mirror devices.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm raid1: implement mirror_flush
Mikulas Patocka [Thu, 10 Dec 2009 23:52:02 +0000 (23:52 +0000)]
dm raid1: implement mirror_flush

Implement flush callee. It uses dm_io to send zero-size barrier synchronously
and concurrently to all the mirror legs.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm log: use flush callback fn
Mikulas Patocka [Thu, 10 Dec 2009 23:52:01 +0000 (23:52 +0000)]
dm log: use flush callback fn

Call the flush callback from the log.

If flush failed, we have no alternative but to mark the whole log as dirty.
Also we set the variable flush_failed to prevent any bits ever being marked as
clean again.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm log: add flush callback fn
Mikulas Patocka [Thu, 10 Dec 2009 23:52:01 +0000 (23:52 +0000)]
dm log: add flush callback fn

Introduce a callback pointer from the log to dm-raid1 layer.

Before some region is set as "in-sync", we need to flush hardware cache on
all the disks. But the log module doesn't have access to the mirror_set
structure. So it will use this callback.

So far the callback is unused, it will be used in further patches.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm log: introduce flush_failed variable
Mikulas Patocka [Thu, 10 Dec 2009 23:52:00 +0000 (23:52 +0000)]
dm log: introduce flush_failed variable

Introduce "flush failed" variable.  When a flush before clearing a bit
in the log fails, we don't know anything about which which regions are
in-sync and which not.

So we need to set all regions as not-in-sync and set the variable
"flush_failed" to prevent setting the in-sync bit in the future.

A target reload is the only way to get out of this situation.

The variable will be set in following patches.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm log: add flush_header function
Mikulas Patocka [Thu, 10 Dec 2009 23:52:00 +0000 (23:52 +0000)]
dm log: add flush_header function

Introduce flush_header and use it to flush the log device.

Note that we don't have to flush if all the regions transition
from "dirty" to "clean" state.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm raid1: split touched state into two
Mikulas Patocka [Thu, 10 Dec 2009 23:51:59 +0000 (23:51 +0000)]
dm raid1: split touched state into two

Split the variable "touched" into two, "touched_dirtied" and
"touched_cleaned", set when some region was dirtied or cleaned.

This will be used to optimize flushes.

After a transition from "dirty" to "clean" state we don't have flush hardware
cache on the log device. After a transition from "clean" to "dirty" the cache
must be flushed.

Before a transition from "clean" to "dirty" state we don't have to flush all
the raid legs. Before a transition from "dirty" to "clean" we must flush all
the legs to make sure that they are really in sync.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm raid1: support flush
Mikulas Patocka [Thu, 10 Dec 2009 23:51:59 +0000 (23:51 +0000)]
dm raid1: support flush

Flush support for dm-raid1.

When it receives an empty barrier, submit it to all the devices via dm-io.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm io: remove extra bi_io_vec region hack
Mikulas Patocka [Thu, 10 Dec 2009 23:51:58 +0000 (23:51 +0000)]
dm io: remove extra bi_io_vec region hack

Remove the hack where we allocate an extra bi_io_vec to store additional
private data.  This hack prevents us from supporting barriers in
dm-raid1 without first making another little block layer change.
Instead of doing that, this patch eliminates the bi_io_vec abuse by
storing the region number directly in the low bits of bi_private.

We need to store two things for each bio, the pointer to the main io
structure and, if parallel writes were requested, an index indicating
which of these writes this bio belongs to.  There can be at most
BITS_PER_LONG regions - 32 or 64.

The index (region number) was stored in the last (hidden) bio vector and
the pointer to struct io was stored in bi_private.

This patch now aligns "struct io" on BITS_PER_LONG bytes and stores the
region number in the low BITS_PER_LONG bits of bi_private.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm io: use slab for struct io
Mikulas Patocka [Thu, 10 Dec 2009 23:51:57 +0000 (23:51 +0000)]
dm io: use slab for struct io

Allocate "struct io" from a slab.

This patch changes dm-io, so that "struct io" is allocated from a slab cache.
It used to be allocated with kmalloc. Allocating from a slab will be needed
for the next patch, because it requires a special alignment of "struct io"
and kmalloc cannot meet this alignment.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm crypt: make wipe message also wipe essiv key
Milan Broz [Thu, 10 Dec 2009 23:51:57 +0000 (23:51 +0000)]
dm crypt: make wipe message also wipe essiv key

The "wipe key" message is used to wipe the volume key from memory
temporarily, for example when suspending to RAM.

But the initialisation vector in ESSIV mode is calculated from the
hashed volume key, so the wipe message should wipe this IV key too and
reinitialise it when the volume key is reinstated.

This patch adds an IV wipe method called from a wipe message callback.
ESSIV is then reinitialised using the init function added by the
last patch.

Cc: stable@kernel.org
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm crypt: separate essiv allocation from initialisation
Milan Broz [Thu, 10 Dec 2009 23:51:56 +0000 (23:51 +0000)]
dm crypt: separate essiv allocation from initialisation

This patch separates the construction of IV from its initialisation.
(For ESSIV it is a hash calculation based on volume key.)

Constructor code now preallocates hash tfm and salt array
and saves it in a private IV structure.

The next patch requires this to reinitialise the wiped IV
without reallocating memory when resuming a suspended device.

Cc: stable@kernel.org
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm crypt: restructure essiv error path
Milan Broz [Thu, 10 Dec 2009 23:51:56 +0000 (23:51 +0000)]
dm crypt: restructure essiv error path

Use kzfree for salt deallocation because it is derived from the volume
key.  Use a common error path in ESSIV constructor.

Required by a later patch which fixes the way key material is wiped
from memory.

Cc: stable@kernel.org
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm crypt: move private iv fields to structs
Milan Broz [Thu, 10 Dec 2009 23:51:55 +0000 (23:51 +0000)]
dm crypt: move private iv fields to structs

Define private structures for IV so it's easy to add further attributes
in a following patch which fixes the way key material is wiped from
memory.  Also move ESSIV destructor and remove unnecessary 'status'
operation.

There are no functional changes in this patch.

Cc: stable@kernel.org
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm crypt: make wipe message also wipe tfm key
Milan Broz [Thu, 10 Dec 2009 23:51:55 +0000 (23:51 +0000)]
dm crypt: make wipe message also wipe tfm key

The "wipe key" message is used to wipe a volume key from memory
temporarily, for example when suspending to RAM.

There are two instances of the key in memory (inside crypto tfm)
but only one got wiped.  This patch wipes them both.

Cc: stable@kernel.org
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: cope with chunk size larger than origin
Mikulas Patocka [Thu, 10 Dec 2009 23:51:54 +0000 (23:51 +0000)]
dm snapshot: cope with chunk size larger than origin

Under some special conditions the snapshot hash_size is calculated as zero.
This patch instead sets a minimum value of 64, the same as for the
pending exception table.

rounddown_pow_of_two(0) is an undefined operation (it expands to shift
by -1).  init_exception_table with an argument of 0 would fail with -ENOMEM.

The way to trigger the problem is to create a snapshot with a chunk size
that is larger than the origin device.

Cc: stable@kernel.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm snapshot: only take lock for statustype info not table
Mikulas Patocka [Thu, 10 Dec 2009 23:51:53 +0000 (23:51 +0000)]
dm snapshot: only take lock for statustype info not table

Take snapshot lock only for STATUSTYPE_INFO, not STATUSTYPE_TABLE.

Commit 4c6fff445d7aa753957856278d4d93bcad6e2c14
(dm-snapshot-lock-snapshot-while-supplying-status.patch)
introduced this use of the lock, but userspace applications using
libdevmapper have been found to request STATUSTYPE_TABLE while the device
is suspended and the lock is already held, leading to deadlock.  Since
the lock is not necessary in this case, don't try to take it.

Cc: stable@kernel.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: sysfs add empty release function to avoid debug warning
Milan Broz [Thu, 10 Dec 2009 23:51:53 +0000 (23:51 +0000)]
dm: sysfs add empty release function to avoid debug warning

This patch just removes an unnecessary warning:
 kobject: 'dm': does not have a release() function,
 it is broken and must be fixed.

The kobject is embedded in mapped device struct, so
code does not need to release memory explicitly here.

Cc: stable@kernel.org
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm exception store: free tmp_store on persistent flag error
Julia Lawall [Thu, 10 Dec 2009 23:51:52 +0000 (23:51 +0000)]
dm exception store: free tmp_store on persistent flag error

Error handling code following a kmalloc should free the allocated data.

Cc: stable@kernel.org
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agodm: avoid _hash_lock deadlock
Mikulas Patocka [Thu, 10 Dec 2009 23:51:52 +0000 (23:51 +0000)]
dm: avoid _hash_lock deadlock

Fix a reported deadlock if there are still unprocessed multipath events
on a device that is being removed.

_hash_lock is held during dev_remove while trying to send the
outstanding events.  Sending the events requests the _hash_lock
again in dm_copy_name_and_uuid.

This patch introduces a separate lock around regions that modify the
link to the hash table (dm_set_mdptr) or the name or uuid so that
dm_copy_name_and_uuid no longer needs _hash_lock.

Additionally, dm_copy_name_and_uuid can only be called if md exists
so we can drop the dm_get() and dm_put() which can lead to a BUG()
while md is being freed.

The deadlock:
 #0 [ffff8106298dfb48] schedule at ffffffff80063035
 #1 [ffff8106298dfc20] __down_read at ffffffff8006475d
 #2 [ffff8106298dfc60] dm_copy_name_and_uuid at ffffffff8824f740
 #3 [ffff8106298dfc90] dm_send_uevents at ffffffff88252685
 #4 [ffff8106298dfcd0] event_callback at ffffffff8824c678
 #5 [ffff8106298dfd00] dm_table_event at ffffffff8824dd01
 #6 [ffff8106298dfd10] __hash_remove at ffffffff882507ad
 #7 [ffff8106298dfd30] dev_remove at ffffffff88250865
 #8 [ffff8106298dfd60] ctl_ioctl at ffffffff88250d80
 #9 [ffff8106298dfee0] do_ioctl at ffffffff800418c4
#10 [ffff8106298dff00] vfs_ioctl at ffffffff8002fab9
#11 [ffff8106298dff40] sys_ioctl at ffffffff8004bdaf
#12 [ffff8106298dff80] tracesys at ffffffff8005d28d (via system_call)

Cc: stable@kernel.org
Reported-by: guy keren <choo@actcom.co.il>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
11 years agoMerge branch 'acpica' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux...
Linus Torvalds [Thu, 10 Dec 2009 03:57:06 +0000 (19:57 -0800)]
Merge branch 'acpica' of git://git./linux/kernel/git/lenb/linux-acpi-2.6

* 'acpica' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux-acpi-2.6:
  ACPICA: Update version to 20091112.
  ACPICA: Add additional module-level code support
  ACPICA: Deploy new create integer interface where appropriate
  ACPICA: New internal utility function to create Integer objects
  ACPICA: Add repair for predefined methods that must return sorted lists
  ACPICA: Fix possible fault if return Package objects contain NULL elements
  ACPICA: Add post-order callback to acpi_walk_namespace
  ACPICA: Change package length error message to an info message
  ACPICA: Reduce severity of predefined repair messages, Warning to Info
  ACPICA: Update version to 20091013
  ACPICA: Fix possible memory leak for Scope ASL operator
  ACPICA: Remove possibility of executing _REG methods twice
  ACPICA: Add repair for bad _MAT buffers
  ACPICA: Add repair for bad _BIF/_BIX packages

11 years agoMerge branch 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jdelv...
Linus Torvalds [Thu, 10 Dec 2009 03:53:11 +0000 (19:53 -0800)]
Merge branch 'hwmon-for-linus' of git://git./linux/kernel/git/jdelvare/staging

* 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jdelvare/staging: (41 commits)
  hwmon: (adt7475) Add VID support for the ADT7476
  hwmon: (adt7475) Add an entry in MAINTAINERS
  hwmon: (adt7475) Add support for the ADT7476
  hwmon: (adt7475) Voltage attenuators can be bypassed
  hwmon: (adt7475) Print device information on probe
  hwmon: (adt7475) Handle alternative pin functions
  hwmon: (adt7475) Move sysfs files removal to a separate function
  hwmon: (adt7475) Add support for the ADT7490
  hwmon: (adt7475) Improve device detection
  hwmon: (adt7475) Add missing static marker
  hwmon: (adt7475) Rework voltage inputs handling
  hwmon: (adt7475) Implement pwm_use_point2_pwm_at_crit
  hwmon: (adt7475) New documentation
  hwmon: (adt7475) Add support for the ADT7473
  hwmon: (f71882fg) Add support for the f71889fg (version 2)
  hwmon: (f71882fg) Fix sysfs file removal
  hwmon: (f71882fg) Cleanup sysfs attr creation 2/2
  hwmon: (f71882fg) Cleanup sysfs attr creation 1/2
  hwmon: (thmc50) Stop using I2C_CLIENT_MODULE_PARM
  hwmon: Add Freescale MC13783 ADC driver
  ...

11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6
Linus Torvalds [Thu, 10 Dec 2009 03:52:13 +0000 (19:52 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/tiwai/sound-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6:
  ALSA: ice1724 - aureon - modify WM8770 Master & DAC volume
  ALSA: hda/realtek: quirk for D945GCLF2 mainboard
  ALSA: hda - Terradici HDA controllers does not support 64-bit mode
  ALSA: document: Add direct git link to grub hda-analyzer
  ALSA: radio/sound/miro: fix build, cleanup depends/selects
  ALSA: hda - Generalize EAPD inversion check in patch_analog.c
  ASoC: Wrong variable returned on error
  ALSA: snd-usb-us122l: add product IDs of US-122MKII and US-144MKII
  ALSA: hda - Exclude unusable ADCs for ALC88x
  ALSA: hda - Add missing Line-Out and PCM switches as slave
  ALSA: hda - iMac 9,1 sound patch.
  ALSA: opti93x: set MC indirect registers base from PnP data

11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Thu, 10 Dec 2009 03:52:01 +0000 (19:52 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input: (51 commits)
  Input: appletouch - give up maintainership
  Input: dm355evm_kbd - switch to using sparse keymap library
  Input: wistron_btns - switch to using sparse keymap library
  Input: add generic support for sparse keymaps
  Input: fix memory leak in force feedback core
  Input: wistron - remove identification strings from DMI table
  Input: psmouse - remove identification strings from DMI tables
  Input: atkbd - remove identification strings from DMI table
  Input: i8042 - remove identification strings from DMI tables
  DMI: allow omitting ident strings in DMI tables
  Input: psmouse - do not carry DMI data around
  Input: matrix-keypad - switch to using dev_pm_ops
  Input: keyboard - fix lack of locking when traversing handler->h_list
  Input: gpio_keys - scan gpio state at probe and resume time
  Input: keyboard - add locking around event handling
  Input: usbtouchscreen - add support for ET&T TC5UH touchscreen controller
  Input: xpad - add two new Xbox 360 devices
  Input: polled device - do not start polling if interval is zero
  Input: polled device - schedule first poll immediately
  Input: add S3C24XX touchscreen driver
  ...

11 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Thu, 10 Dec 2009 03:50:49 +0000 (19:50 -0800)]
Merge branch 'for_linus' of git://git./linux/kernel/git/mchehab/linux-2.6

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-2.6: (345 commits)
  V4L/DVB (13542): ir-keytable: Allow dynamic table change
  V4L/DVB (13541): atbm8830: replace 64-bit division and floating point usage
  V4L/DVB (13540): ir-common: Cleanup get key evdev code
  V4L/DVB (13539): ir-common: add __func__ for debug messages
  V4L/DVB (13538): ir-common: Use a dynamic keycode table
  V4L/DVB (13537): ir: Prepare the code for dynamic keycode table allocation
  V4L/DVB (13536): em28xx: Use the full RC5 code on HVR-950 Remote Controller
  V4L/DVB (13535): ir-common: Add a hauppauge new table with the complete RC5 code
  V4L/DVB (13534): ir-common: Remove some unused fields/structs
  V4L/DVB (13533): ir: use dynamic tables, instead of static ones
  V4L/DVB (13532): ir-common: Add infrastructure to use a dynamic keycode table
  V4L/DVB (13531): ir-common: rename the debug routine to allow exporting it
  V4L/DVB (13458): go7007: subdev conversion
  V4L/DVB (13457): s2250: subdev conversion
  V4L/DVB (13456): s2250: Change module structure
  V4L/DVB (13528): em28xx: add support for em2800 VC211A card
  em28xx: don't reduce scale to half size for em2800
  em28xx: don't load audio modules when AC97 is mis-detected
  em28xx: em2800 chips support max width of 640
  V4L/DVB (13523): dvb-bt8xx: fix compile warning
  ...

Fix up trivial conflicts due to spelling fixes from the trivial tree in
Documentation/video4linux/gspca.txt
drivers/media/video/cx18/cx18-mailbox.h

11 years agoMerge branch 'next-i2c' of git://git.fluff.org/bjdooks/linux
Linus Torvalds [Thu, 10 Dec 2009 03:46:12 +0000 (19:46 -0800)]
Merge branch 'next-i2c' of git://git.fluff.org/bjdooks/linux

* 'next-i2c' of git://git.fluff.org/bjdooks/linux: (25 commits)
  i2c-pnx: Map I2C adapter number to platform ID number
  i2c-omap: OMAP3: PM: (re)init for every transfer to support off-mode
  i2c-designware: i2c_dw_handle_tx_abort: Use dev_dbg() for NOACK cases
  i2c-designware: Cosmetic cleanups
  i2c-designware: Tx abort cleanups
  i2c-designware: Skip RX_FULL and TX_EMPTY bits on tx abort errors
  i2c-designware: i2c_dw_xfer_msg: Fix error handling procedures
  i2c-designware: Disable TX_EMPTY when all i2c_msg msgs has been processed
  i2c-designware: Process all i2c_msg messages in the interrupt handler
  i2c-designware: i2c_dw_read: Remove redundant target address checker
  i2c-designware: i2c_dw_func: Set I2C_FUNC_SMBUS_foo bits
  i2c-designware: Initialize byte count variables just prior to being used
  i2c-designware: i2c_dw_xfer_msg: Introduce a local "buf" pointer
  i2c-designware: Divide i2c_dw_xfer_msg into two functions
  i2c-designware: Enable RX_FULL interrupt
  i2c-designware: Set Tx/Rx FIFO threshold levels
  i2c-designware: Process i2c_msg messages in the interrupt handler
  i2c-designware: i2c_dw_xfer_msg: Fix i2c_msg search bug
  i2c-designware: Improved _HCNT/_LCNT calculation
  i2c-designware: Remove an useless local variable "num"
  ...

11 years agoMerge branch 's3c24xx-updates' of git://git.fluff.org/bjdooks/linux
Linus Torvalds [Thu, 10 Dec 2009 03:45:58 +0000 (19:45 -0800)]
Merge branch 's3c24xx-updates' of git://git.fluff.org/bjdooks/linux

* 's3c24xx-updates' of git://git.fluff.org/bjdooks/linux:
  ARM: S3C24XX: DMA: Use valid index when accessing array
  ARM: S3C: move s3c_pwm_remove to .devexit.text
  ARM: S3C24XX: Export s3c24xx_set_fiq for modules.
  ARM: S3C: move s3c_adc_remove to .devexit.text

11 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/ide-next-2.6
Linus Torvalds [Thu, 10 Dec 2009 03:45:38 +0000 (19:45 -0800)]
Merge git://git./linux/kernel/git/davem/ide-next-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/ide-next-2.6:
  sl82c105: remove no longer needed debugging code
  sis5513: remove stale TODO
  pdc202xx_old: remove no longer needed debugging code
  cy82c693: remove no longer needed debugging code
  cmd64x: remove no longer needed debugging code
  alim15x3: remove obsolete and dangerous wdc_udma parameter
  ide: Increase WAIT_DRQ to accomodate some CF cards and SSD drives.
  cs5535: add pci id for AMD based CS5535 controllers
  slc90e66: fix UDMA handling
  drivers/ide/tx4938ide.c: use resource_size()
  drivers/ide/ide_platform.c: use resource_size()
  drivers/ide/au1xxx-ide.c: use resource_size()
  hpt366: remove dead old timing tables
  ide: update Kconfig text to mark as deprecated
  ide-tape: remove the BKL
  hpt366: kill unused #define's

11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/trivial
Linus Torvalds [Thu, 10 Dec 2009 03:43:33 +0000 (19:43 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jikos/trivial

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/trivial: (42 commits)
  tree-wide: fix misspelling of "definition" in comments
  reiserfs: fix misspelling of "journaled"
  doc: Fix a typo in slub.txt.
  inotify: remove superfluous return code check
  hdlc: spelling fix in find_pvc() comment
  doc: fix regulator docs cut-and-pasteism
  mtd: Fix comment in Kconfig
  doc: Fix IRQ chip docs
  tree-wide: fix assorted typos all over the place
  drivers/ata/libata-sff.c: comment spelling fixes
  fix typos/grammos in Documentation/edac.txt
  sysctl: add missing comments
  fs/debugfs/inode.c: fix comment typos
  sgivwfb: Make use of ARRAY_SIZE.
  sky2: fix sky2_link_down copy/paste comment error
  tree-wide: fix typos "couter" -> "counter"
  tree-wide: fix typos "offest" -> "offset"
  fix kerneldoc for set_irq_msi()
  spidev: fix double "of of" in comment
  comment typo fix: sybsystem -> subsystem
  ...

11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid
Linus Torvalds [Thu, 10 Dec 2009 03:43:19 +0000 (19:43 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jikos/hid

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid:
  HID: add multi-input quirk for NextWindow Touchscreen.
  HID: add support for Acan FG-8100 barcode reader
  HID: make Media key on Logitech DiNovo Mini work
  HID: support Logitech/3DConnexion SpaceTraveler and SpaceNavigator
  HID: remove BKL from hiddev_ioctl_usage()
  HID: fixup quirk for NCR devices
  HID: pidff - fix unnecessary loop iterations on reset
  HID: add NOGET quirk for another device from CH Products
  HID: remove useless DRIVER_VERSION macro
  HID: fix MODULE_AUTHOR usage in HID modules
  HID: blacklist Acer Ferrari 4005 optical mouse

11 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6
Linus Torvalds [Thu, 10 Dec 2009 03:42:25 +0000 (19:42 -0800)]
Merge git://git./linux/kernel/git/jejb/scsi-misc-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6: (222 commits)
  [SCSI] zfcp: Remove flag ZFCP_STATUS_FSFREQ_TMFUNCNOTSUPP
  [SCSI] zfcp: Activate fc4s attributes for zfcp in FC transport class
  [SCSI] zfcp: Block scsi_eh thread for rport state BLOCKED
  [SCSI] zfcp: Update FSF error reporting
  [SCSI] zfcp: Improve ELS ADISC handling
  [SCSI] zfcp: Simplify handling of ct and els requests
  [SCSI] zfcp: Remove ZFCP_DID_MASK
  [SCSI] zfcp: Move WKA port to zfcp FC code
  [SCSI] zfcp: Use common code definitions for FC CT structs
  [SCSI] zfcp: Use common code definitions for FC ELS structs
  [SCSI] zfcp: Update FCP protocol related code
  [SCSI] zfcp: Dont fail SCSI commands when transitioning to blocked fc_rport
  [SCSI] zfcp: Assign scheduled work to driver queue
  [SCSI] zfcp: Remove STATUS_COMMON_REMOVE flag as it is not required anymore
  [SCSI] zfcp: Implement module unloading
  [SCSI] zfcp: Merge trace code for fsf requests in one function
  [SCSI] zfcp: Access ports and units with container_of in sysfs code
  [SCSI] zfcp: Remove suspend callback
  [SCSI] zfcp: Remove global config_mutex
  [SCSI] zfcp: Replace local reference counting with common kref
  ...

11 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/gerg/m68knommu
Linus Torvalds [Thu, 10 Dec 2009 03:09:42 +0000 (19:09 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/gerg/m68knommu

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/gerg/m68knommu:
  m68knommu: export clk_* symbols in clk.c
  m68knommu: Split the .init section into INIT_TEXT_SECTION and INIT_DATA_SECTION.
  m68knommu: Move __init_end out of the .init section.
  m68knommu: Move __init_begin out of the .init section.
  m68knommu: Use more macros inside the .init section.
  m68knommu: Use INIT_TASK_DATA and CACHELINE_ALIGNED_DATA.
  m68knommu: Make THREAD_SIZE available to assembly files.
  m68knommu: Don't hardcode the value of PAGE_SIZE in the linker script.
  m68knommu: rename BSS define in linker script
  m68knommu: add a task_pt_regs() macro
  m68knommu: define arch_has_single_step() and friends
  m68knommu: add uboot commandline argument passing support
  m68knommu: Coldfire GPIO corrections
  m68knommu: move mcf_remove to .devexit.text

Fixed up (?) conflict in arch/m68k/include/asm/ptrace.h

11 years agoMerge branch 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6
Linus Torvalds [Thu, 10 Dec 2009 03:03:49 +0000 (19:03 -0800)]
Merge branch 'release' of git://git./linux/kernel/git/aegl/linux-2.6

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6:
  [IA64] Fix cut/paste detritus from unistd.h

11 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/lethal/sh-2.6
Linus Torvalds [Thu, 10 Dec 2009 03:03:16 +0000 (19:03 -0800)]
Merge git://git./linux/kernel/git/lethal/sh-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/lethal/sh-2.6: (137 commits)
  sh: include empty zero page in romImage
  sh: Make associative cache writes fatal on all SH-4A parts.
  sh: Drop associative writes for SH-4 cache flushes.
  sh: Partial revert of copy/clear_user_highpage() optimizations.
  sh: Add default uImage rule for se7724, ap325rxa, and migor.
  sh: allow runtime pm without suspend/resume callbacks
  sh: mach-ecovec24: Remove un-defined settings for VPU
  sh: mach-ecovec24: LCDC drive ability become high
  sh: fix sh7724 VEU3F resource size
  serial: sh-sci: Fix too early port disabling.
  sh: pfc: pr_info() -> pr_debug() cleanups.
  sh: pfc: Convert from ctrl_xxx() to __raw_xxx() I/O routines.
  sh: Improve kfr2r09 serial port setup code
  sh: Break out SuperH PFC code
  sh: Move KEYSC header file
  sh: convert /proc/cpu/aligmnent, /proc/cpu/kernel_alignment to seq_file
  sh: Add CPG save/restore code for sh7724 R-standby
  sh: Add SDHI power control support to Ecovec
  mfd: Add power control platform data to SDHI driver
  sh: mach-ecovec24: modify address map
  ...

11 years agoMerge branch 'davinci-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 10 Dec 2009 03:02:46 +0000 (19:02 -0800)]
Merge branch 'davinci-for-linus' of git://git./linux/kernel/git/khilman/linux-davinci

* 'davinci-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/khilman/linux-davinci: (69 commits)
  davinci: Initial support for Neuros OSD2 platform.
  davinci: remove unused variable in arch/arm/mach-davinci/board-sffsdr.c
  davinci: fix section mismatch warning in arch/arm/mach-davinci/board-dm646x-evm.c
  DaVinci: DM365: Enable DaVinci RTC support for DM365 EVM
  DA8xx/OMAP-L1xx: Add high speed SD/MMC capabilities
  davinci: DA8XX/OMAP-L1XX: enable cpuidle and regulator in defconfig
  davinci: DA850/OMAP-L138: avoid using separate initcall for initializing regulator
  davinci: DA850/OMAP-L138 EVM: register for cpuidle support
  davinci: DA8XX/OMAP-L1XX: add support for cpuidle driver register
  davinci: add CPU idle driver
  davinci: DA8XX/OMAP-L1XX: fix compiler warning
  davinci: DA850/OMAP-L138: eliminate static function declaration
  davinci: DA850/OMAP-L138 EVM: simplify configuration of emac in MII/RMII mode
  davinci: DA850/OMAP-L138 EVM: get rid of DA850_UI_EXP config option
  davinci: DA850/OMAP-L138 EVM: implement autodetect of RMII PHY
  davinci: DA830/OMAP-L137 EVM: do not configure NAND on UI card when MMC/SD is selected
  davinci: DA830/OMAP-L137 EVM: use runtime detection for UI card
  davinci: DA830/OMAP-L137 EVM: remove ifdefs inside da830_evm_init()
  davinci: DA830/OMAP-L137 EVM: fix warning with default config
  davinci: Add NAND support for DA830/OMAP-L137 EVM platform
  ...

11 years agoMerge branch 'for-linus' of git://git390.marist.edu/pub/scm/linux-2.6
Linus Torvalds [Thu, 10 Dec 2009 03:01:47 +0000 (19:01 -0800)]
Merge branch 'for-linus' of git://git390.marist.edu/linux-2.6

* 'for-linus' of git://git390.marist.edu/pub/scm/linux-2.6: (72 commits)
  [S390] 3215/3270 console: remove wrong comment
  [S390] dasd: remove BKL from extended error reporting code
  [S390] vmlogrdr: remove BKL
  [S390] vmur: remove BKL
  [S390] zcrypt: remove BKL
  [S390] 3270: remove BKL
  [S390] vmwatchdog: remove lock_kernel() from open() function
  [S390] monwriter: remove lock_kernel() from open() function
  [S390] monreader: remove lock_kernel() from open() function
  [S390] s390: remove unused nfsd #includes
  [S390] ftrace: build ftrace.o when CONFIG_FTRACE_SYSCALLS is set for s390
  [S390] etr/stp: put correct per cpu variable
  [S390] tty3270: move keyboard compat ioctls
  [S390] sclp: improve servicability setting
  [S390] s390: use change recording override for kernel mapping
  [S390] MAINTAINERS: Add s390 drivers block
  [S390] use generic sockios.h header file
  [S390] use generic termbits.h header file
  [S390] smp: remove unused typedef and defines
  [S390] cmm: free pages on hibernate.
  ...

11 years agoDoc: use misc-devices/ dir for drivers
Jean Delvare [Mon, 7 Dec 2009 02:30:44 +0000 (18:30 -0800)]
Doc: use misc-devices/ dir for drivers

We have a directory for misc drivers documentation, let's use it.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: "Darrick J. Wong" <djwong@us.ibm.com>
Cc: Rodolfo Giometti <giometti@linux.it>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoDoc: better explanation of procs_running
Luis Garces-Erice [Mon, 7 Dec 2009 02:30:44 +0000 (18:30 -0800)]
Doc: better explanation of procs_running

the description in Documentation/filesystems/proc.txt of the
procs_running entry in /proc/stat is confusing (according to that
description, it looks as if procs_running could only be a number
between 0 and the number of CPUs).

Changed it to a more accurate description in the patch attached.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agocorrect gpio.txt typos
Dominik D. Geyer [Mon, 7 Dec 2009 02:30:44 +0000 (18:30 -0800)]
correct gpio.txt typos

Corrected sysfs gpio chip node name and fixed punctuation.

Signed-off-by: Dominik D. Geyer <dominik.geyer@gmx.de>
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agodocbook: fix signal_pending() argument
Shawn Bohrer [Mon, 7 Dec 2009 02:30:44 +0000 (18:30 -0800)]
docbook: fix signal_pending() argument

Since signal_pending() takes a task_struct pointer as an argument, update
the example to pass in 'current'.

Signed-off-by: Shawn Bohrer <shawn.bohrer@gmail.com>
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>