Commit Graph

3225 Commits

Author SHA1 Message Date
Linus Torvalds 7f87307818 Just a few md patches for the 3.15 merge window.
-----BEGIN PGP SIGNATURE-----
 Version: GnuPG v2.0.22 (GNU/Linux)
 
 iQIVAwUAU0h5fjnsnt1WYoG5AQKmkw/8DUn0vV4q5UbLp0m2Yy6o6EOxwiSJUH/p
 6EEUmSwyXou75w9OWOJKDX2lI7z1yAtzqiuCQ19ekD5lA6gAosja+D0jKjv0SA01
 rQm2rMjnwOIxZUIRx/7Z+w/H1ZxjIAc8uxOcCBP6DOynWt/YAyVz1SzLLtwCxELL
 N9vjgb+4lVt0E9bBYvVRNiJmtVXpDcmn1YySd6Dqyj9t+Mmysnv8QuIStrT3CE7k
 apss3ew6bBbtbiJHuCno/Q4FDWVAhUH+9GMvksdajw8QW52oHV+RRBB5IpCU6hOx
 OKCT4MVdzmTgi6GRhSr86Dt+KMOLWZmbx7pK7aRQPiL6uFNhqAlJDb/u0xfaHohG
 DiRclZBbsHkEpejHaZcJCkyKFHQTEiia3JVk426FAhtiK1qIBuyxEc65RmKf6dsA
 1KlZVeclD3wYWKG4hWk/0W3qIPOWBMll+Ely5Zg6s2X3gGy9u5TU+tUsfJaL1aDU
 NOY+5D0+hg7o21kK7WgTaP2upexC/iaBrVrdlasM2KYXJVDrsfCAQr1/BwTl4qLq
 Lm/OkIg+WtrQ95RvsI85Hm4PJVxBd1HeyDlKNCcz47kc3Xxqabeq8KnwERyOh4hU
 U4EmAeCZmSGOOETIWQQxlIn8XdM1+dF4olUH9viEAXrQfGgUyrg6Vcc7BOdTTZCY
 8ek3CWG3TwE=
 =kHl3
 -----END PGP SIGNATURE-----

Merge tag 'md/3.15' of git://neil.brown.name/md

Pull md updates from Neil Brown:
 "Just a few md patches for the 3.15 merge window.

  Not much happening in md/raid at the moment.  Just a few bug fixes
  (one for -stable) and a couple of performance tweaks"

* tag 'md/3.15' of git://neil.brown.name/md:
  raid5: get_active_stripe avoids device_lock
  raid5: make_request does less prepare wait
  md: avoid oops on unload if some process is in poll or select.
  md/raid1: r1buf_pool_alloc: free allocate pages when subsequent allocation fails.
  md/bitmap: don't abuse i_writecount for bitmap files.
2014-04-11 17:20:38 -07:00
Shaohua Li e240c1839d raid5: get_active_stripe avoids device_lock
For sequential workload (or request size big workload), get_active_stripe can
find cached stripe. In this case, we always hold device_lock, which exposes a
lot of lock contention for such workload. If stripe count isn't 0, we don't
need hold the lock actually, since we just increase its count. And this is the
hot code path for such workload. Unfortunately we must delete the BUG_ON.

Signed-off-by: Shaohua Li <shli@fusionio.com>
Signed-off-by: NeilBrown <neilb@suse.de>
2014-04-09 14:42:42 +10:00
Shaohua Li 27c0f68f07 raid5: make_request does less prepare wait
In NUMA machine, prepare_to_wait/finish_wait in make_request exposes a
lot of contention for sequential workload (or big request size
workload). For such workload, each bio includes several stripes. So we
can just do prepare_to_wait/finish_wait once for the whold bio instead
of every stripe.  This reduces the lock contention completely for such
workload. Random workload might have the similar lock contention too,
but I didn't see it yet, maybe because my stroage is still not fast
enough.

Signed-off-by: Shaohua Li <shli@fusionio.com>
Signed-off-by: NeilBrown <neilb@suse.de>
2014-04-09 14:42:38 +10:00
NeilBrown e2f23b606b md: avoid oops on unload if some process is in poll or select.
If md-mod is unloaded while some process is in poll() or select(),
then that process maintains a pointer to md_event_waiters, and when
the try to unlink from that list, they will oops.

The procfs infrastructure ensures that ->poll won't be called after
remove_proc_entry, but doesn't provide a wait_queue_head for us to
use, and the waitqueue code doesn't provide a way to remove all
listeners from a waitqueue.

So we need to:
 1/ make sure no further references to md_event_waiters are taken (by
    setting md_unloading)
 2/ wake up all processes currently waiting, and
 3/ wait until all those processes have disconnected from our
    wait_queue_head.

Reported-by: "majianpeng" <majianpeng@gmail.com>
Signed-off-by: NeilBrown <neilb@suse.de>
2014-04-09 14:42:34 +10:00
NeilBrown da1aab3dca md/raid1: r1buf_pool_alloc: free allocate pages when subsequent allocation fails.
When performing a user-request check/repair (MD_RECOVERY_REQUEST is set)
on a raid1, we allocate multiple bios each with their own set of pages.

If the page allocations for one bio fails, we currently do *not* free
the pages allocated for the previous bios, nor do we free the bio itself.

This patch frees all the already-allocate pages, and makes sure that
all the bios are freed as well.

This bug can cause a memory leak which can ultimately OOM a machine.
It was introduced in 3.10-rc1.

Fixes: a07876064a
Cc: Kent Overstreet <koverstreet@google.com>
Cc: stable@vger.kernel.org (3.10+)
Reported-by: Russell King - ARM Linux <linux@arm.linux.org.uk>
Signed-off-by: NeilBrown <neilb@suse.de>
2014-04-09 14:42:23 +10:00
NeilBrown 035328c202 md/bitmap: don't abuse i_writecount for bitmap files.
md bitmap code currently tries to use i_writecount to stop any other
process from writing to out bitmap file.  But that is really an abuse
and has bit-rotted so locking is all wrong.

So discard that - root should be allowed to shoot self in foot.

Still use it in a much less intrusive way to stop the same file being
used as bitmap on two different array, and apply other checks to
ensure the file is at least vaguely usable for bitmap storage
(is regular, is open for write.  Support for ->bmap is already checked
elsewhere).

Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: NeilBrown <neilb@suse.de>
2014-04-09 12:26:59 +10:00
Linus Torvalds 04535d273e . Fix dm-cache corruption caused by discard_block_size >
cache_block_size
 
 . Fix a lock-inversion detected by LOCKDEP in dm-cache
 
 . Fix a dangling bio bug in the dm-thinp target's process_deferred_bios
   error path
 
 . Fix corruption due to non-atomic transaction commit which allowed a
   metadata superblock to be written before all other metadata was
   successfully written -- this is common to all targets that use the
   persistent-data library's transaction manager (dm-thinp, dm-cache and
   dm-era).
 
 . Various small cleanups in the DM core
 
 . Add the dm-era target which is useful for keeping track of which
   blocks were written within a user defined period of time called an
   'era'.  Use cases include tracking changed blocks for backup software,
   and partially invalidating the contents of a cache to restore cache
   coherency after rolling back a vendor snapshot.
 
 . Improve the on-disk layout of multithreaded writes to the dm-thin-pool
   by splitting the pool's deferred bio list to be a per-thin device list
   and then sorting that list using an rb_tree.  The subsequent read
   throughput of the data written via multiple threads improved by ~70%.
 
 . Simplify the multipath target's handling of queuing IO by pushing
   requests back to the request queue rather than queueing the IO
   internally.
 -----BEGIN PGP SIGNATURE-----
 Version: GnuPG v1
 
 iQEcBAABAgAGBQJTPv/6AAoJEMUj8QotnQNagQYH/3EkB2f66TRfjRQpVAZuchw/
 U0IbVWcMJKMdhj3uaSNzIkAbTgF+QsZUOLHP/7Q6zLq0M2J3WGrJn2ELqq53MenF
 E0+rJ8duKnJ5oLhhVT62ukBDh3XHWT0JyijXPWNa2gUoYwJqM9BAlXbC/OTfUNaZ
 mBCxvUWGME8k3ht310GhwvzBQjYuxIXhw8XlbGvakb9S83SZwNpCh231iumOEzPe
 Vzfx/xTto0fH3R5/knNV/H9xt0Dv4vt4Aqbqqys9UbQvPzj9qN/mxUZIFg+LZh/w
 WuvHHw6HcAiNNrQGFcm6i1AK2jJ+F61K3afMlYsiamTxMNM+0q/B9HemkX/0ieU=
 =lY8m
 -----END PGP SIGNATURE-----

Merge tag 'dm-3.15-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm

Pull device mapper changes from Mike Snitzer:

 - Fix dm-cache corruption caused by discard_block_size > cache_block_size

 - Fix a lock-inversion detected by LOCKDEP in dm-cache

 - Fix a dangling bio bug in the dm-thinp target's process_deferred_bios
   error path

 - Fix corruption due to non-atomic transaction commit which allowed a
   metadata superblock to be written before all other metadata was
   successfully written -- this is common to all targets that use the
   persistent-data library's transaction manager (dm-thinp, dm-cache and
   dm-era).

 - Various small cleanups in the DM core

 - Add the dm-era target which is useful for keeping track of which
   blocks were written within a user defined period of time called an
   'era'.  Use cases include tracking changed blocks for backup
   software, and partially invalidating the contents of a cache to
   restore cache coherency after rolling back a vendor snapshot.

 - Improve the on-disk layout of multithreaded writes to the
   dm-thin-pool by splitting the pool's deferred bio list to be a
   per-thin device list and then sorting that list using an rb_tree.
   The subsequent read throughput of the data written via multiple
   threads improved by ~70%.

 - Simplify the multipath target's handling of queuing IO by pushing
   requests back to the request queue rather than queueing the IO
   internally.

* tag 'dm-3.15-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm: (24 commits)
  dm cache: fix a lock-inversion
  dm thin: sort the per thin deferred bios using an rb_tree
  dm thin: use per thin device deferred bio lists
  dm thin: simplify pool_is_congested
  dm thin: fix dangling bio in process_deferred_bios error path
  dm mpath: print more useful warnings in multipath_message()
  dm-mpath: do not activate failed paths
  dm mpath: remove extra nesting in map function
  dm mpath: remove map_io()
  dm mpath: reduce memory pressure when requeuing
  dm mpath: remove process_queued_ios()
  dm mpath: push back requests instead of queueing
  dm table: add dm_table_run_md_queue_async
  dm mpath: do not call pg_init when it is already running
  dm: use RCU_INIT_POINTER instead of rcu_assign_pointer in __unbind
  dm: stop using bi_private
  dm: remove dm_get_mapinfo
  dm: make dm_table_alloc_md_mempools static
  dm: take care to copy the space map roots before locking the superblock
  dm transaction manager: fix corruption due to non-atomic transaction commit
  ...
2014-04-05 18:49:31 -07:00
Joe Thornber 0596661f0a dm cache: fix a lock-inversion
When suspending a cache the policy is walked and the individual policy
hints written to the metadata via sync_metadata().  This led to this
lock order:

      policy->lock
        cache_metadata->root_lock

When loading the cache target the policy is populated while the metadata
lock is held:

      cache_metadata->root_lock
         policy->lock

Fix this potential lock-inversion (ABBA) deadlock in sync_metadata() by
ensuring the cache_metadata root_lock is held whilst all the hints are
written, rather than being repeatedly locked while policy->lock is held
(as was the case with each callout that policy_walk_mappings() made to
the old save_hint() method).

Found by turning on the CONFIG_PROVE_LOCKING ("Lock debugging: prove
locking correctness") build option.  However, it is not clear how the
LOCKDEP reported paths can lead to a deadlock since the two paths,
suspending a target and loading a target, never occur at the same time.
But that doesn't mean the same lock-inversion couldn't have occurred
elsewhere.

Reported-by: Marian Csontos <mcsontos@redhat.com>
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org
2014-04-04 14:53:05 -04:00
Mike Snitzer 67324ea188 dm thin: sort the per thin deferred bios using an rb_tree
A thin-pool will allocate blocks using FIFO order for all thin devices
which share the thin-pool.  Because of this simplistic allocation the
thin-pool's space can become fragmented quite easily; especially when
multiple threads are requesting blocks in parallel.

Sort each thin device's deferred_bio_list based on logical sector to
help reduce fragmentation of the thin-pool's ondisk layout.

The following tables illustrate the realized gains/potential offered by
sorting each thin device's deferred_bio_list.  An "io size"-sized random
read of the device would result in "seeks/io" fragments being read, with
an average "distance/seek" between each fragment.

Data was written to a single thin device using multiple threads via
iozone (8 threads, 64K for both the block_size and io_size).

unsorted:

     io size   seeks/io distance/seek
  --------------------------------------
          4k    0.000   0b
         16k    0.013   11m
         64k    0.065   11m
        256k    0.274   10m
          1m    1.109   10m
          4m    4.411   10m
         16m    17.097  11m
         64m    60.055  13m
        256m    148.798 25m
          1g    809.929 21m

sorted:

     io size   seeks/io distance/seek
  --------------------------------------
          4k    0.000   0b
         16k    0.000   1g
         64k    0.001   1g
        256k    0.003   1g
          1m    0.011   1g
          4m    0.045   1g
         16m    0.181   1g
         64m    0.747   1011m
        256m    3.299   1g
          1g    14.373  1g

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <ejt@redhat.com>
2014-04-04 14:53:03 -04:00
Linus Torvalds b33ce44299 Merge branch 'for-3.15/drivers' of git://git.kernel.dk/linux-block
Pull block driver update from Jens Axboe:
 "On top of the core pull request, here's the pull request for the
  driver related changes for 3.15.  It contains:

   - Improvements for msi-x registration for block drivers (mtip32xx,
     skd, cciss, nvme) from Alexander Gordeev.

   - A round of cleanups and improvements for drbd from Andreas
     Gruenbacher and Rashika Kheria.

   - A round of clanups and improvements for bcache from Kent.

   - Removal of sleep_on() and friends in DAC960, ataflop, swim3 from
     Arnd Bergmann.

   - Bug fix for a bug in the mtip32xx async completion code from Sam
     Bradshaw.

   - Bug fix for accidentally bouncing IO on 32-bit platforms with
     mtip32xx from Felipe Franciosi"

* 'for-3.15/drivers' of git://git.kernel.dk/linux-block: (103 commits)
  bcache: remove nested function usage
  bcache: Kill bucket->gc_gen
  bcache: Kill unused freelist
  bcache: Rework btree cache reserve handling
  bcache: Kill btree_io_wq
  bcache: btree locking rework
  bcache: Fix a race when freeing btree nodes
  bcache: Add a real GC_MARK_RECLAIMABLE
  bcache: Add bch_keylist_init_single()
  bcache: Improve priority_stats
  bcache: Better alloc tracepoints
  bcache: Kill dead cgroup code
  bcache: stop moving_gc marking buckets that can't be moved.
  bcache: Fix moving_pred()
  bcache: Fix moving_gc deadlocking with a foreground write
  bcache: Fix discard granularity
  bcache: Fix another bug recovering from unclean shutdown
  bcache: Fix a bug recovering from unclean shutdown
  bcache: Fix a journalling reclaim after recovery bug
  bcache: Fix a null ptr deref in journal replay
  ...
2014-04-01 19:43:53 -07:00
Linus Torvalds 675c354a95 Char/Misc driver patches for 3.15-rc1
Here's the big char/misc driver updates for 3.15-rc1.
 
 Lots of various things here, including the new mcb driver subsystem.
 
 All of these have been in linux-next for a while.
 
 Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
 -----BEGIN PGP SIGNATURE-----
 Version: GnuPG v2.0.22 (GNU/Linux)
 
 iEYEABECAAYFAlM7ArIACgkQMUfUDdst+ylS+gCfcJr0Zo2v5aWnqD7rFtFETmFI
 LhcAoNTQ4cvlVdxnI0driWCWFYxLj6at
 =aj+L
 -----END PGP SIGNATURE-----

Merge tag 'char-misc-3.15-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc

Pull char/misc driver patches from Greg KH:
 "Here's the big char/misc driver updates for 3.15-rc1.

  Lots of various things here, including the new mcb driver subsystem.

  All of these have been in linux-next for a while"

* tag 'char-misc-3.15-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc: (118 commits)
  extcon: Move OF helper function to extcon core and change function name
  extcon: of: Remove unnecessary function call by using the name of device_node
  extcon: gpio: Use SIMPLE_DEV_PM_OPS macro
  extcon: palmas: Use SIMPLE_DEV_PM_OPS macro
  mei: don't use deprecated DEFINE_PCI_DEVICE_TABLE macro
  mei: amthif: fix checkpatch error
  mei: client.h fix checkpatch errors
  mei: use cl_dbg where appropriate
  mei: fix Unnecessary space after function pointer name
  mei: report consistently copy_from/to_user failures
  mei: drop pr_fmt macros
  mei: make me hw headers private to me hw.
  mei: fix memory leak of pending write cb objects
  mei: me: do not reset when less than expected data is received
  drivers: mcb: Fix build error discovered by 0-day bot
  cs5535-mfgpt: Simplify dependencies
  spmi: pm: drop bus-level PM suspend/resume routines
  spmi: pmic_arb: make selectable on ARCH_QCOM
  Drivers: hv: vmbus: Increase the limit on the number of pfns we can handle
  pch_phub: Report error writing MAC back to user
  ...
2014-04-01 16:13:21 -07:00
Mike Snitzer c140e1c4e2 dm thin: use per thin device deferred bio lists
The thin-pool previously only had a single deferred_bios list that would
collect bios for all thin devices in the pool.  Split this per-pool
deferred_bios list out to per-thin deferred_bios_list -- doing so
enables increased parallelism when processing deferred bios.  And now
that each thin device has it's own deferred_bios_list we can sort all
bios in the list using logical sector.  The requeue code in error
handling path is also cleaner as a side-effect.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <ejt@redhat.com>
2014-03-31 14:14:15 -04:00
Mike Snitzer 760fe67e53 dm thin: simplify pool_is_congested
The pool is congested if the pool is in PM_OUT_OF_DATA_SPACE mode.  This
is more explicit/clear/efficient than inferring whether or not the pool
is congested by checking if retry_on_resume_list is empty.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <ejt@redhat.com>
2014-03-31 10:05:51 -04:00
Mike Snitzer fe76cd88e6 dm thin: fix dangling bio in process_deferred_bios error path
If unable to ensure_next_mapping() we must add the current bio, which
was removed from the @bios list via bio_list_pop, back to the
deferred_bios list before all the remaining @bios.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <ejt@redhat.com>
Cc: stable@vger.kernel.org
2014-03-28 14:37:02 -04:00
Jose Castillo a356e42620 dm mpath: print more useful warnings in multipath_message()
The warning message "Unrecognised multipath message received" is
displayed in two different situations in multipath_message(): when the
number of arguments passed is invalid and when the string passed in
argv[0] is not recognized.

Make it easier to identify where the problem is by making these warnings
more specific with additional context for each case.

Signed-off-by: Jose Castillo <jcastillo@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-03-27 16:56:25 -04:00
Hannes Reinecke 3a01750964 dm-mpath: do not activate failed paths
activate_path() is run without a lock, so the path might be
set to failed before activate_path() had a chance to run.
This patch add a check for ->active in activate_path() to
avoid unnecessary overhead by calling functions which are known
to be failing.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-03-27 16:56:25 -04:00
Mike Snitzer 9bf59a611a dm mpath: remove extra nesting in map function
Return early for case when no path exists, and when the
pathgroup isn't ready. This eliminates the need for
extra nesting for the the common case.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Hannes Reinecke <hare@suse.de>
2014-03-27 16:56:25 -04:00
Hannes Reinecke 36fcffcc65 dm mpath: remove map_io()
multipath_map() is now just a wrapper around map_io(), so we
can rename map_io() to multipath_map().

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Reviewed-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
2014-03-27 16:56:25 -04:00
Hannes Reinecke e3bde04f1e dm mpath: reduce memory pressure when requeuing
When multipath needs to requeue I/O in the block layer the per-request
context shouldn't be allocated, as it will be freed immediately
afterwards anyway.  Avoiding this memory allocation will reduce memory
pressure during requeuing.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Reviewed-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
2014-03-27 16:56:25 -04:00
Hannes Reinecke 3e9f1be1b4 dm mpath: remove process_queued_ios()
process_queued_ios() has served 3 functions:
  1) select pg and pgpath if none is selected
  2) start pg_init if requested
  3) dispatch queued IOs when pg is ready

Basically, a call to queue_work(process_queued_ios) can be replaced by
dm_table_run_md_queue_async(), which runs request queue and ends up
calling map_io(), which does 1), 2) and 3).

Exception is when !pg_ready() (which means either pg_init is running or
requested), then multipath_busy() prevents map_io() being called from
request_fn.

If pg_init is running, it should be ok as long as pg_init_done() does
the right thing when pg_init is completed, I.e.: restart pg_init if
!pg_ready() or call dm_table_run_md_queue_async() to kick map_io().

If pg_init is requested, we have to make sure the request is detected
and pg_init will be started.  pg_init is requested in 3 places:
  a) __choose_pgpath() in map_io()
  b) __choose_pgpath() in multipath_ioctl()
  c) pg_init retry in pg_init_done()
a) is ok because map_io() calls __pg_init_all_paths(), which does 2).
b) needs a call to __pg_init_all_paths(), which does 2).
c) needs a call to __pg_init_all_paths(), which does 2).

So this patch removes process_queued_ios() and ensures that
__pg_init_all_paths() is called at the appropriate locations.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Reviewed-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
2014-03-27 16:56:24 -04:00
Hannes Reinecke e809917735 dm mpath: push back requests instead of queueing
There is no reason why multipath needs to queue requests internally for
queue_if_no_path or pg_init; we should rather push them back onto the
request queue.

And while we're at it we can simplify the conditional statement in
map_io() to make it easier to read.

Since mpath no longer does internal queuing of I/O the table info no
longer emits the internal queue_size.  Instead it displays 1 if queuing
is being used or 0 if it is not.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Reviewed-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
2014-03-27 16:56:24 -04:00
Mike Snitzer 9974fa2c6a dm table: add dm_table_run_md_queue_async
Introduce dm_table_run_md_queue_async() to run the request_queue of the
mapped_device associated with a request-based DM table.

Also add dm_md_get_queue() wrapper to extract the request_queue from a
mapped_device.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
2014-03-27 16:56:24 -04:00
Hannes Reinecke 17f4ff45b5 dm mpath: do not call pg_init when it is already running
This patch moves condition checks as a preparation of following
patches and has no effect on behaviour.
process_queued_ios() is the only caller of __pg_init_all_paths()
and 2 condition checks are moved from outside to inside without
side effects.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Reviewed-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
2014-03-27 16:56:24 -04:00
Monam Agarwal 9cdb852004 dm: use RCU_INIT_POINTER instead of rcu_assign_pointer in __unbind
Replace rcu_assign_pointer(p, NULL) with RCU_INIT_POINTER(p, NULL).

The rcu_assign_pointer() ensures that the initialization of a structure
is carried out before storing a pointer to that structure.  And in the
case of the NULL pointer, there is no structure to initialize.  So,
rcu_assign_pointer(p, NULL) can be safely converted to
RCU_INIT_POINTER(p, NULL).

Signed-off-by: Monam Agarwal <monamagarwal123@gmail.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-03-27 16:56:24 -04:00
Mikulas Patocka bfc6d41cee dm: stop using bi_private
Device mapper uses the bio structure's bi_private field as a pointer
to dm_target_io or dm_rq_clone_bio_info.  But a bio structure is
embedded in the dm_target_io and dm_rq_clone_bio_info structures, so the
pointer to the structure that contains the bio can be found with the
container_of() macro.

Remove the use of bi_private and use container_of() instead.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-03-27 16:56:24 -04:00
Mikulas Patocka d70ab4fb72 dm: remove dm_get_mapinfo
Remove dm_get_mapinfo() because no target uses it.  Targets can allocate
per-bio data using ti->per_bio_data_size, this is much more flexible
than union map_info.

Leave union map_info only for the request-based multipath target's use.
Also delete the unused "unsigned long long ll" field of union map_info.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-03-27 16:56:24 -04:00
Mikulas Patocka 473c36dfee dm: make dm_table_alloc_md_mempools static
Make the function dm_table_alloc_md_mempools static because it is not
called from another file.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-03-27 16:56:23 -04:00
Joe Thornber 5a32083d03 dm: take care to copy the space map roots before locking the superblock
In theory copying the space map root can fail, but in practice it never
does because we're careful to check what size buffer is needed.

But make certain we're able to copy the space map roots before
locking the superblock.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org # drop dm-era and dm-cache changes as needed
2014-03-27 16:56:23 -04:00
Joe Thornber a9d45396f5 dm transaction manager: fix corruption due to non-atomic transaction commit
The persistent-data library used by dm-thin, dm-cache, etc is
transactional.  If anything goes wrong, such as an io error when writing
new metadata or a power failure, then we roll back to the last
transaction.

Atomicity when committing a transaction is achieved by:

a) Never overwriting data from the previous transaction.
b) Writing the superblock last, after all other metadata has hit the
   disk.

This commit and the following commit ("dm: take care to copy the space
map roots before locking the superblock") fix a bug associated with (b).
When committing it was possible for the superblock to still be written
in spite of an io error occurring during the preceeding metadata flush.
With these commits we're careful not to take the write lock out on the
superblock until after the metadata flush has completed.

Change the transaction manager's semantics for dm_tm_commit() to assume
all data has been flushed _before_ the single superblock that is passed
in.

As a prerequisite, split the block manager's block unlocking and
flushing by simplifying dm_bm_flush_and_unlock() to dm_bm_flush().  Now
the unlocking must be done separately.

This issue was discovered by forcing io errors at the crucial time
using dm-flakey.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org
2014-03-27 16:56:23 -04:00
Heinz Mauelshagen 64ab346a36 dm cache: remove remainder of distinct discard block size
Discard block size not being equal to cache block size causes data
corruption by erroneously avoiding migrations in issue_copy() because
the discard state is being cleared for a group of cache blocks when it
should not.

Completely remove all code that enabled a distinction between the
cache block size and discard block size.

Signed-off-by: Heinz Mauelshagen <heinzm@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-03-27 16:56:23 -04:00
Mike Snitzer d132cc6d9e dm cache: prevent corruption caused by discard_block_size > cache_block_size
If the discard block size is larger than the cache block size we will
not properly quiesce IO to a region that is about to be discarded.  This
results in a race between a cache migration where no copy is needed, and
a write to an adjacent cache block that's within the same large discard
block.

Workaround this by limiting the discard_block_size to cache_block_size.
Also limit the max_discard_sectors to cache_block_size.

A more comprehensive fix that introduces range locking support in the
bio_prison and proper quiescing of a discard range that spans multiple
cache blocks is already in development.

Reported-by: Morgan Mears <Morgan.Mears@netapp.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <ejt@redhat.com>
Acked-by: Heinz Mauelshagen <heinzm@redhat.com>
Cc: stable@vger.kernel.org
2014-03-27 16:56:23 -04:00
Joe Thornber 428e469864 dm bitset: only flush the current word if it has been dirtied
This change offers a big performance boost for dm-era.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-03-27 16:56:23 -04:00
Joe Thornber eec40579d8 dm: add era target
dm-era is a target that behaves similar to the linear target.  In
addition it keeps track of which blocks were written within a user
defined period of time called an 'era'.  Each era target instance
maintains the current era as a monotonically increasing 32-bit
counter.

Use cases include tracking changed blocks for backup software, and
partially invalidating the contents of a cache to restore cache
coherency after rolling back a vendor snapshot.

dm-era is primarily expected to be paired with the dm-cache target.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-03-27 16:56:23 -04:00
John Sheu cb85114956 bcache: remove nested function usage
Uninlined nested functions can cause crashes when using ftrace, as they don't
follow the normal calling convention and confuse the ftrace function graph
tracer as it examines the stack.

Also, nested functions are supported as a gcc extension, but may fail on other
compilers (e.g. llvm).

Signed-off-by: John Sheu <john.sheu@gmail.com>
2014-03-18 12:39:28 -07:00
Kent Overstreet 3a2fd9d509 bcache: Kill bucket->gc_gen
gc_gen was a temporary used to recalculate last_gc, but since we only need
bucket->last_gc when gc isn't running (gc_mark_valid = 1), we can just update
last_gc directly.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
2014-03-18 12:24:54 -07:00
Kent Overstreet 2531d9ee61 bcache: Kill unused freelist
This was originally added as at optimization that for various reasons isn't
needed anymore, but it does add a lot of nasty corner cases (and it was
responsible for some recently fixed bugs). Just get rid of it now.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
2014-03-18 12:23:36 -07:00
Kent Overstreet 0a63b66db5 bcache: Rework btree cache reserve handling
This changes the bucket allocation reserves to use _real_ reserves - separate
freelists - instead of watermarks, which if nothing else makes the current code
saner to reason about and is going to be important in the future when we add
support for multiple btrees.

It also adds btree_check_reserve(), which checks (and locks) the reserves for
both bucket allocation and memory allocation for btree nodes; the old code just
kinda sorta assumed that since (e.g. for btree node splits) it had the root
locked and that meant no other threads could try to make use of the same
reserve; this technically should have been ok for memory allocation (we should
always have a reserve for memory allocation (the btree node cache is used as a
reserve and we preallocate it)), but multiple btrees will mean that locking the
root won't be sufficient anymore, and for the bucket allocation reserve it was
technically possible for the old code to deadlock.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
2014-03-18 12:23:35 -07:00
Kent Overstreet 56b30770b2 bcache: Kill btree_io_wq
With the locking rework in the last patch, this shouldn't be needed anymore -
btree_node_write_work() only takes b->write_lock which is never held for very
long.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
2014-03-18 12:23:35 -07:00
Kent Overstreet 2a285686c1 bcache: btree locking rework
Add a new lock, b->write_lock, which is required to actually modify - or write -
a btree node; this lock is only held for short durations.

This means we can write out a btree node without taking b->lock, which _is_ held
for long durations - solving a deadlock when btree_flush_write() (from the
journalling code) is called with a btree node locked.

Right now just occurs in bch_btree_set_root(), but with an upcoming journalling
rework is going to happen a lot more.

This also turns b->lock is now more of a read/intent lock instead of a
read/write lock - but not completely, since it still blocks readers. May turn it
into a real intent lock at some point in the future.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
2014-03-18 12:23:35 -07:00
Kent Overstreet 05335cff9f bcache: Fix a race when freeing btree nodes
This isn't a bulletproof fix; btree_node_free() -> bch_bucket_free() puts the
bucket on the unused freelist, where it can be reused right away without any
ordering requirements. It would be better to wait on at least a journal write to
go down before reusing the bucket. bch_btree_set_root() does this, and inserting
into non leaf nodes is completely synchronous so we should be ok, but future
patches are just going to get rid of the unused freelist - it was needed in the
past for various reasons but shouldn't be anymore.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
2014-03-18 12:23:34 -07:00
Kent Overstreet 4fe6a81670 bcache: Add a real GC_MARK_RECLAIMABLE
This means the garbage collection code can better check for data and metadata
pointers to the same buckets.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
2014-03-18 12:22:36 -07:00
Kent Overstreet c13f3af924 bcache: Add bch_keylist_init_single()
This will potentially save us an allocation when we've got inode/dirent bkeys
that don't fit in the keylist's inline keys.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
2014-03-18 12:22:36 -07:00
Kent Overstreet 1575402052 bcache: Improve priority_stats
Break down data into clean data/dirty data/metadata.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
2014-03-18 12:22:35 -07:00
Kent Overstreet 7159b1ad3d bcache: Better alloc tracepoints
Change the invalidate tracepoint to indicate how much data we're invalidating,
and change the alloc tracepoints to indicate what offset they're for.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
2014-03-18 12:22:35 -07:00
Kent Overstreet 3f5e0a34da bcache: Kill dead cgroup code
This hasn't been used or even enabled in ages.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
2014-03-18 12:22:35 -07:00
Nicholas Swenson 3f6ef38110 bcache: stop moving_gc marking buckets that can't be moved.
Signed-off-by: Nicholas Swenson <nks@daterainc.com>
2014-03-18 12:22:34 -07:00
Kent Overstreet 10d9dcf6ee bcache: Fix moving_pred()
Avoid a potential null pointer deref (e.g. from check keys for cache misses)

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
2014-03-18 12:22:34 -07:00
Nicholas Swenson da415a096f bcache: Fix moving_gc deadlocking with a foreground write
Deadlock happened because a foreground write slept, waiting for a bucket
to be allocated. Normally the gc would mark buckets available for invalidation.
But the moving_gc was stuck waiting for outstanding writes to complete.
These writes used the bcache_wq, the same queue foreground writes used.

This fix gives moving_gc its own work queue, so it was still finish moving
even if foreground writes are stuck waiting for allocation. It also makes
work queue a parameter to the data_insert path, so moving_gc can use its
workqueue for writes.

Signed-off-by: Nicholas Swenson <nks@daterainc.com>
Signed-off-by: Kent Overstreet <kmo@daterainc.com>
2014-03-18 12:22:33 -07:00
Kent Overstreet 90db6919f5 bcache: Fix discard granularity
blk_stack_limits() doesn't like a discard granularity of 0.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
2014-03-18 12:22:33 -07:00
Kent Overstreet 487dded86e bcache: Fix another bug recovering from unclean shutdown
The on disk bucket gens are allowed to be out of date, when we reuse buckets
that didn't have any live data in them. To deal with this, the initial gc has to
update the bucket gen when we find a pointer gen newer than the bucket's gen.

Unfortunately we weren't doing this for pointers in the journal that we're about
to replay.

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
2014-03-18 12:22:33 -07:00