linux-toradex.git/drivers/md/dm-bufio.c, branch v4.12-rc1

mm, vmalloc: use __GFP_HIGHMEM implicitly

2017-05-09T00:15:13+00:00

__vmalloc* allows users to provide gfp flags for the underlying
allocation.  This API is quite popular

  $ git grep "=[[:space:]]__vmalloc\|return[[:space:]]*__vmalloc" | wc -l
  77

The only problem is that many people are not aware that they really want
to give __GFP_HIGHMEM along with other flags because there is really no
reason to consume precious lowmemory on CONFIG_HIGHMEM systems for pages
which are mapped to the kernel vmalloc space.  About half of users don't
use this flag, though.  This signals that we make the API unnecessarily
too complex.

This patch simply uses __GFP_HIGHMEM implicitly when allocating pages to
be mapped to the vmalloc space.  Current users which add __GFP_HIGHMEM
are simplified and drop the flag.

Link: http://lkml.kernel.org/r/20170307141020.29107-1-mhocko@kernel.org
Signed-off-by: Michal Hocko 
Reviewed-by: Matthew Wilcox 
Cc: Al Viro 
Cc: Vlastimil Babka 
Cc: David Rientjes 
Cc: Cristopher Lameter 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

dm bufio: check new buffer allocation watermark every 30 seconds

2017-05-01T19:21:42+00:00

dm-bufio checks a watermark when it allocates a new buffer in
__bufio_new().  However, it doesn't check the watermark when the user
changes /sys/module/dm_bufio/parameters/max_cache_size_bytes.

This may result in a problem - if the watermark is high enough so that
all possible buffers are allocated and if the user lowers the value of
"max_cache_size_bytes", the watermark will never be checked against the
new value because no new buffer would be allocated.

To fix this, change __evict_old_buffers() so that it checks the
watermark.  __evict_old_buffers() is called every 30 seconds, so if the
user reduces "max_cache_size_bytes", dm-bufio will react to this change
within 30 seconds and decrease memory consumption.

Depends-on: 1b0fb5a5b2 ("dm bufio: avoid a possible ABBA deadlock")
Cc: stable@vger.kernel.org
Signed-off-by: Mikulas Patocka 
Signed-off-by: Mike Snitzer

dm bufio: avoid a possible ABBA deadlock

2017-05-01T19:18:13+00:00

__get_memory_limit() tests if dm_bufio_cache_size changed and calls
__cache_size_refresh() if it did.  It takes dm_bufio_clients_lock while
it already holds the client lock.  However, lock ordering is violated
because in cleanup_old_buffers() dm_bufio_clients_lock is taken before
the client lock.

This results in a possible deadlock and lockdep engine warning.

Fix this deadlock by changing mutex_lock() to mutex_trylock().  If the
lock can't be taken, it will be re-checked next time when a new buffer
is allocated.

Also add "unlikely" to the if condition, so that the optimizer assumes
that the condition is false.

Cc: stable@vger.kernel.org
Signed-off-by: Mikulas Patocka 
Signed-off-by: Mike Snitzer

dm bufio: add sector start offset to dm-bufio interface

2017-03-07T18:28:33+00:00

Introduce dm_bufio_set_sector_offset() interface to allow setting a
sector offset for a dm-bufio client.  This is a prereq for the DM
integrity target.

Signed-off-by: Mikulas Patocka 
Signed-off-by: Milan Broz 
Signed-off-by: Mike Snitzer

sched/headers: Prepare to move the memalloc_noio_*() APIs to

2017-03-02T07:42:33+00:00

Update the .c files that depend on these APIs.

Acked-by: Linus Torvalds 
Cc: Mike Galbraith 
Cc: Peter Zijlstra 
Cc: Thomas Gleixner 
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar

sched/core: Remove set_task_state()

2017-01-14T10:14:16+00:00

This is a nasty interface and setting the state of a foreign task must
not be done. As of the following commit:

  be628be0956 ("bcache: Make gc wakeup sane, remove set_task_state()")

... everyone in the kernel calls set_task_state() with current, allowing
the helper to be removed.

However, as the comment indicates, it is still around for those archs
where computing current is more expensive than using a pointer, at least
in theory. An important arch that is affected is arm64, however this has
been addressed now [1] and performance is up to par making no difference
with either calls.

Of all the callers, if any, it's the locking bits that would care most
about this -- ie: we end up passing a tsk pointer to a lot of the lock
slowpath, and setting ->state on that. The following numbers are based
on two tests: a custom ad-hoc microbenchmark that just measures
latencies (for ~65 million calls) between get_task_state() vs
get_current_state().

Secondly for a higher overview, an unlink microbenchmark was used,
which pounds on a single file with open, close,unlink combos with
increasing thread counts (up to 4x ncpus). While the workload is quite
unrealistic, it does contend a lot on the inode mutex or now rwsem.

[1] https://lkml.kernel.org/r/1483468021-8237-1-git-send-email-mark.rutland@arm.com

== 1. x86-64 ==

Avg runtime set_task_state():    601 msecs
Avg runtime set_current_state(): 552 msecs

                                            vanilla                 dirty
Hmean    unlink1-processes-2      36089.26 (  0.00%)    38977.33 (  8.00%)
Hmean    unlink1-processes-5      28555.01 (  0.00%)    29832.55 (  4.28%)
Hmean    unlink1-processes-8      37323.75 (  0.00%)    44974.57 ( 20.50%)
Hmean    unlink1-processes-12     43571.88 (  0.00%)    44283.01 (  1.63%)
Hmean    unlink1-processes-21     34431.52 (  0.00%)    38284.45 ( 11.19%)
Hmean    unlink1-processes-30     34813.26 (  0.00%)    37975.17 (  9.08%)
Hmean    unlink1-processes-48     37048.90 (  0.00%)    39862.78 (  7.59%)
Hmean    unlink1-processes-79     35630.01 (  0.00%)    36855.30 (  3.44%)
Hmean    unlink1-processes-110    36115.85 (  0.00%)    39843.91 ( 10.32%)
Hmean    unlink1-processes-141    32546.96 (  0.00%)    35418.52 (  8.82%)
Hmean    unlink1-processes-172    34674.79 (  0.00%)    36899.21 (  6.42%)
Hmean    unlink1-processes-203    37303.11 (  0.00%)    36393.04 ( -2.44%)
Hmean    unlink1-processes-224    35712.13 (  0.00%)    36685.96 (  2.73%)

== 2. ppc64le ==

Avg runtime set_task_state():  938 msecs
Avg runtime set_current_state: 940 msecs

                                            vanilla                 dirty
Hmean    unlink1-processes-2      19269.19 (  0.00%)    30704.50 ( 59.35%)
Hmean    unlink1-processes-5      20106.15 (  0.00%)    21804.15 (  8.45%)
Hmean    unlink1-processes-8      17496.97 (  0.00%)    17243.28 ( -1.45%)
Hmean    unlink1-processes-12     14224.15 (  0.00%)    17240.21 ( 21.20%)
Hmean    unlink1-processes-21     14155.66 (  0.00%)    15681.23 ( 10.78%)
Hmean    unlink1-processes-30     14450.70 (  0.00%)    15995.83 ( 10.69%)
Hmean    unlink1-processes-48     16945.57 (  0.00%)    16370.42 ( -3.39%)
Hmean    unlink1-processes-79     15788.39 (  0.00%)    14639.27 ( -7.28%)
Hmean    unlink1-processes-110    14268.48 (  0.00%)    14377.40 (  0.76%)
Hmean    unlink1-processes-141    14023.65 (  0.00%)    16271.69 ( 16.03%)
Hmean    unlink1-processes-172    13417.62 (  0.00%)    16067.55 ( 19.75%)
Hmean    unlink1-processes-203    15293.08 (  0.00%)    15440.40 (  0.96%)
Hmean    unlink1-processes-234    13719.32 (  0.00%)    16190.74 ( 18.01%)
Hmean    unlink1-processes-265    16400.97 (  0.00%)    16115.22 ( -1.74%)
Hmean    unlink1-processes-296    14388.60 (  0.00%)    16216.13 ( 12.70%)
Hmean    unlink1-processes-320    15771.85 (  0.00%)    15905.96 (  0.85%)

x86-64 (known to be fast for get_current()/this_cpu_read_stable() caching)
and ppc64 (with paca) show similar improvements in the unlink microbenches.
The small delta for ppc64 (2ms), does not represent the gains on the unlink
runs. In the case of x86, there was a decent amount of variation in the
latency runs, but always within a 20 to 50ms increase), ppc was more constant.

Signed-off-by: Davidlohr Bueso 
Signed-off-by: Peter Zijlstra (Intel) 
Cc: Andrew Morton 
Cc: Linus Torvalds 
Cc: Paul E. McKenney 
Cc: Peter Zijlstra 
Cc: Thomas Gleixner 
Cc: dave@stgolabs.net
Cc: mark.rutland@arm.com
Link: http://lkml.kernel.org/r/1483479794-14013-5-git-send-email-dave@stgolabs.net
Signed-off-by: Ingo Molnar

Merge tag 'dm-4.10-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm

2016-12-14T19:01:00+00:00

Pull device mapper updates from Mike Snitzer:

 - various fixes and improvements to request-based DM and DM multipath

 - some locking improvements in DM bufio

 - add Kconfig option to disable the DM block manager's extra locking
   which mainly serves as a developer tool

 - a few bug fixes to DM's persistent-data

 - a couple changes to prepare for multipage biovec support in the block
   layer

 - various improvements and cleanups in the DM core, DM cache, DM raid
   and DM crypt

 - add ability to have DM crypt use keys from the kernel key retention
   service

 - add a new "error_writes" feature to the DM flakey target, reads are
   left unchanged in this mode

* tag 'dm-4.10-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm: (40 commits)
  dm flakey: introduce "error_writes" feature
  dm cache policy smq: use hash_32() instead of hash_32_generic()
  dm crypt: reject key strings containing whitespace chars
  dm space map: always set ev if sm_ll_mutate() succeeds
  dm space map metadata: skip useless memcpy in metadata_ll_init_index()
  dm space map metadata: fix 'struct sm_metadata' leak on failed create
  Documentation: dm raid: define data_offset status field
  dm raid: fix discard support regression
  dm raid: don't allow "write behind" with raid4/5/6
  dm mpath: use hw_handler_params if attached hw_handler is same as requested
  dm crypt: add ability to use keys from the kernel key retention service
  dm array: remove a dead assignment in populate_ablock_with_values()
  dm ioctl: use offsetof() instead of open-coding it
  dm rq: simplify use_blk_mq initialization
  dm: use blk_set_queue_dying() in __dm_destroy()
  dm bufio: drop the lock when doing GFP_NOIO allocation
  dm bufio: don't take the lock in dm_bufio_shrink_count
  dm bufio: avoid sleeping while holding the dm_bufio lock
  dm table: simplify dm_table_determine_type()
  dm table: an 'all_blk_mq' table must be loaded for a blk-mq DM device
  ...

dm bufio: drop the lock when doing GFP_NOIO allocation

2016-12-08T19:13:05+00:00

If the first allocation attempt using GFP_NOWAIT fails, drop the lock
and retry using GFP_NOIO allocation (lock is dropped because the
allocation can take some time).

Note that we won't do GFP_NOIO allocation when we loop for the second
time, because the lock shouldn't be dropped between __wait_for_free_buffer
and __get_unclaimed_buffer.

Signed-off-by: Mikulas Patocka 
Signed-off-by: Mike Snitzer

dm bufio: don't take the lock in dm_bufio_shrink_count

2016-12-08T19:13:04+00:00

dm_bufio_shrink_count() is called from do_shrink_slab to find out how many
freeable objects are there. The reported value doesn't have to be precise,
so we don't need to take the dm-bufio lock.

Suggested-by: David Rientjes 
Signed-off-by: Mikulas Patocka 
Signed-off-by: Mike Snitzer

dm bufio: avoid sleeping while holding the dm_bufio lock

2016-12-08T19:13:04+00:00

We've seen in-field reports showing _lots_ (18 in one case, 41 in
another) of tasks all sitting there blocked on:

  mutex_lock+0x4c/0x68
  dm_bufio_shrink_count+0x38/0x78
  shrink_slab.part.54.constprop.65+0x100/0x464
  shrink_zone+0xa8/0x198

In the two cases analyzed, we see one task that looks like this:

  Workqueue: kverityd verity_prefetch_io

  __switch_to+0x9c/0xa8
  __schedule+0x440/0x6d8
  schedule+0x94/0xb4
  schedule_timeout+0x204/0x27c
  schedule_timeout_uninterruptible+0x44/0x50
  wait_iff_congested+0x9c/0x1f0
  shrink_inactive_list+0x3a0/0x4cc
  shrink_lruvec+0x418/0x5cc
  shrink_zone+0x88/0x198
  try_to_free_pages+0x51c/0x588
  __alloc_pages_nodemask+0x648/0xa88
  __get_free_pages+0x34/0x7c
  alloc_buffer+0xa4/0x144
  __bufio_new+0x84/0x278
  dm_bufio_prefetch+0x9c/0x154
  verity_prefetch_io+0xe8/0x10c
  process_one_work+0x240/0x424
  worker_thread+0x2fc/0x424
  kthread+0x10c/0x114

...and that looks to be the one holding the mutex.

The problem has been reproduced on fairly easily:
0. Be running Chrome OS w/ verity enabled on the root filesystem
1. Pick test patch: http://crosreview.com/412360
2. Install launchBalloons.sh and balloon.arm from
     http://crbug.com/468342
   ...that's just a memory stress test app.
3. On a 4GB rk3399 machine, run
     nice ./launchBalloons.sh 4 900 100000
   ...that tries to eat 4 * 900 MB of memory and keep accessing.
4. Login to the Chrome web browser and restore many tabs

With that, I've seen printouts like:
  DOUG: long bufio 90758 ms
...and stack trace always show's we're in dm_bufio_prefetch().

The problem is that we try to allocate memory with GFP_NOIO while
we're holding the dm_bufio lock.  Instead we should be using
GFP_NOWAIT.  Using GFP_NOIO can cause us to sleep while holding the
lock and that causes the above problems.

The current behavior explained by David Rientjes:

  It will still try reclaim initially because __GFP_WAIT (or
  __GFP_KSWAPD_RECLAIM) is set by GFP_NOIO.  This is the cause of
  contention on dm_bufio_lock() that the thread holds.  You want to
  pass GFP_NOWAIT instead of GFP_NOIO to alloc_buffer() when holding a
  mutex that can be contended by a concurrent slab shrinker (if
  count_objects didn't use a trylock, this pattern would trivially
  deadlock).

This change significantly increases responsiveness of the system while
in this state.  It makes a real difference because it unblocks kswapd.
In the bug report analyzed, kswapd was hung:

   kswapd0         D ffffffc000204fd8     0    72      2 0x00000000
   Call trace:
   [] __switch_to+0x9c/0xa8
   [] __schedule+0x440/0x6d8
   [] schedule+0x94/0xb4
   [] schedule_preempt_disabled+0x28/0x44
   [] __mutex_lock_slowpath+0x120/0x1ac
   [] mutex_lock+0x4c/0x68
   [] dm_bufio_shrink_count+0x38/0x78
   [] shrink_slab.part.54.constprop.65+0x100/0x464
   [] shrink_zone+0xa8/0x198
   [] balance_pgdat+0x328/0x508
   [] kswapd+0x424/0x51c
   [] kthread+0x10c/0x114
   [] ret_from_fork+0x10/0x40

By unblocking kswapd memory pressure should be reduced.

Suggested-by: David Rientjes 
Reviewed-by: Guenter Roeck 
Signed-off-by: Douglas Anderson 
Signed-off-by: Mike Snitzer