Linux-Fsdevel Archive on lore.kernel.org
 help / color / Atom feed
From: Kirill Tkhai <ktkhai@virtuozzo.com>
To: akpm@linux-foundation.org, gregkh@linuxfoundation.org,
	rafael@kernel.org, viro@zeniv.linux.org.uk,
	darrick.wong@oracle.com, paulmck@linux.vnet.ibm.com,
	josh@joshtriplett.org, rostedt@goodmis.org,
	mathieu.desnoyers@efficios.com, jiangshanlai@gmail.com,
	hughd@google.com, shuah@kernel.org, robh@kernel.org,
	ulf.hansson@linaro.org, aspriel@gmail.com,
	vivek.gautam@codeaurora.org, robin.murphy@arm.com,
	joe@perches.com, heikki.krogerus@linux.intel.com,
	ktkhai@virtuozzo.com, sfr@canb.auug.org.au,
	vdavydov.dev@gmail.com, mhocko@suse.com,
	chris@chris-wilson.co.uk, penguin-kernel@I-love.SAKURA.ne.jp,
	aryabinin@virtuozzo.com, willy@infradead.org,
	ying.huang@intel.com, shakeelb@google.com, jbacik@fb.com,
	mingo@kernel.org, mhiramat@kernel.org,
	linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org,
	linux-mm@kvack.org
Subject: [PATCH RFC 00/10] Introduce lockless shrink_slab()
Date: Tue, 07 Aug 2018 18:37:19 +0300
Message-ID: <153365347929.19074.12509495712735843805.stgit@localhost.localdomain> (raw)

After bitmaps of not-empty memcg shrinkers were implemented
(see "[PATCH v9 00/17] Improve shrink_slab() scalability..."
series, which is already in mm tree), all the evil in perf
trace has moved from shrink_slab() to down_read_trylock().
As reported by Shakeel Butt:

     > I created 255 memcgs, 255 ext4 mounts and made each memcg create a
     > file containing few KiBs on corresponding mount. Then in a separate
     > memcg of 200 MiB limit ran a fork-bomb.
     >
     > I ran the "perf record -ag -- sleep 60" and below are the results:
     > +  47.49%            fb.sh  [kernel.kallsyms]    [k] down_read_trylock
     > +  30.72%            fb.sh  [kernel.kallsyms]    [k] up_read
     > +   9.51%            fb.sh  [kernel.kallsyms]    [k] mem_cgroup_iter
     > +   1.69%            fb.sh  [kernel.kallsyms]    [k] shrink_node_memcg
     > +   1.35%            fb.sh  [kernel.kallsyms]    [k] mem_cgroup_protected
     > +   1.05%            fb.sh  [kernel.kallsyms]    [k] queued_spin_lock_slowpath
     > +   0.85%            fb.sh  [kernel.kallsyms]    [k] _raw_spin_lock
     > +   0.78%            fb.sh  [kernel.kallsyms]    [k] lruvec_lru_size
     > +   0.57%            fb.sh  [kernel.kallsyms]    [k] shrink_node
     > +   0.54%            fb.sh  [kernel.kallsyms]    [k] queue_work_on
     > +   0.46%            fb.sh  [kernel.kallsyms]    [k] shrink_slab_memcg

The patchset continues to improve shrink_slab() scalability and makes
it lockless completely. Here are several steps for that:

1)Use SRCU to synchronize shrink_slab() and unregister_shrinker().
  Nothing exiting is here, just srcu_read_lock() in shrink_slab()
  and shrink_slab_memcg() and synchronize_srcu() in unregister_shrinker().
  See [2/10] for details.
2)The above requires to make SRCU unconditional enabled.
  [1/10] makes this. Note, that if we can't always enable
  SRCU, we may use percpu_rw_semaphore instead of this.
  See comment to [2/10] for details.
3)Convert shrinker_rwsem to mutex. Just cleanup.

4)Further patches make possible to speed up unregister_shrinker()
  by splitting it in two stages. The first stage unlinks shrinker
  from shrinker_list and shrinker_idr, while the second finalizes
  the thing by calling synchronize_srcu() and freeing memory.
  Patch [4/10] actually splits unregister_shrinker(), while
  [10/10] makes superblock shrinker to use the new helpers
  (unregister_shrinker_delayed_{initiate,finalize}().

5)Patches [5-9/10] are preparations on fs, which make possible
  to split superblock unregistration in two stages. They sequentially
  make super_cache_count() and super_cache_scan() safe to be called
  on unregistering shrinker:

  [cpu0]                                           [cpu1]
  unregister_shrinker_delayed_initiate(shrinker)
  ...                                              shrink_slab(shrinker) (OK!)
  unregister_shrinker_delayed_finalize(shrinker)

After all of this, shrink_slab() becomes lockless, while unregister_shrinker()
remains fast at least for superblock shrinker (another shrinkers also
can be made to unregister in the same delayed manner).

(This requires "mm: Use special value SHRINKER_REGISTERING instead list_empty() check"
 from https://lkml.org/lkml/2018/8/6/276, which is on the way to -mm tree, as said
 by -mm tree notification message from Andrew).

---

Kirill Tkhai (10):
      rcu: Make CONFIG_SRCU unconditionally enabled
      mm: Make shrink_slab() lockless
      mm: Convert shrinker_rwsem to mutex
      mm: Split unregister_shrinker()
      fs: Move list_lru_destroy() to destroy_super_work()
      fs: Shrink only (SB_ACTIVE|SB_BORN) superblocks in super_cache_scan()
      fs: Introduce struct super_operations::destroy_super() callback.
      xfs: Introduce xfs_fs_destroy_super()
      shmem: Implement shmem_destroy_super()
      fs: Use unregister_shrinker_delayed_{initiate,finalize} for super_block shrinker


 drivers/base/core.c                                |   42 ----------
 fs/super.c                                         |   32 ++++----
 fs/xfs/xfs_super.c                                 |   14 +++
 include/linux/device.h                             |    2 
 include/linux/fs.h                                 |    6 +
 include/linux/rcutiny.h                            |    4 -
 include/linux/shrinker.h                           |    2 
 include/linux/srcu.h                               |    5 -
 kernel/notifier.c                                  |    3 -
 kernel/rcu/Kconfig                                 |   12 ---
 kernel/rcu/tree.h                                  |    5 -
 kernel/rcu/update.c                                |    4 -
 mm/shmem.c                                         |    8 ++
 mm/vmscan.c                                        |   82 ++++++++++----------
 .../selftests/rcutorture/doc/TREE_RCU-kconfig.txt  |    5 -
 15 files changed, 89 insertions(+), 137 deletions(-)

--
Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>

             reply index

Thread overview: 40+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-08-07 15:37 Kirill Tkhai [this message]
2018-08-07 15:37 ` [PATCH RFC 01/10] rcu: Make CONFIG_SRCU unconditionally enabled Kirill Tkhai
2018-08-08  0:55   ` Steven Rostedt
2018-08-08  1:05   ` Stephen Rothwell
2018-08-08  9:46     ` Kirill Tkhai
2018-08-08  1:08   ` Stephen Rothwell
2018-08-08  9:59     ` Kirill Tkhai
2018-08-08 11:04       ` Stephen Rothwell
2018-08-08  7:20   ` Michal Hocko
2018-08-08 10:17     ` Kirill Tkhai
2018-08-08 10:27       ` Michal Hocko
2018-08-08 21:31         ` Dave Chinner
2018-08-09  0:07           ` Matthew Wilcox
2018-08-09  7:45             ` Greg KH
2018-08-09 10:22           ` Kirill Tkhai
2018-08-08 16:13       ` Josh Triplett
2018-08-08 16:23         ` Kirill Tkhai
2018-08-08 16:30           ` Kirill Tkhai
2018-08-08 18:01             ` Josh Triplett
2018-08-08 23:02               ` Shakeel Butt
2018-08-08 23:09                 ` Josh Triplett
2018-08-07 15:37 ` [PATCH RFC 02/10] mm: Make shrink_slab() lockless Kirill Tkhai
2018-08-08 11:51   ` Kirill Tkhai
2018-08-08 13:20     ` [PATCH RFC v2 " Kirill Tkhai
2018-08-09  7:14       ` Michal Hocko
2018-08-09  9:21         ` Kirill Tkhai
2018-08-09 10:37           ` Tetsuo Handa
2018-08-09 10:58             ` Kirill Tkhai
2018-08-09 11:23         ` Kirill Tkhai
2018-08-07 15:38 ` [PATCH RFC 03/10] mm: Convert shrinker_rwsem to mutex Kirill Tkhai
2018-08-07 15:38 ` [PATCH RFC 04/10] mm: Split unregister_shrinker() Kirill Tkhai
2018-08-07 15:38 ` [PATCH RFC 05/10] fs: Move list_lru_destroy() to destroy_super_work() Kirill Tkhai
2018-08-07 15:38 ` [PATCH RFC 06/10] fs: Shrink only (SB_ACTIVE|SB_BORN) superblocks in super_cache_scan() Kirill Tkhai
2018-08-07 15:38 ` [PATCH RFC 07/10] fs: Introduce struct super_operations::destroy_super() callback Kirill Tkhai
2018-08-07 15:39 ` [PATCH RFC 08/10] xfs: Introduce xfs_fs_destroy_super() Kirill Tkhai
2018-08-07 15:39 ` [PATCH RFC 09/10] shmem: Implement shmem_destroy_super() Kirill Tkhai
2018-08-07 15:39 ` [PATCH RFC 10/10] fs: Use unregister_shrinker_delayed_{initiate, finalize} for super_block shrinker Kirill Tkhai
2018-08-08  1:12 ` [PATCH RFC 00/10] Introduce lockless shrink_slab() Stephen Rothwell
2018-08-08  5:39   ` Shakeel Butt
2018-08-08 10:18     ` Kirill Tkhai

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=153365347929.19074.12509495712735843805.stgit@localhost.localdomain \
    --to=ktkhai@virtuozzo.com \
    --cc=akpm@linux-foundation.org \
    --cc=aryabinin@virtuozzo.com \
    --cc=aspriel@gmail.com \
    --cc=chris@chris-wilson.co.uk \
    --cc=darrick.wong@oracle.com \
    --cc=gregkh@linuxfoundation.org \
    --cc=heikki.krogerus@linux.intel.com \
    --cc=hughd@google.com \
    --cc=jbacik@fb.com \
    --cc=jiangshanlai@gmail.com \
    --cc=joe@perches.com \
    --cc=josh@joshtriplett.org \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=mathieu.desnoyers@efficios.com \
    --cc=mhiramat@kernel.org \
    --cc=mhocko@suse.com \
    --cc=mingo@kernel.org \
    --cc=paulmck@linux.vnet.ibm.com \
    --cc=penguin-kernel@I-love.SAKURA.ne.jp \
    --cc=rafael@kernel.org \
    --cc=robh@kernel.org \
    --cc=robin.murphy@arm.com \
    --cc=rostedt@goodmis.org \
    --cc=sfr@canb.auug.org.au \
    --cc=shakeelb@google.com \
    --cc=shuah@kernel.org \
    --cc=ulf.hansson@linaro.org \
    --cc=vdavydov.dev@gmail.com \
    --cc=viro@zeniv.linux.org.uk \
    --cc=vivek.gautam@codeaurora.org \
    --cc=willy@infradead.org \
    --cc=ying.huang@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux-Fsdevel Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-fsdevel/0 linux-fsdevel/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-fsdevel linux-fsdevel/ https://lore.kernel.org/linux-fsdevel \
		linux-fsdevel@vger.kernel.org
	public-inbox-index linux-fsdevel

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-fsdevel


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git