Commit 3812c8c8 ("mm: memcg: do not trap chargers with full
callstack on OOM") assumed that only a few places that can trigger a
memcg OOM situation do not return VM_FAULT_OOM, like optional page cache
readahead.  But there are many more and it's impractical to annotate
them all.

First of all, we don't want to invoke the OOM killer when the failed
allocation is gracefully handled, so defer the actual kill to the end of
the fault handling as well.  This simplifies the code quite a bit for
added bonus.

Second, since a failed allocation might not be the abrupt end of the
fault, the memcg OOM handler needs to be re-entrant until the fault
finishes for subsequent allocation attempts.  If an allocation is
attempted after the task already OOMed, allow it to bypass the limit so
that it can quickly finish the fault and invoke the OOM killer.
Reported-by: NazurIt <azurit@pobox.sk>
Signed-off-by: NJohannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: <stable@kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

for_each_online_cpu() needs the protection of {get,put}_online_cpus() so
cpu_online_mask doesn't change during the iteration.

cpu_hotplug.lock is held while a cpu is going down, it's a coarse lock
that is used kernel-wide to synchronize cpu hotplug activity.  Memcg has
a cpu hotplug notifier, called while there may not be any cpu hotplug
refcounts, which drains per-cpu event counts to memcg->nocpu_base.events
to maintain a cumulative event count as cpus disappear.  Without
get_online_cpus() in mem_cgroup_read_events(), it's possible to account
for the event count on a dying cpu twice, and this value may be
significantly large.

In fact, all memcg->pcp_counter_lock use should be nested by
{get,put}_online_cpus().

This fixes that issue and ensures the reported statistics are not vastly
over-reported during cpu hotplug.
Signed-off-by: NDavid Rientjes <rientjes@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: NKOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Revert commit 3b38722e ("memcg, vmscan: integrate soft reclaim
tighter with zone shrinking code")

I merged this prematurely - Michal and Johannes still disagree about the
overall design direction and the future remains unclear.

Cc: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Revert commit e883110a ("memcg: get rid of soft-limit tree
infrastructure")

I merged this prematurely - Michal and Johannes still disagree about the
overall design direction and the future remains unclear.

Cc: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Revert commit a5b7c87f ("vmscan, memcg: do softlimit reclaim also
for targeted reclaim")

I merged this prematurely - Michal and Johannes still disagree about the
overall design direction and the future remains unclear.

Cc: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Revert commit de57780d ("memcg: enhance memcg iterator to support
predicates")

I merged this prematurely - Michal and Johannes still disagree about the
overall design direction and the future remains unclear.

Cc: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Revert commit 7d910c05 ("memcg: track children in soft limit excess
to improve soft limit")

I merged this prematurely - Michal and Johannes still disagree about the
overall design direction and the future remains unclear.

Cc: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Revert commit e839b6a1 ("memcg, vmscan: do not attempt soft limit
reclaim if it would not scan anything")

I merged this prematurely - Michal and Johannes still disagree about the
overall design direction and the future remains unclear.

Cc: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Revert commit 1be171d6 ("memcg: track all children over limit in the
root")

I merged this prematurely - Michal and Johannes still disagree about the
overall design direction and the future remains unclear.

Cc: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Add memcg routines to count writeback pages, later dirty pages will also
be accounted.

After Kame's commit 89c06bd5 ("memcg: use new logic for page stat
accounting"), we can use 'struct page' flag to test page state instead
of per page_cgroup flag.  But memcg has a feature to move a page from a
cgroup to another one and may have race between "move" and "page stat
accounting".  So in order to avoid the race we have designed a new lock:

         mem_cgroup_begin_update_page_stat()
         modify page information        -->(a)
         mem_cgroup_update_page_stat()  -->(b)
         mem_cgroup_end_update_page_stat()

It requires both (a) and (b)(writeback pages accounting) to be pretected
in mem_cgroup_{begin/end}_update_page_stat().  It's full no-op for
!CONFIG_MEMCG, almost no-op if memcg is disabled (but compiled in), rcu
read lock in the most cases (no task is moving), and spin_lock_irqsave
on top in the slow path.

There're two writeback interfaces to modify: test_{clear/set}_page_writeback().
And the lock order is:
	--> memcg->move_lock
	  --> mapping->tree_lock
Signed-off-by: NSha Zhengju <handai.szj@taobao.com>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Reviewed-by: NGreg Thelen <gthelen@google.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

We should call mem_cgroup_begin_update_page_stat() before
mem_cgroup_update_page_stat() to get proper locks, however the latter
doesn't do any checking that we use proper locking, which would be hard.
Suggested by Michal Hock we could at least test for rcu_read_lock_held()
because RCU is held if !mem_cgroup_disabled().
Signed-off-by: NSha Zhengju <handai.szj@taobao.com>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Reviewed-by: NGreg Thelen <gthelen@google.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

While accounting memcg page stat, it's not worth to use
MEMCG_NR_FILE_MAPPED as an extra layer of indirection because of the
complexity and presumed performance overhead.  We can use
MEM_CGROUP_STAT_FILE_MAPPED directly.
Signed-off-by: NSha Zhengju <handai.szj@taobao.com>
Acked-by: NKAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Acked-by: NFengguang Wu <fengguang.wu@intel.com>
Reviewed-by: NGreg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

RESOURCE_MAX is far too general name, change it to RES_COUNTER_MAX.
Signed-off-by: NSha Zhengju <handai.szj@taobao.com>
Signed-off-by: NQiang Huang <h.huangqiang@huawei.com>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Cc: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Cc: Jeff Liu <jeff.liu@oracle.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The memcg OOM handling is incredibly fragile and can deadlock.  When a
task fails to charge memory, it invokes the OOM killer and loops right
there in the charge code until it succeeds.  Comparably, any other task
that enters the charge path at this point will go to a waitqueue right
then and there and sleep until the OOM situation is resolved.  The problem
is that these tasks may hold filesystem locks and the mmap_sem; locks that
the selected OOM victim may need to exit.

For example, in one reported case, the task invoking the OOM killer was
about to charge a page cache page during a write(), which holds the
i_mutex.  The OOM killer selected a task that was just entering truncate()
and trying to acquire the i_mutex:

OOM invoking task:
  mem_cgroup_handle_oom+0x241/0x3b0
  mem_cgroup_cache_charge+0xbe/0xe0
  add_to_page_cache_locked+0x4c/0x140
  add_to_page_cache_lru+0x22/0x50
  grab_cache_page_write_begin+0x8b/0xe0
  ext3_write_begin+0x88/0x270
  generic_file_buffered_write+0x116/0x290
  __generic_file_aio_write+0x27c/0x480
  generic_file_aio_write+0x76/0xf0           # takes ->i_mutex
  do_sync_write+0xea/0x130
  vfs_write+0xf3/0x1f0
  sys_write+0x51/0x90
  system_call_fastpath+0x18/0x1d

OOM kill victim:
  do_truncate+0x58/0xa0              # takes i_mutex
  do_last+0x250/0xa30
  path_openat+0xd7/0x440
  do_filp_open+0x49/0xa0
  do_sys_open+0x106/0x240
  sys_open+0x20/0x30
  system_call_fastpath+0x18/0x1d

The OOM handling task will retry the charge indefinitely while the OOM
killed task is not releasing any resources.

A similar scenario can happen when the kernel OOM killer for a memcg is
disabled and a userspace task is in charge of resolving OOM situations.
In this case, ALL tasks that enter the OOM path will be made to sleep on
the OOM waitqueue and wait for userspace to free resources or increase
the group's limit.  But a userspace OOM handler is prone to deadlock
itself on the locks held by the waiting tasks.  For example one of the
sleeping tasks may be stuck in a brk() call with the mmap_sem held for
writing but the userspace handler, in order to pick an optimal victim,
may need to read files from /proc/<pid>, which tries to acquire the same
mmap_sem for reading and deadlocks.

This patch changes the way tasks behave after detecting a memcg OOM and
makes sure nobody loops or sleeps with locks held:

1. When OOMing in a user fault, invoke the OOM killer and restart the
   fault instead of looping on the charge attempt.  This way, the OOM
   victim can not get stuck on locks the looping task may hold.

2. When OOMing in a user fault but somebody else is handling it
   (either the kernel OOM killer or a userspace handler), don't go to
   sleep in the charge context.  Instead, remember the OOMing memcg in
   the task struct and then fully unwind the page fault stack with
   -ENOMEM.  pagefault_out_of_memory() will then call back into the
   memcg code to check if the -ENOMEM came from the memcg, and then
   either put the task to sleep on the memcg's OOM waitqueue or just
   restart the fault.  The OOM victim can no longer get stuck on any
   lock a sleeping task may hold.

Debugged by Michal Hocko.
Signed-off-by: NJohannes Weiner <hannes@cmpxchg.org>
Reported-by: NazurIt <azurit@pobox.sk>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The memcg OOM handler open-codes a sleeping lock for OOM serialization
(trylock, wait, repeat) because the required locking is so specific to
memcg hierarchies.  However, it would be nice if this construct would be
clearly recognizable and not be as obfuscated as it is right now.  Clean
up as follows:

1. Remove the return value of mem_cgroup_oom_unlock()

2. Rename mem_cgroup_oom_lock() to mem_cgroup_oom_trylock().

3. Pull the prepare_to_wait() out of the memcg_oom_lock scope.  This
   makes it more obvious that the task has to be on the waitqueue
   before attempting to OOM-trylock the hierarchy, to not miss any
   wakeups before going to sleep.  It just didn't matter until now
   because it was all lumped together into the global memcg_oom_lock
   spinlock section.

4. Pull the mem_cgroup_oom_notify() out of the memcg_oom_lock scope.
   It is proctected by the hierarchical OOM-lock.

5. The memcg_oom_lock spinlock is only required to propagate the OOM
   lock in any given hierarchy atomically.  Restrict its scope to
   mem_cgroup_oom_(trylock|unlock).

6. Do not wake up the waitqueue unconditionally at the end of the
   function.  Only the lockholder has to wake up the next in line
   after releasing the lock.

   Note that the lockholder kicks off the OOM-killer, which in turn
   leads to wakeups from the uncharges of the exiting task.  But a
   contender is not guaranteed to see them if it enters the OOM path
   after the OOM kills but before the lockholder releases the lock.
   Thus there has to be an explicit wakeup after releasing the lock.

7. Put the OOM task on the waitqueue before marking the hierarchy as
   under OOM as that is the point where we start to receive wakeups.
   No point in listening before being on the waitqueue.

8. Likewise, unmark the hierarchy before finishing the sleep, for
   symmetry.
Signed-off-by: NJohannes Weiner <hannes@cmpxchg.org>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: azurIt <azurit@pobox.sk>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

System calls and kernel faults (uaccess, gup) can handle an out of memory
situation gracefully and just return -ENOMEM.

Enable the memcg OOM killer only for user faults, where it's really the
only option available.
Signed-off-by: NJohannes Weiner <hannes@cmpxchg.org>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: azurIt <azurit@pobox.sk>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Clean up some mess made by the "Soft limit rework" series, and a few other
things.

Cc: Michal Hocko <mhocko@suse.cz>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Children in soft limit excess are currently tracked up the hierarchy in
memcg->children_in_excess.  Nevertheless there still might exist tons of
groups that are not in hierarchy relation to the root cgroup (e.g.  all
first level groups if root_mem_cgroup->use_hierarchy == false).

As the whole tree walk has to be done when the iteration starts at
root_mem_cgroup the iterator should be able to skip the walk if there is
no child above the limit without iterating them.  This can be done
easily if the root tracks all children rather than only hierarchical
children.  This is done by this patch which updates root_mem_cgroup
children_in_excess if root_mem_cgroup->use_hierarchy == false so the
root knows about all children in excess.

Please note that this is not an issue for inner memcgs which have
use_hierarchy == false because then only the single group is visited so
no special optimization is necessary.
Signed-off-by: NMichal Hocko <mhocko@suse.cz>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Greg Thelen <gthelen@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Ying Han <yinghan@google.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

mem_cgroup_should_soft_reclaim controls whether soft reclaim pass is
done and it always says yes currently.  Memcg iterators are clever to
skip nodes that are not soft reclaimable quite efficiently but
mem_cgroup_should_soft_reclaim can be more clever and do not start the
soft reclaim pass at all if it knows that nothing would be scanned
anyway.

In order to do that, simply reuse mem_cgroup_soft_reclaim_eligible for
the target group of the reclaim and allow the pass only if the whole
subtree wouldn't be skipped.
Signed-off-by: NMichal Hocko <mhocko@suse.cz>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Greg Thelen <gthelen@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Ying Han <yinghan@google.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Soft limit reclaim has to check the whole reclaim hierarchy while doing
the first pass of the reclaim.  This leads to a higher system time which
can be visible especially when there are many groups in the hierarchy.

This patch adds a per-memcg counter of children in excess.  It also
restores MEM_CGROUP_TARGET_SOFTLIMIT into mem_cgroup_event_ratelimit for a
proper batching.

If a group crosses soft limit for the first time it increases parent's
children_in_excess up the hierarchy.  The similarly if a group gets below
the limit it will decrease the counter.  The transition phase is recorded
in soft_contributed flag.

mem_cgroup_soft_reclaim_eligible then uses this information to better
decide whether to skip the node or the whole subtree.  The rule is simple.
 Skip the node with a children in excess or skip the whole subtree
otherwise.

This has been tested by a stream IO (dd if=/dev/zero of=file with
4*MemTotal size) which is quite sensitive to overhead during reclaim.  The
load is running in a group with soft limit set to 0 and without any limit.
 Apart from that there was a hierarchy with ~500, 2k and 8k groups (two
groups on each level) without any pages in them.  base denotes to the
kernel on which the whole series is based on, rework is the kernel before
this patch and reworkoptim is with this patch applied:

* Run with soft limit set to 0
Elapsed
0-0-limit/base: min: 88.21 max: 94.61 avg: 91.73 std: 2.65 runs: 3
0-0-limit/rework: min: 76.05 [86.2%] max: 79.08 [83.6%] avg: 77.84 [84.9%] std: 1.30 runs: 3
0-0-limit/reworkoptim: min: 77.98 [88.4%] max: 80.36 [84.9%] avg: 78.92 [86.0%] std: 1.03 runs: 3
System
0.5k-0-limit/base: min: 34.86 max: 36.42 avg: 35.89 std: 0.73 runs: 3
0.5k-0-limit/rework: min: 43.26 [124.1%] max: 48.95 [134.4%] avg: 46.09 [128.4%] std: 2.32 runs: 3
0.5k-0-limit/reworkoptim: min: 46.98 [134.8%] max: 50.98 [140.0%] avg: 48.49 [135.1%] std: 1.77 runs: 3
Elapsed
0.5k-0-limit/base: min: 88.50 max: 97.52 avg: 93.92 std: 3.90 runs: 3
0.5k-0-limit/rework: min: 75.92 [85.8%] max: 78.45 [80.4%] avg: 77.34 [82.3%] std: 1.06 runs: 3
0.5k-0-limit/reworkoptim: min: 75.79 [85.6%] max: 79.37 [81.4%] avg: 77.55 [82.6%] std: 1.46 runs: 3
System
2k-0-limit/base: min: 34.57 max: 37.65 avg: 36.34 std: 1.30 runs: 3
2k-0-limit/rework: min: 64.17 [185.6%] max: 68.20 [181.1%] avg: 66.21 [182.2%] std: 1.65 runs: 3
2k-0-limit/reworkoptim: min: 49.78 [144.0%] max: 52.99 [140.7%] avg: 51.00 [140.3%] std: 1.42 runs: 3
Elapsed
2k-0-limit/base: min: 92.61 max: 97.83 avg: 95.03 std: 2.15 runs: 3
2k-0-limit/rework: min: 78.33 [84.6%] max: 84.08 [85.9%] avg: 81.09 [85.3%] std: 2.35 runs: 3
2k-0-limit/reworkoptim: min: 75.72 [81.8%] max: 78.57 [80.3%] avg: 76.73 [80.7%] std: 1.30 runs: 3
System
8k-0-limit/base: min: 39.78 max: 42.09 avg: 41.09 std: 0.97 runs: 3
8k-0-limit/rework: min: 200.86 [504.9%] max: 265.42 [630.6%] avg: 241.80 [588.5%] std: 29.06 runs: 3
8k-0-limit/reworkoptim: min: 53.70 [135.0%] max: 54.89 [130.4%] avg: 54.43 [132.5%] std: 0.52 runs: 3
Elapsed
8k-0-limit/base: min: 95.11 max: 98.61 avg: 96.81 std: 1.43 runs: 3
8k-0-limit/rework: min: 246.96 [259.7%] max: 331.47 [336.1%] avg: 301.32 [311.2%] std: 38.52 runs: 3
8k-0-limit/reworkoptim: min: 76.79 [80.7%] max: 81.71 [82.9%] avg: 78.97 [81.6%] std: 2.05 runs: 3

System time is increased by 30-40% but it is reduced a lot comparing to
kernel without this patch.  The higher time can be explained by the fact
that the original soft reclaim scanned at priority 0 so it was much more
effective for this workload (which is basically touch once and writeback).
 The Elapsed time looks better though (~20%).

* Run with no soft limit set
System
0-no-limit/base: min: 42.18 max: 50.38 avg: 46.44 std: 3.36 runs: 3
0-no-limit/rework: min: 40.57 [96.2%] max: 47.04 [93.4%] avg: 43.82 [94.4%] std: 2.64 runs: 3
0-no-limit/reworkoptim: min: 40.45 [95.9%] max: 45.28 [89.9%] avg: 42.10 [90.7%] std: 2.25 runs: 3
Elapsed
0-no-limit/base: min: 75.97 max: 78.21 avg: 76.87 std: 0.96 runs: 3
0-no-limit/rework: min: 75.59 [99.5%] max: 80.73 [103.2%] avg: 77.64 [101.0%] std: 2.23 runs: 3
0-no-limit/reworkoptim: min: 77.85 [102.5%] max: 82.42 [105.4%] avg: 79.64 [103.6%] std: 1.99 runs: 3
System
0.5k-no-limit/base: min: 44.54 max: 46.93 avg: 46.12 std: 1.12 runs: 3
0.5k-no-limit/rework: min: 42.09 [94.5%] max: 46.16 [98.4%] avg: 43.92 [95.2%] std: 1.69 runs: 3
0.5k-no-limit/reworkoptim: min: 42.47 [95.4%] max: 45.67 [97.3%] avg: 44.06 [95.5%] std: 1.31 runs: 3
Elapsed
0.5k-no-limit/base: min: 78.26 max: 81.49 avg: 79.65 std: 1.36 runs: 3
0.5k-no-limit/rework: min: 77.01 [98.4%] max: 80.43 [98.7%] avg: 78.30 [98.3%] std: 1.52 runs: 3
0.5k-no-limit/reworkoptim: min: 76.13 [97.3%] max: 77.87 [95.6%] avg: 77.18 [96.9%] std: 0.75 runs: 3
System
2k-no-limit/base: min: 62.96 max: 69.14 avg: 66.14 std: 2.53 runs: 3
2k-no-limit/rework: min: 76.01 [120.7%] max: 81.06 [117.2%] avg: 78.17 [118.2%] std: 2.12 runs: 3
2k-no-limit/reworkoptim: min: 62.57 [99.4%] max: 66.10 [95.6%] avg: 64.53 [97.6%] std: 1.47 runs: 3
Elapsed
2k-no-limit/base: min: 76.47 max: 84.22 avg: 79.12 std: 3.60 runs: 3
2k-no-limit/rework: min: 89.67 [117.3%] max: 93.26 [110.7%] avg: 91.10 [115.1%] std: 1.55 runs: 3
2k-no-limit/reworkoptim: min: 76.94 [100.6%] max: 79.21 [94.1%] avg: 78.45 [99.2%] std: 1.07 runs: 3
System
8k-no-limit/base: min: 104.74 max: 151.34 avg: 129.21 std: 19.10 runs: 3
8k-no-limit/rework: min: 205.23 [195.9%] max: 285.94 [188.9%] avg: 258.98 [200.4%] std: 38.01 runs: 3
8k-no-limit/reworkoptim: min: 161.16 [153.9%] max: 184.54 [121.9%] avg: 174.52 [135.1%] std: 9.83 runs: 3
Elapsed
8k-no-limit/base: min: 125.43 max: 181.00 avg: 154.81 std: 22.80 runs: 3
8k-no-limit/rework: min: 254.05 [202.5%] max: 355.67 [196.5%] avg: 321.46 [207.6%] std: 47.67 runs: 3
8k-no-limit/reworkoptim: min: 193.77 [154.5%] max: 222.72 [123.0%] avg: 210.18 [135.8%] std: 12.13 runs: 3

Both System and Elapsed are in stdev with the base kernel for all
configurations except for 8k where both System and Elapsed are up by 35%.
I do not have a good explanation for this because there is no soft reclaim
pass going on as no group is above the limit which is checked in
mem_cgroup_should_soft_reclaim.

Then I have tested kernel build with the same configuration to see the
behavior with a more general behavior.

* Soft limit set to 0 for the build
System
0-0-limit/base: min: 242.70 max: 245.17 avg: 243.85 std: 1.02 runs: 3
0-0-limit/rework min: 237.86 [98.0%] max: 240.22 [98.0%] avg: 239.00 [98.0%] std: 0.97 runs: 3
0-0-limit/reworkoptim: min: 241.11 [99.3%] max: 243.53 [99.3%] avg: 242.01 [99.2%] std: 1.08 runs: 3
Elapsed
0-0-limit/base: min: 348.48 max: 360.86 avg: 356.04 std: 5.41 runs: 3
0-0-limit/rework min: 286.95 [82.3%] max: 290.26 [80.4%] avg: 288.27 [81.0%] std: 1.43 runs: 3
0-0-limit/reworkoptim: min: 286.55 [82.2%] max: 289.00 [80.1%] avg: 287.69 [80.8%] std: 1.01 runs: 3
System
0.5k-0-limit/base: min: 251.77 max: 254.41 avg: 252.70 std: 1.21 runs: 3
0.5k-0-limit/rework min: 286.44 [113.8%] max: 289.30 [113.7%] avg: 287.60 [113.8%] std: 1.23 runs: 3
0.5k-0-limit/reworkoptim: min: 252.18 [100.2%] max: 253.16 [99.5%] avg: 252.62 [100.0%] std: 0.41 runs: 3
Elapsed
0.5k-0-limit/base: min: 347.83 max: 353.06 avg: 350.04 std: 2.21 runs: 3
0.5k-0-limit/rework min: 290.19 [83.4%] max: 295.62 [83.7%] avg: 293.12 [83.7%] std: 2.24 runs: 3
0.5k-0-limit/reworkoptim: min: 293.91 [84.5%] max: 294.87 [83.5%] avg: 294.29 [84.1%] std: 0.42 runs: 3
System
2k-0-limit/base: min: 263.05 max: 271.52 avg: 267.94 std: 3.58 runs: 3
2k-0-limit/rework min: 458.99 [174.5%] max: 468.31 [172.5%] avg: 464.45 [173.3%] std: 3.97 runs: 3
2k-0-limit/reworkoptim: min: 267.10 [101.5%] max: 279.38 [102.9%] avg: 272.78 [101.8%] std: 5.05 runs: 3
Elapsed
2k-0-limit/base: min: 372.33 max: 379.32 avg: 375.47 std: 2.90 runs: 3
2k-0-limit/rework min: 334.40 [89.8%] max: 339.52 [89.5%] avg: 337.44 [89.9%] std: 2.20 runs: 3
2k-0-limit/reworkoptim: min: 301.47 [81.0%] max: 319.19 [84.1%] avg: 307.90 [82.0%] std: 8.01 runs: 3
System
8k-0-limit/base: min: 320.50 max: 332.10 avg: 325.46 std: 4.88 runs: 3
8k-0-limit/rework min: 1115.76 [348.1%] max: 1165.66 [351.0%] avg: 1132.65 [348.0%] std: 23.34 runs: 3
8k-0-limit/reworkoptim: min: 403.75 [126.0%] max: 409.22 [123.2%] avg: 406.16 [124.8%] std: 2.28 runs: 3
Elapsed
8k-0-limit/base: min: 475.48 max: 585.19 avg: 525.54 std: 45.30 runs: 3
8k-0-limit/rework min: 616.25 [129.6%] max: 625.90 [107.0%] avg: 620.68 [118.1%] std: 3.98 runs: 3
8k-0-limit/reworkoptim: min: 420.18 [88.4%] max: 428.28 [73.2%] avg: 423.05 [80.5%] std: 3.71 runs: 3

Apart from 8k the system time is comparable with the base kernel while
Elapsed is up to 20% better with all configurations.

* No soft limit set
System
0-no-limit/base: min: 234.76 max: 237.42 avg: 236.25 std: 1.11 runs: 3
0-no-limit/rework min: 233.09 [99.3%] max: 238.65 [100.5%] avg: 236.09 [99.9%] std: 2.29 runs: 3
0-no-limit/reworkoptim: min: 236.12 [100.6%] max: 240.53 [101.3%] avg: 237.94 [100.7%] std: 1.88 runs: 3
Elapsed
0-no-limit/base: min: 288.52 max: 295.42 avg: 291.29 std: 2.98 runs: 3
0-no-limit/rework min: 283.17 [98.1%] max: 284.33 [96.2%] avg: 283.78 [97.4%] std: 0.48 runs: 3
0-no-limit/reworkoptim: min: 288.50 [100.0%] max: 290.79 [98.4%] avg: 289.78 [99.5%] std: 0.95 runs: 3
System
0.5k-no-limit/base: min: 286.51 max: 293.23 avg: 290.21 std: 2.78 runs: 3
0.5k-no-limit/rework min: 291.69 [101.8%] max: 294.38 [100.4%] avg: 292.97 [101.0%] std: 1.10 runs: 3
0.5k-no-limit/reworkoptim: min: 277.05 [96.7%] max: 288.76 [98.5%] avg: 284.17 [97.9%] std: 5.11 runs: 3
Elapsed
0.5k-no-limit/base: min: 294.94 max: 298.92 avg: 296.47 std: 1.75 runs: 3
0.5k-no-limit/rework min: 292.55 [99.2%] max: 294.21 [98.4%] avg: 293.55 [99.0%] std: 0.72 runs: 3
0.5k-no-limit/reworkoptim: min: 294.41 [99.8%] max: 301.67 [100.9%] avg: 297.78 [100.4%] std: 2.99 runs: 3
System
2k-no-limit/base: min: 443.41 max: 466.66 avg: 457.66 std: 10.19 runs: 3
2k-no-limit/rework min: 490.11 [110.5%] max: 516.02 [110.6%] avg: 501.42 [109.6%] std: 10.83 runs: 3
2k-no-limit/reworkoptim: min: 435.25 [98.2%] max: 458.11 [98.2%] avg: 446.73 [97.6%] std: 9.33 runs: 3
Elapsed
2k-no-limit/base: min: 330.85 max: 333.75 avg: 332.52 std: 1.23 runs: 3
2k-no-limit/rework min: 343.06 [103.7%] max: 349.59 [104.7%] avg: 345.95 [104.0%] std: 2.72 runs: 3
2k-no-limit/reworkoptim: min: 330.01 [99.7%] max: 333.92 [100.1%] avg: 332.22 [99.9%] std: 1.64 runs: 3
System
8k-no-limit/base: min: 1175.64 max: 1259.38 avg: 1222.39 std: 34.88 runs: 3
8k-no-limit/rework min: 1226.31 [104.3%] max: 1241.60 [98.6%] avg: 1233.74 [100.9%] std: 6.25 runs: 3
8k-no-limit/reworkoptim: min: 1023.45 [87.1%] max: 1056.74 [83.9%] avg: 1038.92 [85.0%] std: 13.69 runs: 3
Elapsed
8k-no-limit/base: min: 613.36 max: 619.60 avg: 616.47 std: 2.55 runs: 3
8k-no-limit/rework min: 627.56 [102.3%] max: 642.33 [103.7%] avg: 633.44 [102.8%] std: 6.39 runs: 3
8k-no-limit/reworkoptim: min: 545.89 [89.0%] max: 555.36 [89.6%] avg: 552.06 [89.6%] std: 4.37 runs: 3

and these numbers look good as well.  System time is around 100%
(suprisingly better for the 8k case) and Elapsed is copies that trend.
Signed-off-by: NMichal Hocko <mhocko@suse.cz>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Greg Thelen <gthelen@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Ying Han <yinghan@google.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The caller of the iterator might know that some nodes or even subtrees
should be skipped but there is no way to tell iterators about that so the
only choice left is to let iterators to visit each node and do the
selection outside of the iterating code.  This, however, doesn't scale
well with hierarchies with many groups where only few groups are
interesting.

This patch adds mem_cgroup_iter_cond variant of the iterator with a
callback which gets called for every visited node.  There are three
possible ways how the callback can influence the walk.  Either the node is
visited, it is skipped but the tree walk continues down the tree or the
whole subtree of the current group is skipped.

[hughd@google.com: fix memcg-less page reclaim]
Signed-off-by: NMichal Hocko <mhocko@suse.cz>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Ying Han <yinghan@google.com>
Signed-off-by: NHugh Dickins <hughd@google.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Soft reclaim has been done only for the global reclaim (both background
and direct).  Since "memcg: integrate soft reclaim tighter with zone
shrinking code" there is no reason for this limitation anymore as the soft
limit reclaim doesn't use any special code paths and it is a part of the
zone shrinking code which is used by both global and targeted reclaims.

From the semantic point of view it is natural to consider soft limit
before touching all groups in the hierarchy tree which is touching the
hard limit because soft limit tells us where to push back when there is a
memory pressure.  It is not important whether the pressure comes from the
limit or imbalanced zones.

This patch simply enables soft reclaim unconditionally in
mem_cgroup_should_soft_reclaim so it is enabled for both global and
targeted reclaim paths.  mem_cgroup_soft_reclaim_eligible needs to learn
about the root of the reclaim to know where to stop checking soft limit
state of parents up the hierarchy.  Say we have

A (over soft limit)
 \
  B (below s.l., hit the hard limit)
 / \
C   D (below s.l.)

B is the source of the outside memory pressure now for D but we shouldn't
soft reclaim it because it is behaving well under B subtree and we can
still reclaim from C (pressumably it is over the limit).
mem_cgroup_soft_reclaim_eligible should therefore stop climbing up the
hierarchy at B (root of the memory pressure).
Signed-off-by: NMichal Hocko <mhocko@suse.cz>
Reviewed-by: NGlauber Costa <glommer@openvz.org>
Reviewed-by: NTejun Heo <tj@kernel.org>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: Ying Han <yinghan@google.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Now that the soft limit is integrated to the reclaim directly the whole
soft-limit tree infrastructure is not needed anymore.  Rip it out.
Signed-off-by: NMichal Hocko <mhocko@suse.cz>
Reviewed-by: NGlauber Costa <glommer@openvz.org>
Reviewed-by: NTejun Heo <tj@kernel.org>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: Ying Han <yinghan@google.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This patchset is sitting out of tree for quite some time without any
objections.  I would be really happy if it made it into 3.12.  I do not
want to push it too hard but I think this work is basically ready and
waiting more doesn't help.

The basic idea is quite simple.  Pull soft reclaim into shrink_zone in the
first step and get rid of the previous soft reclaim infrastructure.
shrink_zone is done in two passes now.  First it tries to do the soft
limit reclaim and it falls back to reclaim-all mode if no group is over
the limit or no pages have been scanned.  The second pass happens at the
same priority so the only time we waste is the memcg tree walk which has
been updated in the third step to have only negligible overhead.

As a bonus we will get rid of a _lot_ of code by this and soft reclaim
will not stand out like before when it wasn't integrated into the zone
shrinking code and it reclaimed at priority 0 (the testing results show
that some workloads suffers from such an aggressive reclaim).  The clean
up is in a separate patch because I felt it would be easier to review that
way.

The second step is soft limit reclaim integration into targeted reclaim.
It should be rather straight forward.  Soft limit has been used only for
the global reclaim so far but it makes sense for any kind of pressure
coming from up-the-hierarchy, including targeted reclaim.

The third step (patches 4-8) addresses the tree walk overhead by enhancing
memcg iterators to enable skipping whole subtrees and tracking number of
over soft limit children at each level of the hierarchy.  This information
is updated same way the old soft limit tree was updated (from
memcg_check_events) so we shouldn't see an additional overhead.  In fact
mem_cgroup_update_soft_limit is much simpler than tree manipulation done
previously.

__shrink_zone uses mem_cgroup_soft_reclaim_eligible as a predicate for
mem_cgroup_iter so the decision whether a particular group should be
visited is done at the iterator level which allows us to decide to skip
the whole subtree as well (if there is no child in excess).  This reduces
the tree walk overhead considerably.

* TEST 1
========

My primary test case was a parallel kernel build with 2 groups (make is
running with -j8 with a distribution .config in a separate cgroup without
any hard limit) on a 32 CPU machine booted with 1GB memory and both builds
run taskset to Node 0 cpus.

I was mostly interested in 2 setups.  Default - no soft limit set and -
and 0 soft limit set to both groups.  The first one should tell us whether
the rework regresses the default behavior while the second one should show
us improvements in an extreme case where both workloads are always over
the soft limit.

/usr/bin/time -v has been used to collect the statistics and each
configuration had 3 runs after fresh boot without any other load on the
system.

base is mmotm-2013-07-18-16-40
rework all 8 patches applied on top of base

* No-limit
User
no-limit/base: min: 651.92 max: 672.65 avg: 664.33 std: 8.01 runs: 6
no-limit/rework: min: 657.34 [100.8%] max: 668.39 [99.4%] avg: 663.13 [99.8%] std: 3.61 runs: 6
System
no-limit/base: min: 69.33 max: 71.39 avg: 70.32 std: 0.79 runs: 6
no-limit/rework: min: 69.12 [99.7%] max: 71.05 [99.5%] avg: 70.04 [99.6%] std: 0.59 runs: 6
Elapsed
no-limit/base: min: 398.27 max: 422.36 avg: 408.85 std: 7.74 runs: 6
no-limit/rework: min: 386.36 [97.0%] max: 438.40 [103.8%] avg: 416.34 [101.8%] std: 18.85 runs: 6

The results are within noise. Elapsed time has a bigger variance but the
average looks good.

* 0-limit
User
0-limit/base: min: 573.76 max: 605.63 avg: 585.73 std: 12.21 runs: 6
0-limit/rework: min: 645.77 [112.6%] max: 666.25 [110.0%] avg: 656.97 [112.2%] std: 7.77 runs: 6
System
0-limit/base: min: 69.57 max: 71.13 avg: 70.29 std: 0.54 runs: 6
0-limit/rework: min: 68.68 [98.7%] max: 71.40 [100.4%] avg: 69.91 [99.5%] std: 0.87 runs: 6
Elapsed
0-limit/base: min: 1306.14 max: 1550.17 avg: 1430.35 std: 90.86 runs: 6
0-limit/rework: min: 404.06 [30.9%] max: 465.94 [30.1%] avg: 434.81 [30.4%] std: 22.68 runs: 6

The improvement is really huge here (even bigger than with my previous
testing and I suspect that this highly depends on the storage).  Page
fault statistics tell us at least part of the story:

Minor
0-limit/base: min: 37180461.00 max: 37319986.00 avg: 37247470.00 std: 54772.71 runs: 6
0-limit/rework: min: 36751685.00 [98.8%] max: 36805379.00 [98.6%] avg: 36774506.33 [98.7%] std: 17109.03 runs: 6
Major
0-limit/base: min: 170604.00 max: 221141.00 avg: 196081.83 std: 18217.01 runs: 6
0-limit/rework: min: 2864.00 [1.7%] max: 10029.00 [4.5%] avg: 5627.33 [2.9%] std: 2252.71 runs: 6

Same as with my previous testing Minor faults are more or less within
noise but Major fault count is way bellow the base kernel.

While this looks as a nice win it is fair to say that 0-limit
configuration is quite artificial. So I was playing with 0-no-limit
loads as well.

* TEST 2
========

The following results are from 2 groups configuration on a 16GB machine
(single NUMA node).

- A running stream IO (dd if=/dev/zero of=local.file bs=1024) with
  2*TotalMem with 0 soft limit.
- B running a mem_eater which consumes TotalMem-1G without any limit. The
  mem_eater consumes the memory in 100 chunks with 1s nap after each
  mmap+poppulate so that both loads have chance to fight for the memory.

The expected result is that B shouldn't be reclaimed and A shouldn't see
a big dropdown in elapsed time.

User
base: min: 2.68 max: 2.89 avg: 2.76 std: 0.09 runs: 3
rework: min: 3.27 [122.0%] max: 3.74 [129.4%] avg: 3.44 [124.6%] std: 0.21 runs: 3
System
base: min: 86.26 max: 88.29 avg: 87.28 std: 0.83 runs: 3
rework: min: 81.05 [94.0%] max: 84.96 [96.2%] avg: 83.14 [95.3%] std: 1.61 runs: 3
Elapsed
base: min: 317.28 max: 332.39 avg: 325.84 std: 6.33 runs: 3
rework: min: 281.53 [88.7%] max: 298.16 [89.7%] avg: 290.99 [89.3%] std: 6.98 runs: 3

System time improved slightly as well as Elapsed. My previous testing
has shown worse numbers but this again seem to depend on the storage
speed.

My theory is that the writeback doesn't catch up and prio-0 soft reclaim
falls into wait on writeback page too often in the base kernel. The
patched kernel doesn't do that because the soft reclaim is done from the
kswapd/direct reclaim context. This can be seen on the following graph
nicely. The A's group usage_in_bytes regurarly drops really low very often.

All 3 runs
http://labs.suse.cz/mhocko/soft_limit_rework/stream_io-vs-mem_eater/stream.png
resp. a detail of the single run
http://labs.suse.cz/mhocko/soft_limit_rework/stream_io-vs-mem_eater/stream-one-run.png

mem_eater seems to be doing better as well. It gets to the full
allocation size faster as can be seen on the following graph:
http://labs.suse.cz/mhocko/soft_limit_rework/stream_io-vs-mem_eater/mem_eater-one-run.png

/proc/meminfo collected during the test also shows that rework kernel
hasn't swapped that much (well almost not at all):
base: max: 123900 K avg: 56388.29 K
rework: max: 300 K avg: 128.68 K

kswapd and direct reclaim statistics are of no use unfortunatelly because
soft reclaim is not accounted properly as the counters are hidden by
global_reclaim() checks in the base kernel.

* TEST 3
========

Another test was the same configuration as TEST2 except the stream IO was
replaced by a single kbuild (16 parallel jobs bound to Node0 cpus same as
in TEST1) and mem_eater allocated TotalMem-200M so kbuild had only 200MB
left.

Kbuild did better with the rework kernel here as well:
User
base: min: 860.28 max: 872.86 avg: 868.03 std: 5.54 runs: 3
rework: min: 880.81 [102.4%] max: 887.45 [101.7%] avg: 883.56 [101.8%] std: 2.83 runs: 3
System
base: min: 84.35 max: 85.06 avg: 84.79 std: 0.31 runs: 3
rework: min: 85.62 [101.5%] max: 86.09 [101.2%] avg: 85.79 [101.2%] std: 0.21 runs: 3
Elapsed
base: min: 135.36 max: 243.30 avg: 182.47 std: 45.12 runs: 3
rework: min: 110.46 [81.6%] max: 116.20 [47.8%] avg: 114.15 [62.6%] std: 2.61 runs: 3
Minor
base: min: 36635476.00 max: 36673365.00 avg: 36654812.00 std: 15478.03 runs: 3
rework: min: 36639301.00 [100.0%] max: 36695541.00 [100.1%] avg: 36665511.00 [100.0%] std: 23118.23 runs: 3
Major
base: min: 14708.00 max: 53328.00 avg: 31379.00 std: 16202.24 runs: 3
rework: min: 302.00 [2.1%] max: 414.00 [0.8%] avg: 366.33 [1.2%] std: 47.22 runs: 3

Again we can see a significant improvement in Elapsed (it also seems to
be more stable), there is a huge dropdown for the Major page faults and
much more swapping:
base: max: 583736 K avg: 112547.43 K
rework: max: 4012 K avg: 124.36 K

Graphs from all three runs show the variability of the kbuild quite
nicely.  It even seems that it took longer after every run with the base
kernel which would be quite surprising as the source tree for the build is
removed and caches are dropped after each run so the build operates on a
freshly extracted sources everytime.
http://labs.suse.cz/mhocko/soft_limit_rework/stream_io-vs-mem_eater/kbuild-mem_eater.png

My other testing shows that this is just a matter of timing and other runs
behave differently the std for Elapsed time is similar ~50.  Example of
other three runs:
http://labs.suse.cz/mhocko/soft_limit_rework/stream_io-vs-mem_eater/kbuild-mem_eater2.png

So to wrap this up.  The series is still doing good and improves the soft
limit.

The testing results for bunch of cgroups with both stream IO and kbuild
loads can be found in "memcg: track children in soft limit excess to
improve soft limit".

This patch:

Memcg soft reclaim has been traditionally triggered from the global
reclaim paths before calling shrink_zone.  mem_cgroup_soft_limit_reclaim
then picked up a group which exceeds the soft limit the most and reclaimed
it with 0 priority to reclaim at least SWAP_CLUSTER_MAX pages.

The infrastructure requires per-node-zone trees which hold over-limit
groups and keep them up-to-date (via memcg_check_events) which is not cost
free.  Although this overhead hasn't turned out to be a bottle neck the
implementation is suboptimal because mem_cgroup_update_tree has no idea
which zones consumed memory over the limit so we could easily end up
having a group on a node-zone tree having only few pages from that
node-zone.

This patch doesn't try to fix node-zone trees management because it seems
that integrating soft reclaim into zone shrinking sounds much easier and
more appropriate for several reasons.  First of all 0 priority reclaim was
a crude hack which might lead to big stalls if the group's LRUs are big
and hard to reclaim (e.g.  a lot of dirty/writeback pages).  Soft reclaim
should be applicable also to the targeted reclaim which is awkward right
now without additional hacks.  Last but not least the whole infrastructure
eats quite some code.

After this patch shrink_zone is done in 2 passes.  First it tries to do
the soft reclaim if appropriate (only for global reclaim for now to keep
compatible with the original state) and fall back to ignoring soft limit
if no group is eligible to soft reclaim or nothing has been scanned during
the first pass.  Only groups which are over their soft limit or any of
their parents up the hierarchy is over the limit are considered eligible
during the first pass.

Soft limit tree which is not necessary anymore will be removed in the
follow up patch to make this patch smaller and easier to review.
Signed-off-by: NMichal Hocko <mhocko@suse.cz>
Reviewed-by: NGlauber Costa <glommer@openvz.org>
Reviewed-by: NTejun Heo <tj@kernel.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Ying Han <yinghan@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Glauber Costa <glommer@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

vfs guarantees the cgroup won't be destroyed, so it's redundant to get a
css reference.
Signed-off-by: NLi Zefan <lizefan@huawei.com>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

A memory cgroup with (1) multiple threshold notifications and (2) at least
one threshold >=2G was not reliable.  Specifically the notifications would
either not fire or would not fire in the proper order.

The __mem_cgroup_threshold() signaling logic depends on keeping 64 bit
thresholds in sorted order.  mem_cgroup_usage_register_event() sorts them
with compare_thresholds(), which returns the difference of two 64 bit
thresholds as an int.  If the difference is positive but has bit[31] set,
then sort() treats the difference as negative and breaks sort order.

This fix compares the two arbitrary 64 bit thresholds returning the
classic -1, 0, 1 result.

The test below sets two notifications (at 0x1000 and 0x81001000):
  cd /sys/fs/cgroup/memory
  mkdir x
  for x in 4096 2164264960; do
    cgroup_event_listener x/memory.usage_in_bytes $x | sed "s/^/$x listener:/" &
  done
  echo $$ > x/cgroup.procs
  anon_leaker 500M

v3.11-rc7 fails to signal the 4096 event listener:
  Leaking...
  Done leaking pages.

Patched v3.11-rc7 properly notifies:
  Leaking...
  4096 listener:2013:8:31:14:13:36
  Done leaking pages.

The fixed bug is old.  It appears to date back to the introduction of
memcg threshold notifications in v2.6.34-rc1-116-g2e72b634 "memcg:
implement memory thresholds"
Signed-off-by: NGreg Thelen <gthelen@google.com>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Acked-by: NKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: NJohannes Weiner <hannes@cmpxchg.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The memcg_cache_params structure contains the common part and the union,
which represents two different types of data: one for root cashes and
another for child caches.

The size of child data is fixed.  The size of the memcg_caches array is
calculated in runtime.

Currently the size of memcg_cache_params for root caches is calculated
incorrectly, because it includes the size of parameters for child caches.

ssize_t size = memcg_caches_array_size(num_groups);
size *= sizeof(void *);

size += sizeof(struct memcg_cache_params);

v2: Fix a typo in calculations
Signed-off-by: NAndrey Vagin <avagin@openvz.org>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The swapaccount kernel parameter without any values has been removed by
commit a2c8990a ("memsw: remove noswapaccount kernel parameter") but
it seems that we didn't get rid of all the left overs.

Make sure that menuconfig help text and kernel-parameters.txt are clear
about value for the paramter and remove the stalled comment which is not
very much useful on its own.
Signed-off-by: NMichal Hocko <mhocko@suse.cz>
Reported-by: NGergely Risko <gergely@risko.hu>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

struct memcg_cache_params has a union.  Different parts of this union
are used for root and non-root caches.  A part with destroying work is
used only for non-root caches.

I fixed the same problem in another place v3.9-rc1-16204-gf101a946, but
didn't notice this one.

This patch fixes the kernel panic:

[   46.848187] BUG: unable to handle kernel paging request at 000000fffffffeb8
[   46.849026] IP: [<ffffffff811a484c>] kmem_cache_destroy_memcg_children+0x6c/0xc0
[   46.849092] PGD 0
[   46.849092] Oops: 0000 [#1] SMP
...
Signed-off-by: NAndrey Vagin <avagin@openvz.org>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Cc: <stable@vger.kernel.org>    [3.9.x]
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Previously, all css descendant iterators didn't include the origin
(root of subtree) css in the iteration.  The reasons were maintaining
consistency with css_for_each_child() and that at the time of
introduction more use cases needed skipping the origin anyway;
however, given that css_is_descendant() considers self to be a
descendant, omitting the origin css has become more confusing and
looking at the accumulated use cases rather clearly indicates that
including origin would result in simpler code overall.

While this is a change which can easily lead to subtle bugs, cgroup
API including the iterators has recently gone through major
restructuring and no out-of-tree changes will be applicable without
adjustments making this a relatively acceptable opportunity for this
type of change.

The conversions are mostly straight-forward.  If the iteration block
had explicit origin handling before or after, it's moved inside the
iteration.  If not, if (pos == origin) continue; is added.  Some
conversions add extra reference get/put around origin handling by
consolidating origin handling and the rest.  While the extra ref
operations aren't strictly necessary, this shouldn't cause any
noticeable difference.
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NLi Zefan <lizefan@huawei.com>
Acked-by: NVivek Goyal <vgoyal@redhat.com>
Acked-by: NAristeu Rozanski <aris@redhat.com>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Matt Helsley <matthltc@us.ibm.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Balbir Singh <bsingharora@gmail.com>

cgroup is in the process of converting to css (cgroup_subsys_state)
from cgroup as the principal subsystem interface handle.  This is
mostly to prepare for the unified hierarchy support where css's will
be created and destroyed dynamically but also helps cleaning up
subsystem implementations as css is usually what they are interested
in anyway.

cftype->[un]register_event() is among the remaining couple interfaces
which still use struct cgroup.  Convert it to cgroup_subsys_state.
The conversion is mostly mechanical and removes the last users of
mem_cgroup_from_cont() and cg_to_vmpressure(), which are removed.

v2: indentation update as suggested by Li Zefan.
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NLi Zefan <lizefan@huawei.com>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Balbir Singh <bsingharora@gmail.com>

cgroup is in the process of converting to css (cgroup_subsys_state)
from cgroup as the principal subsystem interface handle.  This is
mostly to prepare for the unified hierarchy support where css's will
be created and destroyed dynamically but also helps cleaning up
subsystem implementations as css is usually what they are interested
in anyway.

This patch converts task iterators to deal with css instead of cgroup.
Note that under unified hierarchy, different sets of tasks will be
considered belonging to a given cgroup depending on the subsystem in
question and making the iterators deal with css instead cgroup
provides them with enough information about the iteration.

While at it, fix several function comment formats in cpuset.c.

This patch doesn't introduce any behavior differences.
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NLi Zefan <lizefan@huawei.com>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Matt Helsley <matthltc@us.ibm.com>

Currently all cgroup_task_iter functions require @cgrp to be passed
in, which is superflous and increases chance of usage error.  Make
cgroup_task_iter remember the cgroup being iterated and drop @cgrp
argument from next and end functions.

This patch doesn't introduce any behavior differences.
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NLi Zefan <lizefan@huawei.com>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Cc: Matt Helsley <matthltc@us.ibm.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Balbir Singh <bsingharora@gmail.com>

cgroup now has multiple iterators and it's quite confusing to have
something which walks over tasks of a single cgroup named cgroup_iter.
Let's rename it to cgroup_task_iter.

While at it, reformat / update comments and replace the overview
comment above the interface function decls with proper function
comments.  Such overview can be useful but function comments should be
more than enough here.

This is pure rename and doesn't introduce any functional changes.
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NLi Zefan <lizefan@huawei.com>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Cc: Matt Helsley <matthltc@us.ibm.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Balbir Singh <bsingharora@gmail.com>

cgroup is currently in the process of transitioning to using css
(cgroup_subsys_state) as the primary handle instead of cgroup in
subsystem API.  For hierarchy iterators, this is beneficial because

* In most cases, css is the only thing subsystems care about anyway.

* On the planned unified hierarchy, iterations for different
  subsystems will need to skip over different subtrees of the
  hierarchy depending on which subsystems are enabled on each cgroup.
  Passing around css makes it unnecessary to explicitly specify the
  subsystem in question as css is intersection between cgroup and
  subsystem

* For the planned unified hierarchy, css's would need to be created
  and destroyed dynamically independent from cgroup hierarchy.  Having
  cgroup core manage css iteration makes enforcing deref rules a lot
  easier.

Most subsystem conversions are straight-forward.  Noteworthy changes
are

* blkio: cgroup_to_blkcg() is no longer used.  Removed.

* freezer: cgroup_freezer() is no longer used.  Removed.

* devices: cgroup_to_devcgroup() is no longer used.  Removed.
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NLi Zefan <lizefan@huawei.com>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Acked-by: NVivek Goyal <vgoyal@redhat.com>
Acked-by: NAristeu Rozanski <aris@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Matt Helsley <matthltc@us.ibm.com>
Cc: Jens Axboe <axboe@kernel.dk>

cgroup is currently in the process of transitioning to using struct
cgroup_subsys_state * as the primary handle instead of struct cgroup.
Please see the previous commit which converts the subsystem methods
for rationale.

This patch converts all cftype file operations to take @css instead of
@cgroup.  cftypes for the cgroup core files don't have their subsytem
pointer set.  These will automatically use the dummy_css added by the
previous patch and can be converted the same way.

Most subsystem conversions are straight forwards but there are some
interesting ones.

* freezer: update_if_frozen() is also converted to take @css instead
  of @cgroup for consistency.  This will make the code look simpler
  too once iterators are converted to use css.

* memory/vmpressure: mem_cgroup_from_css() needs to be exported to
  vmpressure while mem_cgroup_from_cont() can be made static.
  Updated accordingly.

* cpu: cgroup_tg() doesn't have any user left.  Removed.

* cpuacct: cgroup_ca() doesn't have any user left.  Removed.

* hugetlb: hugetlb_cgroup_form_cgroup() doesn't have any user left.
  Removed.

* net_cls: cgrp_cls_state() doesn't have any user left.  Removed.
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NLi Zefan <lizefan@huawei.com>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Acked-by: NVivek Goyal <vgoyal@redhat.com>
Acked-by: NAristeu Rozanski <aris@redhat.com>
Acked-by: NDaniel Wagner <daniel.wagner@bmw-carit.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Matt Helsley <matthltc@us.ibm.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Steven Rostedt <rostedt@goodmis.org>

cgroup is currently in the process of transitioning to using struct
cgroup_subsys_state * as the primary handle instead of struct cgroup *
in subsystem implementations for the following reasons.

* With unified hierarchy, subsystems will be dynamically bound and
  unbound from cgroups and thus css's (cgroup_subsys_state) may be
  created and destroyed dynamically over the lifetime of a cgroup,
  which is different from the current state where all css's are
  allocated and destroyed together with the associated cgroup.  This
  in turn means that cgroup_css() should be synchronized and may
  return NULL, making it more cumbersome to use.

* Differing levels of per-subsystem granularity in the unified
  hierarchy means that the task and descendant iterators should behave
  differently depending on the specific subsystem the iteration is
  being performed for.

* In majority of the cases, subsystems only care about its part in the
  cgroup hierarchy - ie. the hierarchy of css's.  Subsystem methods
  often obtain the matching css pointer from the cgroup and don't
  bother with the cgroup pointer itself.  Passing around css fits
  much better.

This patch converts all cgroup_subsys methods to take @css instead of
@cgroup.  The conversions are mostly straight-forward.  A few
noteworthy changes are

* ->css_alloc() now takes css of the parent cgroup rather than the
  pointer to the new cgroup as the css for the new cgroup doesn't
  exist yet.  Knowing the parent css is enough for all the existing
  subsystems.

* In kernel/cgroup.c::offline_css(), unnecessary open coded css
  dereference is replaced with local variable access.

This patch shouldn't cause any behavior differences.

v2: Unnecessary explicit cgrp->subsys[] deref in css_online() replaced
    with local variable @css as suggested by Li Zefan.

    Rebased on top of new for-3.12 which includes for-3.11-fixes so
    that ->css_free() invocation added by da0a12ca ("cgroup: fix a
    leak when percpu_ref_init() fails") is converted too.  Suggested
    by Li Zefan.
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NLi Zefan <lizefan@huawei.com>
Acked-by: NMichal Hocko <mhocko@suse.cz>
Acked-by: NVivek Goyal <vgoyal@redhat.com>
Acked-by: NAristeu Rozanski <aris@redhat.com>
Acked-by: NDaniel Wagner <daniel.wagner@bmw-carit.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Matt Helsley <matthltc@us.ibm.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Steven Rostedt <rostedt@goodmis.org>

Currently, controllers have to explicitly follow the cgroup hierarchy
to find the parent of a given css.  cgroup is moving towards using
cgroup_subsys_state as the main controller interface construct, so
let's provide a way to climb the hierarchy using just csses.

This patch implements css_parent() which, given a css, returns its
parent.  The function is guarnateed to valid non-NULL parent css as
long as the target css is not at the top of the hierarchy.

freezer, cpuset, cpu, cpuacct, hugetlb, memory, net_cls and devices
are converted to use css_parent() instead of accessing cgroup->parent
directly.

* __parent_ca() is dropped from cpuacct and its usage is replaced with
  parent_ca().  The only difference between the two was NULL test on
  cgroup->parent which is now embedded in css_parent() making the
  distinction moot.  Note that eventually a css->parent field will be
  added to css and the NULL check in css_parent() will go away.

This patch shouldn't cause any behavior differences.
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NLi Zefan <lizefan@huawei.com>

css (cgroup_subsys_state) is usually embedded in a subsys specific
data structure.  Subsystems either use container_of() directly to cast
from css to such data structure or has an accessor function wrapping
such cast.  As cgroup as whole is moving towards using css as the main
interface handle, add and update such accessors to ease dealing with
css's.

All accessors explicitly handle NULL input and return NULL in those
cases.  While this looks like an extra branch in the code, as all
controllers specific data structures have css as the first field, the
casting doesn't involve any offsetting and the compiler can trivially
optimize out the branch.

* blkio, freezer, cpuset, cpu, cpuacct and net_cls didn't have such
  accessor.  Added.

* memory, hugetlb and devices already had one but didn't explicitly
  handle NULL input.  Updated.
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NLi Zefan <lizefan@huawei.com>

The names of the two struct cgroup_subsys_state accessors -
cgroup_subsys_state() and task_subsys_state() - are somewhat awkward.
The former clashes with the type name and the latter doesn't even
indicate it's somehow related to cgroup.

We're about to revamp large portion of cgroup API, so, let's rename
them so that they're less awkward.  Most per-controller usages of the
accessors are localized in accessor wrappers and given the amount of
scheduled changes, this isn't gonna add any noticeable headache.

Rename cgroup_subsys_state() to cgroup_css() and task_subsys_state()
to task_css().  This patch is pure rename.
Signed-off-by: NTejun Heo <tj@kernel.org>
Acked-by: NLi Zefan <lizefan@huawei.com>