Skip to content

cloud-kernel
cloud-kernel

openanolis / cloud-kernel
1 年多 前同步成功

通知 160
Star 36
Fork 7
cloud-kernel
cloud-kernel
切换分支/标签
  1. 17 1月, 2020 10 次提交
  3. 15 1月, 2020 30 次提交
    • Y
      alinux: mm: thp: move deferred split queue to memcg's nodeinfo · e6ca020b
      由 Yang Shi 提交于 
      The commit 87eaceb3faa59b9b4d940ec9554ce251325d83fe ("mm: thp: make
deferred split shrinker memcg aware") makes deferred split queue per
memcg to resolve memcg pre-mature OOM problem.  But, all nodes end up
sharing the same queue instead of one queue per-node before the commit.
It is not a big deal for memcg limit reclaim, but it may cause global
kswapd shrink THPs from a different node.

And, 0-day testing reported -19.6% regression of stress-ng's madvise
test [1].  I didn't see that much regression on my test box (24 threads,
48GB memory, 2 nodes), with the same test (stress-ng --timeout 1
--metrics-brief --sequential 72  --class vm --exclude spawn,exec), I saw
average -3% (run the same test 10 times then calculate the average since
the test itself may have most 15% variation according to my test)
regression sometimes (not every time, sometimes I didn't see regression
at all).

This might be caused by deferred split queue lock contention.  With some
configuration (i.e. just one root memcg) the lock contention my be worse
than before (given 2 nodes, two locks are reduced to one lock).

So, moving deferred split queue to memcg's nodeinfo to make it NUMA
aware again.

With this change stress-ng's madvise test shows average 4% improvement
sometimes and I didn't see degradation anymore.

[1]: https://lore.kernel.org/lkml/20190930084604.GC17687@shao2-debian/

Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: NYang Shi <yang.shi@linux.alibaba.com>
Reviewed-by: NXunlei Pang <xlpang@linux.alibaba.com>
      e6ca020b
    • Y
      mm: thp: make deferred split shrinker memcg aware · d651fcbb
      由 Yang Shi 提交于 
      commit 87eaceb3faa59b9b4d940ec9554ce251325d83fe upstream

Currently THP deferred split shrinker is not memcg aware, this may cause
premature OOM with some configuration.  For example the below test would
run into premature OOM easily:

$ cgcreate -g memory:thp
$ echo 4G > /sys/fs/cgroup/memory/thp/memory/limit_in_bytes
$ cgexec -g memory:thp transhuge-stress 4000

transhuge-stress comes from kernel selftest.

It is easy to hit OOM, but there are still a lot THP on the deferred
split queue, memcg direct reclaim can't touch them since the deferred split
shrinker is not memcg aware.

Convert deferred split shrinker memcg aware by introducing per memcg
deferred split queue.  The THP should be on either per node or per memcg
deferred split queue if it belongs to a memcg.  When the page is
immigrated to the other memcg, it will be immigrated to the target
memcg's deferred split queue too.

Reuse the second tail page's deferred_list for per memcg list since the
same THP can't be on multiple deferred split queues.

[yang.shi@linux.alibaba.com: simplify deferred split queue dereference per Kirill Tkhai]
  Link: http://lkml.kernel.org/r/1566496227-84952-5-git-send-email-yang.shi@linux.alibaba.com
Link: http://lkml.kernel.org/r/1565144277-36240-5-git-send-email-yang.shi@linux.alibaba.comSigned-off-by: NYang Shi <yang.shi@linux.alibaba.com>
Acked-by: NKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: NKirill Tkhai <ktkhai@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Qian Cai <cai@lca.pw>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>
Reviewed-by: NXunlei Pang <xlpang@linux.alibaba.com>
      d651fcbb
    • Y
      mm: shrinker: make shrinker not depend on memcg kmem · bd5596b4
      由 Yang Shi 提交于 
      commit 0a432dcbeb32edcd211a5d8f7847d0da7642a8b4 upstream

Currently shrinker is just allocated and can work when memcg kmem is
enabled.  But, THP deferred split shrinker is not slab shrinker, it
doesn't make too much sense to have such shrinker depend on memcg kmem.
It should be able to reclaim THP even though memcg kmem is disabled.

Introduce a new shrinker flag, SHRINKER_NONSLAB, for non-slab shrinker.
When memcg kmem is disabled, just such shrinkers can be called in
shrinking memcg slab.

[yang.shi@linux.alibaba.com: add comment]
  Link: http://lkml.kernel.org/r/1566496227-84952-4-git-send-email-yang.shi@linux.alibaba.com
Link: http://lkml.kernel.org/r/1565144277-36240-4-git-send-email-yang.shi@linux.alibaba.comSigned-off-by: NYang Shi <yang.shi@linux.alibaba.com>
Acked-by: NKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: NKirill Tkhai <ktkhai@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Qian Cai <cai@lca.pw>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>
Reviewed-by: NXunlei Pang <xlpang@linux.alibaba.com>
      bd5596b4
    • Y
      mm: move mem_cgroup_uncharge out of __page_cache_release() · 9b78918c
      由 Yang Shi 提交于 
      commit 7ae88534cdd96235cd775c03b32a75009355740b upstream

A later patch makes THP deferred split shrinker memcg aware, but it
needs page->mem_cgroup information in THP destructor, which is called after
mem_cgroup_uncharge() now.

So move mem_cgroup_uncharge() from __page_cache_release() to compound
page destructor, which is called by both THP and other compound pages except
HugeTLB.  And call it in __put_single_page() for single order page.

Link: http://lkml.kernel.org/r/1565144277-36240-3-git-send-email-yang.shi@linux.alibaba.comSigned-off-by: NYang Shi <yang.shi@linux.alibaba.com>
Suggested-by: N"Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Acked-by: NKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: NKirill Tkhai <ktkhai@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Qian Cai <cai@lca.pw>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>
Reviewed-by: NXunlei Pang <xlpang@linux.alibaba.com>
      9b78918c
    • Y
      mm: thp: extract split_queue_* into a struct · e65b6961
      由 Yang Shi 提交于 
      commit 364c1eebe453f06f0c1e837eb155a5725c9cd272 upstream

Patch series "Make deferred split shrinker memcg aware", v6.

Currently THP deferred split shrinker is not memcg aware, this may cause
premature OOM with some configuration.  For example the below test would
run into premature OOM easily:

$ cgcreate -g memory:thp
$ echo 4G > /sys/fs/cgroup/memory/thp/memory/limit_in_bytes
$ cgexec -g memory:thp transhuge-stress 4000

transhuge-stress comes from kernel selftest.

It is easy to hit OOM, but there are still a lot THP on the deferred
split queue, memcg direct reclaim can't touch them since the deferred split
shrinker is not memcg aware.

Convert deferred split shrinker memcg aware by introducing per memcg
deferred split queue.  The THP should be on either per node or per memcg
deferred split queue if it belongs to a memcg.  When the page is
immigrated to the other memcg, it will be immigrated to the target
memcg's deferred split queue too.

Reuse the second tail page's deferred_list for per memcg list since the
same THP can't be on multiple deferred split queues.

Make deferred split shrinker not depend on memcg kmem since it is not
slab.  It doesn't make sense to not shrink THP even though memcg kmem is
disabled.

With the above change the test demonstrated above doesn't trigger OOM
even though with cgroup.memory=nokmem.

This patch (of 4):

Put split_queue, split_queue_lock and split_queue_len into a struct in
order to reduce code duplication when we convert deferred_split to memcg
aware in the later patches.

Link: http://lkml.kernel.org/r/1565144277-36240-2-git-send-email-yang.shi@linux.alibaba.comSigned-off-by: NYang Shi <yang.shi@linux.alibaba.com>
Suggested-by: N"Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Acked-by: NKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: NKirill Tkhai <ktkhai@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Qian Cai <cai@lca.pw>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>
Reviewed-by: NXunlei Pang <xlpang@linux.alibaba.com>
      e65b6961
    • G
      alinux: mm: Support kidled · a29243e2
      由 Gavin Shan 提交于 
      This enables scanning pages in fixed interval to determine their access
frequency (hot/cold). The result is exported to user land on basis of
memory cgroup by "memory.idle_page_stats". The design is highlighted as
below:

   * A kernel thread is spawn when this feature is enabled by writing
     non-zero value to "/sys/kernel/mm/kidled/scan_period_in_seconds".
     The thread sequentially scans the nodes and their pages that have
     been chained up in LRU list.

   * For each page, its corresponding age information is stored in the
     page flags or array in node. The age represents the scanning intervals
     in which the page isn't accessed. Also, the page flag (PG_idle) is
     leveraged. The page's age is increased by one if the idle flag isn't
     cleared in two consective scans. Otherwise, the page's age is cleared out.
     Also, the page's age information is cleared when it's free'd so that
     the stale age information won't be fetched when it's allocated.

   * Initially, the flag is set, while the access bit in its PTE is cleared
     out by the thread. In next scanning period, its PTE access bit is
     synchronized with the page flag: clear the flag if access bit is set.
     The flag is kept otherwise. For unmapped pages, the flag is cleared
     when it's accessed.

   * Eventually, the page's aging information is updated to the unstable
     bucket of its corresponding memory cgroup, taking as statistics. The
     unstable bucket (statistics) is copied to stable bucket when all pages
     in all nodes are scanned for once. The stable bucket (statistics) is
     exported to user land through "memory.idle_page_stats".

TESTING
=======

   * cgroup1, unmapped pagecache

     # dd if=/dev/zero of=/ext4/test.data oflag=direct bs=1M count=128
     #
     # echo 1 > /sys/kernel/mm/kidled/use_hierarchy
     # echo 15 > /sys/kernel/mm/kidled/scan_period_in_seconds
     # mkdir -p /cgroup/memory
     # mount -tcgroup -o memory /cgroup/memory
     # echo 1 > /cgroup/memory/memory.use_hierarchy
     # mkdir -p /cgroup/memory/test
     # echo 1 > /cgroup/memory/test/memory.use_hierarchy
     #
     # echo $$ > /cgroup/memory/test/cgroup.procs
     # dd if=/ext4/test.data of=/dev/null bs=1M count=128
     # < wait a few minutes >
     # cat /cgroup/memory/test/memory.idle_page_stats | grep cfei
     # cat /cgroup/memory/test/memory.idle_page_stats | grep cfei
       cfei   0   0   0   134217728   0   0   0   0
     # cat /cgroup/memory/memory.idle_page_stats | grep cfei
       cfei   0   0   0   134217728   0   0   0   0

   * cgroup1, mapped pagecache

     # < create same file and memory cgroups as above >
     #
     # echo $$ > /cgroup/memory/test/cgroup.procs
     # < run program to mmap the whole created file and access the area >
     # < wait a few minutes >
     # cat /cgroup/memory/test/memory.idle_page_stats | grep cfei
       cfei   0   134217728   0   0   0   0   0   0
     # cat /cgroup/memory/memory.idle_page_stats | grep cfei
       cfei   0   134217728   0   0   0   0   0   0

   * cgroup1, mapped and locked pagecache

     # < create same file and memory cgroups as above >
     #
     # echo $$ > /cgroup/memory/test/cgroup.procs
     # < run program to mmap the whole created file and mlock the area >
     # < wait a few minutes >
     # cat /cgroup/memory/test/memory.idle_page_stats | grep cfui
       cfui   0   134217728   0   0   0   0   0   0
     # cat /cgroup/memory/memory.idle_page_stats | grep cfui
       cfui   0   134217728   0   0   0   0   0   0

   * cgroup1, anonymous and locked area

     # < create memory cgroups as above >
     #
     # echo $$ > /cgroup/memory/test/cgroup.procs
     # < run program to mmap anonymous area and mlock it >
     # < wait a few minutes >
     # cat /cgroup/memory/test/memory.idle_page_stats | grep csui
       csui   0   0   134217728   0   0   0   0   0
     # cat /cgroup/memory/memory.idle_page_stats | grep csui
       csui   0   0   134217728   0   0   0   0   0

   * Rerun above test cases in cgroup2 and the results are no exceptional.
     However, the cgroups are populated in different way as below:

     # mkdir -p /cgroup
     # mount -tcgroup2 none /cgroup
     # echo "+memory" > /cgroup/cgroup.subtree_control
     # mkdir -p /cgroup/test
Signed-off-by: NGavin Shan <shan.gavin@linux.alibaba.com>
Reviewed-by: NYang Shi <yang.shi@linux.alibaba.com>
Reviewed-by: NXunlei Pang <xlpang@linux.alibaba.com>
      a29243e2
    • Y
      alinux: mm: memcontrol: make distance between wmark_low and wmark_high configurable · bbaee3af
      由 Yang Shi 提交于 
      Introduce a new interface, wmark_scale_factor, which defines the
distance between wmark_high and wmark_low.  The unit is in fractions of
10,000. The default value of 50 means the distance between wmark_high
and wmark_low is 0.5% of the max limit of the cgroup.  The maximum value
is 1000, or 10% of the max limit.

The distance between wmark_low and wmark_high have impact on how hard
memcg kswapd would reclaim.
Reviewed-by: NGavin Shan <shan.gavin@linux.alibaba.com>
Reviewed-by: NXunlei Pang <xlpang@linux.alibaba.com>
Signed-off-by: NYang Shi <yang.shi@linux.alibaba.com>
      bbaee3af
    • Y
      alinux: mm: vmscan: make memcg kswapd set memcg state to dirty or writeback · c69c12cc
      由 Yang Shi 提交于 
      The global kswapd could set memory node to dirty or writeback if current
scan find all pages are unqueued dirty or writeback.  Then kswapd would
write out dirty pages or wait for writeback done.  The memcg kswapd
behaves like global kswapd, and it should set dirty or writeback state
to memcg too if the same condition is met.

Since direct reclaim can't write out page caches, the system depends on
kswapd to write out dirty pages if scan finds too many dirty pages in
order to avoid pre-mature OOM.  But, if page cache is dirtied too fast,
writing out pages definitely can't catch up with dirtying pages.  It is
the responsibility of dirty page balance to throttle dirtying pages.
Reviewed-by: NGavin Shan <shan.gavin@linux.alibaba.com>
Reviewed-by: NXunlei Pang <xlpang@linux.alibaba.com>
Signed-off-by: NYang Shi <yang.shi@linux.alibaba.com>
      c69c12cc
    • Y
      alinux: mm: memcontrol: treat memcg wmark reclaim work as kswapd · 6332d4e3
      由 Yang Shi 提交于 
      Since background water mark reclaim is scheduled by workqueue, it could
do more work than direct reclaim, i.e. write out dirty page, etc.

So, add PF_KSWAPD flag, so that current_is_kswapd() would return true
for memcg background reclaim.  The condition "current_is_kswapd() &&
!global_reclaim(sc)" is good enough to tell current is global kswapd or
memcg background reclaim.

And, kswapd is not allowed to break memory.low protection for now, memcg
kswapd should not break it either.
Reviewed-by: NGavin Shan <shan.gavin@linux.alibaba.com>
Reviewed-by: NXunlei Pang <xlpang@linux.alibaba.com>
Signed-off-by: NYang Shi <yang.shi@linux.alibaba.com>
      6332d4e3
    • Y
      alinux: mm: memcontrol: add background reclaim support for cgroupv2 · 0149d7b9
      由 Yang Shi 提交于 
      Like v1, add background reclaim support for cgroup v2.  The interfaces
are exactly same with v1.  However, if high limit is setup for v2, the
water mark would be calculated by high limit instead of max limit.
Reviewed-by: NGavin Shan <shan.gavin@linux.alibaba.com>
Reviewed-by: NXunlei Pang <xlpang@linux.alibaba.com>
Signed-off-by: NYang Shi <yang.shi@linux.alibaba.com>
      0149d7b9
    • Y
      alinux: mm: memcontrol: support background async page reclaim · 6967792f
      由 Yang Shi 提交于 
      Currently when memory usage exceeds memory cgroup limit, memory cgroup
just can do sync direct reclaim.  This may incur unexpected stall on
some applications which are sensitive to latency.  Introduce background
async page reclaim mechanism, like what kswapd does.

Define memcg memory usage water mark by introducing wmark_ratio interface,
which is from 0 to 100 and represents percentage of max limit.  The
wmark_high is calculated by (max * wmark_ratio / 100), the wmark_low is
(wmark_high - wmark_high >> 8), which is an empirical value.  If wmark_ratio
is 0, it means water mark is disabled, both wmark_low and wmark_high is max,
which is the default value.

If wmark_ratio is setup, when charging page, if usage is greater than
wmark_high, which means the available memory of memcg is low, a work
would be scheduled to do background page reclaim until memory usage is
reduced to wmark_low if possible.

Define a dedicated unbound workqueue for scheduling water mark reclaim
works.
Reviewed-by: NGavin Shan <shan.gavin@linux.alibaba.com>
Reviewed-by: NXunlei Pang <xlpang@linux.alibaba.com>
Signed-off-by: NYang Shi <yang.shi@linux.alibaba.com>
      6967792f
    • Y
      alinux: mm: vmscan: make it sane reclaim if cgwb_v1 is enabled · 49a3b465
      由 Yang Shi 提交于 
      AliOS Cloud Kernel has cgroup writeback support for v1, so the reclaim could be
treated as sane reclaim if cgwb_v1 is enabled.
Reviewed-by: NGavin Shan <shan.gavin@linux.alibaba.com>
Reviewed-by: NXunlei Pang <xlpang@linux.alibaba.com>
Reviewed-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Reviewed-by: NCaspar Zhang <caspar@linux.alibaba.com>
Signed-off-by: NYang Shi <yang.shi@linux.alibaba.com>
      49a3b465
    • J
      alinux: iocost: rename weight to cost.weight to avoid conflict with cfq · 7d59e27f
      由 Jiufei Xue 提交于 
      Signed-off-by: NJiufei Xue <jiufei.xue@linux.alibaba.com>
Reviewed-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
      7d59e27f
    • J
      alinux: ovl: implement async IO routines · 3e5dd02b
      由 Jiufei Xue 提交于 
      A performance regression is observed since linux v4.19 when we do aio
test using fio with iodepth 128 on overlayfs. And we found that queue
depth of the device is always 1 which is unexpected.

After investigation, it is found that commit 16914e6f
("ovl: add ovl_read_iter()") and commit 2a92e07e
("ovl: add ovl_write_iter()") use do_iter_readv_writev() to submit
requests to real filesystem. Async IOs are converted to sync IOs here
and cause performance regression.

So implement async IO for stacked reading and writing.

Changes since v1:
  - add a cleanup helper for completion/error handling
  - handle the case when aio_req allocation failed
Signed-off-by: NJiufei Xue <jiufei.xue@linux.alibaba.com>
Reviewed-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
      3e5dd02b
    • J
      alinux: vfs: add vfs_iocb_iter_[read|write] helper functions · 6011bef7
      由 Jiufei Xue 提交于 
      This isn't cause any behavior changes and will be used by overlay
async IO implementation.
Signed-off-by: NJiufei Xue <jiufei.xue@linux.alibaba.com>
Reviewed-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
      6011bef7
    • X
      alinux: mm, memcg: fix possible soft lockup in try_charge · 2ced6afd
      由 Xu Yu 提交于 
      When events such as direct reclaim and oom occur intensively, soft
lockup is very likely to happen in the instances with 1 vcpu and with
kernel preempt disabled.

The example soft lockup is as follows.

[  160.555984] watchdog: BUG: soft lockup - CPU#0 stuck for 112s! [malloc:2188]
[  160.557975] Modules linked in: button
[  160.559495] CPU: 0 PID: 2188 Comm: malloc Not tainted 4.19.57-15.457.al7.x86_64 #1
[  160.561546] Hardware name: Alibaba Cloud Alibaba Cloud ECS, BIOS 3288b3c 04/01/2014
[  160.563707] RIP: 0010:shrink_node+0x1ae/0x450
[  160.565391] Code: 00 00 00 49 8b 4f 20 ba 01 00 00 00 4c 8b 74 24 10 4d 8b 47 28 49 8b 77 10 48 2b 4c 24 08 41 8b 7f 1c 4d8
[  160.570747] RSP: 0000:ffff9d0ec07a3b58 EFLAGS: 00000286 ORIG_RAX: ffffffffffffff13
[  160.572889] RAX: ffff982ab6014330 RBX: ffff982ab6014000 RCX: 0000000000000000
[  160.574992] RDX: 0000000000000001 RSI: ffff982ab6014000 RDI: ffff982ab6014000
[  160.577106] RBP: ffff982afffb6000 R08: 0000000000000000 R09: ffff982ab6014000
[  160.579219] R10: 0000000000000004 R11: 0000000000aaaaaa R12: 0000000000000000
[  160.581326] R13: 0000000000000000 R14: 0000000000000000 R15: ffff9d0ec07a3c50
[  160.583450] FS:  00007f8b414f7740(0000) GS:ffff982afda00000(0000) knlGS:0000000000000000
[  160.585704] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  160.587662] CR2: 00007f8adb800010 CR3: 000000007ac9e001 CR4: 00000000003606b0
[  160.589835] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  160.591971] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[  160.594133] Call Trace:
[  160.595602]  do_try_to_free_pages+0xcc/0x390
[  160.597356]  try_to_free_mem_cgroup_pages+0xf9/0x1d0
[  160.599198]  ? out_of_memory+0xb5/0x4a0
[  160.600882]  try_charge+0x244/0x750
[  160.602522]  ? __pagevec_lru_add_fn+0x1d0/0x330
[  160.604310]  mem_cgroup_try_charge+0xb4/0x1d0
[  160.606085]  mem_cgroup_try_charge_delay+0x1c/0x40
[  160.607892]  do_anonymous_page+0xf7/0x540
[  160.609574]  __handle_mm_fault+0x665/0xa00
[  160.611233]  ? __switch_to_asm+0x35/0x70
[  160.612838]  handle_mm_fault+0x122/0x1e0
[  160.614407]  __do_page_fault+0x1b7/0x470
[  160.615962]  do_page_fault+0x32/0x140
[  160.617474]  ? async_page_fault+0x8/0x30
[  160.619012]  async_page_fault+0x1e/0x30
[  160.620526] RIP: 0033:0x40068e

Fix it by adding cond_resched() in try_charge(), just before goto retry
after OOM_SUCCESS, in order to let OOM free some memory first.
Signed-off-by: NXu Yu <xuyu@linux.alibaba.com>
Reviewed-by: NYang Shi <yang.shi@linux.alibaba.com>
Reviewed-by: NXunlei Pang <xlpang@linux.alibaba.com>
      2ced6afd
    • J
      alinux: iocost: add ioc_gq stat · 0670363c
      由 Jiufei Xue 提交于 
      Add a stat file to monitor the ioc_gq stat.
Signed-off-by: NJiufei Xue <jiufei.xue@linux.alibaba.com>
Reviewed-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
      0670363c
    • J
      dm thin: wakeup worker only when deferred bios exist · 88c0b3cc
      由 Jeffle Xu 提交于 
      commit d256d796279de0bdc227ff4daef565aa7e80c898 upstream.

Single thread fio test (read, bs=4k, ioengine=libaio, iodepth=128,
numjobs=1) over dm-thin device has poor performance versus bare nvme
device.

Further investigation with perf indicates that queue_work_on() consumes
over 20% CPU time when doing IO over dm-thin device. The call stack is
as follows.

- 40.57% thin_map
    + 22.07% queue_work_on
    + 9.95% dm_thin_find_block
    + 2.80% cell_defer_no_holder
      1.91% inc_all_io_entry.isra.33.part.34
    + 1.78% bio_detain.isra.35

In cell_defer_no_holder(), wakeup_worker() is always called, no matter
whether the tc->deferred_bio_list list is empty or not. In single thread
IO model, this list is most likely empty. So skip waking up worker thread
if tc->deferred_bio_list list is empty.

Single thread IO performance improves from 448 MiB/s to 646 MiB/s (+44%)
once the needless wake_worker() calls are properly skipped.
Signed-off-by: NJeffle Xu <jefflexu@linux.alibaba.com>
Signed-off-by: NMike Snitzer <snitzer@redhat.com>
Reviewed-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Signed-off-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
      88c0b3cc
    • X
      alinux: blk-throttle: limit bios to fix amount of pages entering writeback prematurely · 0fd4aa6d
      由 Xiaoguang Wang 提交于 
      Currently in blk_throtl_bio(), if one bio exceeds its throtl_grp's bps
or iops limit, this bio will be queued throtl_grp's throtl_service_queue,
then obviously mm subsys will submit more pages, even underlying device
can not handle these io requests, also this will make large amount of pages
entering writeback prematurely, later if some process writes some of these
pages, it will wait for long time.

I have done some tests: one process does buffered writes on a 1GB file,
and make this process's blkcg max bps limit be 10MB/s, I observe this:
	#cat /proc/meminfo  | grep -i back
	Writeback:        900024 kB
	WritebackTmp:          0 kB

I think this Writeback value is just too big, indeed many bios have been
queued in throtl_grp's throtl_service_queue, if one process try to write
the last bio's page in this queue, it will call wait_on_page_writeback(page),
which must wait the previous bios to finish and will take long time, we
have also see 120s hung task warning in our server.

 INFO: task kworker/u128:0:30072 blocked for more than 120 seconds.
       Tainted: G            E 4.9.147-013.ali3000_015_test.alios7.x86_64 #1
 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
 kworker/u128:0  D    0 30072      2 0x00000000
 Workqueue: writeback wb_workfn (flush-8:16)
  ffff882ddd066b40 0000000000000000 ffff882e5cad3400 ffff882fbe959e80
  ffff882fa50b1a00 ffffc9003a5a3768 ffffffff8173325d ffffc9003a5a3780
  00ff882e5cad3400 ffff882fbe959e80 ffffffff81360b49 ffff882e5cad3400
 Call Trace:
  [<ffffffff8173325d>] ? __schedule+0x23d/0x6d0
  [<ffffffff81360b49>] ? alloc_request_struct+0x19/0x20
  [<ffffffff81733726>] schedule+0x36/0x80
  [<ffffffff81736c56>] schedule_timeout+0x206/0x4b0
  [<ffffffff81036c69>] ? sched_clock+0x9/0x10
  [<ffffffff81363073>] ? get_request+0x403/0x810
  [<ffffffff8110ca10>] ? ktime_get+0x40/0xb0
  [<ffffffff81732f8a>] io_schedule_timeout+0xda/0x170
  [<ffffffff81733f90>] ? bit_wait+0x60/0x60
  [<ffffffff81733fab>] bit_wait_io+0x1b/0x60
  [<ffffffff81733b28>] __wait_on_bit+0x58/0x90
  [<ffffffff811b0d91>] ? find_get_pages_tag+0x161/0x2e0
  [<ffffffff811aff62>] wait_on_page_bit+0x82/0xa0
  [<ffffffff810d47f0>] ? wake_atomic_t_function+0x60/0x60
  [<ffffffffa02fc181>] mpage_prepare_extent_to_map+0x2d1/0x310 [ext4]
  [<ffffffff8121ff65>] ? kmem_cache_alloc+0x185/0x1a0
  [<ffffffffa0305a2f>] ? ext4_init_io_end+0x1f/0x40 [ext4]
  [<ffffffffa0300294>] ext4_writepages+0x404/0xef0 [ext4]
  [<ffffffff81508c64>] ? scsi_init_io+0x44/0x200
  [<ffffffff81398a0f>] ? fprop_fraction_percpu+0x2f/0x80
  [<ffffffff811c139e>] do_writepages+0x1e/0x30
  [<ffffffff8127c0f5>] __writeback_single_inode+0x45/0x320
  [<ffffffff8127c942>] writeback_sb_inodes+0x272/0x600
  [<ffffffff8127cf6b>] wb_writeback+0x10b/0x300
  [<ffffffff8127d884>] wb_workfn+0xb4/0x380
  [<ffffffff810b85e9>] ? try_to_wake_up+0x59/0x3e0
  [<ffffffff810a5759>] process_one_work+0x189/0x420
  [<ffffffff810a5a3e>] worker_thread+0x4e/0x4b0
  [<ffffffff810a59f0>] ? process_one_work+0x420/0x420
  [<ffffffff810ac026>] kthread+0xe6/0x100
  [<ffffffff810abf40>] ? kthread_park+0x60/0x60
  [<ffffffff81738499>] ret_from_fork+0x39/0x50

To fix this issue, we can simply limit throtl_service_queue's max queued
bios, currently we limit it to throtl_grp's bps_limit or iops limit, if it
still exteeds, we just sleep for a while.
Signed-off-by: NXiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
Reviewed-by: NLiu Bo <bo.liu@linux.alibaba.com>
Signed-off-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Acked-by: NCaspar Zhang <caspar@linux.alibaba.com>
      0fd4aa6d
    • J
      alinux: block-throttle: add counters for completed io · 33ed5f09
      由 Jiufei Xue 提交于 
      Now we have counters for wait_time and service_time, but no completed
ios, so the average latency can not be measured.
Signed-off-by: NJiufei Xue <jiufei.xue@linux.alibaba.com>
Signed-off-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Reviewed-by: NXiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
Acked-by: NCaspar Zhang <caspar@linux.alibaba.com>
      33ed5f09
    • J
      alinux: block-throttle: code cleanup · b03ba65b
      由 Jiufei Xue 提交于 
      This patch does the code cleanup because the seq_show handlers for tg
counters are the same. No functional changes.
Signed-off-by: NJiufei Xue <jiufei.xue@linux.alibaba.com>
Signed-off-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Acked-by: NCaspar Zhang <caspar@linux.alibaba.com>
      b03ba65b
    • J
      alinux: blk-throttle: add throttled io/bytes counter · 766cfe98
      由 Joseph Qi 提交于 
      Add 2 interfaces to stat io throttle information:
  blkio.throttle.total_io_queued
  blkio.throttle.total_bytes_queued

These interfaces are used for monitoring throttled io/bytes and
analyzing if delay has relation with io throttle.
Signed-off-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Reviewed-by: NGavin Shan <shan.gavin@linux.alibaba.com>
Reviewed-by: NJiufei Xue <jiufei.xue@linux.alibaba.com>
Reviewed-by: NXiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
Acked-by: NCaspar Zhang <caspar@linux.alibaba.com>
      766cfe98
    • J
      alinux: blk-throttle: fix tg NULL pointer dereference · bc0cc360
      由 Joseph Qi 提交于 
      io throtl stats will blkg_get at the beginning of throttle and then
blkg_put at the new introduced bi_tg_end_io. This will cause blkg to be
freed if end_io is called twice like dm-thin, which will save origin
end_io first, and call its overwrite end_io and then the saved end_io.
After that, access blkg is invalid and finally BUG:

[ 4417.235048] BUG: unable to handle kernel NULL pointer dereference at 00000000000001e0
[ 4417.236475] IP: [<ffffffff812e7c71>] throtl_update_dispatch_stats+0x21/0xb0
[ 4417.237865] PGD 98395067 PUD 362e1067 PMD 0
[ 4417.239232] Oops: 0000 [#1] SMP
......
[ 4417.274070] Call Trace:
[ 4417.275407]  [<ffffffff812ea93d>] blk_throtl_bio+0xfd/0x630
[ 4417.276760]  [<ffffffff810b3613>] ? wake_up_process+0x23/0x40
[ 4417.278079]  [<ffffffff81094c04>] ? wake_up_worker+0x24/0x30
[ 4417.279387]  [<ffffffff81095772>] ? insert_work+0x62/0xa0
[ 4417.280697]  [<ffffffff8116c2c7>] ? mempool_free_slab+0x17/0x20
[ 4417.282019]  [<ffffffff8116c6c9>] ? mempool_free+0x49/0x90
[ 4417.283326]  [<ffffffff812c9acf>] generic_make_request_checks+0x16f/0x360
[ 4417.284637]  [<ffffffffa0340d97>] ? thin_map+0x227/0x2c0 [dm_thin_pool]
[ 4417.285951]  [<ffffffff812c9ce7>] generic_make_request+0x27/0x130
[ 4417.287240]  [<ffffffffa0230b3d>] __map_bio+0xad/0x100 [dm_mod]
[ 4417.288503]  [<ffffffffa023257e>] __clone_and_map_data_bio+0x15e/0x240 [dm_mod]
[ 4417.289778]  [<ffffffffa02329ea>] __split_and_process_bio+0x38a/0x500 [dm_mod]
[ 4417.291062]  [<ffffffffa0232c91>] dm_make_request+0x131/0x1a0 [dm_mod]
[ 4417.292344]  [<ffffffff812c9da2>] generic_make_request+0xe2/0x130
[ 4417.293626]  [<ffffffff812c9e61>] submit_bio+0x71/0x150
[ 4417.294909]  [<ffffffff8121ab1d>] ? bio_alloc_bioset+0x20d/0x360
[ 4417.296195]  [<ffffffff81215acb>] _submit_bh+0x14b/0x220
[ 4417.297484]  [<ffffffff81215bb0>] submit_bh+0x10/0x20
[ 4417.298744]  [<ffffffffa016d8d8>] jbd2_journal_commit_transaction+0x6c8/0x19a0 [jbd2]
[ 4417.300014]  [<ffffffff810135b8>] ? __switch_to+0xf8/0x4c0
[ 4417.301268]  [<ffffffffa01731e9>] kjournald2+0xc9/0x270 [jbd2]
[ 4417.302524]  [<ffffffff810a0fd0>] ? wake_up_atomic_t+0x30/0x30
[ 4417.303753]  [<ffffffffa0173120>] ? commit_timeout+0x10/0x10 [jbd2]
[ 4417.304950]  [<ffffffff8109ffef>] kthread+0xcf/0xe0
[ 4417.306107]  [<ffffffff8109ff20>] ? kthread_create_on_node+0x140/0x140
[ 4417.307255]  [<ffffffff81647f18>] ret_from_fork+0x58/0x90
[ 4417.308349]  [<ffffffff8109ff20>] ? kthread_create_on_node+0x140/0x140
......

Now we introduce a new bio flag BIO_THROTL_STATED to make sure
blkg_get/put only get called once for the same bio.
Signed-off-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Reviewed-by: NJiufei Xue <jiufei.xue@linux.alibaba.com>
Reviewed-by: NXiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
Acked-by: NCaspar Zhang <caspar@linux.alibaba.com>
      bc0cc360
    • J
      alinux: blk-throttle: support io delay stats · dc61ad52
      由 Joseph Qi 提交于 
      Add blkio.throttle.io_service_time and blkio.throttle.io_wait_time to
get per-cgroup io delay statistics.
io_service_time represents the time spent after io throttle to io
completion, while io_wait_time represents the time spent on throttle
queue.
Signed-off-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Reviewed-by: NJiufei Xue <jiufei.xue@linux.alibaba.com>
Reviewed-by: NXiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
Acked-by: NCaspar Zhang <caspar@linux.alibaba.com>
      dc61ad52
    • W
      alinux: nvme-pci: Disable dicard zero-out functionality on Intel's P3600 NVMe disk drive · 2cde0dfb
      由 Wenwei Tao 提交于 
      We found huge performance lost on below particular Intel's disk drive
when discard zeroout functionality is enabled on it. The issue was
found when we have ext4 filesystem mounted on the disk drive and
started regular FIO testing. With it disabled, we don't observe
performance lost any more.

81:00.0 Non-Volatile memory controller: Intel Corporation \
             PCIe Data Center SSD (rev 01)

This imposes to disable the discard zero-out functionality on above
disk drive in order to regain the high performance that NVMe disk
driver supposes to provide.

Differential Revision: https://aone.alibaba-inc.com/code/D377540Signed-off-by: NWenwei Tao <wenwei.tao@linux.alibaba.com>
Reviewed-by: NGavin Shan <shan.gavin@linux.alibaba.com>
Signed-off-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Reviewed-by: NXiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
Acked-by: NCaspar Zhang <caspar@linux.alibaba.com>
      2cde0dfb
    • X
      alinux: memcg: Point wb to root memcg/blkcg when offlining to avoid zombie · cc00f21e
      由 Xunlei Pang 提交于 
      After turning off the memcg kmem charging, we still suffer
from various zombie memcg problems on production environment
because of its non-zero reference count from both page caches
and per-memcg writeback related structure(bdi_writeback takes
a reference).

After we reclaimed all the page caches of the zombie memcg,
it still can't be dropped due to its bdi_writeback.

bdi_writeback is further referenced by the inodes of files,
so the memcg can't be truely released until the inodes are
destroyed afterwards which is quite unlikely in short term.

When memcg is offlining, change it's bdi_writeback to root,
and call css_put to formally release it. We've tested on
product environment, it yields pretty good effect.

Ditto for wb_blkcg_offline().
Reviewed-by: NYang Shi <yang.shi@linux.alibaba.com>
Reviewed-by: NGavin Shan <shan.gavin@linux.alibaba.com>
Signed-off-by: NXunlei Pang <xlpang@linux.alibaba.com>
Signed-off-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Acked-by: NCaspar Zhang <caspar@linux.alibaba.com>
      cc00f21e
    • X
      alinux: block: add counter to track io request's d2c time · ba2896ac
      由 Xiaoguang Wang 提交于 
      Indeed tool iostat's await is not good enough, which is somewhat sketchy
and could not show request's latency on device driver's side.

Here we add a new counter to track io request's d2c time, also with this
patch, we can extend iostat to show this value easily.

Note:
I had checked how iostat is implemented, it just reads fields it needs,
so iostat won't be affected by this change, so does tsar.
Reviewed-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Signed-off-by: NXiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
Signed-off-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Acked-by: NCaspar Zhang <caspar@linux.alibaba.com>
      ba2896ac
    • M
      alinux: fuse: add sysfs api to flush processing queue requests · fc0a9b55
      由 Ma Jie Yue 提交于 
      The failover of fuse userspace daemon will reuse the existing fuse conn,
without unmounting it, during daemon crashing and recovery procedure.
But some requests might be in process in the daemon before sending out reply,
when the crash happens. This will stuck the application since it will
never get the reply after the failover.

We add the sysfs api to flush these requests, after the daemon crash, before
recovery. It is easy to reproduce the issue in the fuse userspace daemon,
just exit after receiving the request and before sending the reply back.
The application will hang up in some read/write operation, before
echo 1 > /sys/fs/fuse/connection/xxx/flush. The flush operation will make
the io fail and return the error to the application.
Signed-off-by: NMa Jie Yue <majieyue@linux.alibaba.com>
Reviewed-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Signed-off-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Reviewed-by: NXiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
Acked-by: NCaspar Zhang <caspar@linux.alibaba.com>
      fc0a9b55
    • X
      alinux: jbd2: add proc entry to control whether doing buffer copy-out · 1ced8a5c
      由 Xiaoguang Wang 提交于 
      When jbd2 tries to get write access to one buffer, and if this buffer
is under writeback with BH_Shadow flag, jbd2 will wait until this buffer
has been written to disk, but sometimes the time taken to wait may be
much long, especially disk capacity is almost full.

Here add a proc entry "force-copy", if its value is not zero, jbd2 will
always do meta buffer copy-cout, then we can eliminate the unnecessary
wating time here, and reduce long tail latency for buffered-write.

I construct such test case below:

$cat offline.fio
; fio-rand-RW.job for fiotest

[global]
name=fio-rand-RW
filename=fio-rand-RW
rw=randrw
rwmixread=60
rwmixwrite=40
bs=4K
direct=0
numjobs=4
time_based=1
runtime=900

[file1]
size=60G
ioengine=sync
iodepth=16

$cat online.fio
; fio-seq-write.job for fiotest

[global]
name=fio-seq-write
filename=fio-seq-write
rw=write
bs=256K
direct=0
numjobs=1
time_based=1
runtime=60

[file1]
rate=50m
size=10G
ioengine=sync
iodepth=16

With this patch:
$cat /proc/fs/jbd2/sda5-8/force_copy
0

online fio almost always get such long tail latency:

Jobs: 1 (f=1), 0B/s-0B/s: [W(1)][100.0%][w=50.0MiB/s][w=200 IOPS][eta
00m:00s]
file1: (groupid=0, jobs=1): err= 0: pid=17855: Thu Nov 15 09:45:57 2018
  write: IOPS=200, BW=50.0MiB/s (52.4MB/s)(3000MiB/60001msec)
    clat (usec): min=135, max=4086.6k, avg=867.21, stdev=50338.22
     lat (usec): min=139, max=4086.6k, avg=871.16, stdev=50338.22
    clat percentiles (usec):
     |  1.00th=[    141],  5.00th=[    143], 10.00th=[    145],
     | 20.00th=[    147], 30.00th=[    147], 40.00th=[    149],
     | 50.00th=[    149], 60.00th=[    151], 70.00th=[    153],
     | 80.00th=[    155], 90.00th=[    159], 95.00th=[    163],
     | 99.00th=[    255], 99.50th=[    273], 99.90th=[    429],
     | 99.95th=[    441], 99.99th=[3640656]

$cat /proc/fs/jbd2/sda5-8/force_copy
1

online fio latency is much better.

Jobs: 1 (f=1), 0B/s-0B/s: [W(1)][100.0%][w=50.0MiB/s][w=200 IOPS][eta
00m:00s]
file1: (groupid=0, jobs=1): err= 0: pid=8084: Thu Nov 15 09:31:15 2018
  write: IOPS=200, BW=50.0MiB/s (52.4MB/s)(3000MiB/60001msec)
    clat (usec): min=137, max=545, avg=151.35, stdev=16.22
     lat (usec): min=140, max=548, avg=155.31, stdev=16.65
    clat percentiles (usec):
     |  1.00th=[  143],  5.00th=[  145], 10.00th=[  145], 20.00th=[
147],
     | 30.00th=[  147], 40.00th=[  147], 50.00th=[  149], 60.00th=[
149],
     | 70.00th=[  151], 80.00th=[  155], 90.00th=[  157], 95.00th=[
161],
     | 99.00th=[  239], 99.50th=[  269], 99.90th=[  420], 99.95th=[
429],
     | 99.99th=[  537]

As to the cost: because we'll always need to copy meta buffer, will
consume minor cpu time and some memory(at most 32MB for 128MB journal
size).
Signed-off-by: NXiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
Reviewed-by: NLiu Bo <bo.liu@linux.alibaba.com>
Signed-off-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Acked-by: NCaspar Zhang <caspar@linux.alibaba.com>
      1ced8a5c
    • X
      alinux: ext4: don't submit unwritten extent while holding active jbd2 handle · c7c8cb0e
      由 Xiaoguang Wang 提交于 
      In ext4_writepages(), for every iteration, mpage_prepare_extent_to_map()
will try to find 2048 pages to map and normally one bio can contain 256
pages at most. If we really found 2048 pages to map, there will be 4 bios
and 4 ext4_io_submit() calls which are called both in ext4_writepages()
and mpage_map_and_submit_extent().

But note that in mpage_map_and_submit_extent(), we hold a valid jbd2 handle,
when dioread_nolock is enabled and extent is unwritten, jbd2 commit thread
will wait this handle to finish, so wait the unwritten extent is written to
disk, this will introduce unnecessary stall time, especially longer when
the writeback operation is io throttled, need to fix this issue.

Here for this scene, we accumulate bios in ext4_io_submit's io_bio, and
only submit these bios after dropping the jbd2 handle.
Signed-off-by: NXiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
Reviewed-by: NLiu Bo <bo.liu@linux.alibaba.com>
Signed-off-by: NJoseph Qi <joseph.qi@linux.alibaba.com>
Acked-by: NCaspar Zhang <caspar@linux.alibaba.com>
      c7c8cb0e