linux-toradex.git/mm, branch v4.12-rc4

mm: consider memblock reservations for deferred memory initialization sizing

2017-06-02T22:07:38+00:00

We have seen an early OOM killer invocation on ppc64 systems with
crashkernel=4096M:

	kthreadd invoked oom-killer: gfp_mask=0x16040c0(GFP_KERNEL|__GFP_COMP|__GFP_NOTRACK), nodemask=7, order=0, oom_score_adj=0
	kthreadd cpuset=/ mems_allowed=7
	CPU: 0 PID: 2 Comm: kthreadd Not tainted 4.4.68-1.gd7fe927-default #1
	Call Trace:
	  dump_stack+0xb0/0xf0 (unreliable)
	  dump_header+0xb0/0x258
	  out_of_memory+0x5f0/0x640
	  __alloc_pages_nodemask+0xa8c/0xc80
	  kmem_getpages+0x84/0x1a0
	  fallback_alloc+0x2a4/0x320
	  kmem_cache_alloc_node+0xc0/0x2e0
	  copy_process.isra.25+0x260/0x1b30
	  _do_fork+0x94/0x470
	  kernel_thread+0x48/0x60
	  kthreadd+0x264/0x330
	  ret_from_kernel_thread+0x5c/0xa4

	Mem-Info:
	active_anon:0 inactive_anon:0 isolated_anon:0
	 active_file:0 inactive_file:0 isolated_file:0
	 unevictable:0 dirty:0 writeback:0 unstable:0
	 slab_reclaimable:5 slab_unreclaimable:73
	 mapped:0 shmem:0 pagetables:0 bounce:0
	 free:0 free_pcp:0 free_cma:0
	Node 7 DMA free:0kB min:0kB low:0kB high:0kB active_anon:0kB inactive_anon:0kB active_file:0kB inactive_file:0kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:52428800kB managed:110016kB mlocked:0kB dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:320kB slab_unreclaimable:4672kB kernel_stack:1152kB pagetables:0kB unstable:0kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? yes
	lowmem_reserve[]: 0 0 0 0
	Node 7 DMA: 0*64kB 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB 0*8192kB 0*16384kB = 0kB
	0 total pagecache pages
	0 pages in swap cache
	Swap cache stats: add 0, delete 0, find 0/0
	Free swap  = 0kB
	Total swap = 0kB
	819200 pages RAM
	0 pages HighMem/MovableOnly
	817481 pages reserved
	0 pages cma reserved
	0 pages hwpoisoned

the reason is that the managed memory is too low (only 110MB) while the
rest of the the 50GB is still waiting for the deferred intialization to
be done.  update_defer_init estimates the initial memoty to initialize
to 2GB at least but it doesn't consider any memory allocated in that
range.  In this particular case we've had

	Reserving 4096MB of memory at 128MB for crashkernel (System RAM: 51200MB)

so the low 2GB is mostly depleted.

Fix this by considering memblock allocations in the initial static
initialization estimation.  Move the max_initialise to
reset_deferred_meminit and implement a simple memblock_reserved_memory
helper which iterates all reserved blocks and sums the size of all that
start below the given address.  The cumulative size is than added on top
of the initial estimation.  This is still not ideal because
reset_deferred_meminit doesn't consider holes and so reservation might
be above the initial estimation whihch we ignore but let's make the
logic simpler until we really need to handle more complicated cases.

Fixes: 3a80a7fa7989 ("mm: meminit: initialise a subset of struct pages if CONFIG_DEFERRED_STRUCT_PAGE_INIT is set")
Link: http://lkml.kernel.org/r/20170531104010.GI27783@dhcp22.suse.cz
Signed-off-by: Michal Hocko 
Acked-by: Mel Gorman 
Tested-by: Srikar Dronamraju 
Cc: 	[4.2+]
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm/hugetlb: report -EHWPOISON not -EFAULT when FOLL_HWPOISON is specified

2017-06-02T22:07:38+00:00

KVM uses get_user_pages() to resolve its stage2 faults.  KVM sets the
FOLL_HWPOISON flag causing faultin_page() to return -EHWPOISON when it
finds a VM_FAULT_HWPOISON.  KVM handles these hwpoison pages as a
special case.  (check_user_page_hwpoison())

When huge pages are involved, this doesn't work so well.
get_user_pages() calls follow_hugetlb_page(), which stops early if it
receives VM_FAULT_HWPOISON from hugetlb_fault(), eventually returning
-EFAULT to the caller.  The step to map this to -EHWPOISON based on the
FOLL_ flags is missing.  The hwpoison special case is skipped, and
-EFAULT is returned to user-space, causing Qemu or kvmtool to exit.

Instead, move this VM_FAULT_ to errno mapping code into a header file
and use it from faultin_page() and follow_hugetlb_page().

With this, KVM works as expected.

This isn't a problem for arm64 today as we haven't enabled
MEMORY_FAILURE, but I can't see any reason this doesn't happen on x86
too, so I think this should be a fix.  This doesn't apply earlier than
stable's v4.11.1 due to all sorts of cleanup.

[james.morse@arm.com: add vm_fault_to_errno() call to faultin_page()]
suggested.
  Link: http://lkml.kernel.org/r/20170525171035.16359-1-james.morse@arm.com
[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/20170524160900.28786-1-james.morse@arm.com
Signed-off-by: James Morse 
Acked-by: Punit Agrawal 
Acked-by: Naoya Horiguchi 
Cc: "Kirill A . Shutemov" 
Cc: 	[4.11.1+]
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mlock: fix mlock count can not decrease in race condition

2017-06-02T22:07:38+00:00

Kefeng reported that when running the follow test, the mlock count in
meminfo will increase permanently:

 [1] testcase
 linux:~ # cat test_mlockal
 grep Mlocked /proc/meminfo
  for j in `seq 0 10`
  do
 	for i in `seq 4 15`
 	do
 		./p_mlockall >> log &
 	done
 	sleep 0.2
 done
 # wait some time to let mlock counter decrease and 5s may not enough
 sleep 5
 grep Mlocked /proc/meminfo

 linux:~ # cat p_mlockall.c
 #include 
 #include 
 #include 

 #define SPACE_LEN	4096

 int main(int argc, char ** argv)
 {
	 	int ret;
	 	void *adr = malloc(SPACE_LEN);
	 	if (!adr)
	 		return -1;

	 	ret = mlockall(MCL_CURRENT | MCL_FUTURE);
	 	printf("mlcokall ret = %d\n", ret);

	 	ret = munlockall();
	 	printf("munlcokall ret = %d\n", ret);

	 	free(adr);
	 	return 0;
	 }

In __munlock_pagevec() we should decrement NR_MLOCK for each page where
we clear the PageMlocked flag.  Commit 1ebb7cc6a583 ("mm: munlock: batch
NR_MLOCK zone state updates") has introduced a bug where we don't
decrement NR_MLOCK for pages where we clear the flag, but fail to
isolate them from the lru list (e.g.  when the pages are on some other
cpu's percpu pagevec).  Since PageMlocked stays cleared, the NR_MLOCK
accounting gets permanently disrupted by this.

Fix it by counting the number of page whose PageMlock flag is cleared.

Fixes: 1ebb7cc6a583 (" mm: munlock: batch NR_MLOCK zone state updates")
Link: http://lkml.kernel.org/r/1495678405-54569-1-git-send-email-xieyisheng1@huawei.com
Signed-off-by: Yisheng Xie 
Reported-by: Kefeng Wang 
Tested-by: Kefeng Wang 
Cc: Vlastimil Babka 
Cc: Joern Engel 
Cc: Mel Gorman 
Cc: Michel Lespinasse 
Cc: Hugh Dickins 
Cc: Rik van Riel 
Cc: Johannes Weiner 
Cc: Michal Hocko 
Cc: Xishi Qiu 
Cc: zhongjiang 
Cc: Hanjun Guo 
Cc: 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm/migrate: fix refcount handling when !hugepage_migration_supported()

2017-06-02T22:07:38+00:00

On failing to migrate a page, soft_offline_huge_page() performs the
necessary update to the hugepage ref-count.

But when !hugepage_migration_supported() , unmap_and_move_hugepage()
also decrements the page ref-count for the hugepage.  The combined
behaviour leaves the ref-count in an inconsistent state.

This leads to soft lockups when running the overcommitted hugepage test
from mce-tests suite.

  Soft offlining pfn 0x83ed600 at process virtual address 0x400000000000
  soft offline: 0x83ed600: migration failed 1, type 1fffc00000008008 (uptodate|head)
  INFO: rcu_preempt detected stalls on CPUs/tasks:
   Tasks blocked on level-0 rcu_node (CPUs 0-7): P2715
    (detected by 7, t=5254 jiffies, g=963, c=962, q=321)
    thugetlb_overco R  running task        0  2715   2685 0x00000008
    Call trace:
      dump_backtrace+0x0/0x268
      show_stack+0x24/0x30
      sched_show_task+0x134/0x180
      rcu_print_detail_task_stall_rnp+0x54/0x7c
      rcu_check_callbacks+0xa74/0xb08
      update_process_times+0x34/0x60
      tick_sched_handle.isra.7+0x38/0x70
      tick_sched_timer+0x4c/0x98
      __hrtimer_run_queues+0xc0/0x300
      hrtimer_interrupt+0xac/0x228
      arch_timer_handler_phys+0x3c/0x50
      handle_percpu_devid_irq+0x8c/0x290
      generic_handle_irq+0x34/0x50
      __handle_domain_irq+0x68/0xc0
      gic_handle_irq+0x5c/0xb0

Address this by changing the putback_active_hugepage() in
soft_offline_huge_page() to putback_movable_pages().

This only triggers on systems that enable memory failure handling
(ARCH_SUPPORTS_MEMORY_FAILURE) but not hugepage migration
(!ARCH_ENABLE_HUGEPAGE_MIGRATION).

I imagine this wasn't triggered as there aren't many systems running
this configuration.

[akpm@linux-foundation.org: remove dead comment, per Naoya]
Link: http://lkml.kernel.org/r/20170525135146.32011-1-punit.agrawal@arm.com
Reported-by: Manoj Iyer 
Tested-by: Manoj Iyer 
Suggested-by: Naoya Horiguchi 
Signed-off-by: Punit Agrawal 
Cc: Joonsoo Kim 
Cc: Wanpeng Li 
Cc: Christoph Lameter 
Cc: Mel Gorman 
Cc: 	[3.14+]
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm: avoid spurious 'bad pmd' warning messages

2017-06-02T22:07:37+00:00

When the pmd_devmap() checks were added by 5c7fb56e5e3f ("mm, dax:
dax-pmd vs thp-pmd vs hugetlbfs-pmd") to add better support for DAX huge
pages, they were all added to the end of if() statements after existing
pmd_trans_huge() checks.  So, things like:

  -       if (pmd_trans_huge(*pmd))
  +       if (pmd_trans_huge(*pmd) || pmd_devmap(*pmd))

When further checks were added after pmd_trans_unstable() checks by
commit 7267ec008b5c ("mm: postpone page table allocation until we have
page to map") they were also added at the end of the conditional:

  +       if (pmd_trans_unstable(fe->pmd) || pmd_devmap(*fe->pmd))

This ordering is fine for pmd_trans_huge(), but doesn't work for
pmd_trans_unstable().  This is because DAX huge pages trip the bad_pmd()
check inside of pmd_none_or_trans_huge_or_clear_bad() (called by
pmd_trans_unstable()), which prints out a warning and returns 1.  So, we
do end up doing the right thing, but only after spamming dmesg with
suspicious looking messages:

  mm/pgtable-generic.c:39: bad pmd ffff8808daa49b88(84000001006000a5)

Reorder these checks in a helper so that pmd_devmap() is checked first,
avoiding the error messages, and add a comment explaining why the
ordering is important.

Fixes: commit 7267ec008b5c ("mm: postpone page table allocation until we have page to map")
Link: http://lkml.kernel.org/r/20170522215749.23516-1-ross.zwisler@linux.intel.com
Signed-off-by: Ross Zwisler 
Reviewed-by: Jan Kara 
Cc: Pawel Lebioda 
Cc: "Darrick J. Wong" 
Cc: Alexander Viro 
Cc: Christoph Hellwig 
Cc: Dan Williams 
Cc: Dave Hansen 
Cc: Matthew Wilcox 
Cc: "Kirill A . Shutemov" 
Cc: Dave Jiang 
Cc: Xiong Zhou 
Cc: Eryu Guan 
Cc: 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm/page_alloc.c: make sure OOM victim can try allocations with no watermarks once

2017-06-02T22:07:37+00:00

Roman Gushchin has reported that the OOM killer can trivially selects
next OOM victim when a thread doing memory allocation from page fault
path was selected as first OOM victim.

    allocate invoked oom-killer: gfp_mask=0x14280ca(GFP_HIGHUSER_MOVABLE|__GFP_ZERO), nodemask=(null),  order=0, oom_score_adj=0
    allocate cpuset=/ mems_allowed=0
    CPU: 1 PID: 492 Comm: allocate Not tainted 4.12.0-rc1-mm1+ #181
    Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
    Call Trace:
     oom_kill_process+0x219/0x3e0
     out_of_memory+0x11d/0x480
     __alloc_pages_slowpath+0xc84/0xd40
     __alloc_pages_nodemask+0x245/0x260
     alloc_pages_vma+0xa2/0x270
     __handle_mm_fault+0xca9/0x10c0
     handle_mm_fault+0xf3/0x210
     __do_page_fault+0x240/0x4e0
     trace_do_page_fault+0x37/0xe0
     do_async_page_fault+0x19/0x70
     async_page_fault+0x28/0x30
    ...
    Out of memory: Kill process 492 (allocate) score 899 or sacrifice child
    Killed process 492 (allocate) total-vm:2052368kB, anon-rss:1894576kB, file-rss:4kB, shmem-rss:0kB
    allocate: page allocation failure: order:0, mode:0x14280ca(GFP_HIGHUSER_MOVABLE|__GFP_ZERO), nodemask=(null)
    allocate cpuset=/ mems_allowed=0
    CPU: 1 PID: 492 Comm: allocate Not tainted 4.12.0-rc1-mm1+ #181
    Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
    Call Trace:
     __alloc_pages_slowpath+0xd32/0xd40
     __alloc_pages_nodemask+0x245/0x260
     alloc_pages_vma+0xa2/0x270
     __handle_mm_fault+0xca9/0x10c0
     handle_mm_fault+0xf3/0x210
     __do_page_fault+0x240/0x4e0
     trace_do_page_fault+0x37/0xe0
     do_async_page_fault+0x19/0x70
     async_page_fault+0x28/0x30
    ...
    oom_reaper: reaped process 492 (allocate), now anon-rss:0kB, file-rss:0kB, shmem-rss:0kB
    ...
    allocate invoked oom-killer: gfp_mask=0x0(), nodemask=(null),  order=0, oom_score_adj=0
    allocate cpuset=/ mems_allowed=0
    CPU: 1 PID: 492 Comm: allocate Not tainted 4.12.0-rc1-mm1+ #181
    Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
    Call Trace:
     oom_kill_process+0x219/0x3e0
     out_of_memory+0x11d/0x480
     pagefault_out_of_memory+0x68/0x80
     mm_fault_error+0x8f/0x190
     ? handle_mm_fault+0xf3/0x210
     __do_page_fault+0x4b2/0x4e0
     trace_do_page_fault+0x37/0xe0
     do_async_page_fault+0x19/0x70
     async_page_fault+0x28/0x30
    ...
    Out of memory: Kill process 233 (firewalld) score 10 or sacrifice child
    Killed process 233 (firewalld) total-vm:246076kB, anon-rss:20956kB, file-rss:0kB, shmem-rss:0kB

There is a race window that the OOM reaper completes reclaiming the
first victim's memory while nothing but mutex_trylock() prevents the
first victim from calling out_of_memory() from pagefault_out_of_memory()
after memory allocation for page fault path failed due to being selected
as an OOM victim.

This is a side effect of commit 9a67f6488eca926f ("mm: consolidate
GFP_NOFAIL checks in the allocator slowpath") because that commit
silently changed the behavior from

    /* Avoid allocations with no watermarks from looping endlessly */

to

    /*
     * Give up allocations without trying memory reserves if selected
     * as an OOM victim
     */

in __alloc_pages_slowpath() by moving the location to check TIF_MEMDIE
flag.  I have noticed this change but I didn't post a patch because I
thought it is an acceptable change other than noise by warn_alloc()
because !__GFP_NOFAIL allocations are allowed to fail.  But we
overlooked that failing memory allocation from page fault path makes
difference due to the race window explained above.

While it might be possible to add a check to pagefault_out_of_memory()
that prevents the first victim from calling out_of_memory() or remove
out_of_memory() from pagefault_out_of_memory(), changing
pagefault_out_of_memory() does not suppress noise by warn_alloc() when
allocating thread was selected as an OOM victim.  There is little point
with printing similar backtraces and memory information from both
out_of_memory() and warn_alloc().

Instead, if we guarantee that current thread can try allocations with no
watermarks once when current thread looping inside
__alloc_pages_slowpath() was selected as an OOM victim, we can follow "who
can use memory reserves" rules and suppress noise by warn_alloc() and
prevent memory allocations from page fault path from calling
pagefault_out_of_memory().

If we take the comment literally, this patch would do

  -    if (test_thread_flag(TIF_MEMDIE))
  -        goto nopage;
  +    if (alloc_flags == ALLOC_NO_WATERMARKS || (gfp_mask & __GFP_NOMEMALLOC))
  +        goto nopage;

because gfp_pfmemalloc_allowed() returns false if __GFP_NOMEMALLOC is
given.  But if I recall correctly (I couldn't find the message), the
condition is meant to apply to only OOM victims despite the comment.
Therefore, this patch preserves TIF_MEMDIE check.

Fixes: 9a67f6488eca926f ("mm: consolidate GFP_NOFAIL checks in the allocator slowpath")
Link: http://lkml.kernel.org/r/201705192112.IAF69238.OQOHSJLFOFFMtV@I-love.SAKURA.ne.jp
Signed-off-by: Tetsuo Handa 
Reported-by: Roman Gushchin 
Tested-by: Roman Gushchin 
Acked-by: Michal Hocko 
Cc: Johannes Weiner 
Cc: Vladimir Davydov 
Cc: 	[4.11]
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

slub/memcg: cure the brainless abuse of sysfs attributes

2017-06-02T22:07:37+00:00

memcg_propagate_slab_attrs() abuses the sysfs attribute file functions
to propagate settings from the root kmem_cache to a newly created
kmem_cache.  It does that with:

     attr->show(root, buf);
     attr->store(new, buf, strlen(bug);

Aside of being a lazy and absurd hackery this is broken because it does
not check the return value of the show() function.

Some of the show() functions return 0 w/o touching the buffer.  That
means in such a case the store function is called with the stale content
of the previous show().  That causes nonsense like invoking
kmem_cache_shrink() on a newly created kmem_cache.  In the worst case it
would cause handing in an uninitialized buffer.

This should be rewritten proper by adding a propagate() callback to
those slub_attributes which must be propagated and avoid that insane
conversion to and from ASCII, but that's too large for a hot fix.

Check at least the return value of the show() function, so calling
store() with stale content is prevented.

Steven said:
 "It can cause a deadlock with get_online_cpus() that has been uncovered
  by recent cpu hotplug and lockdep changes that Thomas and Peter have
  been doing.

     Possible unsafe locking scenario:

           CPU0                    CPU1
           ----                    ----
      lock(cpu_hotplug.lock);
                                   lock(slab_mutex);
                                   lock(cpu_hotplug.lock);
      lock(slab_mutex);

     *** DEADLOCK ***"

Link: http://lkml.kernel.org/r/alpine.DEB.2.20.1705201244540.2255@nanos
Signed-off-by: Thomas Gleixner 
Reported-by: Steven Rostedt 
Acked-by: David Rientjes 
Cc: Johannes Weiner 
Cc: Michal Hocko 
Cc: Peter Zijlstra 
Cc: Christoph Lameter 
Cc: Pekka Enberg 
Cc: Joonsoo Kim 
Cc: Christoph Hellwig 
Cc: 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm: clarify why we want kmalloc before falling backto vmallock

2017-06-02T22:07:37+00:00

While converting drm_[cm]alloc* helpers to kvmalloc* variants Chris
Wilson has wondered why we want to try kmalloc before vmalloc fallback
even for larger allocations requests.  Let's clarify that one larger
physically contiguous block is less likely to fragment memory than many
scattered pages which can prevent more large blocks from being created.

[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/20170517080932.21423-1-mhocko@kernel.org
Signed-off-by: Michal Hocko 
Suggested-by: Chris Wilson 
Reviewed-by: Chris Wilson 
Acked-by: Vlastimil Babka 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

ksm: prevent crash after write_protect_page fails

2017-06-02T22:07:37+00:00

"err" needs to be left set to -EFAULT if split_huge_page succeeds.
Otherwise if "err" gets clobbered with zero and write_protect_page
fails, try_to_merge_one_page() will succeed instead of returning -EFAULT
and then try_to_merge_with_ksm_page() will continue thinking kpage is a
PageKsm when in fact it's still an anonymous page.  Eventually it'll
crash in page_add_anon_rmap.

This has been reproduced on Fedora25 kernel but I can reproduce with
upstream too.

The bug was introduced in commit f765f540598a ("ksm: prepare to new THP
semantics") introduced in v4.5.

    page:fffff67546ce1cc0 count:4 mapcount:2 mapping:ffffa094551e36e1 index:0x7f0f46673
    flags: 0x2ffffc0004007c(referenced|uptodate|dirty|lru|active|swapbacked)
    page dumped because: VM_BUG_ON_PAGE(!PageLocked(page))
    page->mem_cgroup:ffffa09674bf0000
    ------------[ cut here ]------------
    kernel BUG at mm/rmap.c:1222!
    CPU: 1 PID: 76 Comm: ksmd Not tainted 4.9.3-200.fc25.x86_64 #1
    RIP: do_page_add_anon_rmap+0x1c4/0x240
    Call Trace:
      page_add_anon_rmap+0x18/0x20
      try_to_merge_with_ksm_page+0x50b/0x780
      ksm_scan_thread+0x1211/0x1410
      ? prepare_to_wait_event+0x100/0x100
      ? try_to_merge_with_ksm_page+0x780/0x780
      kthread+0xd9/0xf0
      ? kthread_park+0x60/0x60
      ret_from_fork+0x25/0x30

Fixes: f765f54059 ("ksm: prepare to new THP semantics")
Link: http://lkml.kernel.org/r/20170513131040.21732-1-aarcange@redhat.com
Signed-off-by: Andrea Arcangeli 
Reported-by: Federico Simoncelli 
Acked-by: Kirill A. Shutemov 
Cc: Hugh Dickins 
Cc: 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm: vmscan: scan until it finds eligible pages

2017-05-12T22:57:16+00:00

Although there are a ton of free swap and anonymous LRU page in elgible
zones, OOM happened.

  balloon invoked oom-killer: gfp_mask=0x17080c0(GFP_KERNEL_ACCOUNT|__GFP_ZERO|__GFP_NOTRACK), nodemask=(null),  order=0, oom_score_adj=0
  CPU: 7 PID: 1138 Comm: balloon Not tainted 4.11.0-rc6-mm1-zram-00289-ge228d67e9677-dirty #17
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
  Call Trace:
   oom_kill_process+0x21d/0x3f0
   out_of_memory+0xd8/0x390
   __alloc_pages_slowpath+0xbc1/0xc50
   __alloc_pages_nodemask+0x1a5/0x1c0
   pte_alloc_one+0x20/0x50
   __pte_alloc+0x1e/0x110
   __handle_mm_fault+0x919/0x960
   handle_mm_fault+0x77/0x120
   __do_page_fault+0x27a/0x550
   trace_do_page_fault+0x43/0x150
   do_async_page_fault+0x2c/0x90
   async_page_fault+0x28/0x30
  Mem-Info:
  active_anon:424716 inactive_anon:65314 isolated_anon:0
   active_file:52 inactive_file:46 isolated_file:0
   unevictable:0 dirty:27 writeback:0 unstable:0
   slab_reclaimable:3967 slab_unreclaimable:4125
   mapped:133 shmem:43 pagetables:1674 bounce:0
   free:4637 free_pcp:225 free_cma:0
  Node 0 active_anon:1698864kB inactive_anon:261256kB active_file:208kB inactive_file:184kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:532kB dirty:108kB writeback:0kB shmem:172kB writeback_tmp:0kB unstable:0kB all_unreclaimable? no
  DMA free:7316kB min:32kB low:44kB high:56kB active_anon:8064kB inactive_anon:0kB active_file:0kB inactive_file:0kB unevictable:0kB writepending:0kB present:15992kB managed:15908kB mlocked:0kB slab_reclaimable:464kB slab_unreclaimable:40kB kernel_stack:0kB pagetables:24kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
  lowmem_reserve[]: 0 992 992 1952
  DMA32 free:9088kB min:2048kB low:3064kB high:4080kB active_anon:952176kB inactive_anon:0kB active_file:36kB inactive_file:0kB unevictable:0kB writepending:88kB present:1032192kB managed:1019388kB mlocked:0kB slab_reclaimable:13532kB slab_unreclaimable:16460kB kernel_stack:3552kB pagetables:6672kB bounce:0kB free_pcp:56kB local_pcp:24kB free_cma:0kB
  lowmem_reserve[]: 0 0 0 959
  Movable free:3644kB min:1980kB low:2960kB high:3940kB active_anon:738560kB inactive_anon:261340kB active_file:188kB inactive_file:640kB unevictable:0kB writepending:20kB present:1048444kB managed:1010816kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:0kB kernel_stack:0kB pagetables:0kB bounce:0kB free_pcp:832kB local_pcp:60kB free_cma:0kB
  lowmem_reserve[]: 0 0 0 0
  DMA: 1*4kB (E) 0*8kB 18*16kB (E) 10*32kB (E) 10*64kB (E) 9*128kB (ME) 8*256kB (E) 2*512kB (E) 2*1024kB (E) 0*2048kB 0*4096kB = 7524kB
  DMA32: 417*4kB (UMEH) 181*8kB (UMEH) 68*16kB (UMEH) 48*32kB (UMEH) 14*64kB (MH) 3*128kB (M) 1*256kB (H) 1*512kB (M) 2*1024kB (M) 0*2048kB 0*4096kB = 9836kB
  Movable: 1*4kB (M) 1*8kB (M) 1*16kB (M) 1*32kB (M) 0*64kB 1*128kB (M) 2*256kB (M) 4*512kB (M) 1*1024kB (M) 0*2048kB 0*4096kB = 3772kB
  378 total pagecache pages
  17 pages in swap cache
  Swap cache stats: add 17325, delete 17302, find 0/27
  Free swap  = 978940kB
  Total swap = 1048572kB
  524157 pages RAM
  0 pages HighMem/MovableOnly
  12629 pages reserved
  0 pages cma reserved
  0 pages hwpoisoned
  [ pid ]   uid  tgid total_vm      rss nr_ptes nr_pmds swapents oom_score_adj name
  [  433]     0   433     4904        5      14       3       82             0 upstart-udev-br
  [  438]     0   438    12371        5      27       3      191         -1000 systemd-udevd

With investigation, skipping page of isolate_lru_pages makes reclaim
void because it returns zero nr_taken easily so LRU shrinking is
effectively nothing and just increases priority aggressively.  Finally,
OOM happens.

The problem is that get_scan_count determines nr_to_scan with eligible
zones so although priority drops to zero, it couldn't reclaim any pages
if the LRU contains mostly ineligible pages.

get_scan_count:

        size = lruvec_lru_size(lruvec, lru, sc->reclaim_idx);
	size = size >> sc->priority;

Assumes sc->priority is 0 and LRU list is as follows.

	N-N-N-N-H-H-H-H-H-H-H-H-H-H-H-H-H-H-H-H

(Ie, small eligible pages are in the head of LRU but others are
 almost ineligible pages)

In that case, size becomes 4 so VM want to scan 4 pages but 4 pages from
tail of the LRU are not eligible pages.  If get_scan_count counts
skipped pages, it doesn't reclaim any pages remained after scanning 4
pages so it ends up OOM happening.

This patch makes isolate_lru_pages try to scan pages until it encounters
eligible zones's pages.

[akpm@linux-foundation.org: clean up mind-bending `for' statement.  Tweak comment text]
Fixes: 3db65812d688 ("Revert "mm, vmscan: account for skipped pages as a partial scan"")
Link: http://lkml.kernel.org/r/1494457232-27401-1-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim 
Acked-by: Michal Hocko 
Acked-by: Johannes Weiner 
Cc: Mel Gorman 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds