linux/mm
Vlastimil Babka 9e5e366172 mm, vmscan: prevent kswapd livelock due to pfmemalloc-throttled process being killed
Charles Shirron and Paul Cassella from Cray Inc have reported kswapd
stuck in a busy loop with nothing left to balance, but
kswapd_try_to_sleep() failing to sleep.  Their analysis found the cause
to be a combination of several factors:

1. A process is waiting in throttle_direct_reclaim() on pgdat->pfmemalloc_wait

2. The process has been killed (by OOM in this case), but has not yet been
   scheduled to remove itself from the waitqueue and die.

3. kswapd checks for throttled processes in prepare_kswapd_sleep():

        if (waitqueue_active(&pgdat->pfmemalloc_wait)) {
                wake_up(&pgdat->pfmemalloc_wait);
		return false; // kswapd will not go to sleep
	}

   However, for a process that was already killed, wake_up() does not remove
   the process from the waitqueue, since try_to_wake_up() checks its state
   first and returns false when the process is no longer waiting.

4. kswapd is running on the same CPU as the only CPU that the process is
   allowed to run on (through cpus_allowed, or possibly single-cpu system).

5. CONFIG_PREEMPT_NONE=y kernel is used. If there's nothing to balance, kswapd
   encounters no voluntary preemption points and repeatedly fails
   prepare_kswapd_sleep(), blocking the process from running and removing
   itself from the waitqueue, which would let kswapd sleep.

So, the source of the problem is that we prevent kswapd from going to
sleep until there are processes waiting on the pfmemalloc_wait queue,
and a process waiting on a queue is guaranteed to be removed from the
queue only when it gets scheduled.  This was done to make sure that no
process is left sleeping on pfmemalloc_wait when kswapd itself goes to
sleep.

However, it isn't necessary to postpone kswapd sleep until the
pfmemalloc_wait queue actually empties.  To prevent processes from being
left sleeping, it's actually enough to guarantee that all processes
waiting on pfmemalloc_wait queue have been woken up by the time we put
kswapd to sleep.

This patch therefore fixes this issue by substituting 'wake_up' with
'wake_up_all' and removing 'return false' in the code snippet from
prepare_kswapd_sleep() above.  Note that if any process puts itself in
the queue after this waitqueue_active() check, or after the wake up
itself, it means that the process will also wake up kswapd - and since
we are under prepare_to_wait(), the wake up won't be missed.  Also we
update the comment prepare_kswapd_sleep() to hopefully more clearly
describe the races it is preventing.

Fixes: 5515061d22 ("mm: throttle direct reclaimers if PF_MEMALLOC reserves are low and swap is backed by network storage")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: <stable@vger.kernel.org>	[3.6+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-01-08 15:10:52 -08:00
..
backing-dev.c Merge branch 'for-3.18/core' of git://git.kernel.dk/linux-block 2014-10-18 11:53:51 -07:00
balloon_compaction.c mm/balloon_compaction: fix deflation when compaction is disabled 2014-10-29 16:33:15 -07:00
bootmem.c mem-hotplug: reset node managed pages when hot-adding a new pgdat 2014-11-13 16:17:06 -08:00
cleancache.c
cma.c mm: cma: split cma-reserved in dmesg log 2014-12-18 19:08:10 -08:00
compaction.c mm, compaction: more focused lru and pcplists draining 2014-12-10 17:41:06 -08:00
debug-pagealloc.c mm/debug-pagealloc: make debug-pagealloc boottime configurable 2014-12-13 12:42:48 -08:00
debug.c mm: move page->mem_cgroup bad page handling into generic code 2014-12-10 17:41:09 -08:00
dmapool.c mm/dmapool.c: fixed a brace coding style issue 2014-10-09 22:26:00 -04:00
early_ioremap.c
fadvise.c mm: fadvise: document the fadvise(FADV_DONTNEED) behaviour for partial pages 2014-12-13 12:42:49 -08:00
failslab.c
filemap_xip.c mm/xip: share the i_mmap_rwsem 2014-12-13 12:42:45 -08:00
filemap.c mm: get rid of radix tree gfp mask for pagecache_get_page 2014-12-29 12:45:45 -08:00
fremap.c Merge branch 'drm-next' of git://people.freedesktop.org/~airlied/linux 2014-12-15 15:52:01 -08:00
frontswap.c mm/frontswap.c: fix the condition in BUG_ON 2014-12-10 17:41:08 -08:00
gup.c kernel: Provide READ_ONCE and ASSIGN_ONCE 2014-12-20 16:48:59 -08:00
highmem.c mm/highmem: make kmap cache coloring aware 2014-08-06 18:01:22 -07:00
huge_memory.c Merge branch 'drm-next' of git://people.freedesktop.org/~airlied/linux 2014-12-15 15:52:01 -08:00
hugetlb_cgroup.c mm: hugetlb_cgroup: convert to lockless page counters 2014-12-10 17:41:04 -08:00
hugetlb.c Merge branch 'drm-next' of git://people.freedesktop.org/~airlied/linux 2014-12-15 15:52:01 -08:00
hwpoison-inject.c mm/hwpoison-inject.c: remove unnecessary null test before debugfs_remove_recursive 2014-08-06 18:01:19 -07:00
init-mm.c
internal.h mm, compaction: always update cached scanner positions 2014-12-10 17:41:06 -08:00
interval_tree.c mm: convert a few VM_BUG_ON callers to VM_BUG_ON_VMA 2014-10-09 22:25:57 -04:00
iov_iter.c copy_from_iter_nocache() 2014-12-08 20:25:23 -05:00
Kconfig mm/balloon_compaction: add vmstat counters and kpageflags bit 2014-10-09 22:26:01 -04:00
Kconfig.debug mm/debug_pagealloc: remove obsolete Kconfig options 2015-01-08 15:10:52 -08:00
kmemcheck.c mm/slab_common: move kmem_cache definition to internal header 2014-10-09 22:25:50 -04:00
kmemleak-test.c
kmemleak.c
ksm.c mmu_notifier: call mmu_notifier_invalidate_range() from VMM 2014-11-13 13:46:09 +11:00
list_lru.c
maccess.c
madvise.c VFS: Rename do_fallocate() to vfs_fallocate() 2014-11-07 16:17:44 -05:00
Makefile mm/page_owner: keep track of page owners 2014-12-13 12:42:48 -08:00
memblock.c mm/memblock.c: refactor functions to set/clear MEMBLOCK_HOTPLUG 2014-12-13 12:42:46 -08:00
memcontrol.c memcg: fix destination cgroup leak on task charges migration 2015-01-08 15:10:52 -08:00
memory_hotplug.c mm, memory_hotplug/failure: drain single zone pcplists 2014-12-10 17:41:05 -08:00
memory-failure.c mm: vmscan: invoke slab shrinkers from shrink_zone() 2014-12-13 12:42:48 -08:00
memory.c mm: protect set_page_dirty() from ongoing truncation 2015-01-08 15:10:51 -08:00
mempolicy.c Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2014-12-19 18:19:19 -08:00
mempool.c
migrate.c vm_area_operations: kill ->migrate() 2014-12-17 08:26:51 -05:00
mincore.c mm: mincore: add hwpoison page handle 2014-12-13 12:42:46 -08:00
mlock.c Merge branch 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip 2014-10-13 15:44:12 +02:00
mm_init.c
mmap.c mm: export find_extend_vma() and handle_mm_fault() for driver use 2014-12-13 12:42:47 -08:00
mmu_context.c
mmu_notifier.c mmu_notifier: add the callback for mmu_notifier_invalidate_range() 2014-11-13 13:46:09 +11:00
mmzone.c
mprotect.c mm: softdirty: enable write notifications on VMAs after VM_SOFTDIRTY cleared 2014-10-14 02:18:28 +02:00
mremap.c Merge git://git.kvack.org/~bcrl/aio-next 2014-12-14 13:36:57 -08:00
msync.c
nobootmem.c mem-hotplug: reset node managed pages when hot-adding a new pgdat 2014-11-13 16:17:06 -08:00
nommu.c mm/nommu: use alloc_pages_exact() rather than its own implementation 2014-12-13 12:42:48 -08:00
oom_kill.c oom: kill the insufficient and no longer needed PT_TRACE_EXIT check 2014-12-13 12:42:49 -08:00
page_alloc.c mm: cma: split cma-reserved in dmesg log 2014-12-18 19:08:10 -08:00
page_counter.c mm: memcontrol: remove obsolete kmemcg pinning tricks 2014-12-10 17:41:05 -08:00
page_ext.c mm/page_owner: keep track of page owners 2014-12-13 12:42:48 -08:00
page_io.c
page_isolation.c mm, page_isolation: drain single zone pcplists 2014-12-10 17:41:05 -08:00
page_owner.c mm/page_owner: correct owner information for early allocated pages 2014-12-13 12:42:48 -08:00
page-writeback.c mm: protect set_page_dirty() from ongoing truncation 2015-01-08 15:10:51 -08:00
pagewalk.c mm: use VM_BUG_ON_MM where possible 2014-10-09 22:25:58 -04:00
percpu-km.c percpu: implmeent pcpu_nr_empty_pop_pages and chunk->nr_populated 2014-09-02 14:46:05 -04:00
percpu-vm.c percpu: move region iterations out of pcpu_[de]populate_chunk() 2014-09-02 14:46:02 -04:00
percpu.c percpu: off by one in BUG_ON() 2014-10-29 10:34:34 -04:00
pgtable-generic.c mm: actually clear pmd_numa before invalidating 2014-08-29 16:28:15 -07:00
process_vm_access.c
quicklist.c
readahead.c mm/readahead.c: remove unused file_ra_state from count_history_pages 2014-08-06 18:01:15 -07:00
rmap.c mm: prevent endless growth of anon_vma hierarchy 2015-01-08 15:10:51 -08:00
shmem.c new helper: iter_is_iovec() 2014-12-17 06:43:56 -05:00
slab_common.c memcg: use generic slab iterators for showing slabinfo 2014-12-10 17:41:07 -08:00
slab.c slab: fix cpuset check in fallback_alloc 2014-12-13 12:42:53 -08:00
slab.h memcg: use generic slab iterators for showing slabinfo 2014-12-10 17:41:07 -08:00
slob.c mm/sl[ao]b: always track caller in kmalloc_(node_)track_caller() 2014-10-09 22:25:50 -04:00
slub.c slub: fix cpuset check in get_any_partial 2014-12-13 12:42:53 -08:00
sparse-vmemmap.c
sparse.c
swap_cgroup.c mm: page_cgroup: rename file to mm/swap_cgroup.c 2014-12-10 17:41:09 -08:00
swap_state.c mm: page_cgroup: rename file to mm/swap_cgroup.c 2014-12-10 17:41:09 -08:00
swap.c mm: memcontrol: do not kill uncharge batching in free_pages_and_swap_cache 2014-10-09 22:25:59 -04:00
swapfile.c mm: page_cgroup: rename file to mm/swap_cgroup.c 2014-12-10 17:41:09 -08:00
truncate.c mm: Fix comment before truncate_setsize() 2014-11-07 08:29:25 +11:00
util.c proc/maps: make vm_is_stack() logic namespace-friendly 2014-10-09 22:25:50 -04:00
vmacache.c mm,vmacache: count number of system-wide flushes 2014-12-13 12:42:48 -08:00
vmalloc.c mm/vmalloc.c: fix memory ordering bug 2014-12-13 12:42:49 -08:00
vmpressure.c mm/vmpressure.c: fix race in vmpressure_work_fn() 2014-12-02 17:32:07 -08:00
vmscan.c mm, vmscan: prevent kswapd livelock due to pfmemalloc-throttled process being killed 2015-01-08 15:10:52 -08:00
vmstat.c mm,vmacache: count number of system-wide flushes 2014-12-13 12:42:48 -08:00
workingset.c
zbud.c mm/zbud: init user ops only when it is needed 2014-12-13 12:42:51 -08:00
zpool.c mm/zpool: use prefixed module loading 2014-08-29 16:28:16 -07:00
zsmalloc.c mm/zsmalloc: adjust order of functions 2014-12-18 19:08:11 -08:00
zswap.c mm/zswap: delete unnecessary check before calling free_percpu() 2014-12-13 12:42:50 -08:00