linux_dsm_epyc7002/mm
Hugh Dickins 8bdd638091 mm: fix direct reclaim writeback regression
Shortly before 3.16-rc1, Dave Jones reported:

  WARNING: CPU: 3 PID: 19721 at fs/xfs/xfs_aops.c:971
           xfs_vm_writepage+0x5ce/0x630 [xfs]()
  CPU: 3 PID: 19721 Comm: trinity-c61 Not tainted 3.15.0+ #3
  Call Trace:
    xfs_vm_writepage+0x5ce/0x630 [xfs]
    shrink_page_list+0x8f9/0xb90
    shrink_inactive_list+0x253/0x510
    shrink_lruvec+0x563/0x6c0
    shrink_zone+0x3b/0x100
    shrink_zones+0x1f1/0x3c0
    try_to_free_pages+0x164/0x380
    __alloc_pages_nodemask+0x822/0xc90
    alloc_pages_vma+0xaf/0x1c0
    handle_mm_fault+0xa31/0xc50
  etc.

 970   if (WARN_ON_ONCE((current->flags & (PF_MEMALLOC|PF_KSWAPD)) ==
 971                   PF_MEMALLOC))

I did not respond at the time, because a glance at the PageDirty block
in shrink_page_list() quickly shows that this is impossible: we don't do
writeback on file pages (other than tmpfs) from direct reclaim nowadays.
Dave was hallucinating, but it would have been disrespectful to say so.

However, my own /var/log/messages now shows similar complaints

  WARNING: CPU: 1 PID: 28814 at fs/ext4/inode.c:1881 ext4_writepage+0xa7/0x38b()
  WARNING: CPU: 0 PID: 27347 at fs/ext4/inode.c:1764 ext4_writepage+0xa7/0x38b()

from stressing some mmotm trees during July.

Could a dirty xfs or ext4 file page somehow get marked PageSwapBacked,
so fail shrink_page_list()'s page_is_file_cache() test, and so proceed
to mapping->a_ops->writepage()?

Yes, 3.16-rc1's commit 68711a7463 ("mm, migration: add destination
page freeing callback") has provided such a way to compaction: if
migrating a SwapBacked page fails, its newpage may be put back on the
list for later use with PageSwapBacked still set, and nothing will clear
it.

Whether that can do anything worse than issue WARN_ON_ONCEs, and get
some statistics wrong, is unclear: easier to fix than to think through
the consequences.

Fixing it here, before the put_new_page(), addresses the bug directly,
but is probably the worst place to fix it.  Page migration is doing too
many parts of the job on too many levels: fixing it in
move_to_new_page() to complement its SetPageSwapBacked would be
preferable, except why is it (and newpage->mapping and newpage->index)
done there, rather than down in migrate_page_move_mapping(), once we are
sure of success? Not a cleanup to get into right now, especially not
with memcg cleanups coming in 3.17.

Reported-by: Dave Jones <davej@redhat.com>
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-07-26 14:38:50 -07:00
..
backing-dev.c
balloon_compaction.c
bootmem.c
cleancache.c
compaction.c mm, compaction: properly signal and act upon lock and need_sched() contention 2014-06-04 16:54:11 -07:00
debug-pagealloc.c
dmapool.c mm/dmapool.c: reuse devres_release() to free resources 2014-06-04 16:54:08 -07:00
early_ioremap.c
fadvise.c
failslab.c
filemap_xip.c
filemap.c Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2014-06-12 10:30:18 -07:00
fremap.c mm: mark remap_file_pages() syscall as deprecated 2014-06-06 16:08:17 -07:00
frontswap.c swap: change swap_list_head to plist, add swap_avail_head 2014-06-04 16:54:07 -07:00
gup.c mm: cleanup __get_user_pages() 2014-06-04 16:54:05 -07:00
highmem.c
huge_memory.c mm: let mm_find_pmd fix buggy race with THP fault 2014-06-23 16:47:44 -07:00
hugetlb_cgroup.c
hugetlb.c mm: hugetlb: fix copy_hugetlb_page_range() 2014-07-23 15:10:55 -07:00
hwpoison-inject.c
init-mm.c
internal.h mm, compaction: properly signal and act upon lock and need_sched() contention 2014-06-04 16:54:11 -07:00
interval_tree.c
iov_iter.c
Kconfig mm/zsmalloc: make zsmalloc module-buildable 2014-06-04 16:54:14 -07:00
Kconfig.debug
kmemcheck.c
kmemleak-test.c mm/kmemleak-test.c: use pr_fmt for logging 2014-06-06 16:08:18 -07:00
kmemleak.c mm: introduce kmemleak_update_trace() 2014-06-06 16:08:17 -07:00
ksm.c mm: let mm_find_pmd fix buggy race with THP fault 2014-06-23 16:47:44 -07:00
list_lru.c
maccess.c
madvise.c
Makefile mm: move get_user_pages()-related code to separate file 2014-06-04 16:54:04 -07:00
memblock.c mm/memblock.c: call kmemleak directly from memblock_(alloc|free) 2014-06-06 16:08:17 -07:00
memcontrol.c Merge branch 'for-3.16' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup 2014-06-09 15:03:33 -07:00
memory_hotplug.c mm, migration: add destination page freeing callback 2014-06-04 16:54:06 -07:00
memory-failure.c mm/rmap.c: fix pgoff calculation to handle hugepage correctly 2014-07-23 15:10:54 -07:00
memory.c mm: do not call do_fault_around for non-linear fault 2014-07-23 15:10:54 -07:00
mempolicy.c Merge branch 'for-3.16-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup 2014-07-10 11:38:23 -07:00
mempool.c mm/mempool.c: update the kmemleak stack trace for mempool allocations 2014-06-06 16:08:17 -07:00
migrate.c mm: fix direct reclaim writeback regression 2014-07-26 14:38:50 -07:00
mincore.c
mlock.c
mm_init.c
mmap.c mm: convert some level-less printks to pr_* 2014-06-06 16:08:18 -07:00
mmu_context.c
mmu_notifier.c
mmzone.c
mprotect.c
mremap.c
msync.c msync: fix incorrect fstart calculation 2014-07-03 09:21:53 -07:00
nobootmem.c mm/memblock.c: call kmemleak directly from memblock_(alloc|free) 2014-06-06 16:08:17 -07:00
nommu.c mm: nommu: per-thread vma cache fix 2014-06-23 16:47:43 -07:00
oom_kill.c
page_alloc.c mm: page_alloc: fix CMA area initialisation when pageblock > MAX_ORDER 2014-07-03 09:21:53 -07:00
page_cgroup.c
page_io.c fix __swap_writepage() compile failure on old gcc versions 2014-06-14 19:30:48 -05:00
page_isolation.c
page-writeback.c Clean ups and miscellaneous bug fixes, in particular for the new 2014-06-08 13:03:35 -07:00
pagewalk.c
percpu-km.c
percpu-vm.c
percpu.c
pgtable-generic.c
process_vm_access.c
quicklist.c
readahead.c
rmap.c mm/rmap.c: fix pgoff calculation to handle hugepage correctly 2014-07-23 15:10:54 -07:00
shmem.c shmem: fix splicing from a hole while it's punched 2014-07-23 15:10:55 -07:00
slab_common.c Merge branch 'slab/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/penberg/linux into for-3.16-rcX 2014-07-22 18:38:27 -04:00
slab.c slab: fix oops when reading /proc/slab_allocators 2014-06-23 16:47:44 -07:00
slab.h memcg, slab: merge memcg_{bind,release}_pages to memcg_{un}charge_slab 2014-06-04 16:54:01 -07:00
slob.c slab: get_online_mems for kmem_cache_{create,destroy,shrink} 2014-06-04 16:53:59 -07:00
slub.c slub: fix off by one in number of slab tests 2014-07-03 09:21:53 -07:00
sparse-vmemmap.c
sparse.c
swap_state.c mm: page_alloc: convert hot/cold parameter and immediate callers to bool 2014-06-04 16:54:09 -07:00
swap.c mm: non-atomically mark page accessed during page cache allocation where possible 2014-06-04 16:54:10 -07:00
swapfile.c mm/swapfile.c: delete the "last_in_cluster < scan_base" loop in the body of scan_swap_map() 2014-06-04 16:54:12 -07:00
truncate.c mm/fs: fix pessimization in hole-punching pagecache 2014-07-23 15:10:55 -07:00
util.c
vmacache.c mm,vmacache: optimize overflow system-wide flushing 2014-06-04 16:53:57 -07:00
vmalloc.c mm/vmalloc.c: export unmap_kernel_range() 2014-06-04 16:54:14 -07:00
vmpressure.c
vmscan.c Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2014-06-12 10:30:18 -07:00
vmstat.c mm: use the light version __mod_zone_page_state in mlocked_vma_newpage() 2014-06-04 16:54:07 -07:00
workingset.c
zbud.c mm/zbud.c: make size unsigned like unique callsite 2014-06-04 16:54:13 -07:00
zsmalloc.c zsmalloc: fixup trivial zs size classes value in comments 2014-06-04 16:54:13 -07:00
zswap.c mm/zswap: NUMA aware allocation for zswap_dstmem 2014-06-04 16:54:14 -07:00