linux/mm
Linus Torvalds 80d47f5de5 mm: don't try to NUMA-migrate COW pages that have other uses
Oded Gabbay reports that enabling NUMA balancing causes corruption with
his Gaudi accelerator test load:

 "All the details are in the bug, but the bottom line is that somehow,
  this patch causes corruption when the numa balancing feature is
  enabled AND we don't use process affinity AND we use GUP to pin pages
  so our accelerator can DMA to/from system memory.

  Either disabling numa balancing, using process affinity to bind to
  specific numa-node or reverting this patch causes the bug to
  disappear"

and Oded bisected the issue to commit 09854ba94c ("mm: do_wp_page()
simplification").

Now, the NUMA balancing shouldn't actually be changing the writability
of a page, and as such shouldn't matter for COW.  But it appears it
does.  Suspicious.

However, regardless of that, the condition for enabling NUMA faults in
change_pte_range() is nonsensical.  It uses "page_mapcount(page)" to
decide if a COW page should be NUMA-protected or not, and that makes
absolutely no sense.

The number of mappings a page has is irrelevant: not only does GUP get a
reference to a page as in Oded's case, but the other mappings migth be
paged out and the only reference to them would be in the page count.

Since we should never try to NUMA-balance a page that we can't move
anyway due to other references, just fix the code to use 'page_count()'.
Oded confirms that that fixes his issue.

Now, this does imply that something in NUMA balancing ends up changing
page protections (other than the obvious one of making the page
inaccessible to get the NUMA faulting information).  Otherwise the COW
simplification wouldn't matter - since doing the GUP on the page would
make sure it's writable.

The cause of that permission change would be good to figure out too,
since it clearly results in spurious COW events - but fixing the
nonsensical test that just happened to work before is obviously the
CorrectThing(tm) to do regardless.

Fixes: 09854ba94c ("mm: do_wp_page() simplification")
Link: https://bugzilla.kernel.org/show_bug.cgi?id=215616
Link: https://lore.kernel.org/all/CAFCwf10eNmwq2wD71xjUhqkvv5+_pJMR1nPug2RqNDcFT4H86Q@mail.gmail.com/
Reported-and-tested-by: Oded Gabbay <oded.gabbay@gmail.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-02-17 08:57:47 -08:00
..
damon mm/damon: hide kernel pointer from tracepoint event 2022-01-15 16:30:33 +02:00
kasan lib/stackdepot: always do filter_irq_stacks() in stack_depot_save() 2022-01-22 08:33:38 +02:00
kfence kfence: make test case compatible with run time set sample interval 2022-02-11 17:55:00 -08:00
backing-dev.c mm: bdi: initialize bdi_min_ratio when bdi is unregistered 2021-12-10 17:10:56 -08:00
balloon_compaction.c
bootmem_info.c bootmem: Use page->index instead of page->freelist 2022-01-06 12:27:03 +01:00
cma_debug.c
cma_sysfs.c
cma.c memblock: rename memblock_free to memblock_phys_free 2021-11-06 13:30:41 -07:00
cma.h
compaction.c mm: compaction: fix the migration stats in trace_mm_compaction_migratepages() 2022-01-15 16:30:30 +02:00
debug_page_ref.c
debug_vm_pgtable.c mm/debug_vm_pgtable: remove pte entry from the page table 2022-02-04 09:25:04 -08:00
debug.c mm,fs: split dump_mapping() out from dump_page() 2022-01-15 16:30:26 +02:00
dmapool.c mm/dmapool.c: revert "make dma pool to use kmalloc_node" 2022-01-15 16:30:28 +02:00
early_ioremap.c mm/early_ioremap.c: remove redundant early_ioremap_shutdown() 2021-09-08 11:50:24 -07:00
fadvise.c
failslab.c
filemap.c Merge branch 'akpm' (patches from Andrew) 2022-01-22 11:28:23 +02:00
folio-compat.c filemap: Add filemap_release_folio() 2022-01-04 13:15:34 -05:00
frontswap.c frontswap: remove support for multiple ops 2022-01-22 08:33:38 +02:00
gup_test.c
gup_test.h
gup.c Revert "mm/gup: small refactoring: simplify try_grab_page()" 2022-02-03 06:51:42 -08:00
highmem.c Fixes for 5.16 folios: 2021-11-25 10:13:56 -08:00
hmm.c mm/hmm.c: allow VM_MIXEDMAP to work with hmm_range_fault 2022-01-15 16:30:31 +02:00
huge_memory.c Merge branch 'akpm' (patches from Andrew) 2022-01-15 20:37:06 +02:00
hugetlb_cgroup.c hugetlb: add hugetlb.*.numa_stat file 2022-01-15 16:30:29 +02:00
hugetlb_vmemmap.c
hugetlb_vmemmap.h
hugetlb.c mm: change page type prior to adding page table entry 2022-01-15 16:30:28 +02:00
hwpoison-inject.c mm: hwpoison: don't drop slab caches for offlining non-LRU page 2021-09-03 09:58:15 -07:00
init-mm.c mm: add setup_initial_init_mm() helper 2021-07-08 11:48:21 -07:00
internal.h Merge branch 'akpm' (patches from Andrew) 2022-01-15 20:37:06 +02:00
interval_tree.c
io-mapping.c
ioremap.c mm: move ioremap_page_range to vmalloc.c 2021-09-08 11:50:24 -07:00
Kconfig mm: hide the FRONTSWAP Kconfig symbol 2022-01-22 08:33:38 +02:00
Kconfig.debug mm: page table check 2022-01-15 16:30:28 +02:00
khugepaged.c mm/page_table_check: check entries at pmd levels 2022-02-04 09:25:04 -08:00
kmemleak.c mm/kmemleak: avoid scanning potential huge holes 2022-02-04 09:25:05 -08:00
ksm.c mm: ksm: fix use-after-free kasan report in ksm_might_need_to_copy 2022-01-15 16:30:31 +02:00
list_lru.c mm: list_lru: only add memcg-aware lrus to the global lru list 2021-11-06 13:30:35 -07:00
maccess.c ARM: 9115/1: mm/maccess: fix unaligned copy_{from,to}_kernel_nofault 2021-08-20 11:39:25 +01:00
madvise.c mm: move anon_vma declarations to linux/mm_inline.h 2022-01-15 16:30:27 +02:00
Makefile mm: remove cleancache 2022-01-22 08:33:38 +02:00
mapping_dirty_helpers.c mm: move tlb_flush_pending inline helpers to mm_inline.h 2022-01-15 16:30:27 +02:00
memblock.c arm64 fixes for -rc1 2021-11-10 11:29:30 -08:00
memcontrol.c mm: memcg: synchronize objcg lists with a dedicated spinlock 2022-02-11 17:55:00 -08:00
memfd.c mm,hugetlb: remove mlock ulimit for SHM_HUGETLB 2021-11-09 10:02:48 -08:00
memory_hotplug.c treewide: Add missing includes masked by cgroup -> bpf dependency 2021-12-03 10:58:13 -08:00
memory-failure.c memory-failure: fetch compound_head after pgmap_pfn_valid() 2022-01-30 09:56:58 +02:00
memory.c Merge branch 'akpm' (patches from Andrew) 2022-01-20 10:41:01 +02:00
mempolicy.c mm/mempolicy: fix all kernel-doc warnings 2022-01-15 16:30:30 +02:00
mempool.c mm: remove spurious blkdev.h includes 2021-10-18 06:17:01 -06:00
memremap.c Merge branch 'akpm' (patches from Andrew) 2022-01-15 20:37:06 +02:00
memtest.c
migrate.c mm/migrate.c: rework migration_entry_wait() to not take a pageref 2022-01-22 08:33:34 +02:00
mincore.c
mlock.c mm: add a field to store names for private anonymous memory 2022-01-15 16:30:27 +02:00
mm_init.c
mmap_lock.c mm: mmap_lock: fix disabling preemption directly 2021-07-23 17:43:28 -07:00
mmap.c Merge branch 'signal-for-v5.17' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace 2022-01-17 05:49:30 +02:00
mmu_gather.c mm: move tlb_flush_pending inline helpers to mm_inline.h 2022-01-15 16:30:27 +02:00
mmu_notifier.c
mmzone.c
mprotect.c mm: don't try to NUMA-migrate COW pages that have other uses 2022-02-17 08:57:47 -08:00
mremap.c mm, hugepages: add mremap() support for hugepage backed vma 2021-11-06 13:30:39 -07:00
msync.c
nommu.c Merge branch 'akpm' (patches from Andrew) 2021-11-06 14:08:17 -07:00
oom_kill.c Merge branch 'signal-for-v5.17' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace 2022-01-17 05:49:30 +02:00
page_alloc.c Merge branch 'akpm' (patches from Andrew) 2022-01-20 10:41:01 +02:00
page_counter.c mm/page_counter: remove an incorrect call to propagate_protected_usage() 2022-01-15 16:30:27 +02:00
page_ext.c mm: make some vars and functions static or __init 2022-01-15 16:30:31 +02:00
page_idle.c mm/idle_page_tracking: make PG_idle reusable 2021-09-08 11:50:24 -07:00
page_io.c delayacct: support swapin delay accounting for swapping without blkio 2022-01-20 08:52:55 +02:00
page_isolation.c Revert "mm/page_isolation: unset migratetype directly for non Buddy page" 2022-02-04 09:25:04 -08:00
page_owner.c lib/stackdepot: allow optional init and stack_table allocation by kvmalloc() 2022-01-22 08:33:37 +02:00
page_poison.c
page_reporting.c
page_reporting.h
page_table_check.c mm/page_table_check: check entries at pmd levels 2022-02-04 09:25:04 -08:00
page_vma_mapped.c mm: device exclusive memory access 2021-07-01 11:06:03 -07:00
page-writeback.c mm/writeback: Improve __folio_mark_dirty() comment 2022-01-02 20:28:57 -05:00
pagewalk.c
percpu-internal.h mm: memcg/percpu: account extra objcg space to memory cgroups 2022-01-15 16:30:31 +02:00
percpu-km.c percpu: flush tlb in pcpu_reclaim_populated() 2021-07-04 18:30:17 +00:00
percpu-stats.c
percpu-vm.c percpu: flush tlb in pcpu_reclaim_populated() 2021-07-04 18:30:17 +00:00
percpu.c bitmap patches for 5.17-rc1 2022-01-23 06:20:44 +02:00
pgalloc-track.h
pgtable-generic.c mm: move tlb_flush_pending inline helpers to mm_inline.h 2022-01-15 16:30:27 +02:00
process_vm_access.c
ptdump.c
readahead.c readahead: Convert page_cache_ra_unbounded to folios 2022-01-04 13:15:34 -05:00
rmap.c mm/rmap: fix potential batched TLB flush race 2022-01-15 16:30:31 +02:00
rodata_test.c
secretmem.c mm/secretmem: avoid letting secretmem_users drop to zero 2021-10-28 17:18:55 -07:00
shmem.c mm: simplify try_to_unuse 2022-01-22 08:33:38 +02:00
shuffle.c
shuffle.h
slab_common.c Merge branch 'akpm' (patches from Andrew) 2022-01-15 20:37:06 +02:00
slab.c mm/kasan: Convert to struct folio and struct slab 2022-01-06 12:26:14 +01:00
slab.h slab changes for 5.17 - part 2 2022-01-18 06:40:47 +02:00
slob.c mm/slob: Remove unnecessary page_mapcount_reset() function call 2022-01-06 12:27:28 +01:00
slub.c mm/slub: Define struct slab fields for CONFIG_SLUB_CPU_PARTIAL only when enabled 2022-01-06 12:26:53 +01:00
sparse-vmemmap.c mm: remove redundant smp_wmb() 2021-11-06 13:30:36 -07:00
sparse.c bootmem: Use page->index instead of page->freelist 2022-01-06 12:27:03 +01:00
swap_cgroup.c
swap_slots.c treewide: Add missing includes masked by cgroup -> bpf dependency 2021-12-03 10:58:13 -08:00
swap_state.c mm/workingset: Convert workingset_refault() to take a folio 2021-10-18 07:49:40 -04:00
swap.c Merge branch 'akpm' (patches from Andrew) 2022-01-15 20:37:06 +02:00
swapfile.c mm: mark swap_lock and swap_active_head static 2022-01-22 08:33:38 +02:00
truncate.c mm: remove cleancache 2022-01-22 08:33:38 +02:00
usercopy.c mm: Convert check_heap_object() to use struct slab 2022-01-06 12:25:51 +01:00
userfaultfd.c mm: shmem: don't truncate page if memory failure happens 2022-01-15 16:30:26 +02:00
util.c mm: allow !GFP_KERNEL allocations for kvmalloc 2022-01-15 16:30:29 +02:00
vmacache.c
vmalloc.c mm/vmalloc: be more explicit about supported gfp flags. 2022-01-15 16:30:28 +02:00
vmpressure.c mm/vmpressure: fix data-race with memcg->socket_pressure 2021-11-06 13:30:40 -07:00
vmscan.c mm: vmscan: remove deadlock due to throttling failing to make progress 2022-02-11 17:55:00 -08:00
vmstat.c mm/vmstat: add events for THP max_ptes_* exceeds 2022-01-15 16:30:29 +02:00
workingset.c Merge branch 'akpm' (patches from Andrew) 2021-11-09 10:11:53 -08:00
z3fold.c mm/z3fold: add kerneldoc fields for z3fold_pool 2021-07-01 11:06:03 -07:00
zbud.c mm/zbud: add kerneldoc fields for zbud_pool 2021-07-01 11:06:03 -07:00
zpool.c zpool: remove the list of pools_head 2022-01-15 16:30:31 +02:00
zsmalloc.c zsmalloc: replace get_cpu_var with local_lock 2022-01-22 08:33:37 +02:00
zswap.c frontswap: remove support for multiple ops 2022-01-22 08:33:38 +02:00