Re: [syzbot] [mm?] KCSAN: data-race in try_to_migrate_one / zap_page_range_single_batched

[Lance Yang <lance.yang@linux.dev>]

On Fri, Sep 26, 2025 at 2:55 PM Marco Elver <elver@google.com> wrote:
>
> On Fri, 26 Sept 2025 at 08:44, syzbot
> <syzbot+60192c8877d0bc92a92b@syzkaller.appspotmail.com> wrote:
> >
> > Hello,
> >
> > syzbot found the following issue on:
> >
> > HEAD commit:    cec1e6e5d1ab Merge tag 'sched_ext-for-6.17-rc7-fixes' of g..
> > git tree:       upstream
> > console output: https://syzkaller.appspot.com/x/log.txt?x=145d4f12580000
> > kernel config:  https://syzkaller.appspot.com/x/.config?x=6e0c213d0735f5dd
> > dashboard link: https://syzkaller.appspot.com/bug?extid=60192c8877d0bc92a92b
> > compiler:       Debian clang version 20.1.8 (++20250708063551+0c9f909b7976-1~exp1~20250708183702.136), Debian LLD 20.1.8
> >
> > Unfortunately, I don't have any reproducer for this issue yet.
> >
> > Downloadable assets:
> > disk image: https://storage.googleapis.com/syzbot-assets/10b7c8fdfdec/disk-cec1e6e5.raw.xz
> > vmlinux: https://storage.googleapis.com/syzbot-assets/cbecc36962db/vmlinux-cec1e6e5.xz
> > kernel image: https://storage.googleapis.com/syzbot-assets/214f107d0a3e/bzImage-cec1e6e5.xz
> >
> > IMPORTANT: if you fix the issue, please add the following tag to the commit:
> > Reported-by: syzbot+60192c8877d0bc92a92b@syzkaller.appspotmail.com
> >
> > ==================================================================
> > BUG: KCSAN: data-race in try_to_migrate_one / zap_page_range_single_batched
> >
> > write to 0xffff88810adfd798 of 8 bytes by task 13594 on cpu 1:
> >  update_hiwater_rss include/linux/mm.h:2657 [inline]
> >  try_to_migrate_one+0x918/0x16e0 mm/rmap.c:2455
> >  __rmap_walk_file+0x1ec/0x2b0 mm/rmap.c:2905
> >  try_to_migrate+0x1db/0x210 mm/rmap.c:-1
> >  migrate_folio_unmap mm/migrate.c:1324 [inline]
> >  migrate_pages_batch+0x6e1/0x1ae0 mm/migrate.c:1873
> >  migrate_pages_sync mm/migrate.c:1996 [inline]
> >  migrate_pages+0xf5f/0x1770 mm/migrate.c:2105
> >  do_mbind mm/mempolicy.c:1539 [inline]
> >  kernel_mbind mm/mempolicy.c:1682 [inline]
> >  __do_sys_mbind mm/mempolicy.c:1756 [inline]
> >  __se_sys_mbind+0x975/0xac0 mm/mempolicy.c:1752
> >  __x64_sys_mbind+0x78/0x90 mm/mempolicy.c:1752
> >  x64_sys_call+0x2932/0x2ff0 arch/x86/include/generated/asm/syscalls_64.h:238
> >  do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
> >  do_syscall_64+0xd2/0x200 arch/x86/entry/syscall_64.c:94
> >  entry_SYSCALL_64_after_hwframe+0x77/0x7f
> >
> > write to 0xffff88810adfd798 of 8 bytes by task 13595 on cpu 0:
> >  update_hiwater_rss include/linux/mm.h:2657 [inline]
> >  zap_page_range_single_batched+0x182/0x450 mm/memory.c:2007
> >  zap_page_range_single mm/memory.c:2041 [inline]
> >  unmap_mapping_range_vma mm/memory.c:4020 [inline]
> >  unmap_mapping_range_tree+0xfd/0x160 mm/memory.c:4037
> >  unmap_mapping_pages mm/memory.c:4103 [inline]
> >  unmap_mapping_range+0xe4/0xf0 mm/memory.c:4140
> >  shmem_fallocate+0x262/0x840 mm/shmem.c:3746
> >  vfs_fallocate+0x3b6/0x400 fs/open.c:342
> >  madvise_remove mm/madvise.c:1049 [inline]
> >  madvise_vma_behavior+0x192d/0x1cf0 mm/madvise.c:1346
> >  madvise_walk_vmas mm/madvise.c:1669 [inline]
> >  madvise_do_behavior+0x5b7/0x970 mm/madvise.c:1885
> >  do_madvise+0x10e/0x190 mm/madvise.c:1978
> >  __do_sys_madvise mm/madvise.c:1987 [inline]
> >  __se_sys_madvise mm/madvise.c:1985 [inline]
> >  __x64_sys_madvise+0x64/0x80 mm/madvise.c:1985
> >  x64_sys_call+0x1f1a/0x2ff0 arch/x86/include/generated/asm/syscalls_64.h:29
> >  do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
> >  do_syscall_64+0xd2/0x200 arch/x86/entry/syscall_64.c:94
> >  entry_SYSCALL_64_after_hwframe+0x77/0x7f
> >
> > value changed: 0x0000000000001645 -> 0x0000000000002165
>
> One of these writes is getting lost. Which means highwater_rss is
> lossy/approximate - does it matter?

Yes. IMO, hiwater_rss being lossy is an acceptable trade-off for a
non-critical accounting value. The patch[1] simply silences the KCSAN warning.

[1] https://lore.kernel.org/linux-mm/20250926092426.43312-1-lance.yang@linux.dev/