On Wed, Jun 22, 2022 at 11:33:48PM +0800, Muchun Song wrote: > On Wed, Jun 22, 2022 at 06:49:31AM -0700, syzbot wrote: > > Hello, > > > > syzbot found the following issue on: > > > > HEAD commit: ac0ba5454ca8 Add linux-next specific files for 20220622 > > git tree: linux-next > > console output: https://syzkaller.appspot.com/x/log.txt?x=14354c18080000 > > kernel config: https://syzkaller.appspot.com/x/.config?x=12809dacb9e7c5e0 > > dashboard link: https://syzkaller.appspot.com/bug?extid=ec972d37869318fc3ffb > > compiler: gcc (Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.2 > > > > Unfortunately, I don't have any reproducer for this issue yet. > > > > IMPORTANT: if you fix the issue, please add the following tag to the commit: > > Reported-by: syzbot+ec972d37869318fc3ffb@xxxxxxxxxxxxxxxxxxxxxxxxx > > > > folio_put include/linux/mm.h:1227 [inline] > > put_page+0x217/0x280 include/linux/mm.h:1279 > > unmap_and_move_huge_page mm/migrate.c:1343 [inline] > > migrate_pages+0x3dc3/0x5a10 mm/migrate.c:1440 > > do_mbind mm/mempolicy.c:1332 [inline] > > kernel_mbind+0x4d7/0x7d0 mm/mempolicy.c:1479 > > do_syscall_x64 arch/x86/entry/common.c:50 [inline] > > do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80 > > entry_SYSCALL_64_after_hwframe+0x46/0xb0 > > page has been migrated, last migrate reason: mempolicy_mbind > > ------------[ cut here ]------------ > > WARNING: CPU: 1 PID: 18925 at include/linux/memcontrol.h:800 folio_lruvec include/linux/memcontrol.h:800 [inline] > > The warning here is "VM_WARN_ON_ONCE_FOLIO(!memcg && !mem_cgroup_disabled(), folio)", > the memcg returned by folio_memcg() seems to be NULL which has 2 possibility, one is > that objcg returned by folio_objcg() is NULL, another is that obj_cgroup_memcg(objcg) > returns NULL. However, obj_cgroup_memcg() always returns a valid memcg. So Most likely > objcg is NULL meaning this page is not charged to memcg. Is this possible for LRU pages? > > I am not sure if this issue is caused by my commit cca700a8e695 ("mm: lru: use lruvec > lock to serialize memcg changes") since I have removed folio_test_clear_lru() check > from folio_batch_move_lru(). We know that a non-lru page may be not charged to memcg. > But is it possible for a non-lru page to be passed to folio_batch_move_lru()? Seems > impossible. Right? I am not very confident about this commit, hopefully, someone can > review it. How about to temporarily drop it? I was about to suggest it anyway during the review: it's a standalone optimization and the main patchset is already quite big and complex, so probably easier to validate it separately. Thanks!
