Re: [syzbot] [cgroups?] [mm?] WARNING in __mod_memcg_lruvec_state

[Yosry Ahmed]

On Thu, Apr 4, 2024 at 6:44 PM Yu Zhao <yuzhao@xxxxxxxxxx> wrote:
>
> On Thu, Apr 4, 2024 at 7:36 PM Yosry Ahmed <yosryahmed@xxxxxxxxxx> wrote:
> >
> > On Wed, Apr 3, 2024 at 6:08 PM Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote:
> > >
> > > On Tue, 02 Apr 2024 01:03:26 -0700 syzbot <syzbot+9319a4268a640e26b72b@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
> > >
> > > > Hello,
> > > >
> > > > syzbot found the following issue on:
> > > >
> > > > HEAD commit:    317c7bc0ef03 Merge tag 'mmc-v6.9-rc1' of git://git.kernel...
> > > > git tree:       upstream
> > > > console output: https://syzkaller.appspot.com/x/log.txt?x=15fd40c5180000
> > > > kernel config:  https://syzkaller.appspot.com/x/.config?x=f64ec427e98bccd7
> > > > dashboard link: https://syzkaller.appspot.com/bug?extid=9319a4268a640e26b72b
> > > > compiler:       gcc (Debian 12.2.0-14) 12.2.0, GNU ld (GNU Binutils for Debian) 2.40
> > > >
> > > > Unfortunately, I don't have any reproducer for this issue yet.
> > > >
> > > > Downloadable assets:
> > > > disk image (non-bootable): https://storage.googleapis.com/syzbot-assets/7bc7510fe41f/non_bootable_disk-317c7bc0.raw.xz
> > > > vmlinux: https://storage.googleapis.com/syzbot-assets/efab473d72c0/vmlinux-317c7bc0.xz
> > > > kernel image: https://storage.googleapis.com/syzbot-assets/5ba3f56d362d/bzImage-317c7bc0.xz
> > > >
> > > > IMPORTANT: if you fix the issue, please add the following tag to the commit:
> > > > Reported-by: syzbot+9319a4268a640e26b72b@xxxxxxxxxxxxxxxxxxxxxxxxx
> > > >
> > > > ------------[ cut here ]------------
> > > > WARNING: CPU: 0 PID: 30105 at mm/memcontrol.c:865 __mod_memcg_lruvec_state+0x3fa/0x550 mm/memcontrol.c:865
> > > > Modules linked in:
> > > > CPU: 0 PID: 30105 Comm: syz-executor.2 Not tainted 6.9.0-rc1-syzkaller-00178-g317c7bc0ef03 #0
> > > > Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.2-debian-1.16.2-1 04/01/2014
> > > > RIP: 0010:__mod_memcg_lruvec_state+0x3fa/0x550 mm/memcontrol.c:865
> > > > Code: 45 85 e4 75 1d 48 83 c4 18 5b 5d 41 5c 41 5d 41 5e 41 5f c3 cc cc cc cc b8 00 04 00 00 e9 80 fd ff ff 89 c6 e9 a0 fd ff ff 90 <0f> 0b 90 e9 a7 fc ff ff 48 c7 c7 18 43 e1 8f e8 32 51 f8 ff e9 5e
> > > > RSP: 0018:ffffc900034beef8 EFLAGS: 00010202
> > > > RAX: 0000000000000292 RBX: 0000000000000001 RCX: 1ffffffff1fc2863
> > > > RDX: 0000000000000000 RSI: 0000000000000001 RDI: ffff888024b92bc8
> > > > RBP: ffff888024b92000 R08: 0000000000000005 R09: 0000000000000000
> > > > R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000001
> > > > R13: ffff88801c326000 R14: 0000000000000001 R15: ffff888024b92000
> > > > FS:  00007f0811bf96c0(0000) GS:ffff88806b000000(0000) knlGS:0000000000000000
> > > > CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > > > CR2: 000000000cfff1dd CR3: 000000003e4e2000 CR4: 0000000000350ef0
> > > > DR0: 0000000000000031 DR1: 0000000000000000 DR2: 0000000000000000
> > > > DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> > > > Call Trace:
> > > >  <TASK>
> > > >  __update_lru_size include/linux/mm_inline.h:47 [inline]
> > > >  lru_gen_update_size include/linux/mm_inline.h:199 [inline]
> > > >  lru_gen_add_folio+0x62d/0xe80 include/linux/mm_inline.h:262
> > > >  lruvec_add_folio include/linux/mm_inline.h:323 [inline]
> > > >  lru_add_fn+0x3fc/0xd80 mm/swap.c:215
> > > >  folio_batch_move_lru+0x243/0x400 mm/swap.c:233
> > >
> > > Well it beats me.  I assume we failed to update for a new case.  I'll
> > > toss this into -next to perhaps shed a bit of light.
> > >
> > > --- a/mm/memcontrol.c~__mod_memcg_lruvec_state-enhance-diagnostics
> > > +++ a/mm/memcontrol.c
> > > @@ -860,10 +860,12 @@ void __mod_memcg_lruvec_state(struct lru
> > >                 case NR_ANON_THPS:
> > >                 case NR_SHMEM_PMDMAPPED:
> > >                 case NR_FILE_PMDMAPPED:
> > > -                       WARN_ON_ONCE(!in_task());
> > > +                       if (WARN_ON_ONCE(!in_task()))
> > > +                               pr_warn("stat item index: %d\n", idx);
> > >                         break;
> > >                 default:
> > > -                       VM_WARN_ON_IRQS_ENABLED();
> > > +                       if (VM_WARN_ON_IRQS_ENABLED())
> > > +                               pr_warn("stat item index: %d\n", idx);
> >
> > Line 865 from this commit should be this warning (i.e. warning because
> > IRQs are enabled). This also makes sense because __update_lru_size()
> > should not be updating any of the above stats.
> >
> > folio_batch_move_lru() in the above call stack should be acquiring the
> > lock with IRQs disabled though, so I am not sure what's going on from
> > a quick look.
> >
> > Adding Yu Zhao here.
>
> Probably an RT build where _irqsave doesn't disable IRQ?

Yeah, that's possible. I don't have enough familiarity with RT to know
for sure. Perhaps lruvec->lru_lock needs to be changed to a
raw_spinlock_t? I am not sure if that's the correct course of action.