[syzbot] [cgroups?] BUG: sleeping function called from invalid context in cgroup_rstat_flush

[syzbot]

Hello,

syzbot found the following issue on:

HEAD commit: 7c16f0a4ed1c Merge tag 'i2c-for-6.10-rc5' of git://git.ker..
git tree: upstream
console output: https://syzkaller.appspot.com/x/log.txt?x=1511528e980000
kernel config: https://syzkaller.appspot.com/x/.config?x=12f98862a3c0c799
dashboard link: https://syzkaller.appspot.com/bug?extid=b7f13b2d0cc156edf61a
compiler: Debian clang version 15.0.6, GNU ld (GNU Binutils for Debian) 2.40

Unfortunately, I don't have any reproducer for this issue yet.

Downloadable assets:
disk image: https://storage.googleapis.com/syzbot-assets/50560e9024e5/disk-7c16f0a4.raw.xz
vmlinux: https://storage.googleapis.com/syzbot-assets/080c27daee72/vmlinux-7c16f0a4.xz
kernel image: https://storage.googleapis.com/syzbot-assets/c528e0da4544/bzImage-7c16f0a4.xz

IMPORTANT: if you fix the issue, please add the following tag to the commit:
Reported-by: syzbot+b7f13b...@syzkaller.appspotmail.com

BUG: sleeping function called from invalid context at kernel/cgroup/rstat.c:351
in_atomic(): 0, irqs_disabled(): 0, non_block: 0, pid: 17332, name: syz-executor.4
preempt_count: 0, expected: 0
RCU nest depth: 1, expected: 0
1 lock held by syz-executor.4/17332:
#0: ffffffff8e333fa0 (rcu_read_lock){....}-{1:2}, at: rcu_lock_acquire include/linux/rcupdate.h:329 [inline]
#0: ffffffff8e333fa0 (rcu_read_lock){....}-{1:2}, at: rcu_read_lock include/linux/rcupdate.h:781 [inline]
#0: ffffffff8e333fa0 (rcu_read_lock){....}-{1:2}, at: filemap_cachestat mm/filemap.c:4251 [inline]
#0: ffffffff8e333fa0 (rcu_read_lock){....}-{1:2}, at: __do_sys_cachestat mm/filemap.c:4407 [inline]
#0: ffffffff8e333fa0 (rcu_read_lock){....}-{1:2}, at: __se_sys_cachestat+0x3ee/0xbb0 mm/filemap.c:4372
CPU: 1 PID: 17332 Comm: syz-executor.4 Not tainted 6.10.0-rc4-syzkaller-00330-g7c16f0a4ed1c #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 06/07/2024
Call Trace:
<TASK>
__dump_stack lib/dump_stack.c:88 [inline]
dump_stack_lvl+0x241/0x360 lib/dump_stack.c:114
__might_resched+0x5d4/0x780 kernel/sched/core.c:10196
cgroup_rstat_flush+0x1e/0x50 kernel/cgroup/rstat.c:351
workingset_test_recent+0x48a/0xa90 mm/workingset.c:473
filemap_cachestat mm/filemap.c:4314 [inline]
__do_sys_cachestat mm/filemap.c:4407 [inline]
__se_sys_cachestat+0x795/0xbb0 mm/filemap.c:4372
do_syscall_x64 arch/x86/entry/common.c:52 [inline]
do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7ff329e7d0a9
Code: 28 00 00 00 75 05 48 83 c4 28 c3 e8 e1 20 00 00 90 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 b0 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007ff32ac600c8 EFLAGS: 00000246 ORIG_RAX: 00000000000001c3
RAX: ffffffffffffffda RBX: 00007ff329fb4120 RCX: 00007ff329e7d0a9
RDX: 0000000020000080 RSI: 0000000020000040 RDI: 0000000000000005
RBP: 00007ff329eec074 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 000000000000006e R14: 00007ff329fb4120 R15: 00007ffd3e0ff4a8
</TASK>


---
This report is generated by a bot. It may contain errors.
See https://goo.gl/tpsmEJ for more information about syzbot.
syzbot engineers can be reached at syzk...@googlegroups.com.

syzbot will keep track of this issue. See:
https://goo.gl/tpsmEJ#status for how to communicate with syzbot.

If the report is already addressed, let syzbot know by replying with:
#syz fix: exact-commit-title

If you want to overwrite report's subsystems, reply with:
#syz set subsystems: new-subsystem
(See the list of subsystem names on the web dashboard)

If the report is a duplicate of another one, reply with:
#syz dup: exact-subject-of-another-report

If you want to undo deduplication, reply with:
#syz undup

[Johannes Weiner]

Ok yeah, cachestat() holds the rcu read lock, so
workingset_test_recent() can't do a sleepable rstat flush.

I think the easiest fix would be to flush rstat from the root down
(NULL) in filemap_cachestat(), before the rcu section, and add a flag
to workingset_test_recent() to forego it. Nhat?

[Nhat Pham]

You're right. I think it's been broken since this commit:

b00684722262 mm: workingset: move the stats flush into workingset_test_recent()

which moves the stats flushing from the refault step (before rcu read
lock section) to inside workingset_test_recent(). I believe that's
6.8, 6.9, and 6.10 we need to fix?

The fix sounds reasonable to me :) Let me whip up something real quick.