Re: WARNING in percpu_ref_exit (2)

[Eric Biggers <ebiggers@kernel.org>]

On Sat, Dec 21, 2019 at 07:02:25AM -0700, Jens Axboe wrote:
> On 12/21/19 6:43 AM, Hillf Danton wrote:
> > 
> > On Sat, 21 Dec 2019 00:05:07 -0800
> >> Hello,
> >>
> >> syzbot found the following crash on:
> >>
> >> HEAD commit:    7ddd09fc Add linux-next specific files for 20191220
> >> git tree:       linux-next
> >> console output: https://syzkaller.appspot.com/x/log.txt?x=12a18cc6e00000
> >> kernel config:  https://syzkaller.appspot.com/x/.config?x=f183b01c3088afc6
> >> dashboard link: https://syzkaller.appspot.com/bug?extid=8c4a14856e657b43487c
> >> compiler:       gcc (GCC) 9.0.0 20181231 (experimental)
> >> syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=14b8f351e00000
> >> C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=14b51925e00000
> >>
> >> IMPORTANT: if you fix the bug, please add the following tag to the commit:
> >> Reported-by: syzbot+8c4a14856e657b43487c@syzkaller.appspotmail.com
> >>
> >> ------------[ cut here ]------------
> >> WARNING: CPU: 1 PID: 11482 at lib/percpu-refcount.c:111  
> >> percpu_ref_exit+0xab/0xd0 lib/percpu-refcount.c:111
> >> Kernel panic - not syncing: panic_on_warn set ...
> >> CPU: 1 PID: 11482 Comm: syz-executor051 Not tainted  
> >> 5.5.0-rc2-next-20191220-syzkaller #0
> >> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS  
> >> Google 01/01/2011
> >> Call Trace:
> >>   __dump_stack lib/dump_stack.c:77 [inline]
> >>   dump_stack+0x197/0x210 lib/dump_stack.c:118
> >>   panic+0x2e3/0x75c kernel/panic.c:221
> >>   __warn.cold+0x2f/0x3e kernel/panic.c:582
> >>   report_bug+0x289/0x300 lib/bug.c:195
> >>   fixup_bug arch/x86/kernel/traps.c:174 [inline]
> >>   fixup_bug arch/x86/kernel/traps.c:169 [inline]
> >>   do_error_trap+0x11b/0x200 arch/x86/kernel/traps.c:267
> >>   do_invalid_op+0x37/0x50 arch/x86/kernel/traps.c:286
> >>   invalid_op+0x23/0x30 arch/x86/entry/entry_64.S:1027
> >> RIP: 0010:percpu_ref_exit+0xab/0xd0 lib/percpu-refcount.c:111
> >> Code: 00 00 00 00 fc ff df 48 c1 ea 03 80 3c 02 00 75 1d 48 c7 43 08 03 00  
> >> 00 00 e8 01 41 e5 fd 5b 41 5c 41 5d 5d c3 e8 f5 40 e5 fd <0f> 0b eb bf 4c  
> >> 89 ef e8 29 2c 23 fe eb d9 e8 82 2b 23 fe eb a7 4c
> >> RSP: 0018:ffffc9000cb17968 EFLAGS: 00010293
> >> RAX: ffff8880a3390640 RBX: ffff8880a83a8010 RCX: ffffffff83901432
> >> RDX: 0000000000000000 RSI: ffffffff8390149b RDI: ffff8880a83a8028
> >> RBP: ffffc9000cb17980 R08: ffff8880a3390640 R09: 0000000000000000
> >> R10: 0000000000000000 R11: 0000000000000000 R12: 0000607f51435750
> >> R13: ffff8880a83a8018 R14: ffff888097b95000 R15: ffff888097b95228
> >>   io_sqe_files_unregister+0x7d/0x2f0 fs/io_uring.c:4623
> >>   io_ring_ctx_free fs/io_uring.c:5575 [inline]
> >>   io_ring_ctx_wait_and_kill+0x430/0x9a0 fs/io_uring.c:5644
> >>   io_uring_release+0x42/0x50 fs/io_uring.c:5652
> >>   __fput+0x2ff/0x890 fs/file_table.c:280
> >>   ____fput+0x16/0x20 fs/file_table.c:313
> >>   task_work_run+0x145/0x1c0 kernel/task_work.c:113
> >>   exit_task_work include/linux/task_work.h:22 [inline]
> >>   do_exit+0x909/0x2f20 kernel/exit.c:797
> >>   do_group_exit+0x135/0x360 kernel/exit.c:895
> >>   get_signal+0x47c/0x24f0 kernel/signal.c:2734
> >>   do_signal+0x87/0x1700 arch/x86/kernel/signal.c:815
> >>   exit_to_usermode_loop+0x286/0x380 arch/x86/entry/common.c:160
> >>   prepare_exit_to_usermode arch/x86/entry/common.c:195 [inline]
> >>   syscall_return_slowpath arch/x86/entry/common.c:278 [inline]
> >>   do_syscall_64+0x676/0x790 arch/x86/entry/common.c:304
> >>   entry_SYSCALL_64_after_hwframe+0x49/0xbe
> > 
> > Flush work before killing.
> > 
> > --- a/fs/io_uring.c
> > +++ b/fs/io_uring.c
> > @@ -4618,10 +4618,10 @@ static int io_sqe_files_unregister(struc
> >  	if (!data)
> >  		return -ENXIO;
> >  
> > +	flush_work(&data->ref_work);
> >  	percpu_ref_kill_and_confirm(&data->refs, io_file_ref_kill);
> >  	wait_for_completion(&data->done);
> >  	percpu_ref_exit(&data->refs);
> > -	flush_work(&data->ref_work);
> >  
> >  	__io_sqe_files_unregister(ctx);
> >  	nr_tables = DIV_ROUND_UP(ctx->nr_user_files, IORING_MAX_FILES_TABLE);
> 
> Oh indeed, good catch! Thanks, I'll fold this in.

Hi Jens, this bug is still occurring.  See the syzbot dashboard:
https://syzkaller.appspot.com/bug?extid=8c4a14856e657b43487c

Here's a crash on upstream commit 7ae77150d94d (Fri Jun 5 12:39:30 2020 -0700):
https://syzkaller.appspot.com/text?tag=CrashReport&x=1754d169100000

- Eric