Re: [Cluster-devel] [syzbot] [gfs2?] kernel panic: hung_task: blocked tasks (2)

Bob Peterson <[email protected]>
Newsgroups com.redhat.cluster-devel
Message-ID <20230728114847.b0jVD4olddgMRV4PFvEG38uw_0ziOBZRxpgqKBQrfTI@z>
On 7/28/23 3:20 AM, David Howells wrote:
> syzbot <[email protected]> wrote:
> 
>> Fixes: 9c8ad7a2ff0b ("uapi, x86: Fix the syscall numbering of the mount API syscalls [ver #2]")
> 
> This would seem unlikely to be the culprit.  It just changes the numbering on
> the fsconfig-related syscalls.
> 
> Running the test program on v6.5-rc3, however, I end up with the test process
> stuck in the D state:
> 
> INFO: task repro-17687f1aa:5551 blocked for more than 120 seconds.
>        Not tainted 6.5.0-rc3-build3+ #1448
> "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> task:repro-17687f1aa state:D stack:0     pid:5551  ppid:5516   flags:0x00004002
> Call Trace:
>   <TASK>
>   __schedule+0x4a7/0x4f1
>   schedule+0x66/0xa1
>   schedule_timeout+0x9d/0xd7
>   ? __next_timer_interrupt+0xf6/0xf6
>   gfs2_gl_hash_clear+0xa0/0xdc
>   ? sugov_irq_work+0x15/0x15
>   gfs2_put_super+0x19f/0x1d3
>   generic_shutdown_super+0x78/0x187
>   kill_block_super+0x1c/0x32
>   deactivate_locked_super+0x2f/0x61
>   cleanup_mnt+0xab/0xcc
>   task_work_run+0x6b/0x80
>   exit_to_user_mode_prepare+0x76/0xfd
>   syscall_exit_to_user_mode+0x14/0x31
>   entry_SYSCALL_64_after_hwframe+0x63/0xcd
> RIP: 0033:0x7f89aac31dab
> RSP: 002b:00007fff43d9b878 EFLAGS: 00000206 ORIG_RAX: 00000000000000a6
> RAX: 0000000000000000 RBX: 00007fff43d9cad8 RCX: 00007f89aac31dab
> RDX: 0000000000000000 RSI: 000000000000000a RDI: 00007fff43d9b920
> RBP: 00007fff43d9c960 R08: 0000000000000000 R09: 0000000000000073
> R10: 0000000000000000 R11: 0000000000000206 R12: 0000000000000000
> R13: 00007fff43d9cae8 R14: 0000000000417e18 R15: 00007f89aad51000
>   </TASK>
> 
> David
> 
Hi David,

This indicates gfs2 is having trouble resolving and freeing all its 
glocks, which usually means a reference counting problem or ail (active 
items list) problem during unmount.

If gfs2_gl_hash_clear gets stuck for a long period of time it is 
supposed to dump the remaining list of glocks that still have not been 
resolved. I think it takes 10 minutes or so. Can you post the console 
messages that follow? That will help us figure out what's happening. Thanks.

Regards,

Bob Peterson
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.