Re: INFO: task hung in io_uring_cancel_sqpoll

[Pavel Begunkov]

On 4/30/21 4:02 PM, Palash Oswal wrote:
> On Fri, Apr 30, 2021 at 8:03 PM Pavel Begunkov <asml.silence@xxxxxxxxx> wrote:
>>
>> On 4/30/21 3:21 PM, Palash Oswal wrote:
>>> On Thursday, March 18, 2021 at 9:40:21 PM UTC+5:30 syzbot wrote:
>>>>
>>>> Hello,
>>>>
>>>> syzbot found the following issue on:
>>>>
>>>> HEAD commit: 0d7588ab riscv: process: Fix no prototype for arch_dup_tas..
>>>> git tree: git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux.git fixes
>>>> console output: https://syzkaller.appspot.com/x/log.txt?x=12dde5aed00000
>>>> kernel config: https://syzkaller.appspot.com/x/.config?x=81c0b708b31626cc
>>>> dashboard link: https://syzkaller.appspot.com/bug?extid=11bf59db879676f59e52
>>>> userspace arch: riscv64
>>>> CC: [asml.s...@xxxxxxxxx ax...@xxxxxxxxx io-u...@xxxxxxxxxxxxxxx linux-...@xxxxxxxxxxxxxxx]
>>>>
>>>> Unfortunately, I don't have any reproducer for this issue yet.
>>
>> There was so many fixes in 5.12 after this revision, including sqpoll
>> cancellation related... Can you try something more up-to-date? Like
>> released 5.12 or for-next
>>
> 
> The reproducer works for 5.12.

Tried 5.12, 

[  245.467397] INFO: task iou-sqp-2018:2019 blocked for more than 122 seconds.
[  245.467424] Call Trace:
[  245.467432]  __schedule+0x365/0x960
[  245.467444]  schedule+0x68/0xe0
[  245.467450]  io_uring_cancel_sqpoll+0xdb/0x110
[  245.467461]  ? wait_woken+0x80/0x80
[  245.467472]  io_sq_thread+0x1c6/0x6c0
[  245.467482]  ? wait_woken+0x80/0x80
[  245.467491]  ? finish_task_switch.isra.0+0xca/0x2e0
[  245.467497]  ? io_wq_submit_work+0x140/0x140
[  245.467506]  ret_from_fork+0x22/0x30
[  245.467520] INFO: task iou-sqp-2052:2053 blocked for more than 122 seconds.
[  245.467536] Call Trace:
[  245.467539]  __schedule+0x365/0x960
[  245.467545]  schedule+0x68/0xe0
[  245.467550]  io_uring_cancel_sqpoll+0xdb/0x110
[  245.467559]  ? wait_woken+0x80/0x80
[  245.467568]  io_sq_thread+0x1c6/0x6c0
[  245.467577]  ? wait_woken+0x80/0x80
[  245.467586]  ? finish_task_switch.isra.0+0xca/0x2e0
[  245.467591]  ? io_wq_submit_work+0x140/0x140

> 
> I tested against the HEAD b1ef997bec4d5cf251bfb5e47f7b04afa49bcdfe

However, there is a bunch patches fixing sqpoll cancellations in
5.13, all are waiting for backporting. and for-next doesn't trigger
the issue for me.

Are you absolutely sure b1ef997bec4d5cf251bfb5e47f7b04afa49bcdfe
does hit it?

> commit on for-next tree
> https://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux-block.git/?h=for-next
> and the reproducer fails.

-- 
Pavel Begunkov