Message-ID: <15161e5f-fe26-23e9-1d0a-ebd579d64d9f@gmail.com>
Date:   Wed, 1 Feb 2023 15:52:47 +0000
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.7.0
From:   Pavel Begunkov <asml.silence@gmail.com>
Subject: Re: [Syzkaller & bisect] There is "io_ring_exit_work" related Call
 Trace in v5.2-rc5 kernel
To:     Pengfei Xu <pengfei.xu@intel.com>, linux-kernel@vger.kernel.org
Cc:     heng.su@intel.com, axboe@kernel.dk
References: <Y9TgUupO5C39V/DW@xpf.sh.intel.com>
 <812a7e2d-14d6-78cd-5c6c-ce508fc7fc0d@gmail.com>
Content-Language: en-US
In-Reply-To: <812a7e2d-14d6-78cd-5c6c-ce508fc7fc0d@gmail.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 1/28/23 14:49, Pavel Begunkov wrote:
> On 1/28/23 08:44, Pengfei Xu wrote:
>> Hi Pavel Begunkov and kernel expert,
>>
>> Greeting!
>>
>> There is "io_ring_exit_work" related Call Trace in v5.2-rc5 kernel in guest
>> on Sapphire Rapids server.
> 
> Thanks for the report, we'll take a look

Not reproducible for me. Apparently, the repro creates a normal ring
and immediately closes it, then io_ring_exit_work() hangs waiting to
the task to execute task_work.

It reminds me a bug we had before when the task was being stopped
for debugging blocking task_work. Maybe some signal interaction, or
vm and/or syz magic specific.

>> INFO: task kworker/u4:2:32 blocked for more than 147 seconds.
>> [  300.212512]       Not tainted 6.2.0-rc5-2241ab53cbb5 #1
>> [  300.213087] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
>> [  300.213921] task:kworker/u4:2    state:D stack:0     pid:32    ppid:2      flags:0x00004000
>> [  300.214849] Workqueue: events_unbound io_ring_exit_work
>> [  300.215506] Call Trace:
>> [  300.215789]  <TASK>
>> [  300.216051]  __schedule+0x385/0xb00
>> [  300.216467]  ? write_comp_data+0x2f/0x90
>> [  300.216949]  ? wait_for_completion+0x7b/0x180
>> [  300.217445]  schedule+0x5b/0xe0
>> [  300.217836]  schedule_timeout+0x561/0x650
>> [  300.218301]  ? _raw_spin_unlock_irq+0x2b/0x60
>> [  300.218816]  ? wait_for_completion+0x7b/0x180
>> [  300.219341]  ? __this_cpu_preempt_check+0x20/0x30
>> [  300.219893]  ? lockdep_hardirqs_on+0x8a/0x110
>> [  300.220384]  ? _raw_spin_unlock_irq+0x2b/0x60
>> [  300.220881]  ? wait_for_completion+0x9e/0x180
>> [  300.221382]  ? trace_hardirqs_on+0x3f/0x100
>> [  300.221868]  ? wait_for_completion+0x7b/0x180
>> [  300.222379]  wait_for_completion+0xa6/0x180
>> [  300.222897]  io_ring_exit_work+0x2f7/0x747
>> [  300.223383]  ? __pfx_io_tctx_exit_cb+0x10/0x10
>> [  300.223947]  process_one_work+0x3b1/0x960
>> [  300.224446]  worker_thread+0x52/0x660
>> [  300.224884]  ? __pfx_worker_thread+0x10/0x10
>> [  300.225376]  kthread+0x161/0x1a0
>> [  300.225782]  ? __pfx_kthread+0x10/0x10
>> [  300.226225]  ret_from_fork+0x29/0x50
>> [  300.226688]  </TASK>
>> [  300.227791]
>> [  300.227791] Showing all locks held in the system:
>> [  300.228501] 1 lock held by rcu_tasks_kthre/11:
>> [  300.229003]  #0: ffffffff83d617b0 (rcu_tasks.tasks_gp_mutex){+.+.}-{3:3}, at: rcu_tasks_one_gp+0x2d/0x3d0
>> [  300.230098] 1 lock held by rcu_tasks_rude_/12:
>> [  300.230601]  #0: ffffffff83d61530 (rcu_tasks_rude.tasks_gp_mutex){+.+.}-{3:3}, at: rcu_tasks_one_gp+0x2d/0x3d0
>> [  300.231733] 1 lock held by rcu_tasks_trace/13:
>> [  300.232244]  #0: ffffffff83d61270 (rcu_tasks_trace.tasks_gp_mutex){+.+.}-{3:3}, at: rcu_tasks_one_gp+0x2d/0x3d0
>> [  300.233377] 1 lock held by khungtaskd/29:
>> [  300.233842]  #0: ffffffff83d621c0 (rcu_read_lock){....}-{1:2}, at: debug_show_all_locks+0x27/0x18c
>> [  300.234866] 2 locks held by kworker/u4:2/32:
>> [  300.235351]  #0: ff11000006045d38 ((wq_completion)events_unbound){+.+.}-{0:0}, at: process_one_work+0x2ff/0x960
>> [  300.236502]  #1: ffa0000000117e60 ((work_completion)(&ctx->exit_work)){+.+.}-{0:0}, at: process_one_work+0x303/0x960
>> [  300.237696]
>> [  300.237882] =============================================
>> [  300.237882]
>>
>> Bisected and found the first bad commit is:
>> eebd2e37e662617a6b8041db75205f0a262ce870
>> io_uring: don't take task ring-file notes
>>
>> But after reverted above commit eebd2e37e on top of v6.2-rc5 kernel and made
>> the kernel failed. So it's just the suspected commit for above problem.
>>
>> Syzkaller reproduced code, bisect info, kconfig and v6.2-rc5 dmesg are in
>> attached.
>>
>> All detailed info is in link:
>> https://github.com/xupengfe/syzkaller_logs/tree/main/230124_055801_io_ring_exit_work
>>
>> I hope it's helpful.
>>
>> Thanks!
>> BR.
> 

-- 
Pavel Begunkov