2022-04-07 08:08:23

by Zheyu Ma

Subject: [BUG] hwmon: (adt7470) warning when removing the module

I found a bug when removing the adt7470 module.

The following log can reveal it:

[ 18.459479] ------------[ cut here ]------------
[ 18.459852] do not call blocking ops when !TASK_RUNNING; state=1
set at [<ffffffffa006052b>] adt7470_update_thread+0x7b/0x130 [adt7470]
[ 18.460857] WARNING: CPU: 7 PID: 365 at kernel/sched/core.c:9660
__might_sleep+0x96/0xb0
[ 18.461523] Modules linked in: adt7470(-)
[ 18.464568] RIP: 0010:__might_sleep+0x96/0xb0
[ 18.472577] Call Trace:
[ 18.472778] <TASK>
[ 18.472957] exit_signals+0x2f/0x3b0
[ 18.473251] do_exit+0xcd/0x18c0
[ 18.473534] ? __mutex_unlock_slowpath+0x1b8/0x740
[ 18.473927] kthread_exit+0x42/0x60
[ 18.474212] ? adt7470_detect+0x140/0x140 [adt7470]
[ 18.474622] kthread+0x272/0x2e0
[ 18.474885] ? adt7470_detect+0x140/0x140 [adt7470]
[ 18.475277] ? kthread_blkcg+0xa0/0xa0
[ 18.475601] ret_from_fork+0x22/0x30
[ 18.475901] </TASK>

Regards,
Zheyu Ma

2022-04-07 09:19:25

by Armin Wolf

Subject: Re: [BUG] hwmon: (adt7470) warning when removing the module

Am 07.04.22 um 05:17 schrieb Zheyu Ma:

> I found a bug when removing the adt7470 module.
>
> The following log can reveal it:
>
> [ 18.459479] ------------[ cut here ]------------
> [ 18.459852] do not call blocking ops when !TASK_RUNNING; state=1
> set at [<ffffffffa006052b>] adt7470_update_thread+0x7b/0x130 [adt7470]
> [ 18.460857] WARNING: CPU: 7 PID: 365 at kernel/sched/core.c:9660
> __might_sleep+0x96/0xb0
> [ 18.461523] Modules linked in: adt7470(-)
> [ 18.464568] RIP: 0010:__might_sleep+0x96/0xb0
> [ 18.472577] Call Trace:
> [ 18.472778] <TASK>
> [ 18.472957] exit_signals+0x2f/0x3b0
> [ 18.473251] do_exit+0xcd/0x18c0
> [ 18.473534] ? __mutex_unlock_slowpath+0x1b8/0x740
> [ 18.473927] kthread_exit+0x42/0x60
> [ 18.474212] ? adt7470_detect+0x140/0x140 [adt7470]
> [ 18.474622] kthread+0x272/0x2e0
> [ 18.474885] ? adt7470_detect+0x140/0x140 [adt7470]
> [ 18.475277] ? kthread_blkcg+0xa0/0xa0
> [ 18.475601] ret_from_fork+0x22/0x30
> [ 18.475901] </TASK>
>
> Regards,
> Zheyu Ma

Please correct me if i am wrong, but i think the issue lies in adt7470_update_thread()
leaving the kthread in TASK_INTERRUPTIBLE state when the second check for kthread_should_stop()
returns true and causes the kthread to exit. Since kthread_exit() might sleep in exit_signals(),
we get this warning.
Maybe moving set_current_state(TASK_INTERRUPTIBLE) below the second check for kthread_should_stop()
would fix the issue?

Armin Wolf

2022-04-07 20:40:05

by Guenter Roeck

Subject: Re: [BUG] hwmon: (adt7470) warning when removing the module

On 4/6/22 22:01, Armin Wolf wrote:
> Am 07.04.22 um 05:17 schrieb Zheyu Ma:
>
>> I found a bug when removing the adt7470 module.
>>
>> The following log can reveal it:
>>
>> [   18.459479] ------------[ cut here ]------------
>> [   18.459852] do not call blocking ops when !TASK_RUNNING; state=1
>> set at [<ffffffffa006052b>] adt7470_update_thread+0x7b/0x130 [adt7470]
>> [   18.460857] WARNING: CPU: 7 PID: 365 at kernel/sched/core.c:9660
>> __might_sleep+0x96/0xb0
>> [   18.461523] Modules linked in: adt7470(-)
>> [   18.464568] RIP: 0010:__might_sleep+0x96/0xb0
>> [   18.472577] Call Trace:
>> [   18.472778]  <TASK>
>> [   18.472957]  exit_signals+0x2f/0x3b0
>> [   18.473251]  do_exit+0xcd/0x18c0
>> [   18.473534]  ? __mutex_unlock_slowpath+0x1b8/0x740
>> [   18.473927]  kthread_exit+0x42/0x60
>> [   18.474212]  ? adt7470_detect+0x140/0x140 [adt7470]
>> [   18.474622]  kthread+0x272/0x2e0
>> [   18.474885]  ? adt7470_detect+0x140/0x140 [adt7470]
>> [   18.475277]  ? kthread_blkcg+0xa0/0xa0
>> [   18.475601]  ret_from_fork+0x22/0x30
>> [   18.475901]  </TASK>
>>
>> Regards,
>> Zheyu Ma
>
> Please correct me if i am wrong, but i think the issue lies in adt7470_update_thread()
> leaving the kthread in TASK_INTERRUPTIBLE state when the second check for kthread_should_stop()
> returns true and causes the kthread to exit. Since kthread_exit() might sleep in exit_signals(),
> we get this warning.
> Maybe moving set_current_state(TASK_INTERRUPTIBLE) below the second check for kthread_should_stop()
> would fix the issue?
>

Agreed. Care to send a patch ?

Thanks,
Guenter