Re: [PATCH] kthread: work could not be queued when worker being destroyed

[Petr Mladek]

On Thu 2020-07-02 12:49:46, qiang.zhang@xxxxxxxxxxxxx wrote:
> From: Zhang Qiang <qiang.zhang@xxxxxxxxxxxxx>
> 
> The queuing_blocked func should returns true when the worker
> being destroyed.

Queuing work when the worker is not running usually means
a bad design on the API user side. The work is never done
and some job has not finished correctly. We should not go
over it quietly.

> Signed-off-by: Zhang Qiang <qiang.zhang@xxxxxxxxxxxxx>
> ---
>  kernel/kthread.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/kernel/kthread.c b/kernel/kthread.c
> index 1166f2043e67..2ca711d0e78a 100644
> --- a/kernel/kthread.c
> +++ b/kernel/kthread.c
> @@ -791,7 +791,7 @@ static inline bool queuing_blocked(struct kthread_worker *worker,
>  {
>  	lockdep_assert_held(&worker->lock);

A better solution would be:

	if (WARN_ON(!worker->task)
		return true;


> -	return !list_empty(&work->node) || work->canceling;
> +	return !list_empty(&work->node) || work->canceling || !worker->task;
>  }
>  
>  static void kthread_insert_work_sanity_check(struct kthread_worker *worker,

Best Regards,
Petr