Re: [PATCH v2] workqueue: Remove the warning in wq_worker_sleeping()

[Daniel Jordan]

Hi Sebastian,

On Sat, Mar 28, 2020 at 12:29:59AM +0100, Sebastian Andrzej Siewior wrote:
> v1âv2: - Drop the warning instead of using cmpxchg_local().
> 	 Tglx pointed out that wq_worker_sleeping() is already invoked
> 	 with disabled preemption so the race described can not happen.

I guess you mean this race:

> However, if the interrupt occurs in wq_worker_sleeping() between reading and
> setting `sleeping' i.e.
> 
> |        if (WARN_ON_ONCE(worker->sleeping))
> |                return;
>  *interrupt*
> |        worker->sleeping = 1;
> 
> then pool->nr_running will be decremented twice in wq_worker_sleeping()
> but it will be incremented only once in wq_worker_running().

Why would preemption prevent it?  Interrupts are still enabled.

What am I missing?  :-)