Re: [patch V2 8/8] posix-timers: Handle exit in do_exit() completely

[Oleg Nesterov <oleg@redhat.com>]

I am still trying to understand this patch, one question for now.

On 09/05, Thomas Gleixner wrote:
>
> +void posix_cpu_timers_exit_task(void)
>  {
> -	cleanup_timers(&tsk->posix_cputimers);
> +	posix_cpu_timers_exit_work();
> +
> +	guard(spinlock_irq)(&current->sighand->siglock);
> +	cleanup_timers(&current->posix_cputimers);
>  }

So it calls posix_cpu_timers_exit_work()

> +static inline void posix_cpu_timers_exit_work(void)
> +{
> +	/*
> +	 * current->flags has PF_EXITING set so this can be done lockless and
> +	 * with interrupts enabled as PF_EXITING prevents the interrupt from
> +	 * scheduling the work.
> +	 */
> +	if (current->posix_cputimers_work.scheduled)
> +		task_work_cancel(current, &current->posix_cputimers_work.work);

... which does not clear ->scheduled

>  void posixtimer_exec(void)
>  {
> -	scoped_guard(spinlock_irq, &current->sighand->siglock)
> -		posix_cpu_timers_exit(current);
> -
> +	posix_cpu_timers_exit_task();

... and this looks obviously wrong for posixtimer_exec() ?

Oleg.