Re: [patch V5 16/26] signal: Replace resched_timer logic

[Frederic Weisbecker]

Le Tue, Oct 01, 2024 at 10:42:21AM +0200, Thomas Gleixner a écrit :
> @@ -604,19 +604,19 @@ static int __dequeue_signal(struct sigpe
>   */
>  int dequeue_signal(sigset_t *mask, kernel_siginfo_t *info, enum pid_type *type)
>  {
> +	struct sigqueue *timer_sigq = NULL;
>  	struct task_struct *tsk = current;
> -	bool resched_timer = false;
>  	int signr;
>  
>  	lockdep_assert_held(&tsk->sighand->siglock);
>  
>  again:
>  	*type = PIDTYPE_PID;
> -	signr = __dequeue_signal(&tsk->pending, mask, info, &resched_timer);
> +	signr = __dequeue_signal(&tsk->pending, mask, info, &timer_sigq);
>  	if (!signr) {
>  		*type = PIDTYPE_TGID;
>  		signr = __dequeue_signal(&tsk->signal->shared_pending,
> -					 mask, info, &resched_timer);
> +					 mask, info, &timer_sigq);
>  
>  		if (unlikely(signr == SIGALRM))
>  			posixtimer_rearm_itimer(tsk);
> @@ -642,8 +642,8 @@ int dequeue_signal(sigset_t *mask, kerne
>  		current->jobctl |= JOBCTL_STOP_DEQUEUED;
>  	}
>  
> -	if (IS_ENABLED(CONFIG_POSIX_TIMERS) && unlikely(resched_timer)) {
> -		if (!posixtimer_deliver_signal(info))
> +	if (IS_ENABLED(CONFIG_POSIX_TIMERS) && unlikely(timer_sigq)) {
> +		if (!posixtimer_deliver_signal(info, timer_sigq))
>  			goto again;

If the signal has been refused, it goes goto again without clearing timer_sigq.
With the upcoming patch it becomes bad because if there is another signal
dequeued on the next iteration, posixtimer_deliver_signal() is called again with
two side effects:

_ The reference is put twice
_ The other dequeued signal is now ignored.

Thanks.