Re: [PATCH 7/9] ptrace: Simplify the wait_task_inactive call in ptrace_check_attach

[Oleg Nesterov]

On 04/28, Peter Zijlstra wrote:
>
> On Thu, Apr 28, 2022 at 04:57:50PM +0200, Oleg Nesterov wrote:
>
> > > Shouldn't we then switch wait_task_inactive() so have & matching instead
> > > of the current ==.
> >
> > Sorry, I don't understand the context...
>
> This.. I've always found it strange to have wti use a different matching
> scheme from ttwu.

Ah. This is what I understood (and I too thought about this), just I meant that
this patch from Eric (assuming wait_task_inactive() still uses __TASK_TRACED) is
fine without your change below.

Oleg.

> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index f259621f4c93..c039aef4c8fe 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -3304,7 +3304,7 @@ unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state
>  		 * is actually now running somewhere else!
>  		 */
>  		while (task_running(rq, p)) {
> -			if (match_state && unlikely(READ_ONCE(p->__state) != match_state))
> +			if (match_state && unlikely(!(READ_ONCE(p->__state) & match_state)))
>  				return 0;
>  			cpu_relax();
>  		}
> @@ -3319,7 +3319,7 @@ unsigned long wait_task_inactive(struct task_struct *p, unsigned int match_state
>  		running = task_running(rq, p);
>  		queued = task_on_rq_queued(p);
>  		ncsw = 0;
> -		if (!match_state || READ_ONCE(p->__state) == match_state)
> +		if (!match_state || (READ_ONCE(p->__state) & match_state))
>  			ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
>  		task_rq_unlock(rq, p, &rf);