Re: [patch] fix uml slowness caused by ptrace preemption bug on host

[Oleg Nesterov <oleg@redhat.com>]

On 03/20, Miklos Szeredi wrote:
>
> I'm no scheduler expert either.

neither me ;)

> --- linux.git.orig/kernel/sched.c	2009-03-18 12:53:47.000000000 +0100
> +++ linux.git/kernel/sched.c	2009-03-20 08:58:13.000000000 +0100
> @@ -4629,7 +4629,8 @@ asmlinkage void __sched preempt_schedule
>  	 * If there is a non-zero preempt_count or interrupts are disabled,
>  	 * we do not want to preempt the current task. Just return..
>  	 */
> -	if (likely(ti->preempt_count || irqs_disabled()))
> +	if (likely(ti->preempt_count || irqs_disabled() ||
> +		   current->state != TASK_RUNNING))

But this was specially designed to allow to preempt !TASK_RUNNING tasks,
note the "if (prev->state && !(preempt_count() & PREEMPT_ACTIVE))" in
schedule().

Perhaps "|| current->state == TASK_TRACED" makes more sense, TASK_TRACED
is special because we know we are going to schedule really soon. But I think
your previous patch is better, imho we should change preempt_schedule() to
fix the very specific problem with ptrace_notify().

Oleg.