Re: [PATCH,RFC] perf: panic due to inclied cpu context task_ctx value

[Oleg Nesterov <oleg@redhat.com>]

On 03/28, Peter Zijlstra wrote:
>
> --- linux-2.6.orig/kernel/perf_event.c
> +++ linux-2.6/kernel/perf_event.c
> @@ -1767,7 +1767,6 @@ static void ctx_sched_out(struct perf_ev
>  	struct perf_event *event;
>
>  	raw_spin_lock(&ctx->lock);
> -	perf_pmu_disable(ctx->pmu);
>  	ctx->is_active = 0;
>  	if (likely(!ctx->nr_events))
>  		goto out;
> @@ -1777,6 +1776,7 @@ static void ctx_sched_out(struct perf_ev
>  	if (!ctx->nr_active)
>  		goto out;
>
> +	perf_pmu_disable(ctx->pmu);
>  	if (event_type & EVENT_PINNED) {
>  		list_for_each_entry(event, &ctx->pinned_groups, group_entry)
>  			group_sched_out(event, cpuctx, ctx);
> @@ -1786,8 +1786,8 @@ static void ctx_sched_out(struct perf_ev
>  		list_for_each_entry(event, &ctx->flexible_groups, group_entry)
>  			group_sched_out(event, cpuctx, ctx);
>  	}
> -out:
>  	perf_pmu_enable(ctx->pmu);
> +out:
>  	raw_spin_unlock(&ctx->lock);

Yes, thanks.

Probably this doesn't matter from the perfomance pov, but imho this
makes the code more understandable. This is important for occasional
readers like me ;)


Could you answer another question? It is not immediately clear why
ctx_sched_in() does not check nr_active != 0 before doing
ctx_XXX_sched_in(). I guess, the only reason is perf_rotate_context()
and the similar logic in perf_event_context_sched_in(). If we are
doing, say, cpu_ctx_sched_out(FLEXIBLE) + cpu_ctx_sched_in(FLEXIBLE)
then ->nr_active can be zero after cpu_ctx_sched_out().

Is my understanding correct? Or is there another reason?

Oleg.