Re: [PATCH v4 5/5] sched: add two functions for att(det)aching a task to(from) a cfs_rq

[Peter Zijlstra]

On Thu, Aug 20, 2015 at 08:35:16PM +0900, Byungchul Park wrote:
> On Thu, Aug 20, 2015 at 08:22:00PM +0900, byungchul.park@xxxxxxx wrote:

> > +	/*
> > +	 * If it's !queued, then only when the task is sleeping it has a
> > +	 * non-normalized vruntime, that is, when the task is being migrated
> > +	 * it has a normailized vruntime.
> > +	 */
> 
> i tried to change your XXX comment. i think it can be explaned like this.
> don't you think so? i want to hear any opinions about this.
> 
> > +	if (p->state == TASK_RUNNING)
> > +		return true;


--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -7943,11 +7943,10 @@ static inline bool vruntime_normalized(s
 		return true;
 
 	/*
-	 * If it's !queued, then only when the task is sleeping it has a
-	 * non-normalized vruntime, that is, when the task is being migrated
-	 * it has a normalized vruntime.
+	 * If it's !queued, sleeping tasks have a normalized vruntime,
+	 * see dequeue_entity().
 	 */
-	if (p->state == TASK_RUNNING)
+	if (!p->se.on_rq)
 		return true;
 
 	return false;


Does that make sense?

I think using p->state for this is fragile, as we could be racy with any
random blocking primitive that does set_current_state() _before_
actually calling into the scheduler.


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/