Re: [PATCH v4 1/6] sched/fair: Fix reschedule which is generated on throttled cfs_rq

[Kirill Tkhai <ktkhai@parallels.com>]

В Пт, 24/10/2014 в 17:32 +0800, Wanpeng Li пишет:
> Hi Kirill,
> 10/24/14, 2:01 PM, Kirill Tkhai:
> > Hi, Wanpeng,
> >
> > the commit commentary confuses, I'm agree. Really it's just a cleanup.
> >
> > On Пт, 2014-10-24 at 07:27 +0800, Wanpeng Li wrote:
> >> Hi Kirill,
> >> 8/6/14, 4:06 PM, Kirill Tkhai:
> >>> (sched_entity::on_rq == 1) does not guarantee the task is pickable;
> >>> changes on throttled cfs_rq must not lead to reschedule.
> >> Why (sched_entity::on_rq == 1) doesn't guarantee the task is pickable
> >> since entity will be dequeued during throttling cfs_rq?
> > Because one of task's (grand)parents in hierarhy may be throtthed and
> > dequeued.
> >
> > But task_struct::on_rq check doesn't guarantee this too. So, just ignore
> > commit commentary; the commentary is wrong.
> >
> >>> Check for task_struct::on_rq instead.
> >> Do you mean task_struct::on_rq will be cleared during throttling cfs_rq?
> >> I can't find codes do this.
> > No, it not cleared. The commit commentary should be:
> > "sched: Cleanup. Check task_struct::on_rq instead of sched_entity::on_rq,
> > because it is the same for a task"
> 
> IIUR, for fair class, sched_entity::on_rq will be set/clear during 
> enqueue/dequeue, task_struct::on_rq will changed during task migration, 
> I'm not sure why they are the same.

prio_changed_fair() and switched_to_fair() can't be called during migration.
They are called under pi_lock, and migration needs this lock too.

> >>> Signed-off-by: Kirill Tkhai <ktkhai@parallels.com>
> >>> ---
> >>>    kernel/sched/fair.c |    6 +++---
> >>>    1 file changed, 3 insertions(+), 3 deletions(-)
> >>>
> >>> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> >>> index bfa3c86..6f0ce2b 100644
> >>> --- a/kernel/sched/fair.c
> >>> +++ b/kernel/sched/fair.c
> >>> @@ -7465,7 +7465,7 @@ static void task_fork_fair(struct task_struct *p)
> >>>    static void
> >>>    prio_changed_fair(struct rq *rq, struct task_struct *p, int oldprio)
> >>>    {
> >>> -	if (!p->se.on_rq)
> >>> +	if (!p->on_rq)
> >>>    		return;
> >>>    
> >>>    	/*
> >>> @@ -7521,15 +7521,15 @@ static void switched_from_fair(struct rq *rq, struct task_struct *p)
> >>>     */
> >>>    static void switched_to_fair(struct rq *rq, struct task_struct *p)
> >>>    {
> >>> -	struct sched_entity *se = &p->se;
> >>>    #ifdef CONFIG_FAIR_GROUP_SCHED
> >>> +	struct sched_entity *se = &p->se;
> >>>    	/*
> >>>    	 * Since the real-depth could have been changed (only FAIR
> >>>    	 * class maintain depth value), reset depth properly.
> >>>    	 */
> >>>    	se->depth = se->parent ? se->parent->depth + 1 : 0;
> >>>    #endif
> >>> -	if (!se->on_rq)
> >>> +	if (!p->on_rq)
> >>>    		return;
> >>>    
> >>>    	/*
> >>>
> >>>
> >>>
> >>> --
> >>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> >>> the body of a message to majordomo@vger.kernel.org
> >>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> >>> Please read the FAQ at  http://www.tux.org/lkml/
> >
>