Re: [PATCH 4/4] sched,fair: Fix PELT integrity for new tasks

[Vincent Guittot]

On 17 June 2016 at 18:02, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> On Fri, Jun 17, 2016 at 04:28:14PM +0200, Peter Zijlstra wrote:
>> On Fri, Jun 17, 2016 at 04:09:01PM +0200, Vincent Guittot wrote:
>> > >
>> > > Scenario 1: switch to fair class
>> > >
>> > >   p->sched_class = fair_class;
>> > >   if (queued)
>> > >     enqueue_task(p);
>> > >       ...
>> > >         enqueue_entity()
>> > >           enqueue_entity_load_avg()
>> > >             migrated = !sa->last_update_time (true)
>> > >             if (migrated)
>> > >               attach_entity_load_avg()
>> > >   check_class_changed()
>> > >     switched_from() (!fair)
>> > >     switched_to()   (fair)
>> > >       switched_to_fair()
>> > >         attach_entity_load_avg()
>>
>> > > @@ -733,18 +737,21 @@ void post_init_entity_util_avg(struct sc
>> > >                 }
>> > >                 sa->util_sum = sa->util_avg * LOAD_AVG_MAX;
>> > >         }
>> > > +
>> > > +       update_cfs_rq_load_avg(cfs_rq_clock_task(cfs_rq), cfs_rq, false);
>> > > +       attach_entity_load_avg(cfs_rq, se);
>> >
>> > A new RT task will be attached and will contribute to the load until
>> > it decays to 0
>
>> > Should we detach it for non cfs task ?
>
> Right, an attach + detach, which basically ends up being:
>
>> > We just want to update
>> > last_update_time of RT task to something different from 0
>
> this. That's the same as starting as fair and then doing
> switched_from_fair().

Yes that's also a possibility to add a way to call switched_from_fair
for non fair task

>
> So yes, ho-humm, how to go about doing that bestest. Lemme have a play.
>