On Thu, 20 Aug 2020 at 14:58, <pet...@infradead.org> wrote: > > On Thu, Aug 20, 2020 at 02:51:06PM +0200, Vincent Guittot wrote: > > On Thu, 20 Aug 2020 at 14:00, Jiang Biao <benbji...@gmail.com> wrote: > > > > > > From: Jiang Biao <benbji...@tencent.com> > > > > > > Vruntime compensation has been down in place_entity() to > > > boot the waking procedure for fair tasks. There is no need to > > > > s/boot/boost/ ? > > > > > do that for SCHED_IDLE task actually. > > > > > > Not compensating vruntime for SCHED_IDLE task could make > > > SCHED_IDLE task more harmless for normal tasks. > > This is rather week. It would be much better if there's some actual data > to support this claim. > > > > Signed-off-by: Jiang Biao <benbji...@tencent.com> > > > --- > > > kernel/sched/fair.c | 2 +- > > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > > > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c > > > index 1a68a0536add..adff77676a0a 100644 > > > --- a/kernel/sched/fair.c > > > +++ b/kernel/sched/fair.c > > > @@ -4115,7 +4115,7 @@ place_entity(struct cfs_rq *cfs_rq, struct > > > sched_entity *se, int initial) > > > vruntime += sched_vslice(cfs_rq, se); > > > > > > /* sleeps up to a single latency don't count. */ > > > - if (!initial) { > > > + if (!initial && likely(!task_has_idle_policy(task_of(se)))) { > > > > What if se is not a task ? > > Then we very much need it, because it might have fair tasks inside. I > suppose you could do something complicated with idle_h_nr_running, but > is all that really worth the effort?
Not sure that Jiang is using cgroups otherwise he would have seen a warning I think. That's been said, not compensating the vruntime for a sched_idle task makes sense for me. Even if that will only help for others task in the same cfs_rq > > > > unsigned long thresh = sysctl_sched_latency; > > > > > > /* > > > -- > > > 2.21.0 > > >