Messages in this thread | | | Date | Wed, 24 Jun 2015 07:41:14 +0800 | From | Yuyang Du <> | Subject | Re: [PATCH] sched: update blocked load of idle cpus |
| |
Hi Vincent,
On Wed, Jun 24, 2015 at 09:11:12AM +0200, Vincent Guittot wrote: > The load and the util of idle cpus must be updated periodically in order to > decay the blocked part. > > If CONFIG_FAIR_GROUP_SCHED is not set, the load and util of idle cpus > are not decayed and stay at the values set before becoming idle. > > Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org> > --- > Hi Yuyang, > > While testing your patchset without CONFIG_FAIR_GROUP_SCHED, i have noticed > that the load of idle cpus stays sometimes to an high value whereas they were > not used for a while because we are not decaying the blocked load. > Futhermore, the peridodic load balance was not pulling tasks onto some idle > cpus because their load stayed high. > > This patchset fixes the issue.
Thanks. We should privide an update_blocked_averages() for the CONFIG_FAIR_GROUP_SCHED=n case. Great.
> > Regards, > Vincent > > kernel/sched/fair.c | 11 +++++++++++ > 1 file changed, 11 insertions(+) > > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c > index c5f18d9..665cc4b 100644 > --- a/kernel/sched/fair.c > +++ b/kernel/sched/fair.c > @@ -5864,6 +5864,17 @@ static unsigned long task_h_load(struct task_struct *p) > #else > static inline void update_blocked_averages(int cpu) > { > + struct rq *rq = cpu_rq(cpu); > + struct cfs_rq *cfs_rq = &rq->cfs; > + unsigned long flags; > + > + raw_spin_lock_irqsave(&rq->lock, flags); > + update_rq_clock(rq); > + > + update_cfs_rq_load_avg(cfs_rq_clock_task(cfs_rq), cfs_rq)) > + > + raw_spin_unlock_irqrestore(&rq->lock, flags); > + > } > > static unsigned long task_h_load(struct task_struct *p)
Reviewed-by: Yuyang Du <yuyang.du@intel.com>
> -- > 1.9.1
| |