Commit b0df0e53 authored May 08, 2018 by Pavankumar Kondeti Committed by Satya Durga Srinivasu Prabhala May 24, 2018

sched/fair: Consider task affinity while skipping a sched group



When placement boost is active, the CPUs in the highest capacity
sched group are only considered for task placement. The
find_best_target()->skip_sg() skips lower capacity sched groups.
This is a problem for the tasks whose affinity allows them to run
only on lower capacity sched group.

Change-Id: I935ff3dc3daa98c636885312216a68323aac3ce1
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
[satyap@codeaurora.org: trivial changes to fix compilations issues]
Signed-off-by: Satya Durga Srinivasu Prabhala <satyap@codeaurora.org>

parent 95f3735c

kernel/sched/fair.c

+7 −0

Original line number	Diff line number	Diff line
		@@ -7053,6 +7053,13 @@ static inline bool skip_sg(struct task_struct p, struct sched_group sg,
		if (!sg->group_weight)
		return true;

		/*
		* Don't skip a group if a task affinity allows it
		* to run only on that group.
		*/
		if (cpumask_subset(&p->cpus_allowed, sched_group_span(sg)))
		return false;

		if (!task_fits_max(p, fcpu))
		return true;