[PATCH v4 4/5] sched: sync a se with its cfs_rq when switching sched class to fair class

[byungchul . park]

From: Byungchul Park <byungchul.park@xxxxxxx>

we have to sync a se with its cfs_rq, when switching sched class to
fair class. current code does not sync it because the se average load
won't be valid any more if it has been dettached for a long time. however
the se's average load would be valid just after being detached from
cfs_rq, so we need to sync it in that case, e.g. priority inheritance.

to solve the problem that a se average load becomes more useless over
time, this patch decays its average load even for the duration that
the se has been detached, when it gets attached to the cfs_rq.

Signed-off-by: Byungchul Park <byungchul.park@xxxxxxx>
---
 kernel/sched/fair.c |   19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index d58e9aa..af6ad5f 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2712,6 +2712,18 @@ static inline void update_load_avg(struct sched_entity *se, int update_tg)
 
 static void attach_entity_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
+	/*
+	 * In case of migration and cgroup-change, more care should be taken
+	 * because se's cfs_rq was changed, that means calling __update_load_avg
+	 * with new cfs_rq->avg.last_update_time is meaningless. So we skip the
+	 * update here. We have to update it with prev cfs_rq just before changing
+	 * se's cfs_rq, and get here soon.
+	 */
+	if (se->avg.last_update_time) {
+		__update_load_avg(cfs_rq->avg.last_update_time, cpu_of(rq_of(cfs_rq)),
+				  &se->avg, 0, 0, NULL);
+	}
+
 	se->avg.last_update_time = cfs_rq->avg.last_update_time;
 	cfs_rq->avg.load_avg += se->avg.load_avg;
 	cfs_rq->avg.load_sum += se->avg.load_sum;
@@ -7945,6 +7957,9 @@ static void switched_to_fair(struct rq *rq, struct task_struct *p)
 	se->depth = se->parent ? se->parent->depth + 1 : 0;
 #endif
 
+	/* Synchronize task with its cfs_rq */
+	attach_entity_load_avg(cfs_rq_of(&p->se), &p->se);
+
 	if (!task_on_rq_queued(p)) {
 
 		/*
@@ -8044,6 +8059,10 @@ static void task_move_group_fair(struct task_struct *p, int queued)
 	/* Synchronize task with its prev cfs_rq */
 	detach_entity_load_avg(cfs_rq, se);
 	set_task_rq(p, task_cpu(p));
+
+	/* Tell se's cfs_rq has been changed -- migrated */
+	p->se.avg.last_update_time = 0;
+
 	se->depth = se->parent ? se->parent->depth + 1 : 0;
 	cfs_rq = cfs_rq_of(se);
 	if (!queued)
-- 
1.7.9.5

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/