[PATCH v2] sched: sync with the prev cfs when changing cgroup within a cpu

[Byungchul Park]

current code seems to be wrong with cfs_rq->blocked_load_avg when changing
a task's cgroup(=cfs_rq) to another. i tested with "echo pid > cgroup" and
found that cfs_rq->blocked_load_avg became larger and larger whenever i
changed a cgroup to another again and again.

it is possible to move between groups within *a* cpu, and each cfs_rq is
tracking its own blocked load. so we have to sync se's average load with
both *prev* cfs_rq and next cfs_rq when changing its group. i also removed
some comments mentioning migration_task_rq_fair().

Signed-off-by: Byungchul Park <byungchul.park@xxxxxxx>
---
 kernel/sched/fair.c |   18 +++++++++++++-----
 1 file changed, 13 insertions(+), 5 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index ffa70dc..759a394 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -8229,8 +8229,18 @@ static void task_move_group_fair(struct task_struct *p, int queued)
 	if (!queued && (!se->sum_exec_runtime || p->state == TASK_WAKING))
 		queued = 1;
 
-	if (!queued)
-		se->vruntime -= cfs_rq_of(se)->min_vruntime;
+	if (!queued) {
+		cfs_rq = cfs_rq_of(se);
+		se->vruntime -= cfs_rq->min_vruntime;
+
+#ifdef CONFIG_SMP
+		/*
+		 * we must synchronize with the prev cfs.
+		 */
+		__synchronize_entity_decay(se);
+		subtract_blocked_load_contrib(cfs_rq, se->avg.load_avg_contrib);
+#endif
+	}
 	set_task_rq(p, task_cpu(p));
 	se->depth = se->parent ? se->parent->depth + 1 : 0;
 	if (!queued) {
@@ -8238,9 +8248,7 @@ static void task_move_group_fair(struct task_struct *p, int queued)
 		se->vruntime += cfs_rq->min_vruntime;
 #ifdef CONFIG_SMP
 		/*
-		 * migrate_task_rq_fair() will have removed our previous
-		 * contribution, but we must synchronize for ongoing future
-		 * decay.
+		 * we must synchronize with the next cfs for ongoing future decay.
 		 */
 		se->avg.decay_count = atomic64_read(&cfs_rq->decay_counter);
 		cfs_rq->blocked_load_avg += se->avg.load_avg_contrib;
-- 
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/