[PATCH v4 0/1] sched/fair: Track current se's EEVDF parameters

[PATCH v4 0/1] sched/fair: Track current se's EEVDF parameters

[Yiwei Lin]

I'm sorry that there was an obvious error in the previous submissions
that caused EEVDF to always be unable to find the eligible se......
This patch should address those issues.

Changelog:
v3 -> v4:
    - Fix incorrect changes which make EEVDF picks the first entity
instead of the eligible one
v2 -> v3:
    - Simplify how we avg_vruntime_add() and avg_vruntime_sub() when
neeed
    - Do not avg_vruntime_add() the contributions of 'curr' again when
put_prev_entity()
v1 -> v2:
    - Consider the contribution to avg_vruntime for 'curr'


 kernel/sched/fair.c | 35 +++++++++--------------------------
 1 file changed, 9 insertions(+), 26 deletions(-)

-- 
2.34.1

[PATCH v4 1/1] sched/fair: Track current se's EEVDF parameters

[Yiwei Lin]

After dequeuing the current-picked scheduling entity with
`__dequeue_entity`, its contribution to the EEVDF parameters
cfs_rq->avg_vruntime and cfs_rq->avg_load are also removed.
Because these should in fact be considered for the EEVDF algorithm,
we took curr as the special case and inserted back the contributions
when requests for cfs_rq->avg_vruntime and cfs_rq->avg_load.

Functions like `entity_eligible` which is called inside a loop may
therefore recalculate these statistics repeatly and require more effort.
Instead, we could just avoid to remove these statistics from
cfs_rq->avg_vruntime and cfs_rq->avg_load directly.

Signed-off-by: Yiwei Lin <s921975628@gmail.com>
---
 kernel/sched/fair.c | 35 +++++++++--------------------------
 1 file changed, 9 insertions(+), 26 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 876798824..b67c6d7d1 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -655,17 +655,9 @@ void avg_vruntime_update(struct cfs_rq *cfs_rq, s64 delta)
  */
 u64 avg_vruntime(struct cfs_rq *cfs_rq)
 {
-	struct sched_entity *curr = cfs_rq->curr;
 	s64 avg = cfs_rq->avg_vruntime;
 	long load = cfs_rq->avg_load;
 
-	if (curr && curr->on_rq) {
-		unsigned long weight = scale_load_down(curr->load.weight);
-
-		avg += entity_key(cfs_rq, curr) * weight;
-		load += weight;
-	}
-
 	if (load) {
 		/* sign flips effective floor / ceil */
 		if (avg < 0)
@@ -722,17 +714,9 @@ static void update_entity_lag(struct cfs_rq *cfs_rq, struct sched_entity *se)
  */
 int entity_eligible(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
-	struct sched_entity *curr = cfs_rq->curr;
 	s64 avg = cfs_rq->avg_vruntime;
 	long load = cfs_rq->avg_load;
 
-	if (curr && curr->on_rq) {
-		unsigned long weight = scale_load_down(curr->load.weight);
-
-		avg += entity_key(cfs_rq, curr) * weight;
-		load += weight;
-	}
-
 	return avg >= entity_key(cfs_rq, se) * load;
 }
 
@@ -815,7 +799,6 @@ RB_DECLARE_CALLBACKS(static, min_deadline_cb, struct sched_entity,
  */
 static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
-	avg_vruntime_add(cfs_rq, se);
 	se->min_deadline = se->deadline;
 	rb_add_augmented_cached(&se->run_node, &cfs_rq->tasks_timeline,
 				__entity_less, &min_deadline_cb);
@@ -825,7 +808,6 @@ static void __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
 {
 	rb_erase_augmented_cached(&se->run_node, &cfs_rq->tasks_timeline,
 				  &min_deadline_cb);
-	avg_vruntime_sub(cfs_rq, se);
 }
 
 struct sched_entity *__pick_first_entity(struct cfs_rq *cfs_rq)
@@ -1137,6 +1119,7 @@ static void update_curr(struct cfs_rq *cfs_rq)
 	struct sched_entity *curr = cfs_rq->curr;
 	u64 now = rq_clock_task(rq_of(cfs_rq));
 	u64 delta_exec;
+	u64 delta_fair;
 
 	if (unlikely(!curr))
 		return;
@@ -1158,7 +1141,9 @@ static void update_curr(struct cfs_rq *cfs_rq)
 	curr->sum_exec_runtime += delta_exec;
 	schedstat_add(cfs_rq->exec_clock, delta_exec);
 
-	curr->vruntime += calc_delta_fair(delta_exec, curr);
+	delta_fair = calc_delta_fair(delta_exec, curr);
+	curr->vruntime += delta_fair;
+	cfs_rq->avg_vruntime += delta_fair * scale_load_down(curr->load.weight);
 	update_deadline(cfs_rq, curr);
 	update_min_vruntime(cfs_rq);
 
@@ -3675,8 +3660,7 @@ static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
 		/* commit outstanding execution time */
 		if (cfs_rq->curr == se)
 			update_curr(cfs_rq);
-		else
-			avg_vruntime_sub(cfs_rq, se);
+		avg_vruntime_sub(cfs_rq, se);
 		update_load_sub(&cfs_rq->load, se->load.weight);
 	}
 	dequeue_load_avg(cfs_rq, se);
@@ -3712,8 +3696,7 @@ static void reweight_entity(struct cfs_rq *cfs_rq, struct sched_entity *se,
 	enqueue_load_avg(cfs_rq, se);
 	if (se->on_rq) {
 		update_load_add(&cfs_rq->load, se->load.weight);
-		if (cfs_rq->curr != se)
-			avg_vruntime_add(cfs_rq, se);
+		avg_vruntime_add(cfs_rq, se);
 	}
 }
 
@@ -5023,7 +5006,6 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 	 * EEVDF: placement strategy #1 / #2
 	 */
 	if (sched_feat(PLACE_LAG) && cfs_rq->nr_running) {
-		struct sched_entity *curr = cfs_rq->curr;
 		unsigned long load;
 
 		lag = se->vlag;
@@ -5081,8 +5063,6 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 		 *   vl_i = (W + w_i)*vl'_i / W
 		 */
 		load = cfs_rq->avg_load;
-		if (curr && curr->on_rq)
-			load += scale_load_down(curr->load.weight);
 
 		lag *= load + scale_load_down(se->load.weight);
 		if (WARN_ON_ONCE(!load))
@@ -5158,8 +5138,10 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 
 	check_schedstat_required();
 	update_stats_enqueue_fair(cfs_rq, se, flags);
+	avg_vruntime_add(cfs_rq, se);
 	if (!curr)
 		__enqueue_entity(cfs_rq, se);
+
 	se->on_rq = 1;
 
 	if (cfs_rq->nr_running == 1) {
@@ -5230,6 +5212,7 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 	update_entity_lag(cfs_rq, se);
 	if (se != cfs_rq->curr)
 		__dequeue_entity(cfs_rq, se);
+	avg_vruntime_sub(cfs_rq, se);
 	se->on_rq = 0;
 	account_entity_dequeue(cfs_rq, se);
 
-- 
2.34.1