Skip to contentSkip to conversation

sched/cfs: Fix pick_next_entity() implementation error

PATCH In review Zijun Hu 2 participants 3 messages Jul 1, 2020 Scheduler

lore
Share & tools
lore

Zijun Hu

author
#0 lore raw

PATCH 1/1 sched/cfs: Fix pick_next_entity() implementation error

sched_entity @se not static @left should be compared to pick up @cfs_rq->next.

---
 kernel/sched/fair.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
kernel/sched/fair.c +2 -2
index 658aa7a2ae6f..4790f2d851ad 100644
@@ -4452,13 +4452,13 @@ pick_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *curr)
 	/*
 	 * Prefer last buddy, try to return the CPU to a preempted task.
 	 */
-	if (cfs_rq->last && wakeup_preempt_entity(cfs_rq->last, left) < 1)
+	if (cfs_rq->last && wakeup_preempt_entity(cfs_rq->last, se) < 1)
 		se = cfs_rq->last;
 
 	/*
 	 * Someone really wants this to run. If it's not unfair, run it.
 	 */
-	if (cfs_rq->next && wakeup_preempt_entity(cfs_rq->next, left) < 1)
+	if (cfs_rq->next && wakeup_preempt_entity(cfs_rq->next, se) < 1)
 		se = cfs_rq->next;
 
 	clear_buddies(cfs_rq, se);
Signature · 1 line
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
maintainer
#1 lore raw
Quoted 3 lines · Zijun Hu

sched_entity @se not static @left should be compared to pick up @cfs_rq->next.

Could you elaborate why ?

left is the leftmost sched_entity and the one that should be used.

se != left means that left should be skipped after a yield and the next se in the rbtree is not "far" from left although it has higher vruntime

if we finally want to use last or next instead of se, we must ensure that they are still not "far" from left otherwise you can promote a sched entity that ends up having a really high vruntime

Quoted 29 lines
---
 kernel/sched/fair.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
kernel/sched/fair.c +2 -3
index 658aa7a2ae6f..4790f2d851ad 100644
@@ -4452,13 +4452,13 @@ pick_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *curr)
        /*
         * Prefer last buddy, try to return the CPU to a preempted task.
         */
-       if (cfs_rq->last && wakeup_preempt_entity(cfs_rq->last, left) < 1)
+       if (cfs_rq->last && wakeup_preempt_entity(cfs_rq->last, se) < 1)
                se = cfs_rq->last;
 
        /*
         * Someone really wants this to run. If it's not unfair, run it.
         */
-       if (cfs_rq->next && wakeup_preempt_entity(cfs_rq->next, left) < 1)
+       if (cfs_rq->next && wakeup_preempt_entity(cfs_rq->next, se) < 1)
                se = cfs_rq->next;
 
        clear_buddies(cfs_rq, se);
--

The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project

author
#2 lore raw

thanks for your explanation.
you are right. @lest should be used as reference point to compare. Please ignore this patch.

Quoted 46 lines · Vincent Guittot
Quoted 3 lines · Zijun Hu

sched_entity @se not static @left should be compared to pick up @cfs_rq->next.

Could you elaborate why ?

left is the leftmost sched_entity and the one that should be used.

se != left means that left should be skipped after a yield and the next se in the rbtree is not "far" from left although it has higher vruntime

if we finally want to use last or next instead of se, we must ensure that they are still not "far" from left otherwise you can promote a sched entity that ends up having a really high vruntime

Quoted 29 lines
---
 kernel/sched/fair.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
kernel/sched/fair.c +2 -3
index 658aa7a2ae6f..4790f2d851ad 100644
@@ -4452,13 +4452,13 @@ pick_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *curr)
        /*
         * Prefer last buddy, try to return the CPU to a preempted task.
         */
-       if (cfs_rq->last && wakeup_preempt_entity(cfs_rq->last, left) < 1)
+       if (cfs_rq->last && wakeup_preempt_entity(cfs_rq->last, se) < 1)
                se = cfs_rq->last;
 
        /*
         * Someone really wants this to run. If it's not unfair, run it.
         */
-       if (cfs_rq->next && wakeup_preempt_entity(cfs_rq->next, left) < 1)
+       if (cfs_rq->next && wakeup_preempt_entity(cfs_rq->next, se) < 1)
                se = cfs_rq->next;
 
        clear_buddies(cfs_rq, se);
--

The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project