Skip to contentSkip to conversation

sched/fair: don't restart enqueued cfs quota slack timer

PATCH In review Liangyan 3 participants 3 messages Jun 3, 2019 Scheduler

lore
Share & tools
lore

Liangyan

author
#0 lore raw

PATCH 1/1 sched/fair: don't restart enqueued cfs quota slack timer

From: "liangyan.ply" <[email protected]>

start_cfs_slack_bandwidth() will restart the quota slack timer, if it is called frequently, this timer will be restarted continuously and may have no chance to expire to unthrottle cfs tasks. This will cause that the throttled tasks can't be unthrottled in time although they have remaining quota.

---
 kernel/sched/fair.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
kernel/sched/fair.c +3 -1
index d90a64620072..fdb03c752f97 100644
@@ -4411,9 +4411,11 @@ static void start_cfs_slack_bandwidth(struct cfs_bandwidth *cfs_b)
 	if (runtime_refresh_within(cfs_b, min_left))
 		return;
 
-	hrtimer_start(&cfs_b->slack_timer,
+	if (!hrtimer_active(&cfs_b->slack_timer)) {
+		hrtimer_start(&cfs_b->slack_timer,
 			ns_to_ktime(cfs_bandwidth_slack_period),
 			HRTIMER_MODE_REL);
+	}
 }
 
 /* we know any runtime found here is valid as update_curr() precedes return */
Signature · 1 line
2.14.4.44.g2045bb6
maintainer
#1 lore raw
Quoted 7 lines · Liangyan

From: "liangyan.ply" <[email protected]>

start_cfs_slack_bandwidth() will restart the quota slack timer, if it is called frequently, this timer will be restarted continuously and may have no chance to expire to unthrottle cfs tasks. This will cause that the throttled tasks can't be unthrottled in time although they have remaining quota.

This looks very similar to the patch from Ben here:

https://lkml.kernel.org/r/[email protected]

Quoted 26 lines
---
 kernel/sched/fair.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
kernel/sched/fair.c +3 -1
index d90a64620072..fdb03c752f97 100644
@@ -4411,9 +4411,11 @@ static void start_cfs_slack_bandwidth(struct cfs_bandwidth *cfs_b)
 	if (runtime_refresh_within(cfs_b, min_left))
 		return;
 
-	hrtimer_start(&cfs_b->slack_timer,
+	if (!hrtimer_active(&cfs_b->slack_timer)) {
+		hrtimer_start(&cfs_b->slack_timer,
 			ns_to_ktime(cfs_bandwidth_slack_period),
 			HRTIMER_MODE_REL);
+	}
 }
 
 /* we know any runtime found here is valid as update_curr() precedes return */
Signature · 1 line
2.14.4.44.g2045bb6
#2 lore raw
Quoted 9 lines · Peter Zijlstra
Quoted 7 lines · Liangyan

From: "liangyan.ply" <[email protected]>

start_cfs_slack_bandwidth() will restart the quota slack timer, if it is called frequently, this timer will be restarted continuously and may have no chance to expire to unthrottle cfs tasks. This will cause that the throttled tasks can't be unthrottled in time although they have remaining quota.

Quoted 3 lines

This looks very similar to the patch from Ben here:

https://lkml.kernel.org/r/[email protected]

Yeah, it should do the same sort of thing, but testing hrtimer_active is racy (we could miss restarting the timer if it's just dropped the cfsb lock but hasn't finished yet), so the extra accounting flag is needed.

Quoted 27 lines
Quoted 27 lines
---
 kernel/sched/fair.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
kernel/sched/fair.c +3 -1
index d90a64620072..fdb03c752f97 100644
@@ -4411,9 +4411,11 @@ static void start_cfs_slack_bandwidth(struct cfs_bandwidth *cfs_b)
 	if (runtime_refresh_within(cfs_b, min_left))
 		return;
 
-	hrtimer_start(&cfs_b->slack_timer,
+	if (!hrtimer_active(&cfs_b->slack_timer)) {
+		hrtimer_start(&cfs_b->slack_timer,
 			ns_to_ktime(cfs_bandwidth_slack_period),
 			HRTIMER_MODE_REL);
+	}
 }
 
 /* we know any runtime found here is valid as update_curr() precedes return */
Signature · 1 line
2.14.4.44.g2045bb6