sched/fair: Fix bandwidth timer clock drift condition

When period gets restarted after some idle time, start_cfs_bandwidth() doesn't update the expiration information, expire_cfs_rq_runtime() will see cfs_rq->runtime_expires smaller than rq clock and go to the clock drift logic, wasting needless cpu cycles on the scheduler hot path.

Update the global expiration in start_cfs_bandwidth() to avoid frequent expire_cfs_rq_runtime() calls once a new period begins.

Signed-off-by Xunlei Pang <[email protected]>

---
 kernel/sched/fair.c | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)

kernel/sched/fair.c +10 -5

index e6bb68d52962..f167aca066cc 100644
@@ -5204,13 +5204,18 @@ static void init_cfs_rq_runtime(struct cfs_rq *cfs_rq)
 
 void start_cfs_bandwidth(struct cfs_bandwidth *cfs_b)
 {
+	u64 overrun;
+
 	lockdep_assert_held(&cfs_b->lock);
 
-	if (!cfs_b->period_active) {
-		cfs_b->period_active = 1;
-		hrtimer_forward_now(&cfs_b->period_timer, cfs_b->period);
-		hrtimer_start_expires(&cfs_b->period_timer, HRTIMER_MODE_ABS_PINNED);
-	}
+	if (cfs_b->period_active)
+		return;
+
+	cfs_b->period_active = 1;
+	overrun = hrtimer_forward_now(&cfs_b->period_timer, cfs_b->period);
+	cfs_b->runtime_expires += (overrun + 1) * ktime_to_ns(cfs_b->period);
+	cfs_b->expires_seq++;
+	hrtimer_start_expires(&cfs_b->period_timer, HRTIMER_MODE_ABS_PINNED);
 }
 
 static void destroy_cfs_bandwidth(struct cfs_bandwidth *cfs_b)

I noticed the group constantly got throttled even it consumed low cpu usage, this caused some jitters on the response time to some of our business containers enabling cpu quota.

It's very simple to reproduce:
mkdir /sys/fs/cgroup/cpu/test
cd /sys/fs/cgroup/cpu/test
echo 100000 > cpu.cfs_quota_us
echo $$ > tasks
then repeat:
cat cpu.stat |grep nr_throttled // nr_throttled will increase

After some analysis, we found that cfs_rq::runtime_remaining will be cleared by expire_cfs_rq_runtime() due to two equal but stale "cfs_{b|q}->runtime_expires" after period timer is re-armed.

The current condition to judge clock drift in expire_cfs_rq_runtime() is wrong, the two runtime_expires are actually the same when clock drift happens, so this condtion can never hit. The orginal design was correctly done by commit a9cf55b28610 ("sched: Expire invalid runtime"), but was changed to be the current one due to its locking issue.

This patch introduces another way, it adds a new field in both structures cfs_rq and cfs_bandwidth to record the expiration update sequence, and use them to figure out if clock drift happens(true if they equal).

Fixes 51f2176d74ac ("sched/fair: Fix unlocked reads of some cfs_b->quota/period")

Cc Ben Segall <[email protected]>

Signed-off-by Xunlei Pang <[email protected]>

---
 kernel/sched/fair.c  | 14 ++++++++------
 kernel/sched/sched.h |  6 ++++--
 2 files changed, 12 insertions(+), 8 deletions(-)

kernel/sched/fair.c +8 -6

index e497c05aab7f..e6bb68d52962 100644
@@ -4590,6 +4590,7 @@ void __refill_cfs_bandwidth_runtime(struct cfs_bandwidth *cfs_b)
 	now = sched_clock_cpu(smp_processor_id());
 	cfs_b->runtime = cfs_b->quota;
 	cfs_b->runtime_expires = now + ktime_to_ns(cfs_b->period);
+	cfs_b->expires_seq++;
 }
 
 static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg)

@@ -4612,6 +4613,7 @@ static int assign_cfs_rq_runtime(struct cfs_rq *cfs_rq)
 	struct task_group *tg = cfs_rq->tg;
 	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(tg);
 	u64 amount = 0, min_amount, expires;
+	int expires_seq;
 
 	/* note: this is a positive sum as runtime_remaining <= 0 */
 	min_amount = sched_cfs_bandwidth_slice() - cfs_rq->runtime_remaining;

@@ -4628,6 +4630,7 @@ static int assign_cfs_rq_runtime(struct cfs_rq *cfs_rq)
 			cfs_b->idle = 0;
 		}
 	}
+	expires_seq = cfs_b->expires_seq;
 	expires = cfs_b->runtime_expires;
 	raw_spin_unlock(&cfs_b->lock);
 

@@ -4637,8 +4640,10 @@ static int assign_cfs_rq_runtime(struct cfs_rq *cfs_rq)
 	 * spread between our sched_clock and the one on which runtime was
 	 * issued.
 	 */
-	if ((s64)(expires - cfs_rq->runtime_expires) > 0)
+	if (cfs_rq->expires_seq != expires_seq) {
+		cfs_rq->expires_seq = expires_seq;
 		cfs_rq->runtime_expires = expires;
+	}
 
 	return cfs_rq->runtime_remaining > 0;
 }

@@ -4664,12 +4669,9 @@ static void expire_cfs_rq_runtime(struct cfs_rq *cfs_rq)
 	 * has not truly expired.
 	 *
 	 * Fortunately we can check determine whether this the case by checking
-	 * whether the global deadline has advanced. It is valid to compare
-	 * cfs_b->runtime_expires without any locks since we only care about
-	 * exact equality, so a partial write will still work.
+	 * whether the global deadline(cfs_b->expires_seq) has advanced.
 	 */
-
-	if (cfs_rq->runtime_expires != cfs_b->runtime_expires) {
+	if (cfs_rq->expires_seq == cfs_b->expires_seq) {
 		/* extend local deadline, drift is bounded above by 2 ticks */
 		cfs_rq->runtime_expires += TICK_NSEC;
 	} else {

kernel/sched/sched.h +4 -2

index 6601baf2361c..e977e04f8daf 100644
@@ -334,9 +334,10 @@ struct cfs_bandwidth {
 	u64			runtime;
 	s64			hierarchical_quota;
 	u64			runtime_expires;
+	int			expires_seq;
 
-	int			idle;
-	int			period_active;
+	short			idle;
+	short			period_active;
 	struct hrtimer		period_timer;
 	struct hrtimer		slack_timer;
 	struct list_head	throttled_cfs_rq;

@@ -551,6 +552,7 @@ struct cfs_rq {
 
 #ifdef CONFIG_CFS_BANDWIDTH
 	int			runtime_enabled;
+	int			expires_seq;
 	u64			runtime_expires;
 	s64			runtime_remaining;

Quoted 22 lines · Xunlei Pang

I noticed the group constantly got throttled even it consumed low cpu usage, this caused some jitters on the response time to some of our business containers enabling cpu quota.

It's very simple to reproduce:
mkdir /sys/fs/cgroup/cpu/test
cd /sys/fs/cgroup/cpu/test
echo 100000 > cpu.cfs_quota_us
echo $$ > tasks
then repeat:
cat cpu.stat |grep nr_throttled // nr_throttled will increase

After some analysis, we found that cfs_rq::runtime_remaining will be cleared by expire_cfs_rq_runtime() due to two equal but stale "cfs_{b|q}->runtime_expires" after period timer is re-armed.

The current condition to judge clock drift in expire_cfs_rq_runtime() is wrong, the two runtime_expires are actually the same when clock drift happens, so this condtion can never hit. The orginal design was correctly done by commit a9cf55b28610 ("sched: Expire invalid runtime"), but was changed to be the current one due to its locking issue.

Quoted 6 lines

This patch introduces another way, it adds a new field in both structures cfs_rq and cfs_bandwidth to record the expiration update sequence, and use them to figure out if clock drift happens(true if they equal).

Fixes 51f2176d74ac ("sched/fair: Fix unlocked reads of some cfs_b->quota/period")

Cc Ben Segall <[email protected]>

Reviewed-by Ben Segall <[email protected]>

Quoted 86 lines

Signed-off-by Xunlei Pang <[email protected]>

---
 kernel/sched/fair.c  | 14 ++++++++------
 kernel/sched/sched.h |  6 ++++--
 2 files changed, 12 insertions(+), 8 deletions(-)

kernel/sched/fair.c +8 -6

index e497c05aab7f..e6bb68d52962 100644
@@ -4590,6 +4590,7 @@ void __refill_cfs_bandwidth_runtime(struct cfs_bandwidth *cfs_b)
 	now = sched_clock_cpu(smp_processor_id());
 	cfs_b->runtime = cfs_b->quota;
 	cfs_b->runtime_expires = now + ktime_to_ns(cfs_b->period);
+	cfs_b->expires_seq++;
 }
 
 static inline struct cfs_bandwidth *tg_cfs_bandwidth(struct task_group *tg)

@@ -4612,6 +4613,7 @@ static int assign_cfs_rq_runtime(struct cfs_rq *cfs_rq)
 	struct task_group *tg = cfs_rq->tg;
 	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(tg);
 	u64 amount = 0, min_amount, expires;
+	int expires_seq;
 
 	/* note: this is a positive sum as runtime_remaining <= 0 */
 	min_amount = sched_cfs_bandwidth_slice() - cfs_rq->runtime_remaining;

@@ -4628,6 +4630,7 @@ static int assign_cfs_rq_runtime(struct cfs_rq *cfs_rq)
 			cfs_b->idle = 0;
 		}
 	}
+	expires_seq = cfs_b->expires_seq;
 	expires = cfs_b->runtime_expires;
 	raw_spin_unlock(&cfs_b->lock);
 

@@ -4637,8 +4640,10 @@ static int assign_cfs_rq_runtime(struct cfs_rq *cfs_rq)
 	 * spread between our sched_clock and the one on which runtime was
 	 * issued.
 	 */
-	if ((s64)(expires - cfs_rq->runtime_expires) > 0)
+	if (cfs_rq->expires_seq != expires_seq) {
+		cfs_rq->expires_seq = expires_seq;
 		cfs_rq->runtime_expires = expires;
+	}
 
 	return cfs_rq->runtime_remaining > 0;
 }

@@ -4664,12 +4669,9 @@ static void expire_cfs_rq_runtime(struct cfs_rq *cfs_rq)
 	 * has not truly expired.
 	 *
 	 * Fortunately we can check determine whether this the case by checking
-	 * whether the global deadline has advanced. It is valid to compare
-	 * cfs_b->runtime_expires without any locks since we only care about
-	 * exact equality, so a partial write will still work.
+	 * whether the global deadline(cfs_b->expires_seq) has advanced.
 	 */
-
-	if (cfs_rq->runtime_expires != cfs_b->runtime_expires) {
+	if (cfs_rq->expires_seq == cfs_b->expires_seq) {
 		/* extend local deadline, drift is bounded above by 2 ticks */
 		cfs_rq->runtime_expires += TICK_NSEC;
 	} else {

kernel/sched/sched.h +4 -2

index 6601baf2361c..e977e04f8daf 100644
@@ -334,9 +334,10 @@ struct cfs_bandwidth {
 	u64			runtime;
 	s64			hierarchical_quota;
 	u64			runtime_expires;
+	int			expires_seq;
 
-	int			idle;
-	int			period_active;
+	short			idle;
+	short			period_active;
 	struct hrtimer		period_timer;
 	struct hrtimer		slack_timer;
 	struct list_head	throttled_cfs_rq;

@@ -551,6 +552,7 @@ struct cfs_rq {
 
 #ifdef CONFIG_CFS_BANDWIDTH
 	int			runtime_enabled;
+	int			expires_seq;
 	u64			runtime_expires;
 	s64			runtime_remaining;

Quoted 4 lines

Quoted 2 lines

Fixes 51f2176d74ac ("sched/fair: Fix unlocked reads of some cfs_b->quota/period")

Cc Ben Segall <[email protected]>

Reviewed-by Ben Segall <[email protected]>

Thanks Ben :-)

Hi Peter, could you please have a look at them?

Cc [email protected]

Quoted 7 lines · Xunlei Pang

Quoted 4 lines

Quoted 2 lines

Fixes 51f2176d74ac ("sched/fair: Fix unlocked reads of some cfs_b->quota/period")

Cc Ben Segall <[email protected]>

Reviewed-by Ben Segall <[email protected]>

Thanks Ben :-)

Quoted 1 line

Hi Peter, could you please have a look at them?

I grabbed them, thanks guys!

Quoted 2 lines

Cc [email protected]

For both or just the first?