[PATCH v5 5/6] sched/fair: Track peak per-entity utilization

[Morten Rasmussen]

When using PELT (per-entity load tracking) utilization to place tasks at
wake-up using the decayed utilization (due to sleep) leads to
under-estimation of true utilization of the task. This could mean
putting the task on a cpu with less available capacity than is actually
needed. This issue can be mitigated by using 'peak' utilization instead
of the decayed utilization for placement decisions, e.g. at task
wake-up.

The 'peak' utilization metric, util_peak, tracks util_avg when the task
is running and retains its previous value while the task is
blocked/waiting on the rq. It is instantly updated to track util_avg
again as soon as the task running again.

cc: Ingo Molnar <mingo@xxxxxxxxxx>
cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>

Signed-off-by: Morten Rasmussen <morten.rasmussen@xxxxxxx>
---
 include/linux/sched.h |  2 +-
 kernel/sched/fair.c   | 23 +++++++++++++++++------
 2 files changed, 18 insertions(+), 7 deletions(-)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index ad51978ff15e..988d7f48604e 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1294,7 +1294,7 @@ struct load_weight {
 struct sched_avg {
 	u64 last_update_time, load_sum;
 	u32 util_sum, period_contrib;
-	unsigned long load_avg, util_avg;
+	unsigned long load_avg, util_avg, util_peak;
 };
 
 #ifdef CONFIG_SCHEDSTATS
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index a5efafda23ef..84b767399d61 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -696,6 +696,7 @@ void init_entity_runnable_average(struct sched_entity *se)
 	 * At this point, util_avg won't be used in select_task_rq_fair anyway
 	 */
 	sa->util_avg = 0;
+	sa->util_peak = 0;
 	sa->util_sum = 0;
 	/* when this task enqueue'ed, it will contribute to its cfs_rq's load_avg */
 }
@@ -747,6 +748,7 @@ void post_init_entity_util_avg(struct sched_entity *se)
 		} else {
 			sa->util_avg = cap;
 		}
+		sa->util_peak = sa->util_avg;
 		sa->util_sum = sa->util_avg * LOAD_AVG_MAX;
 	}
 
@@ -3181,6 +3183,10 @@ dequeue_entity_load_avg(struct cfs_rq *cfs_rq, struct sched_entity *se)
 		max_t(long, cfs_rq->runnable_load_avg - se->avg.load_avg, 0);
 	cfs_rq->runnable_load_sum =
 		max_t(s64,  cfs_rq->runnable_load_sum - se->avg.load_sum, 0);
+
+	/* Save peak PELT utilization for task to help wake-up decisions */
+	if (entity_is_task(se))
+		se->avg.util_peak = se->avg.util_avg;
 }
 
 #ifndef CONFIG_64BIT
@@ -5203,7 +5209,7 @@ static int wake_affine(struct sched_domain *sd, struct task_struct *p,
 	return 1;
 }
 
-static inline int task_util(struct task_struct *p);
+static inline int task_util_peak(struct task_struct *p);
 static int cpu_util_wake(int cpu, struct task_struct *p);
 
 static unsigned long capacity_spare_wake(int cpu, struct task_struct *p)
@@ -5286,14 +5292,14 @@ find_idlest_group(struct sched_domain *sd, struct task_struct *p,
 	/*
 	 * The cross-over point between using spare capacity or least load
 	 * is too conservative for high utilization tasks on partially
-	 * utilized systems if we require spare_capacity > task_util(p),
+	 * utilized systems if we require spare_capacity > task_util_peak(p),
 	 * so we allow for some task stuffing by using
-	 * spare_capacity > task_util(p)/2.
+	 * spare_capacity > task_util_peak(p)/2.
 	 */
-	if (this_spare > task_util(p) / 2 &&
+	if (this_spare > task_util_peak(p) / 2 &&
 	    imbalance*this_spare > 100*most_spare)
 		return NULL;
-	else if (most_spare > task_util(p) / 2)
+	else if (most_spare > task_util_peak(p) / 2)
 		return most_spare_sg;
 
 	if (!idlest || 100*this_load < imbalance*min_load)
@@ -5628,6 +5634,11 @@ static inline int task_util(struct task_struct *p)
 	return p->se.avg.util_avg;
 }
 
+static inline int task_util_peak(struct task_struct *p)
+{
+	return p->se.avg.util_peak;
+}
+
 /*
  * cpu_util_wake: Compute cpu utilization with any contributions from
  * the waking task p removed.
@@ -5667,7 +5678,7 @@ static int wake_cap(struct task_struct *p, int cpu, int prev_cpu)
 	/* Bring task utilization in sync with prev_cpu */
 	sync_entity_load_avg(&p->se);
 
-	return min_cap * 1024 < task_util(p) * capacity_margin;
+	return min_cap * 1024 < task_util_peak(p) * capacity_margin;
 }
 
 /*
-- 
2.7.4