[RFC PATCH 4/5] sched: Add nohz stats balancing

[Peter Zijlstra]

Suggested-by: Vincent Guittot <vincent.guittot@xxxxxxxxxx>
Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
---
 kernel/sched/fair.c |   31 +++++++++++++++++++------------
 1 file changed, 19 insertions(+), 12 deletions(-)

--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -8972,6 +8972,7 @@ static struct {
 	cpumask_var_t idle_cpus_mask;
 	atomic_t nr_cpus;
 	unsigned long next_balance;     /* in jiffy units */
+	unsigned long next_stats;
 } nohz ____cacheline_aligned;
 
 static inline int find_new_ilb(void)
@@ -9006,9 +9007,8 @@ static inline void set_cpu_sd_state_busy
  * nohz_load_balancer CPU (if there is one) otherwise fallback to any idle
  * CPU (if there is one).
  */
-static void kick_ilb(void)
+static void kick_ilb(unsigned int flags)
 {
-	unsigned int flags;
 	int ilb_cpu;
 
 	nohz.next_balance++;
@@ -9018,7 +9018,7 @@ static void kick_ilb(void)
 	if (ilb_cpu >= nr_cpu_ids)
 		return;
 
-	flags = atomic_fetch_or(NOHZ_KICK_MASK, nohz_flags(ilb_cpu));
+	flags = atomic_fetch_or(flags, nohz_flags(ilb_cpu));
 	if (flags & NOHZ_KICK_MASK)
 		return;
 
@@ -9048,7 +9048,7 @@ static void nohz_balancer_kick(struct rq
 	struct sched_domain_shared *sds;
 	struct sched_domain *sd;
 	int nr_busy, i, cpu = rq->cpu;
-	bool kick = false;
+	unsigned int flags = 0;
 
 	if (unlikely(rq->idle_balance))
 		return;
@@ -9067,11 +9067,14 @@ static void nohz_balancer_kick(struct rq
 	if (likely(!atomic_read(&nohz.nr_cpus)))
 		return;
 
+	if (time_after(now, nohz.next_stats))
+		flags = NOHZ_STATS_KICK;
+
 	if (time_before(now, nohz.next_balance))
-		return;
+		goto out;
 
 	if (rq->nr_running >= 2) {
-		kick = true;
+		flags = NOHZ_KICK_MASK;
 		goto out;
 	}
 
@@ -9084,7 +9087,7 @@ static void nohz_balancer_kick(struct rq
 		 */
 		nr_busy = atomic_read(&sds->nr_busy_cpus);
 		if (nr_busy > 1) {
-			kick = true;
+			flags = NOHZ_KICK_MASK
 			goto unlock;
 		}
 
@@ -9094,7 +9097,7 @@ static void nohz_balancer_kick(struct rq
 	if (sd) {
 		if ((rq->cfs.h_nr_running >= 1) &&
 				check_cpu_capacity(rq, sd)) {
-			kick = true;
+			flags = NOHZ_KICK_MASK;
 			goto unlock;
 		}
 	}
@@ -9107,7 +9110,7 @@ static void nohz_balancer_kick(struct rq
 				continue;
 
 			if (sched_asym_prefer(i, cpu)) {
-				kick = true;
+				flags = NOHZ_KICK_MASK;
 				goto unlock;
 			}
 		}
@@ -9115,8 +9118,8 @@ static void nohz_balancer_kick(struct rq
 unlock:
 	rcu_read_unlock();
 out:
-	if (kick)
-		kick_ilb();
+	if (flags)
+		kick_ilb(flags);
 }
 
 void nohz_balance_exit_idle(unsigned int cpu)
@@ -9312,7 +9315,9 @@ static void rebalance_domains(struct rq
 static bool nohz_idle_balance(struct rq *this_rq, enum cpu_idle_type idle)
 {
 	/* Earliest time when we have to do rebalance again */
-	unsigned long next_balance = jiffies + 60*HZ;
+	unsigned long now = jiffies;
+	unsigned long next_balance = now + 60*HZ;
+	unsigned long next_stats = now + msecs_to_jiffies(LOAD_AVG_PERIOD);
 	int update_next_balance = 0;
 	int this_cpu = this_rq->cpu;
 	unsigned int flags;
@@ -9372,6 +9377,8 @@ static bool nohz_idle_balance(struct rq
 	if (flags & NOHZ_BALANCE_KICK)
 		rebalance_domains(this_rq, CPU_IDLE);
 
+	nohz.next_stats = next_stats;
+
 	/*
 	 * next_balance will be updated only when there is a need.
 	 * When the CPU is attached to null domain for ex, it will not be