Re: [PATCH] sched: fix clear NOHZ_BALANCE_KICK

[Frederic Weisbecker]

On Tue, Jun 04, 2013 at 11:36:11AM +0200, Peter Zijlstra wrote:
> 
> The best I can seem to come up with is something like the below; but I think
> its ghastly. Surely we can do something saner with that bit.
> 
> Having to clear it at 3 different places is just wrong.

We could clear the flag early in scheduler_ipi() and set some
specific value in rq->idle_balance that tells we want nohz idle
balancing from the softirq, something like this untested:

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 58453b8..330136b 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -630,15 +630,14 @@ void wake_up_nohz_cpu(int cpu)
 		wake_up_idle_cpu(cpu);
 }
 
-static inline bool got_nohz_idle_kick(void)
+static inline bool got_nohz_idle_kick(int cpu)
 {
-	int cpu = smp_processor_id();
-	return idle_cpu(cpu) && test_bit(NOHZ_BALANCE_KICK, nohz_flags(cpu));
+	return test_and_clear_bit(NOHZ_BALANCE_KICK, nohz_flags(cpu));
 }
 
 #else /* CONFIG_NO_HZ_COMMON */
 
-static inline bool got_nohz_idle_kick(void)
+static inline bool got_nohz_idle_kick(int cpu)
 {
 	return false;
 }
@@ -1393,8 +1392,12 @@ static void sched_ttwu_pending(void)
 
 void scheduler_ipi(void)
 {
-	if (llist_empty(&this_rq()->wake_list) && !got_nohz_idle_kick()
-	    && !tick_nohz_full_cpu(smp_processor_id()))
+	int cpu = smp_processor_id();
+	bool idle_kick = got_nohz_idle_kick(cpu);
+
+	if (!(idle_kick && idle_cpu(cpu))
+	    && llist_empty(&this_rq()->wake_list)
+	    && !tick_nohz_full_cpu(cpu)
 		return;
 
 	/*
@@ -1417,8 +1420,8 @@ void scheduler_ipi(void)
 	/*
 	 * Check if someone kicked us for doing the nohz idle load balance.
 	 */
-	if (unlikely(got_nohz_idle_kick() && !need_resched())) {
-		this_rq()->idle_balance = 1;
+	    if (unlikely(idle_kick && idle_cpu(cpu) && !need_resched())) {
+		this_rq()->idle_balance = IDLE_NOHZ_BALANCE;
 		raise_softirq_irqoff(SCHED_SOFTIRQ);
 	}
 	irq_exit();
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index c61a614..816e7b0 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -5577,15 +5577,14 @@ out:
  * In CONFIG_NO_HZ_COMMON case, the idle balance kickee will do the
  * rebalancing for all the cpus for whom scheduler ticks are stopped.
  */
-static void nohz_idle_balance(int this_cpu, enum cpu_idle_type idle)
+static void nohz_idle_balance(int this_cpu)
 {
 	struct rq *this_rq = cpu_rq(this_cpu);
 	struct rq *rq;
 	int balance_cpu;
 
-	if (idle != CPU_IDLE ||
-	    !test_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu)))
-		goto end;
+	if (this_rq->idle_balance != IDLE_NOHZ_BALANCE)
+		return;
 
 	for_each_cpu(balance_cpu, nohz.idle_cpus_mask) {
 		if (balance_cpu == this_cpu || !idle_cpu(balance_cpu))
@@ -5612,8 +5611,12 @@ static void nohz_idle_balance(int this_cpu, enum cpu_idle_type idle)
 			this_rq->next_balance = rq->next_balance;
 	}
 	nohz.next_balance = this_rq->next_balance;
-end:
-	clear_bit(NOHZ_BALANCE_KICK, nohz_flags(this_cpu));
+
+	/* There could be concurrent updates from irqs but we don't care */
+	if (idle_cpu(this_cpu))
+		this_rq->idle_balance = IDLE_BALANCE;
+	else
+		this_rq->idle_balance = 0;
 }
 
 /*
@@ -5679,7 +5682,7 @@ need_kick:
 	return 1;
 }
 #else
-static void nohz_idle_balance(int this_cpu, enum cpu_idle_type idle) { }
+static void nohz_idle_balance(int this_cpu) { }
 #endif
 
 /*
@@ -5700,7 +5703,7 @@ static void run_rebalance_domains(struct softirq_action *h)
 	 * balancing on behalf of the other idle cpus whose ticks are
 	 * stopped.
 	 */
-	nohz_idle_balance(this_cpu, idle);
+	nohz_idle_balance(this_cpu);
 }
 
 static inline int on_null_domain(int cpu)
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index ce39224..e9de976 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -387,6 +387,11 @@ extern struct root_domain def_root_domain;
 
 #endif /* CONFIG_SMP */
 
+enum idle_balance_type {
+	IDLE_BALANCE = 1,
+	IDLE_NOHZ_BALANCE = 2,
+};
+
 /*
  * This is the main, per-CPU runqueue data structure.
  *
@@ -458,7 +463,7 @@ struct rq {
 
 	unsigned long cpu_power;
 
-	unsigned char idle_balance;
+	enum idle_balance_type idle_balance;
 	/* For active balancing */
 	int post_schedule;
 	int active_balance;
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/