sched/nohz: Clean up nohz enter/exit

The primary observation is that nohz enter/exit is always from the
current CPU, therefore NOHZ_TICK_STOPPED does not in fact need to be
an atomic.

Secondary is that we appear to have 2 nearly identical hooks in the
nohz enter code, set_cpu_sd_state_idle() and
nohz_balance_enter_idle(). Fold the whole set_cpu_sd_state thing into
nohz_balance_{enter,exit}_idle.

Removes an atomic op from both enter and exit paths.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

include/linux/sched/nohz.h

@@ -16,11 +16,9 @@ static inline void cpu_load_update_nohz_stop(void) { }
 
 #if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
 extern void nohz_balance_enter_idle(int cpu);
-extern void set_cpu_sd_state_idle(void);
 extern int get_nohz_timer_target(void);
 #else
 static inline void nohz_balance_enter_idle(int cpu) { }
-static inline void set_cpu_sd_state_idle(void) { }
 #endif
 
 #ifdef CONFIG_NO_HZ_COMMON

kernel/sched/core.c

@@ -5861,7 +5861,7 @@ int sched_cpu_dying(unsigned int cpu)
 
 	calc_load_migrate(rq);
 	update_max_interval();
-	nohz_balance_exit_idle(cpu);
+	nohz_balance_exit_idle(rq);
 	hrtick_clear(rq);
 	return 0;
 }

kernel/sched/fair.c

@@ -9103,23 +9103,6 @@ static inline int find_new_ilb(void)
 	return nr_cpu_ids;
 }
 
-static inline void set_cpu_sd_state_busy(void)
-{
-	struct sched_domain *sd;
-	int cpu = smp_processor_id();
-
-	rcu_read_lock();
-	sd = rcu_dereference(per_cpu(sd_llc, cpu));
-
-	if (!sd || !sd->nohz_idle)
-		goto unlock;
-	sd->nohz_idle = 0;
-
-	atomic_inc(&sd->shared->nr_busy_cpus);
-unlock:
-	rcu_read_unlock();
-}
-
 /*
  * Kick a CPU to do the nohz balancing, if it is time for it. We pick the
  * nohz_load_balancer CPU (if there is one) otherwise fallback to any idle
@@ -9175,8 +9158,7 @@ static void nohz_balancer_kick(struct rq *rq)
 	 * We may be recently in ticked or tickless idle mode. At the first
 	 * busy tick after returning from idle, we will update the busy stats.
 	 */
-	set_cpu_sd_state_busy();
-	nohz_balance_exit_idle(cpu);
+	nohz_balance_exit_idle(rq);
 
 	/*
 	 * None are in tickless mode and hence no need for NOHZ idle load
@@ -9240,27 +9222,39 @@ static void nohz_balancer_kick(struct rq *rq)
 		kick_ilb(flags);
 }
 
-void nohz_balance_exit_idle(unsigned int cpu)
-{
-	unsigned int flags = atomic_read(nohz_flags(cpu));
-
-	if (unlikely(flags & NOHZ_TICK_STOPPED)) {
-		/*
-		 * Completely isolated CPUs don't ever set, so we must test.
-		 */
-		if (likely(cpumask_test_cpu(cpu, nohz.idle_cpus_mask))) {
-			cpumask_clear_cpu(cpu, nohz.idle_cpus_mask);
-			atomic_dec(&nohz.nr_cpus);
-		}
-
-		atomic_andnot(NOHZ_TICK_STOPPED, nohz_flags(cpu));
-	}
-}
-
-void set_cpu_sd_state_idle(void)
+static void set_cpu_sd_state_busy(int cpu)
+{
+	struct sched_domain *sd;
+
+	rcu_read_lock();
+	sd = rcu_dereference(per_cpu(sd_llc, cpu));
+
+	if (!sd || !sd->nohz_idle)
+		goto unlock;
+	sd->nohz_idle = 0;
+
+	atomic_inc(&sd->shared->nr_busy_cpus);
+unlock:
+	rcu_read_unlock();
+}
+
+void nohz_balance_exit_idle(struct rq *rq)
+{
+	SCHED_WARN_ON(rq != this_rq());
+
+	if (likely(!rq->nohz_tick_stopped))
+		return;
+
+	rq->nohz_tick_stopped = 0;
+	cpumask_clear_cpu(rq->cpu, nohz.idle_cpus_mask);
+	atomic_dec(&nohz.nr_cpus);
+
+	set_cpu_sd_state_busy(rq->cpu);
+}
+
+static void set_cpu_sd_state_idle(int cpu)
 {
 	struct sched_domain *sd;
-	int cpu = smp_processor_id();
 
 	rcu_read_lock();
 	sd = rcu_dereference(per_cpu(sd_llc, cpu));
@@ -9280,6 +9274,10 @@ void set_cpu_sd_state_idle(void)
  */
 void nohz_balance_enter_idle(int cpu)
 {
+	struct rq *rq = cpu_rq(cpu);
+
+	SCHED_WARN_ON(cpu != smp_processor_id());
+
 	/* If this CPU is going down, then nothing needs to be done: */
 	if (!cpu_active(cpu))
 		return;
@@ -9288,16 +9286,19 @@ void nohz_balance_enter_idle(int cpu)
 	if (!housekeeping_cpu(cpu, HK_FLAG_SCHED))
 		return;
 
-	if (atomic_read(nohz_flags(cpu)) & NOHZ_TICK_STOPPED)
+	if (rq->nohz_tick_stopped)
 		return;
 
 	/* If we're a completely isolated CPU, we don't play: */
-	if (on_null_domain(cpu_rq(cpu)))
+	if (on_null_domain(rq))
 		return;
 
+	rq->nohz_tick_stopped = 1;
+
 	cpumask_set_cpu(cpu, nohz.idle_cpus_mask);
 	atomic_inc(&nohz.nr_cpus);
-	atomic_or(NOHZ_TICK_STOPPED, nohz_flags(cpu));
+
+	set_cpu_sd_state_idle(cpu);
 }
 #else
 static inline void nohz_balancer_kick(struct rq *rq) { }

kernel/sched/sched.h

@@ -764,6 +764,7 @@ struct rq {
 	unsigned long		last_load_update_tick;
 	unsigned long		last_blocked_load_update_tick;
 #endif /* CONFIG_SMP */
+	unsigned int		nohz_tick_stopped;
 	atomic_t nohz_flags;
 #endif /* CONFIG_NO_HZ_COMMON */
 
@@ -2035,11 +2036,9 @@ extern void cfs_bandwidth_usage_inc(void);
 extern void cfs_bandwidth_usage_dec(void);
 
 #ifdef CONFIG_NO_HZ_COMMON
-#define NOHZ_TICK_STOPPED_BIT	0
-#define NOHZ_BALANCE_KICK_BIT	1
-#define NOHZ_STATS_KICK_BIT	2
+#define NOHZ_BALANCE_KICK_BIT	0
+#define NOHZ_STATS_KICK_BIT	1
 
-#define NOHZ_TICK_STOPPED	BIT(NOHZ_TICK_STOPPED_BIT)
 #define NOHZ_BALANCE_KICK	BIT(NOHZ_BALANCE_KICK_BIT)
 #define NOHZ_STATS_KICK		BIT(NOHZ_STATS_KICK_BIT)
 
@@ -2047,9 +2046,9 @@ extern void cfs_bandwidth_usage_dec(void);
 
 #define nohz_flags(cpu)	(&cpu_rq(cpu)->nohz_flags)
 
-extern void nohz_balance_exit_idle(unsigned int cpu);
+extern void nohz_balance_exit_idle(struct rq *rq);
 #else
-static inline void nohz_balance_exit_idle(unsigned int cpu) { }
+static inline void nohz_balance_exit_idle(struct rq *rq) { }
 #endif
 
 

kernel/time/tick-sched.c

@@ -954,13 +954,6 @@ void tick_nohz_idle_enter(void)
 	struct tick_sched *ts;
 
 	lockdep_assert_irqs_enabled();
-	/*
-	 * Update the idle state in the scheduler domain hierarchy
-	 * when tick_nohz_stop_sched_tick() is called from the idle loop.
-	 * State will be updated to busy during the first busy tick after
-	 * exiting idle.
-	 */
-	set_cpu_sd_state_idle();
 
 	local_irq_disable();