sched: Use cpu_dying() to fix balance_push vs hotplug-rollback

Use the new cpu_dying() state to simplify and fix the balance_push()
vs CPU hotplug rollback state.

Specifically, we currently rely on notifiers sched_cpu_dying() /
sched_cpu_activate() to terminate balance_push, however if the
cpu_down() fails when we're past sched_cpu_deactivate(), it should
terminate balance_push at that point and not wait until we hit
sched_cpu_activate().

Similarly, when cpu_up() fails and we're going back down, balance_push
should be active, where it currently is not.

So instead, make sure balance_push is enabled below SCHED_AP_ACTIVE
(when !cpu_active()), and gate it's utility with cpu_dying().
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Valentin Schneider <valentin.schneider@arm.com>
Link: https://lkml.kernel.org/r/YHgAYef83VQhKdC2@hirez.programming.kicks-ass.net

kernel/sched/core.c

@@ -1811,7 +1811,7 @@ static inline bool is_cpu_allowed(struct task_struct *p, int cpu)
 		return cpu_online(cpu);
 
 	/* Regular kernel threads don't get to stay during offline. */
-	if (cpu_rq(cpu)->balance_push)
+	if (cpu_dying(cpu))
 		return false;
 
 	/* But are allowed during online. */
@@ -7638,6 +7638,9 @@ static DEFINE_PER_CPU(struct cpu_stop_work, push_work);
 
 /*
  * Ensure we only run per-cpu kthreads once the CPU goes !active.
+ *
+ * This is enabled below SCHED_AP_ACTIVE; when !cpu_active(), but only
+ * effective when the hotplug motion is down.
  */
 static void balance_push(struct rq *rq)
 {
@@ -7645,11 +7648,18 @@ static void balance_push(struct rq *rq)
 
 	lockdep_assert_held(&rq->lock);
 	SCHED_WARN_ON(rq->cpu != smp_processor_id());
+
 	/*
 	 * Ensure the thing is persistent until balance_push_set(.on = false);
 	 */
 	rq->balance_callback = &balance_push_callback;
 
+	/*
+	 * Only active while going offline.
+	 */
+	if (!cpu_dying(rq->cpu))
+		return;
+
 	/*
 	 * Both the cpu-hotplug and stop task are in this case and are
 	 * required to complete the hotplug process.
@@ -7703,7 +7713,6 @@ static void balance_push_set(int cpu, bool on)
 	struct rq_flags rf;
 
 	rq_lock_irqsave(rq, &rf);
-	rq->balance_push = on;
 	if (on) {
 		WARN_ON_ONCE(rq->balance_callback);
 		rq->balance_callback = &balance_push_callback;
@@ -7828,8 +7837,8 @@ int sched_cpu_activate(unsigned int cpu)
 	struct rq_flags rf;
 
 	/*
-	 * Make sure that when the hotplug state machine does a roll-back
-	 * we clear balance_push. Ideally that would happen earlier...
+	 * Clear the balance_push callback and prepare to schedule
+	 * regular tasks.
 	 */
 	balance_push_set(cpu, false);
 
@@ -8014,12 +8023,6 @@ int sched_cpu_dying(unsigned int cpu)
 	}
 	rq_unlock_irqrestore(rq, &rf);
 
-	/*
-	 * Now that the CPU is offline, make sure we're welcome
-	 * to new tasks once we come back up.
-	 */
-	balance_push_set(cpu, false);
-
 	calc_load_migrate(rq);
 	update_max_interval();
 	hrtick_clear(rq);
@@ -8204,7 +8207,7 @@ void __init sched_init(void)
 		rq->sd = NULL;
 		rq->rd = NULL;
 		rq->cpu_capacity = rq->cpu_capacity_orig = SCHED_CAPACITY_SCALE;
-		rq->balance_callback = NULL;
+		rq->balance_callback = &balance_push_callback;
 		rq->active_balance = 0;
 		rq->next_balance = jiffies;
 		rq->push_cpu = 0;
@@ -8251,6 +8254,7 @@ void __init sched_init(void)
 
 #ifdef CONFIG_SMP
 	idle_thread_set_boot_cpu();
+	balance_push_set(smp_processor_id(), false);
 #endif
 	init_sched_fair_class();
 

kernel/sched/sched.h

@@ -983,7 +983,6 @@ struct rq {
 	unsigned long		cpu_capacity_orig;
 
 	struct callback_head	*balance_callback;
-	unsigned char		balance_push;
 
 	unsigned char		nohz_idle_balance;
 	unsigned char		idle_balance;