sched/fair: Prepare to fix fairness problems on migration

Mike reported that our recent attempt to fix migration problems:

  3a47d51 ("sched/fair: Fix fairness issue on migration")

broke interactivity and the signal starve test. We reverted that
commit and now let's try it again more carefully, with some other
underlying problems fixed first.

One problem is that I assumed ENQUEUE_WAKING was only set when we do a
cross-cpu wakeup (migration), which isn't true. This means we now
destroy the vruntime history of tasks and wakeup-preemption suffers.

Cure this by making my assumption true, only call
sched_class::task_waking() when we do a cross-cpu wakeup. This avoids
the indirect call in the case we do a local wakeup.

Reported-by: Mike Galbraith <mgalbraith@suse.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Hunter <ahh@google.com>
Cc: Ben Segall <bsegall@google.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Morten Rasmussen <morten.rasmussen@arm.com>
Cc: Paul Turner <pjt@google.com>
Cc: Pavan Kondeti <pkondeti@codeaurora.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: byungchul.park@lge.com
Cc: linux-kernel@vger.kernel.org
Fixes: 3a47d51 ("sched/fair: Fix fairness issue on migration")
Signed-off-by: Ingo Molnar <mingo@kernel.org>

kernel/sched/core.c

@@ -1709,14 +1709,22 @@ static void
 ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags,
 		 struct pin_cookie cookie)
 {
+	int en_flags = ENQUEUE_WAKEUP;
+
 	lockdep_assert_held(&rq->lock);
 
 #ifdef CONFIG_SMP
 	if (p->sched_contributes_to_load)
 		rq->nr_uninterruptible--;
+
+	/*
+	 * If we migrated; we must have called sched_class::task_waking().
+	 */
+	if (wake_flags & WF_MIGRATED)
+		en_flags |= ENQUEUE_WAKING;
 #endif
 
-	ttwu_activate(rq, p, ENQUEUE_WAKEUP | ENQUEUE_WAKING);
+	ttwu_activate(rq, p, en_flags);
 	ttwu_do_wakeup(rq, p, wake_flags, cookie);
 }
 
@@ -1762,7 +1770,11 @@ void sched_ttwu_pending(void)
 	while (llist) {
 		p = llist_entry(llist, struct task_struct, wake_entry);
 		llist = llist_next(llist);
-		ttwu_do_activate(rq, p, 0, cookie);
+		/*
+		 * See ttwu_queue(); we only call ttwu_queue_remote() when
+		 * its a x-cpu wakeup.
+		 */
+		ttwu_do_activate(rq, p, WF_MIGRATED, cookie);
 	}
 
 	lockdep_unpin_lock(&rq->lock, cookie);
@@ -1849,7 +1861,7 @@ bool cpus_share_cache(int this_cpu, int that_cpu)
 }
 #endif /* CONFIG_SMP */
 
-static void ttwu_queue(struct task_struct *p, int cpu)
+static void ttwu_queue(struct task_struct *p, int cpu, int wake_flags)
 {
 	struct rq *rq = cpu_rq(cpu);
 	struct pin_cookie cookie;
@@ -1864,7 +1876,7 @@ static void ttwu_queue(struct task_struct *p, int cpu)
 
 	raw_spin_lock(&rq->lock);
 	cookie = lockdep_pin_lock(&rq->lock);
-	ttwu_do_activate(rq, p, 0, cookie);
+	ttwu_do_activate(rq, p, wake_flags, cookie);
 	lockdep_unpin_lock(&rq->lock, cookie);
 	raw_spin_unlock(&rq->lock);
 }
@@ -2034,17 +2046,18 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
 	p->sched_contributes_to_load = !!task_contributes_to_load(p);
 	p->state = TASK_WAKING;
 
-	if (p->sched_class->task_waking)
-		p->sched_class->task_waking(p);
-
 	cpu = select_task_rq(p, p->wake_cpu, SD_BALANCE_WAKE, wake_flags);
 	if (task_cpu(p) != cpu) {
 		wake_flags |= WF_MIGRATED;
+
+		if (p->sched_class->task_waking)
+			p->sched_class->task_waking(p);
+
 		set_task_cpu(p, cpu);
 	}
 #endif /* CONFIG_SMP */
 
-	ttwu_queue(p, cpu);
+	ttwu_queue(p, cpu, wake_flags);
 stat:
 	if (schedstat_enabled())
 		ttwu_stat(p, cpu, wake_flags);

kernel/sched/fair.c

@@ -3254,6 +3254,37 @@ static inline void check_schedstat_required(void)
 #endif
 }
 
+
+/*
+ * MIGRATION
+ *
+ *	dequeue
+ *	  update_curr()
+ *	    update_min_vruntime()
+ *	  vruntime -= min_vruntime
+ *
+ *	enqueue
+ *	  update_curr()
+ *	    update_min_vruntime()
+ *	  vruntime += min_vruntime
+ *
+ * this way the vruntime transition between RQs is done when both
+ * min_vruntime are up-to-date.
+ *
+ * WAKEUP (remote)
+ *
+ *	->task_waking_fair()
+ *	  vruntime -= min_vruntime
+ *
+ *	enqueue
+ *	  update_curr()
+ *	    update_min_vruntime()
+ *	  vruntime += min_vruntime
+ *
+ * this way we don't have the most up-to-date min_vruntime on the originating
+ * CPU and an up-to-date min_vruntime on the destination CPU.
+ */
+
 static void
 enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 {
@@ -4810,6 +4841,12 @@ static unsigned long cpu_avg_load_per_task(int cpu)
 	return 0;
 }
 
+/*
+ * Called to migrate a waking task; as blocked tasks retain absolute vruntime
+ * the migration needs to deal with this by subtracting the old and adding the
+ * new min_vruntime -- the latter is done by enqueue_entity() when placing
+ * the task on the new runqueue.
+ */
 static void task_waking_fair(struct task_struct *p)
 {
 	struct sched_entity *se = &p->se;