sched: Fix rq->nr_iowait ordering

  schedule()				ttwu()
    deactivate_task();			  if (p->on_rq && ...) // false
					    atomic_dec(&task_rq(p)->nr_iowait);
    if (prev->in_iowait)
      atomic_inc(&rq->nr_iowait);

Allows nr_iowait to be decremented before it gets incremented,
resulting in more dodgy IO-wait numbers than usual.

Note that because we can now do ttwu_queue_wakelist() before
p->on_cpu==0, we lose the natural ordering and have to further delay
the decrement.

Fixes: c6e7bd7 ("sched/core: Optimize ttwu() spinning on p->on_cpu")
Reported-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Link: https://lkml.kernel.org/r/20201117093829.GD3121429@hirez.programming.kicks-ass.net

kernel/sched/core.c

@@ -2501,7 +2501,12 @@ ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags,
 #ifdef CONFIG_SMP
 	if (wake_flags & WF_MIGRATED)
 		en_flags |= ENQUEUE_MIGRATED;
+	else
 #endif
+	if (p->in_iowait) {
+		delayacct_blkio_end(p);
+		atomic_dec(&task_rq(p)->nr_iowait);
+	}
 
 	activate_task(rq, p, en_flags);
 	ttwu_do_wakeup(rq, p, wake_flags, rf);
@@ -2888,11 +2893,6 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
 	if (READ_ONCE(p->on_rq) && ttwu_runnable(p, wake_flags))
 		goto unlock;
 
-	if (p->in_iowait) {
-		delayacct_blkio_end(p);
-		atomic_dec(&task_rq(p)->nr_iowait);
-	}
-
 #ifdef CONFIG_SMP
 	/*
 	 * Ensure we load p->on_cpu _after_ p->on_rq, otherwise it would be
@@ -2963,6 +2963,11 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
 
 	cpu = select_task_rq(p, p->wake_cpu, SD_BALANCE_WAKE, wake_flags);
 	if (task_cpu(p) != cpu) {
+		if (p->in_iowait) {
+			delayacct_blkio_end(p);
+			atomic_dec(&task_rq(p)->nr_iowait);
+		}
+
 		wake_flags |= WF_MIGRATED;
 		psi_ttwu_dequeue(p);
 		set_task_cpu(p, cpu);