sched: Fix pick_next_task_fair() vs try_to_wake_up() race

Syzkaller robot reported KCSAN tripping over the
ASSERT_EXCLUSIVE_WRITER(p->on_rq) in __block_task().

The report noted that both pick_next_task_fair() and try_to_wake_up()
were concurrently trying to write to the same p->on_rq, violating the
assertion -- even though both paths hold rq->__lock.

The logical consequence is that both code paths end up holding a
different rq->__lock. And looking through ttwu(), this is possible
when the __block_task() 'p->on_rq = 0' store is visible to the ttwu()
'p->on_rq' load, which then assumes the task is not queued and
continues to migrate it.

Rearrange things such that __block_task() releases @p with the store
and no code thereafter will use @p again.

Fixes: 152e11f ("sched/fair: Implement delayed dequeue")
Reported-by: [email protected]
Reported-by: Kent Overstreet <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Tested-by: Marco Elver <[email protected]>
Link: https://lkml.kernel.org/r/[email protected]

Author: Peter Zijlstra

kernel/sched/fair.c

@@ -5625,8 +5625,9 @@ pick_next_entity(struct rq *rq, struct cfs_rq *cfs_rq)
 	struct sched_entity *se = pick_eevdf(cfs_rq);
 	if (se->sched_delayed) {
 		dequeue_entities(rq, se, DEQUEUE_SLEEP | DEQUEUE_DELAYED);
-		SCHED_WARN_ON(se->sched_delayed);
-		SCHED_WARN_ON(se->on_rq);
+		/*
+		 * Must not reference @se again, see __block_task().
+		 */
 		return NULL;
 	}
 	return se;
@@ -7176,7 +7177,11 @@ static int dequeue_entities(struct rq *rq, struct sched_entity *se, int flags)
 		/* Fix-up what dequeue_task_fair() skipped */
 		hrtick_update(rq);
 
-		/* Fix-up what block_task() skipped. */
+		/*
+		 * Fix-up what block_task() skipped.
+		 *
+		 * Must be last, @p might not be valid after this.
+		 */
 		__block_task(rq, p);
 	}
 
@@ -7193,12 +7198,14 @@ static bool dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags)
 	if (!(p->se.sched_delayed && (task_on_rq_migrating(p) || (flags & DEQUEUE_SAVE))))
 		util_est_dequeue(&rq->cfs, p);
 
-	if (dequeue_entities(rq, &p->se, flags) < 0) {
-		util_est_update(&rq->cfs, p, DEQUEUE_SLEEP);
+	util_est_update(&rq->cfs, p, flags & DEQUEUE_SLEEP);
+	if (dequeue_entities(rq, &p->se, flags) < 0)
 		return false;
-	}
 
-	util_est_update(&rq->cfs, p, flags & DEQUEUE_SLEEP);
+	/*
+	 * Must not reference @p after dequeue_entities(DEQUEUE_DELAYED).
+	 */
+
 	hrtick_update(rq);
 	return true;
 }

kernel/sched/sched.h

@@ -2769,15 +2769,45 @@ static inline void sub_nr_running(struct rq *rq, unsigned count)
 
 static inline void __block_task(struct rq *rq, struct task_struct *p)
 {
-	WRITE_ONCE(p->on_rq, 0);
-	ASSERT_EXCLUSIVE_WRITER(p->on_rq);
 	if (p->sched_contributes_to_load)
 		rq->nr_uninterruptible++;
 
 	if (p->in_iowait) {
 		atomic_inc(&rq->nr_iowait);
 		delayacct_blkio_start();
 	}
+
+	ASSERT_EXCLUSIVE_WRITER(p->on_rq);
+
+	/*
+	 * The moment this write goes through, ttwu() can swoop in and migrate
+	 * this task, rendering our rq->__lock ineffective.
+	 *
+	 * __schedule()				try_to_wake_up()
+	 *   LOCK rq->__lock			  LOCK p->pi_lock
+	 *   pick_next_task()
+	 *     pick_next_task_fair()
+	 *       pick_next_entity()
+	 *         dequeue_entities()
+	 *           __block_task()
+	 *             RELEASE p->on_rq = 0	  if (p->on_rq && ...)
+	 *					    break;
+	 *
+	 *					  ACQUIRE (after ctrl-dep)
+	 *
+	 *					  cpu = select_task_rq();
+	 *					  set_task_cpu(p, cpu);
+	 *					  ttwu_queue()
+	 *					    ttwu_do_activate()
+	 *					      LOCK rq->__lock
+	 *					      activate_task()
+	 *					        STORE p->on_rq = 1
+	 *   UNLOCK rq->__lock
+	 *
+	 * Callers must ensure to not reference @p after this -- we no longer
+	 * own it.
+	 */
+	smp_store_release(&p->on_rq, 0);
 }
 
 extern void activate_task(struct rq *rq, struct task_struct *p, int flags);