1 files changed, 341 insertions, 125 deletions

diff --git a/kernel/futex.c b/kernel/futex.c
index 408cad5e8968..c2565c3dddcd 100644
--- a/kernel/futex.c
+++ b/kernel/futex.c
@@ -197,6 +197,8 @@ struct futex_pi_state {
  * @rt_waiter:		rt_waiter storage for use with requeue_pi
  * @requeue_pi_key:	the requeue_pi target futex key
  * @bitset:		bitset for the optional bitmasked wakeup
+ * @requeue_state:	State field for futex_requeue_pi()
+ * @requeue_wait:	RCU wait for futex_requeue_pi() (RT only)
  *
  * We use this hashed waitqueue, instead of a normal wait_queue_entry_t, so
  * we can wake only the relevant ones (hashed queues may be shared).
@@ -219,6 +221,10 @@ struct futex_q {
 	struct rt_mutex_waiter *rt_waiter;
 	union futex_key *requeue_pi_key;
 	u32 bitset;
+	atomic_t requeue_state;
+#ifdef CONFIG_PREEMPT_RT
+	struct rcuwait requeue_wait;
+#endif
 } __randomize_layout;
 
 static const struct futex_q futex_q_init = {
@@ -1354,7 +1360,7 @@ static int lock_pi_update_atomic(u32 __user *uaddr, u32 uval, u32 newval)
  *  -  1 - acquired the lock;
  *  - <0 - error
  *
- * The hb->lock and futex_key refs shall be held by the caller.
+ * The hb->lock must be held by the caller.
  *
  * @exiting is only set when the return value is -EBUSY. If so, this holds
  * a refcount on the exiting task on return and the caller needs to drop it
@@ -1493,11 +1499,11 @@ static void mark_wake_futex(struct wake_q_head *wake_q, struct futex_q *q)
  */
 static int wake_futex_pi(u32 __user *uaddr, u32 uval, struct futex_pi_state *pi_state)
 {
+	DEFINE_RT_MUTEX_WAKE_Q_HEAD(wqh);
 	u32 curval, newval;
 	struct rt_mutex_waiter *top_waiter;
 	struct task_struct *new_owner;
 	bool postunlock = false;
-	DEFINE_WAKE_Q(wake_q);
 	int ret = 0;
 
 	top_waiter = rt_mutex_top_waiter(&pi_state->pi_mutex);
@@ -1549,14 +1555,14 @@ static int wake_futex_pi(u32 __user *uaddr, u32 uval, struct futex_pi_state *pi_
 		 * not fail.
 		 */
 		pi_state_update_owner(pi_state, new_owner);
-		postunlock = __rt_mutex_futex_unlock(&pi_state->pi_mutex, &wake_q);
+		postunlock = __rt_mutex_futex_unlock(&pi_state->pi_mutex, &wqh);
 	}
 
 out_unlock:
 	raw_spin_unlock_irq(&pi_state->pi_mutex.wait_lock);
 
 	if (postunlock)
-		rt_mutex_postunlock(&wake_q);
+		rt_mutex_postunlock(&wqh);
 
 	return ret;
 }
@@ -1796,6 +1802,158 @@ void requeue_futex(struct futex_q *q, struct futex_hash_bucket *hb1,
 	q->key = *key2;
 }
 
+/*
+ * On PREEMPT_RT, the hash bucket lock is a 'sleeping' spinlock with an
+ * underlying rtmutex. The task which is about to be requeued could have
+ * just woken up (timeout, signal). After the wake up the task has to
+ * acquire hash bucket lock, which is held by the requeue code.  As a task
+ * can only be blocked on _ONE_ rtmutex at a time, the proxy lock blocking
+ * and the hash bucket lock blocking would collide and corrupt state.
+ *
+ * On !PREEMPT_RT this is not a problem and everything could be serialized
+ * on hash bucket lock, but aside of having the benefit of common code,
+ * this allows to avoid doing the requeue when the task is already on the
+ * way out and taking the hash bucket lock of the original uaddr1 when the
+ * requeue has been completed.
+ *
+ * The following state transitions are valid:
+ *
+ * On the waiter side:
+ *   Q_REQUEUE_PI_NONE		-> Q_REQUEUE_PI_IGNORE
+ *   Q_REQUEUE_PI_IN_PROGRESS	-> Q_REQUEUE_PI_WAIT
+ *
+ * On the requeue side:
+ *   Q_REQUEUE_PI_NONE		-> Q_REQUEUE_PI_INPROGRESS
+ *   Q_REQUEUE_PI_IN_PROGRESS	-> Q_REQUEUE_PI_DONE/LOCKED
+ *   Q_REQUEUE_PI_IN_PROGRESS	-> Q_REQUEUE_PI_NONE (requeue failed)
+ *   Q_REQUEUE_PI_WAIT		-> Q_REQUEUE_PI_DONE/LOCKED
+ *   Q_REQUEUE_PI_WAIT		-> Q_REQUEUE_PI_IGNORE (requeue failed)
+ *
+ * The requeue side ignores a waiter with state Q_REQUEUE_PI_IGNORE as this
+ * signals that the waiter is already on the way out. It also means that
+ * the waiter is still on the 'wait' futex, i.e. uaddr1.
+ *
+ * The waiter side signals early wakeup to the requeue side either through
+ * setting state to Q_REQUEUE_PI_IGNORE or to Q_REQUEUE_PI_WAIT depending
+ * on the current state. In case of Q_REQUEUE_PI_IGNORE it can immediately
+ * proceed to take the hash bucket lock of uaddr1. If it set state to WAIT,
+ * which means the wakeup is interleaving with a requeue in progress it has
+ * to wait for the requeue side to change the state. Either to DONE/LOCKED
+ * or to IGNORE. DONE/LOCKED means the waiter q is now on the uaddr2 futex
+ * and either blocked (DONE) or has acquired it (LOCKED). IGNORE is set by
+ * the requeue side when the requeue attempt failed via deadlock detection
+ * and therefore the waiter q is still on the uaddr1 futex.
+ */
+enum {
+	Q_REQUEUE_PI_NONE		=  0,
+	Q_REQUEUE_PI_IGNORE,
+	Q_REQUEUE_PI_IN_PROGRESS,
+	Q_REQUEUE_PI_WAIT,
+	Q_REQUEUE_PI_DONE,
+	Q_REQUEUE_PI_LOCKED,
+};
+
+static inline bool futex_requeue_pi_prepare(struct futex_q *q,
+					    struct futex_pi_state *pi_state)
+{
+	int cur, res, new;
+
+	/*
+	 * Set state to Q_REQUEUE_PI_IN_PROGRESS unless an early wakeup has
+	 * already set Q_REQUEUE_PI_IGNORE to signal that requeue should
+	 * ignore the waiter.
+	 */
+	for (cur = atomic_read(&q->requeue_state);; cur = res) {
+		if (cur == Q_REQUEUE_PI_IGNORE)
+			return false;
+
+		/*
+		 * futex_proxy_trylock_atomic() might have set it to
+		 * IN_PROGRESS and a interleaved early wake to WAIT.
+		 *
+		 * It was considered to have an extra state for that
+		 * trylock, but that would just add more conditionals
+		 * all over the place for a dubious value.
+		 */
+		if (cur != Q_REQUEUE_PI_NONE)
+			break;
+
+		new = Q_REQUEUE_PI_IN_PROGRESS;
+		res = atomic_cmpxchg(&q->requeue_state, cur, new);
+		if (likely(cur == res))
+			break;
+	}
+	q->pi_state = pi_state;
+	return true;
+}
+
+static inline void futex_requeue_pi_complete(struct futex_q *q, int locked)
+{
+	int cur, res, new;
+
+	for (cur = atomic_read(&q->requeue_state);; cur = res) {
+		if (locked >= 0) {
+			/* Requeue succeeded. Set DONE or LOCKED */
+			new = Q_REQUEUE_PI_DONE + locked;
+		} else if (cur == Q_REQUEUE_PI_IN_PROGRESS) {
+			/* Deadlock, no early wakeup interleave */
+			new = Q_REQUEUE_PI_NONE;
+		} else {
+			/* Deadlock, early wakeup interleave. */
+			new = Q_REQUEUE_PI_IGNORE;
+		}
+
+		res = atomic_cmpxchg(&q->requeue_state, cur, new);
+		if (likely(cur == res))
+			break;
+	}
+
+#ifdef CONFIG_PREEMPT_RT
+	/* If the waiter interleaved with the requeue let it know */
+	if (unlikely(cur == Q_REQUEUE_PI_WAIT))
+		rcuwait_wake_up(&q->requeue_wait);
+#endif
+}
+
+static inline int futex_requeue_pi_wakeup_sync(struct futex_q *q)
+{
+	int cur, new, res;
+
+	for (cur = atomic_read(&q->requeue_state);; cur = res) {
+		/* Is requeue done already? */
+		if (cur >= Q_REQUEUE_PI_DONE)
+			break;
+
+		/*
+		 * If not done, then tell the requeue code to either ignore
+		 * the waiter or to wake it up once the requeue is done.
+		 */
+		new = !cur ? Q_REQUEUE_PI_IGNORE : Q_REQUEUE_PI_WAIT;
+		res = atomic_cmpxchg(&q->requeue_state, cur, new);
+		if (likely(cur == res))
+			break;
+	}
+
+	/* If the requeue was in progress, wait for it to complete */
+	if (cur == Q_REQUEUE_PI_IN_PROGRESS) {
+#ifdef CONFIG_PREEMPT_RT
+		rcuwait_wait_event(&q->requeue_wait,
+				   atomic_read(&q->requeue_state) != Q_REQUEUE_PI_WAIT,
+				   TASK_UNINTERRUPTIBLE);
+#else
+		while (atomic_read(&q->requeue_state) == Q_REQUEUE_PI_WAIT)
+			cpu_relax();
+#endif
+	}
+
+	/*
+	 * Requeue is now either prohibited or complete. Reread state
+	 * because during the wait above it might have changed. Nothing
+	 * will modify q->requeue_state after this point.
+	 */
+	return atomic_read(&q->requeue_state);
+}
+
 /**
  * requeue_pi_wake_futex() - Wake a task that acquired the lock during requeue
  * @q:		the futex_q
@@ -1823,6 +1981,8 @@ void requeue_pi_wake_futex(struct futex_q *q, union futex_key *key,
 
 	q->lock_ptr = &hb->lock;
 
+	/* Signal locked state to the waiter */
+	futex_requeue_pi_complete(q, 1);
 	wake_up_state(q->task, TASK_NORMAL);
 }
 
@@ -1882,10 +2042,18 @@ futex_proxy_trylock_atomic(u32 __user *pifutex, struct futex_hash_bucket *hb1,
 	if (!top_waiter)
 		return 0;
 
+	/* Ensure that this is a waiter sitting in futex_wait_requeue_pi() */
+	if (!top_waiter->rt_waiter || top_waiter->pi_state)
+		ret = -EINVAL;
+
 	/* Ensure we requeue to the expected futex. */
 	if (!match_futex(top_waiter->requeue_pi_key, key2))
 		return -EINVAL;
 
+	/* Ensure that this does not race against an early wakeup */
+	if (!futex_requeue_pi_prepare(top_waiter, NULL))
+		return -EAGAIN;
+
 	/*
 	 * Try to take the lock for top_waiter.  Set the FUTEX_WAITERS bit in
 	 * the contended case or if set_waiters is 1.  The pi_state is returned
@@ -1895,8 +2063,22 @@ futex_proxy_trylock_atomic(u32 __user *pifutex, struct futex_hash_bucket *hb1,
 	ret = futex_lock_pi_atomic(pifutex, hb2, key2, ps, top_waiter->task,
 				   exiting, set_waiters);
 	if (ret == 1) {
+		/* Dequeue, wake up and update top_waiter::requeue_state */
 		requeue_pi_wake_futex(top_waiter, key2, hb2);
 		return vpid;
+	} else if (ret < 0) {
+		/* Rewind top_waiter::requeue_state */
+		futex_requeue_pi_complete(top_waiter, ret);
+	} else {
+		/*
+		 * futex_lock_pi_atomic() did not acquire the user space
+		 * futex, but managed to establish the proxy lock and pi
+		 * state. top_waiter::requeue_state cannot be fixed up here
+		 * because the waiter is not enqueued on the rtmutex
+		 * yet. This is handled at the callsite depending on the
+		 * result of rt_mutex_start_proxy_lock() which is
+		 * guaranteed to be reached with this function returning 0.
+		 */
 	}
 	return ret;
 }
@@ -1956,15 +2138,27 @@ static int futex_requeue(u32 __user *uaddr1, unsigned int flags,
 		 */
 		if (refill_pi_state_cache())
 			return -ENOMEM;
+
 		/*
-		 * requeue_pi must wake as many tasks as it can, up to nr_wake
-		 * + nr_requeue, since it acquires the rt_mutex prior to
-		 * returning to userspace, so as to not leave the rt_mutex with
-		 * waiters and no owner.  However, second and third wake-ups
-		 * cannot be predicted as they involve race conditions with the
-		 * first wake and a fault while looking up the pi_state.  Both
-		 * pthread_cond_signal() and pthread_cond_broadcast() should
-		 * use nr_wake=1.
+		 * futex_requeue() allows the caller to define the number
+		 * of waiters to wake up via the @nr_wake argument. With
+		 * REQUEUE_PI waking up more than one waiter is creating
+		 * more problems than it solves. Waking up a waiter makes
+		 * only sense if the PI futex @uaddr2 is uncontended as
+		 * this allows the requeue code to acquire the futex
+		 * @uaddr2 before waking the waiter. The waiter can then
+		 * return to user space without further action. A secondary
+		 * wakeup would just make the futex_wait_requeue_pi()
+		 * handling more complex because that code would have to
+		 * look up pi_state and do more or less all the handling
+		 * which the requeue code has to do for the to be requeued
+		 * waiters. So restrict the number of waiters to wake to
+		 * one and only wake it up when the PI futex is
+		 * uncontended. Otherwise requeue it and let the unlock of
+		 * the PI futex handle the wakeup.
+		 *
+		 * All REQUEUE_PI users, e.g. pthread_cond_signal() and
+		 * pthread_cond_broadcast() must use nr_wake=1.
 		 */
 		if (nr_wake != 1)
 			return -EINVAL;
@@ -2025,6 +2219,8 @@ retry_private:
 		 * intend to requeue waiters, force setting the FUTEX_WAITERS
 		 * bit.  We force this here where we are able to easily handle
 		 * faults rather in the requeue loop below.
+		 *
+		 * Updates topwaiter::requeue_state if a top waiter exists.
 		 */
 		ret = futex_proxy_trylock_atomic(uaddr2, hb1, hb2, &key1,
 						 &key2, &pi_state,
@@ -2038,6 +2234,27 @@ retry_private:
 		 * vpid of the top waiter task.
 		 * If the lock was not taken, we have pi_state and an initial
 		 * refcount on it. In case of an error we have nothing.
+		 *
+		 * The top waiter's requeue_state is up to date:
+		 *
+		 *  - If the lock was acquired atomically (ret > 0), then
+		 *    the state is Q_REQUEUE_PI_LOCKED. No matter whether
+		 *    the below lookup_pi_state() fails or not requeue_state
+		 *    is correct because that waiter is dequeued and woken
+		 *    up and nothing can hold it up.
+		 *
+		 *  - If the trylock failed with an error (ret < 0) then
+		 *    the state is either Q_REQUEUE_PI_NONE, i.e. "nothing
+		 *    happened", or Q_REQUEUE_PI_IGNORE when there was an
+		 *    interleaved early wakeup.
+		 *
+		 *  - If the trylock did not succeed (ret == 0) then the
+		 *    state is either Q_REQUEUE_PI_IN_PROGRESS or
+		 *    Q_REQUEUE_PI_WAIT if an early wakeup interleaved.
+		 *    This will be cleaned up in the loop below, which
+		 *    cannot fail because futex_proxy_trylock_atomic() did
+		 *    the same sanity checks for requeue_pi as the loop
+		 *    below does.
 		 */
 		if (ret > 0) {
 			WARN_ON(pi_state);
@@ -2063,7 +2280,10 @@ retry_private:
 			/* We hold a reference on the pi state. */
 			break;
 
-			/* If the above failed, then pi_state is NULL */
+		/*
+		 * If the above failed, then pi_state is NULL and
+		 * waiter::requeue_state is correct.
+		 */
 		case -EFAULT:
 			double_unlock_hb(hb1, hb2);
 			hb_waiters_dec(hb2);
@@ -2115,18 +2335,17 @@ retry_private:
 			break;
 		}
 
-		/*
-		 * Wake nr_wake waiters.  For requeue_pi, if we acquired the
-		 * lock, we already woke the top_waiter.  If not, it will be
-		 * woken by futex_unlock_pi().
-		 */
-		if (++task_count <= nr_wake && !requeue_pi) {
-			mark_wake_futex(&wake_q, this);
+		/* Plain futexes just wake or requeue and are done */
+		if (!requeue_pi) {
+			if (++task_count <= nr_wake)
+				mark_wake_futex(&wake_q, this);
+			else
+				requeue_futex(this, hb1, hb2, &key2);
 			continue;
 		}
 
 		/* Ensure we requeue to the expected futex for requeue_pi. */
-		if (requeue_pi && !match_futex(this->requeue_pi_key, &key2)) {
+		if (!match_futex(this->requeue_pi_key, &key2)) {
 			ret = -EINVAL;
 			break;
 		}
@@ -2134,48 +2353,61 @@ retry_private:
 		/*
 		 * Requeue nr_requeue waiters and possibly one more in the case
 		 * of requeue_pi if we couldn't acquire the lock atomically.
+		 *
+		 * Prepare the waiter to take the rt_mutex. Take a refcount
+		 * on the pi_state and store the pointer in the futex_q
+		 * object of the waiter.
 		 */
-		if (requeue_pi) {
+		get_pi_state(pi_state);
+
+		/* Don't requeue when the waiter is already on the way out. */
+		if (!futex_requeue_pi_prepare(this, pi_state)) {
 			/*
-			 * Prepare the waiter to take the rt_mutex. Take a
-			 * refcount on the pi_state and store the pointer in
-			 * the futex_q object of the waiter.
+			 * Early woken waiter signaled that it is on the
+			 * way out. Drop the pi_state reference and try the
+			 * next waiter. @this->pi_state is still NULL.
 			 */
-			get_pi_state(pi_state);
-			this->pi_state = pi_state;
-			ret = rt_mutex_start_proxy_lock(&pi_state->pi_mutex,
-							this->rt_waiter,
-							this->task);
-			if (ret == 1) {
-				/*
-				 * We got the lock. We do neither drop the
-				 * refcount on pi_state nor clear
-				 * this->pi_state because the waiter needs the
-				 * pi_state for cleaning up the user space
-				 * value. It will drop the refcount after
-				 * doing so.
-				 */
-				requeue_pi_wake_futex(this, &key2, hb2);
-				continue;
-			} else if (ret) {
-				/*
-				 * rt_mutex_start_proxy_lock() detected a
-				 * potential deadlock when we tried to queue
-				 * that waiter. Drop the pi_state reference
-				 * which we took above and remove the pointer
-				 * to the state from the waiters futex_q
-				 * object.
-				 */
-				this->pi_state = NULL;
-				put_pi_state(pi_state);
-				/*
-				 * We stop queueing more waiters and let user
-				 * space deal with the mess.
-				 */
-				break;
-			}
+			put_pi_state(pi_state);
+			continue;
+		}
+
+		ret = rt_mutex_start_proxy_lock(&pi_state->pi_mutex,
+						this->rt_waiter,
+						this->task);
+
+		if (ret == 1) {
+			/*
+			 * We got the lock. We do neither drop the refcount
+			 * on pi_state nor clear this->pi_state because the
+			 * waiter needs the pi_state for cleaning up the
+			 * user space value. It will drop the refcount
+			 * after doing so. this::requeue_state is updated
+			 * in the wakeup as well.
+			 */
+			requeue_pi_wake_futex(this, &key2, hb2);
+			task_count++;
+		} else if (!ret) {
+			/* Waiter is queued, move it to hb2 */
+			requeue_futex(this, hb1, hb2, &key2);
+			futex_requeue_pi_complete(this, 0);
+			task_count++;
+		} else {
+			/*
+			 * rt_mutex_start_proxy_lock() detected a potential
+			 * deadlock when we tried to queue that waiter.
+			 * Drop the pi_state reference which we took above
+			 * and remove the pointer to the state from the
+			 * waiters futex_q object.
+			 */
+			this->pi_state = NULL;
+			put_pi_state(pi_state);
+			futex_requeue_pi_complete(this, ret);
+			/*
+			 * We stop queueing more waiters and let user space
+			 * deal with the mess.
+			 */
+			break;
 		}
-		requeue_futex(this, hb1, hb2, &key2);
 	}
 
 	/*
@@ -2617,8 +2849,7 @@ static void futex_wait_queue_me(struct futex_hash_bucket *hb, struct futex_q *q,
  *
  * Setup the futex_q and locate the hash_bucket.  Get the futex value and
  * compare it with the expected value.  Handle atomic faults internally.
- * Return with the hb lock held and a q.key reference on success, and unlocked
- * with no q.key reference on failure.
+ * Return with the hb lock held on success, and unlocked on failure.
  *
  * Return:
  *  -  0 - uaddr contains val and hb has been locked;
@@ -3075,27 +3306,22 @@ pi_faulted:
 }
 
 /**
- * handle_early_requeue_pi_wakeup() - Detect early wakeup on the initial futex
+ * handle_early_requeue_pi_wakeup() - Handle early wakeup on the initial futex
  * @hb:		the hash_bucket futex_q was original enqueued on
  * @q:		the futex_q woken while waiting to be requeued
- * @key2:	the futex_key of the requeue target futex
  * @timeout:	the timeout associated with the wait (NULL if none)
  *
- * Detect if the task was woken on the initial futex as opposed to the requeue
- * target futex.  If so, determine if it was a timeout or a signal that caused
- * the wakeup and return the appropriate error code to the caller.  Must be
- * called with the hb lock held.
+ * Determine the cause for the early wakeup.
  *
  * Return:
- *  -  0 = no early wakeup detected;
- *  - <0 = -ETIMEDOUT or -ERESTARTNOINTR
+ *  -EWOULDBLOCK or -ETIMEDOUT or -ERESTARTNOINTR
  */
 static inline
 int handle_early_requeue_pi_wakeup(struct futex_hash_bucket *hb,
-				   struct futex_q *q, union futex_key *key2,
+				   struct futex_q *q,
 				   struct hrtimer_sleeper *timeout)
 {
-	int ret = 0;
+	int ret;
 
 	/*
 	 * With the hb lock held, we avoid races while we process the wakeup.
@@ -3104,22 +3330,21 @@ int handle_early_requeue_pi_wakeup(struct futex_hash_bucket *hb,
 	 * It can't be requeued from uaddr2 to something else since we don't
 	 * support a PI aware source futex for requeue.
 	 */
-	if (!match_futex(&q->key, key2)) {
-		WARN_ON(q->lock_ptr && (&hb->lock != q->lock_ptr));
-		/*
-		 * We were woken prior to requeue by a timeout or a signal.
-		 * Unqueue the futex_q and determine which it was.
-		 */
-		plist_del(&q->list, &hb->chain);
-		hb_waiters_dec(hb);
+	WARN_ON_ONCE(&hb->lock != q->lock_ptr);
 
-		/* Handle spurious wakeups gracefully */
-		ret = -EWOULDBLOCK;
-		if (timeout && !timeout->task)
-			ret = -ETIMEDOUT;
-		else if (signal_pending(current))
-			ret = -ERESTARTNOINTR;
-	}
+	/*
+	 * We were woken prior to requeue by a timeout or a signal.
+	 * Unqueue the futex_q and determine which it was.
+	 */
+	plist_del(&q->list, &hb->chain);
+	hb_waiters_dec(hb);
+
+	/* Handle spurious wakeups gracefully */
+	ret = -EWOULDBLOCK;
+	if (timeout && !timeout->task)
+		ret = -ETIMEDOUT;
+	else if (signal_pending(current))
+		ret = -ERESTARTNOINTR;
 	return ret;
 }
 
@@ -3172,6 +3397,7 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, unsigned int flags,
 	struct futex_hash_bucket *hb;
 	union futex_key key2 = FUTEX_KEY_INIT;
 	struct futex_q q = futex_q_init;
+	struct rt_mutex *pi_mutex;
 	int res, ret;
 
 	if (!IS_ENABLED(CONFIG_FUTEX_PI))
@@ -3221,32 +3447,22 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, unsigned int flags,
 	/* Queue the futex_q, drop the hb lock, wait for wakeup. */
 	futex_wait_queue_me(hb, &q, to);
 
-	spin_lock(&hb->lock);
-	ret = handle_early_requeue_pi_wakeup(hb, &q, &key2, to);
-	spin_unlock(&hb->lock);
-	if (ret)
-		goto out;
-
-	/*
-	 * In order for us to be here, we know our q.key == key2, and since
-	 * we took the hb->lock above, we also know that futex_requeue() has
-	 * completed and we no longer have to concern ourselves with a wakeup
-	 * race with the atomic proxy lock acquisition by the requeue code. The
-	 * futex_requeue dropped our key1 reference and incremented our key2
-	 * reference count.
-	 */
+	switch (futex_requeue_pi_wakeup_sync(&q)) {
+	case Q_REQUEUE_PI_IGNORE:
+		/* The waiter is still on uaddr1 */
+		spin_lock(&hb->lock);
+		ret = handle_early_requeue_pi_wakeup(hb, &q, to);
+		spin_unlock(&hb->lock);
+		break;
 
-	/*
-	 * Check if the requeue code acquired the second futex for us and do
-	 * any pertinent fixup.
-	 */
-	if (!q.rt_waiter) {
+	case Q_REQUEUE_PI_LOCKED:
+		/* The requeue acquired the lock */
 		if (q.pi_state && (q.pi_state->owner != current)) {
 			spin_lock(q.lock_ptr);
 			ret = fixup_owner(uaddr2, &q, true);
 			/*
-			 * Drop the reference to the pi state which
-			 * the requeue_pi() code acquired for us.
+			 * Drop the reference to the pi state which the
+			 * requeue_pi() code acquired for us.
 			 */
 			put_pi_state(q.pi_state);
 			spin_unlock(q.lock_ptr);
@@ -3256,18 +3472,14 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, unsigned int flags,
 			 */
 			ret = ret < 0 ? ret : 0;
 		}
-	} else {
-		struct rt_mutex *pi_mutex;
+		break;
 
-		/*
-		 * We have been woken up by futex_unlock_pi(), a timeout, or a
-		 * signal.  futex_unlock_pi() will not destroy the lock_ptr nor
-		 * the pi_state.
-		 */
-		WARN_ON(!q.pi_state);
+	case Q_REQUEUE_PI_DONE:
+		/* Requeue completed. Current is 'pi_blocked_on' the rtmutex */
 		pi_mutex = &q.pi_state->pi_mutex;
 		ret = rt_mutex_wait_proxy_lock(pi_mutex, to, &rt_waiter);
 
+		/* Current is not longer pi_blocked_on */
 		spin_lock(q.lock_ptr);
 		if (ret && !rt_mutex_cleanup_proxy_lock(pi_mutex, &rt_waiter))
 			ret = 0;
@@ -3287,17 +3499,21 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, unsigned int flags,
 
 		unqueue_me_pi(&q);
 		spin_unlock(q.lock_ptr);
-	}
 
-	if (ret == -EINTR) {
-		/*
-		 * We've already been requeued, but cannot restart by calling
-		 * futex_lock_pi() directly. We could restart this syscall, but
-		 * it would detect that the user space "val" changed and return
-		 * -EWOULDBLOCK.  Save the overhead of the restart and return
-		 * -EWOULDBLOCK directly.
-		 */
-		ret = -EWOULDBLOCK;
+		if (ret == -EINTR) {
+			/*
+			 * We've already been requeued, but cannot restart
+			 * by calling futex_lock_pi() directly. We could
+			 * restart this syscall, but it would detect that
+			 * the user space "val" changed and return
+			 * -EWOULDBLOCK.  Save the overhead of the restart
+			 * and return -EWOULDBLOCK directly.
+			 */
+			ret = -EWOULDBLOCK;
+		}
+		break;
+	default:
+		BUG();
 	}
 
 out: