epoll: use rwlock in order to reduce ep_poll_callback() contention

The goal of this patch is to reduce contention of ep_poll_callback()
which can be called concurrently from different CPUs in case of high
events rates and many fds per epoll. Problem can be very well
reproduced by generating events (write to pipe or eventfd) from many
threads, while consumer thread does polling. In other words this patch
increases the bandwidth of events which can be delivered from sources to
the poller by adding poll items in a lockless way to the list.

The main change is in replacement of the spinlock with a rwlock, which
is taken on read in ep_poll_callback(), and then by adding poll items to
the tail of the list using xchg atomic instruction. Write lock is taken
everywhere else in order to stop list modifications and guarantee that
list updates are fully completed (I assume that write side of a rwlock
does not starve, it seems qrwlock implementation has these guarantees).

The following are some microbenchmark results based on the test [1]
which starts threads which generate N events each. The test ends when
all events are successfully fetched by the poller thread:

spinlock
========

threads events/ms run-time ms
8 6402 12495
16 7045 22709
32 7395 43268

rwlock + xchg
=============

threads events/ms run-time ms
8 10038 7969
16 12178 13138
32 13223 24199

According to the results bandwidth of delivered events is significantly
increased, thus execution time is reduced.

This patch was tested with different sort of microbenchmarks and
artificial delays (e.g. "udelay(get_random_int() & 0xff)") introduced
in kernel on paths where items are added to lists.

[1] https://github.com/rouming/test-tools/blob/master/stress-epoll.c

Link: http://lkml.kernel.org/r/[e-mail]
Signed-off-by: Roman Penyaev <[e-mail]>
Cc: Davidlohr Bueso <[e-mail]>
Cc: Jason Baron <[e-mail]>
Cc: Al Viro <[e-mail]>
Cc: "Paul E. McKenney" <[e-mail]>
Signed-off-by: Andrew Morton <[e-mail]>
Signed-off-by: Linus Torvalds <[e-mail]>

fs/eventpoll.c

···
  *
  * 1) epmutex (mutex)
  * 2) ep->mtx (mutex)
- * 3) ep->wq.lock (spinlock)
+ * 3) ep->lock (rwlock)
  *
  * The acquire order is the one listed above, from 1 to 3.
- * We need a spinlock (ep->wq.lock) because we manipulate objects
+ * We need a rwlock (ep->lock) because we manipulate objects
  * from inside the poll callback, that might be triggered from
  * a wake_up() that in turn might be called from IRQ context.
  * So we can't sleep inside the poll callback and hence we need
···
  * of epoll file descriptors, we use the current recursion depth as
  * the lockdep subkey.
  * It is possible to drop the "ep->mtx" and to use the global
- * mutex "epmutex" (together with "ep->wq.lock") to have it working,
+ * mutex "epmutex" (together with "ep->lock") to have it working,
  * but having "ep->mtx" will make the interface more scalable.
  * Events that require holding "epmutex" are very rare, while for
  * normal operations the epoll private "ep->mtx" will guarantee
···
  * This structure is stored inside the "private_data" member of the file
  * structure and represents the main data structure for the eventpoll
  * interface.
- *
- * Access to it is protected by the lock inside wq.
  */
 struct eventpoll {
 	/*
···
 	/* List of ready file descriptors */
 	struct list_head rdllist;
 
+	/* Lock which protects rdllist and ovflist */
+	rwlock_t lock;
+
 	/* RB tree root used to store monitored fd structs */
 	struct rb_root_cached rbr;
 
 	/*
 	 * This is a single linked list that chains all the "struct epitem" that
 	 * happened while transferring ready events to userspace w/out
-	 * holding ->wq.lock.
+	 * holding ->lock.
 	 */
 	struct epitem *ovflist;
 
···
 	 * because we want the "sproc" callback to be able to do it
 	 * in a lockless way.
 	 */
-	spin_lock_irq(&ep->wq.lock);
+	write_lock_irq(&ep->lock);
 	list_splice_init(&ep->rdllist, &txlist);
 	WRITE_ONCE(ep->ovflist, NULL);
-	spin_unlock_irq(&ep->wq.lock);
+	write_unlock_irq(&ep->lock);
 
 	/*
 	 * Now call the callback function.
 	 */
 	res = (*sproc)(ep, &txlist, priv);
 
-	spin_lock_irq(&ep->wq.lock);
+	write_lock_irq(&ep->lock);
 	/*
 	 * During the time we spent inside the "sproc" callback, some
 	 * other events might have been queued by the poll callback.
···
 		 * the ->poll() wait list (delayed after we release the lock).
 		 */
 		if (waitqueue_active(&ep->wq))
-			wake_up_locked(&ep->wq);
+			wake_up(&ep->wq);
 		if (waitqueue_active(&ep->poll_wait))
 			pwake++;
 	}
-	spin_unlock_irq(&ep->wq.lock);
+	write_unlock_irq(&ep->lock);
 
 	if (!ep_locked)
 		mutex_unlock(&ep->mtx);
···
 
 	rb_erase_cached(&epi->rbn, &ep->rbr);
 
-	spin_lock_irq(&ep->wq.lock);
+	write_lock_irq(&ep->lock);
 	if (ep_is_linked(epi))
 		list_del_init(&epi->rdllink);
-	spin_unlock_irq(&ep->wq.lock);
+	write_unlock_irq(&ep->lock);
 
 	wakeup_source_unregister(ep_wakeup_source(epi));
 	/*
···
 	 * Walks through the whole tree by freeing each "struct epitem". At this
 	 * point we are sure no poll callbacks will be lingering around, and also by
 	 * holding "epmutex" we can be sure that no file cleanup code will hit
-	 * us during this operation. So we can avoid the lock on "ep->wq.lock".
+	 * us during this operation. So we can avoid the lock on "ep->lock".
 	 * We do not need to lock ep->mtx, either, we only do it to prevent
 	 * a lockdep warning.
 	 */
···
 		goto free_uid;
 
 	mutex_init(&ep->mtx);
+	rwlock_init(&ep->lock);
 	init_waitqueue_head(&ep->wq);
 	init_waitqueue_head(&ep->poll_wait);
 	INIT_LIST_HEAD(&ep->rdllist);
···
 }
 #endif /* CONFIG_CHECKPOINT_RESTORE */
 
+/**
+ * Adds a new entry to the tail of the list in a lockless way, i.e.
+ * multiple CPUs are allowed to call this function concurrently.
+ *
+ * Beware: it is necessary to prevent any other modifications of the
+ *         existing list until all changes are completed, in other words
+ *         concurrent list_add_tail_lockless() calls should be protected
+ *         with a read lock, where write lock acts as a barrier which
+ *         makes sure all list_add_tail_lockless() calls are fully
+ *         completed.
+ *
+ *        Also an element can be locklessly added to the list only in one
+ *        direction i.e. either to the tail either to the head, otherwise
+ *        concurrent access will corrupt the list.
+ *
+ * Returns %false if element has been already added to the list, %true
+ * otherwise.
+ */
+static inline bool list_add_tail_lockless(struct list_head *new,
+					  struct list_head *head)
+{
+	struct list_head *prev;
+
+	/*
+	 * This is simple 'new->next = head' operation, but cmpxchg()
+	 * is used in order to detect that same element has been just
+	 * added to the list from another CPU: the winner observes
+	 * new->next == new.
+	 */
+	if (cmpxchg(&new->next, new, head) != new)
+		return false;
+
+	/*
+	 * Initially ->next of a new element must be updated with the head
+	 * (we are inserting to the tail) and only then pointers are atomically
+	 * exchanged.  XCHG guarantees memory ordering, thus ->next should be
+	 * updated before pointers are actually swapped and pointers are
+	 * swapped before prev->next is updated.
+	 */
+
+	prev = xchg(&head->prev, new);
+
+	/*
+	 * It is safe to modify prev->next and new->prev, because a new element
+	 * is added only to the tail and new->next is updated before XCHG.
+	 */
+
+	prev->next = new;
+	new->prev = prev;
+
+	return true;
+}
+
+/**
+ * Chains a new epi entry to the tail of the ep->ovflist in a lockless way,
+ * i.e. multiple CPUs are allowed to call this function concurrently.
+ *
+ * Returns %false if epi element has been already chained, %true otherwise.
+ */
+static inline bool chain_epi_lockless(struct epitem *epi)
+{
+	struct eventpoll *ep = epi->ep;
+
+	/* Check that the same epi has not been just chained from another CPU */
+	if (cmpxchg(&epi->next, EP_UNACTIVE_PTR, NULL) != EP_UNACTIVE_PTR)
+		return false;
+
+	/* Atomically exchange tail */
+	epi->next = xchg(&ep->ovflist, epi);
+
+	return true;
+}
+
 /*
  * This is the callback that is passed to the wait queue wakeup
  * mechanism. It is called by the stored file descriptors when they
  * have events to report.
+ *
+ * This callback takes a read lock in order not to content with concurrent
+ * events from another file descriptors, thus all modifications to ->rdllist
+ * or ->ovflist are lockless.  Read lock is paired with the write lock from
+ * ep_scan_ready_list(), which stops all list modifications and guarantees
+ * that lists state is seen correctly.
+ *
+ * Another thing worth to mention is that ep_poll_callback() can be called
+ * concurrently for the same @epi from different CPUs if poll table was inited
+ * with several wait queues entries.  Plural wakeup from different CPUs of a
+ * single wait queue is serialized by wq.lock, but the case when multiple wait
+ * queues are used should be detected accordingly.  This is detected using
+ * cmpxchg() operation.
  */
 static int ep_poll_callback(wait_queue_entry_t *wait, unsigned mode, int sync, void *key)
 {
 	int pwake = 0;
-	unsigned long flags;
 	struct epitem *epi = ep_item_from_wait(wait);
 	struct eventpoll *ep = epi->ep;
 	__poll_t pollflags = key_to_poll(key);
+	unsigned long flags;
 	int ewake = 0;
 
-	spin_lock_irqsave(&ep->wq.lock, flags);
+	read_lock_irqsave(&ep->lock, flags);
 
 	ep_set_busy_poll_napi_id(epi);
 
···
 	 * chained in ep->ovflist and requeued later on.
 	 */
 	if (READ_ONCE(ep->ovflist) != EP_UNACTIVE_PTR) {
-		if (epi->next == EP_UNACTIVE_PTR) {
-			epi->next = READ_ONCE(ep->ovflist);
-			WRITE_ONCE(ep->ovflist, epi);
+		if (epi->next == EP_UNACTIVE_PTR &&
+		    chain_epi_lockless(epi))
 			ep_pm_stay_awake_rcu(epi);
-		}
 		goto out_unlock;
 	}
 
 	/* If this file is already in the ready list we exit soon */
-	if (!ep_is_linked(epi)) {
-		list_add_tail(&epi->rdllink, &ep->rdllist);
+	if (!ep_is_linked(epi) &&
+	    list_add_tail_lockless(&epi->rdllink, &ep->rdllist)) {
 		ep_pm_stay_awake_rcu(epi);
 	}
 
···
 				break;
 			}
 		}
-		wake_up_locked(&ep->wq);
+		wake_up(&ep->wq);
 	}
 	if (waitqueue_active(&ep->poll_wait))
 		pwake++;
 
 out_unlock:
-	spin_unlock_irqrestore(&ep->wq.lock, flags);
+	read_unlock_irqrestore(&ep->lock, flags);
 
 	/* We have to call this outside the lock */
 	if (pwake)
···
 		goto error_remove_epi;
 
 	/* We have to drop the new item inside our item list to keep track of it */
-	spin_lock_irq(&ep->wq.lock);
+	write_lock_irq(&ep->lock);
 
 	/* record NAPI ID of new item if present */
 	ep_set_busy_poll_napi_id(epi);
···
 
 		/* Notify waiting tasks that events are available */
 		if (waitqueue_active(&ep->wq))
-			wake_up_locked(&ep->wq);
+			wake_up(&ep->wq);
 		if (waitqueue_active(&ep->poll_wait))
 			pwake++;
 	}
 
-	spin_unlock_irq(&ep->wq.lock);
+	write_unlock_irq(&ep->lock);
 
 	atomic_long_inc(&ep->user->epoll_watches);
 
···
 	 * list, since that is used/cleaned only inside a section bound by "mtx".
 	 * And ep_insert() is called with "mtx" held.
 	 */
-	spin_lock_irq(&ep->wq.lock);
+	write_lock_irq(&ep->lock);
 	if (ep_is_linked(epi))
 		list_del_init(&epi->rdllink);
-	spin_unlock_irq(&ep->wq.lock);
+	write_unlock_irq(&ep->lock);
 
 	wakeup_source_unregister(ep_wakeup_source(epi));
 
···
 	 * 1) Flush epi changes above to other CPUs.  This ensures
 	 *    we do not miss events from ep_poll_callback if an
 	 *    event occurs immediately after we call f_op->poll().
-	 *    We need this because we did not take ep->wq.lock while
+	 *    We need this because we did not take ep->lock while
 	 *    changing epi above (but ep_poll_callback does take
-	 *    ep->wq.lock).
+	 *    ep->lock).
 	 *
 	 * 2) We also need to ensure we do not miss _past_ events
 	 *    when calling f_op->poll().  This barrier also
···
 	 * list, push it inside.
 	 */
 	if (ep_item_poll(epi, &pt, 1)) {
-		spin_lock_irq(&ep->wq.lock);
+		write_lock_irq(&ep->lock);
 		if (!ep_is_linked(epi)) {
 			list_add_tail(&epi->rdllink, &ep->rdllist);
 			ep_pm_stay_awake(epi);
 
 			/* Notify waiting tasks that events are available */
 			if (waitqueue_active(&ep->wq))
-				wake_up_locked(&ep->wq);
+				wake_up(&ep->wq);
 			if (waitqueue_active(&ep->poll_wait))
 				pwake++;
 		}
-		spin_unlock_irq(&ep->wq.lock);
+		write_unlock_irq(&ep->lock);
 	}
 
 	/* We have to call this outside the lock */
···
 		 */
 		timed_out = 1;
 
-		spin_lock_irq(&ep->wq.lock);
+		write_lock_irq(&ep->lock);
 		eavail = ep_events_available(ep);
-		spin_unlock_irq(&ep->wq.lock);
+		write_unlock_irq(&ep->lock);
 
 		goto send_events;
 	}