[PATCH 5/6] [-mm] epoll: clean up ep_modify

[PATCH 5/6] [-mm] epoll: clean up ep_modify

[Tony Battersby]

ep_modify() doesn't need to set event.data from within the ep->lock
spinlock as the comment suggests.  The only place event.data is used
is ep_send_events_proc(), and this is protected by ep->mtx instead
of ep->lock.  Also update the comment for mutex_lock() at the top
of ep_scan_ready_list(), which mentions epoll_ctl(EPOLL_CTL_DEL)
but not epoll_ctl(EPOLL_CTL_MOD).

ep_modify() can also use spin_lock_irq() instead of
spin_lock_irqsave().

Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
---

This patch is against the current -mm tree, and might not apply
to 2.6.29.  Since this patch probably won't go into 2.6.29, you may
apply this version of the patch after the other patches in -mm.

--- a/fs/eventpoll.c	2009-02-23 14:02:49.000000000 -0500
+++ b/fs/eventpoll.c	2009-02-23 14:02:55.000000000 -0500
@@ -454,7 +454,7 @@ static int ep_scan_ready_list(struct eve
 
 	/*
 	 * We need to lock this because we could be hit by
-	 * eventpoll_release_file() and epoll_ctl(EPOLL_CTL_DEL).
+	 * eventpoll_release_file() and epoll_ctl().
 	 */
 	mutex_lock(&ep->mtx);
 
@@ -1037,15 +1037,14 @@ static int ep_modify(struct eventpoll *e
 {
 	int pwake = 0;
 	unsigned int revents;
-	unsigned long flags;
 
 	/*
-	 * Set the new event interest mask before calling f_op->poll(), otherwise
-	 * a potential race might occur. In fact if we do this operation inside
-	 * the lock, an event might happen between the f_op->poll() call and the
-	 * new event set registering.
+	 * Set the new event interest mask before calling f_op->poll();
+	 * otherwise we might miss an event that happens between the
+	 * f_op->poll() call and the new event set registering.
 	 */
 	epi->event.events = event->events;
+	epi->event.data = event->data; /* protected by mtx */
 
 	/*
 	 * Get current event bits. We can safely use the file* here because
@@ -1053,16 +1052,12 @@ static int ep_modify(struct eventpoll *e
 	 */
 	revents = epi->ffd.file->f_op->poll(epi->ffd.file, NULL);
 
-	spin_lock_irqsave(&ep->lock, flags);
-
-	/* Copy the data member from inside the lock */
-	epi->event.data = event->data;
-
 	/*
 	 * If the item is "hot" and it is not registered inside the ready
 	 * list, push it inside.
 	 */
 	if (revents & event->events) {
+		spin_lock_irq(&ep->lock);
 		if (!ep_is_linked(&epi->rdllink)) {
 			list_add_tail(&epi->rdllink, &ep->rdllist);
 
@@ -1072,8 +1067,8 @@ static int ep_modify(struct eventpoll *e
 			if (waitqueue_active(&ep->poll_wait))
 				pwake++;
 		}
+		spin_unlock_irq(&ep->lock);
 	}
-	spin_unlock_irqrestore(&ep->lock, flags);
 
 	/* We have to call this outside the lock */
 	if (pwake)

Re: [PATCH 5/6] [-mm] epoll: clean up ep_modify

[Davide Libenzi]

On Tue, 24 Feb 2009, Tony Battersby wrote:

> ep_modify() doesn't need to set event.data from within the ep->lock
> spinlock as the comment suggests.  The only place event.data is used
> is ep_send_events_proc(), and this is protected by ep->mtx instead
> of ep->lock.  Also update the comment for mutex_lock() at the top
> of ep_scan_ready_list(), which mentions epoll_ctl(EPOLL_CTL_DEL)
> but not epoll_ctl(EPOLL_CTL_MOD).
> 
> ep_modify() can also use spin_lock_irq() instead of
> spin_lock_irqsave().

Pre-mtx stuff.

Acked-by: Davide Libenzi <davidel@xmailserver.org>


- Davide