From: Kevin Corry <kevcorry@us.ibm.com>

Some testing of DM multipath has turned up a problem with the DEVICE_WAIT
command.  In the tests, while performing a DEVICE_WAIT on a multipath device,
the command sometimes returns immediately, even though the event-number is
correct and no path-failure has occurred to trigger an event.  The problem was
tracked down to the call to schedule() in dev_wait(), which would return even
though it was not woken up by a DM table event.

This patch moves the responsibility for waiting from the ioctl interface into
the core driver, and uses wait_event_interruptible() instead of relying on
wait-queues and schedule().

Signed-off-by:  Andrew Morton <akpm@osdl.org>
---

 25-akpm/drivers/md/dm-ioctl.c |    9 +++------
 25-akpm/drivers/md/dm.c       |   37 +++++++------------------------------
 25-akpm/drivers/md/dm.h       |    4 +---
 3 files changed, 11 insertions(+), 39 deletions(-)

diff -puN drivers/md/dm.c~dm-ioctl-replace-dm__wait_queue-with-dm_wait_event drivers/md/dm.c
--- 25/drivers/md/dm.c~dm-ioctl-replace-dm__wait_queue-with-dm_wait_event	Wed May 26 15:42:42 2004
+++ 25-akpm/drivers/md/dm.c	Wed May 26 15:42:42 2004
@@ -80,7 +80,7 @@ struct mapped_device {
 	/*
 	 * Event handling.
 	 */
-	uint32_t event_nr;
+	atomic_t event_nr;
 	wait_queue_head_t eventq;
 
 	/*
@@ -700,6 +700,7 @@ static struct mapped_device *alloc_dev(u
 	init_rwsem(&md->lock);
 	rwlock_init(&md->map_lock);
 	atomic_set(&md->holders, 1);
+	atomic_set(&md->event_nr, 0);
 
 	md->queue = blk_alloc_queue(GFP_KERNEL);
 	if (!md->queue)
@@ -770,10 +771,8 @@ static void event_callback(void *context
 {
 	struct mapped_device *md = (struct mapped_device *) context;
 
-	down_write(&md->lock);
-	md->event_nr++;
+	atomic_inc(&md->event_nr);;
 	wake_up(&md->eventq);
-	up_write(&md->lock);
 }
 
 static void __set_size(struct gendisk *disk, sector_t size)
@@ -1071,35 +1070,13 @@ int dm_resume(struct mapped_device *md)
  *---------------------------------------------------------------*/
 uint32_t dm_get_event_nr(struct mapped_device *md)
 {
-	uint32_t r;
-
-	down_read(&md->lock);
-	r = md->event_nr;
-	up_read(&md->lock);
-
-	return r;
-}
-
-int dm_add_wait_queue(struct mapped_device *md, wait_queue_t *wq,
-		      uint32_t event_nr)
-{
-	down_write(&md->lock);
-	if (event_nr != md->event_nr) {
-		up_write(&md->lock);
-		return 1;
-	}
-
-	add_wait_queue(&md->eventq, wq);
-	up_write(&md->lock);
-
-	return 0;
+	return atomic_read(&md->event_nr);
 }
 
-void dm_remove_wait_queue(struct mapped_device *md, wait_queue_t *wq)
+int dm_wait_event(struct mapped_device *md, int event_nr)
 {
-	down_write(&md->lock);
-	remove_wait_queue(&md->eventq, wq);
-	up_write(&md->lock);
+	return wait_event_interruptible(md->eventq,
+			(event_nr != atomic_read(&md->event_nr)));
 }
 
 /*
diff -puN drivers/md/dm.h~dm-ioctl-replace-dm__wait_queue-with-dm_wait_event drivers/md/dm.h
--- 25/drivers/md/dm.h~dm-ioctl-replace-dm__wait_queue-with-dm_wait_event	Wed May 26 15:42:42 2004
+++ 25-akpm/drivers/md/dm.h	Wed May 26 15:42:42 2004
@@ -81,9 +81,7 @@ struct dm_table *dm_get_table(struct map
  * Event functions.
  */
 uint32_t dm_get_event_nr(struct mapped_device *md);
-int dm_add_wait_queue(struct mapped_device *md, wait_queue_t *wq,
-		      uint32_t event_nr);
-void dm_remove_wait_queue(struct mapped_device *md, wait_queue_t *wq);
+int dm_wait_event(struct mapped_device *md, int event_nr);
 
 /*
  * Info functions.
diff -puN drivers/md/dm-ioctl.c~dm-ioctl-replace-dm__wait_queue-with-dm_wait_event drivers/md/dm-ioctl.c
--- 25/drivers/md/dm-ioctl.c~dm-ioctl-replace-dm__wait_queue-with-dm_wait_event	Wed May 26 15:42:42 2004
+++ 25-akpm/drivers/md/dm-ioctl.c	Wed May 26 15:42:42 2004
@@ -850,7 +850,6 @@ static int dev_wait(struct dm_ioctl *par
 	int r;
 	struct mapped_device *md;
 	struct dm_table *table;
-	DECLARE_WAITQUEUE(wq, current);
 
 	md = find_device(param);
 	if (!md)
@@ -859,12 +858,10 @@ static int dev_wait(struct dm_ioctl *par
 	/*
 	 * Wait for a notification event
 	 */
-	set_current_state(TASK_INTERRUPTIBLE);
-	if (!dm_add_wait_queue(md, &wq, param->event_nr)) {
-		schedule();
-		dm_remove_wait_queue(md, &wq);
+	if (dm_wait_event(md, param->event_nr)) {
+		r = -ERESTARTSYS;
+		goto out;
 	}
- 	set_current_state(TASK_RUNNING);
 
 	/*
 	 * The userland program is going to want to know what
_