md: move revalidate_disk() back outside open_mutex

commit 147e0b6a upstream.

Commit b821eaa5 "md: remove ->changed and related code" moved
revalidate_disk() under open_mutex, and lockdep noticed.

[ INFO: possible circular locking dependency detected ]
2.6.32-mdadm-locking #1
-------------------------------------------------------
mdadm/3640 is trying to acquire lock:
 (&bdev->bd_mutex){+.+.+.}, at: [<ffffffff811acecb>] revalidate_disk+0x5b/0x90

but task is already holding lock:
 (&mddev->open_mutex){+.+...}, at: [<ffffffffa055e07a>] do_md_stop+0x4a/0x4d0 [md_mod]

which lock already depends on the new lock.

It is suitable for 2.6.35.x
Reported-by: Przemyslaw Czarnowski <przemyslaw.hawrylewicz.czarnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

drivers/md/md.c

@@ -4711,7 +4711,7 @@ static int md_set_readonly(mddev_t *mddev, int is_open)
  */
 static int do_md_stop(mddev_t * mddev, int mode, int is_open)
 {
-	int err = 0;
+	int err = 0, revalidate = 0;
 	struct gendisk *disk = mddev->gendisk;
 	mdk_rdev_t *rdev;
 
@@ -4740,7 +4740,7 @@ static int do_md_stop(mddev_t * mddev, int mode, int is_open)
 			}
 
 		set_capacity(disk, 0);
-		revalidate_disk(disk);
+		revalidate = 1;
 
 		if (mddev->ro)
 			mddev->ro = 0;
@@ -4748,6 +4748,8 @@ static int do_md_stop(mddev_t * mddev, int mode, int is_open)
 		err = 0;
 	}
 	mutex_unlock(&mddev->open_mutex);
+	if (revalidate)
+		revalidate_disk(disk);
 	if (err)
 		return err;
 	/*