Revert commit 1ee159f

Fix lock order inversion with zvol_open() as it did not account
for use of zvols as vdevs. The latter use cases resulted in the
lock order inversion deadlocks that involved spa_namespace_lock
and bdev->bd_mutex.

Signed-off-by: Boris Protopopov <boris.protopopov@actifio.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #6065
Issue #6134

module/zfs/zvol.c

@@ -1150,12 +1150,36 @@ static int
 zvol_first_open(zvol_state_t *zv)
 {
 	objset_t *os;
-	int error;
+	int error, locked = 0;
+
+	/*
+	 * In all other cases the spa_namespace_lock is taken before the
+	 * bdev->bd_mutex lock.	 But in this case the Linux __blkdev_get()
+	 * function calls fops->open() with the bdev->bd_mutex lock held.
+	 * This deadlock can be easily observed with zvols used as vdevs.
+	 *
+	 * To avoid a potential lock inversion deadlock we preemptively
+	 * try to take the spa_namespace_lock().  Normally it will not
+	 * be contended and this is safe because spa_open_common() handles
+	 * the case where the caller already holds the spa_namespace_lock.
+	 *
+	 * When it is contended we risk a lock inversion if we were to
+	 * block waiting for the lock.	Luckily, the __blkdev_get()
+	 * function allows us to return -ERESTARTSYS which will result in
+	 * bdev->bd_mutex being dropped, reacquired, and fops->open() being
+	 * called again.  This process can be repeated safely until both
+	 * locks are acquired.
+	 */
+	if (!mutex_owned(&spa_namespace_lock)) {
+		locked = mutex_tryenter(&spa_namespace_lock);
+		if (!locked)
+			return (-SET_ERROR(ERESTARTSYS));
+	}
 
 	/* lie and say we're read-only */
 	error = dmu_objset_own(zv->zv_name, DMU_OST_ZVOL, 1, zv, &os);
 	if (error)
-		return (SET_ERROR(-error));
+		goto out_mutex;
 
 	zv->zv_objset = os;
 
@@ -1166,6 +1190,9 @@ zvol_first_open(zvol_state_t *zv)
 		zv->zv_objset = NULL;
 	}
 
+out_mutex:
+	if (locked)
+		mutex_exit(&spa_namespace_lock);
 	return (SET_ERROR(-error));
 }