X-Git-Url: https://git.camperquake.de/gitweb.cgi?a=blobdiff_plain;f=module%2Fzfs%2Fzvol.c;h=f8e32567609ff3cef794c152120011ef355995e8;hb=b01615d5ac86913da1e092d0378bfb8f0e72af30;hp=7a448f1948fa74436713c61d642d816bba206e8a;hpb=6c2856726fbae681649930d9620d9087080e58fc;p=zfs.git

diff --git a/module/zfs/zvol.c b/module/zfs/zvol.c
index 7a448f1..f8e3256 100644
--- a/module/zfs/zvol.c
+++ b/module/zfs/zvol.c
@@ -453,20 +453,20 @@ zvol_replay_err(zvol_state_t *zv, lr_t *lr, boolean_t byteswap)
  * Callback vectors for replaying records.
  * Only TX_WRITE is needed for zvol.
  */
-zil_replay_func_t *zvol_replay_vector[TX_MAX_TYPE] = {
-	(zil_replay_func_t *)zvol_replay_err,	/* no such transaction type */
-	(zil_replay_func_t *)zvol_replay_err,	/* TX_CREATE */
-	(zil_replay_func_t *)zvol_replay_err,	/* TX_MKDIR */
-	(zil_replay_func_t *)zvol_replay_err,	/* TX_MKXATTR */
-	(zil_replay_func_t *)zvol_replay_err,	/* TX_SYMLINK */
-	(zil_replay_func_t *)zvol_replay_err,	/* TX_REMOVE */
-	(zil_replay_func_t *)zvol_replay_err,	/* TX_RMDIR */
-	(zil_replay_func_t *)zvol_replay_err,	/* TX_LINK */
-	(zil_replay_func_t *)zvol_replay_err,	/* TX_RENAME */
-	(zil_replay_func_t *)zvol_replay_write,	/* TX_WRITE */
-	(zil_replay_func_t *)zvol_replay_err,	/* TX_TRUNCATE */
-	(zil_replay_func_t *)zvol_replay_err,	/* TX_SETATTR */
-	(zil_replay_func_t *)zvol_replay_err,	/* TX_ACL */
+zil_replay_func_t zvol_replay_vector[TX_MAX_TYPE] = {
+	(zil_replay_func_t)zvol_replay_err,	/* no such transaction type */
+	(zil_replay_func_t)zvol_replay_err,	/* TX_CREATE */
+	(zil_replay_func_t)zvol_replay_err,	/* TX_MKDIR */
+	(zil_replay_func_t)zvol_replay_err,	/* TX_MKXATTR */
+	(zil_replay_func_t)zvol_replay_err,	/* TX_SYMLINK */
+	(zil_replay_func_t)zvol_replay_err,	/* TX_REMOVE */
+	(zil_replay_func_t)zvol_replay_err,	/* TX_RMDIR */
+	(zil_replay_func_t)zvol_replay_err,	/* TX_LINK */
+	(zil_replay_func_t)zvol_replay_err,	/* TX_RENAME */
+	(zil_replay_func_t)zvol_replay_write,	/* TX_WRITE */
+	(zil_replay_func_t)zvol_replay_err,	/* TX_TRUNCATE */
+	(zil_replay_func_t)zvol_replay_err,	/* TX_SETATTR */
+	(zil_replay_func_t)zvol_replay_err,	/* TX_ACL */
 };
 
 /*
@@ -891,25 +891,49 @@ zvol_first_open(zvol_state_t *zv)
 {
 	objset_t *os;
 	uint64_t volsize;
+	int locked = 0;
 	int error;
 	uint64_t ro;
 
+	/*
+	 * 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.
+	 *
+	 * 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 (-ERESTARTSYS);
+	}
+
 	/* lie and say we're read-only */
 	error = dmu_objset_own(zv->zv_name, DMU_OST_ZVOL, 1, zvol_tag, &os);
 	if (error)
-		return (-error);
+		goto out_mutex;
 
 	error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
 	if (error) {
-	        dmu_objset_disown(os, zvol_tag);
-	        return (-error);
+		dmu_objset_disown(os, zvol_tag);
+		goto out_mutex;
 	}
 
 	zv->zv_objset = os;
 	error = dmu_bonus_hold(os, ZVOL_OBJ, zvol_tag, &zv->zv_dbuf);
 	if (error) {
-	        dmu_objset_disown(os, zvol_tag);
-	        return (-error);
+		dmu_objset_disown(os, zvol_tag);
+		goto out_mutex;
 	}
 
 	set_capacity(zv->zv_disk, volsize >> 9);
@@ -918,13 +942,17 @@ zvol_first_open(zvol_state_t *zv)
 
 	VERIFY(dsl_prop_get_integer(zv->zv_name, "readonly", &ro, NULL) == 0);
 	if (ro || dmu_objset_is_snapshot(os)) {
-                set_disk_ro(zv->zv_disk, 1);
-	        zv->zv_flags |= ZVOL_RDONLY;
+		set_disk_ro(zv->zv_disk, 1);
+		zv->zv_flags |= ZVOL_RDONLY;
 	} else {
-                set_disk_ro(zv->zv_disk, 0);
-	        zv->zv_flags &= ~ZVOL_RDONLY;
+		set_disk_ro(zv->zv_disk, 0);
+		zv->zv_flags &= ~ZVOL_RDONLY;
 	}
 
+out_mutex:
+	if (locked)
+		mutex_exit(&spa_namespace_lock);
+
 	return (-error);
 }