Avoid reacquiring a mutex in zfs_zinactive

Commit @40d06e3 removed a scheme to avoid reacquiring a mutex in
zfs_zinactive. It turns out this the scheme is necessary. Reinstate it.

Signed-off-by: Chris Dunlop <chris@onthe.net.au>
Issue openzfs#3225
Closes openzfs#3304

module/zfs/zfs_znode.c

@@ -1099,13 +1099,23 @@ zfs_zinactive(znode_t *zp)
 {
 	zfs_sb_t *zsb = ZTOZSB(zp);
 	uint64_t z_id = zp->z_id;
+	boolean_t drop_mutex = 0;
 
 	ASSERT(zp->z_sa_hdl);
 
 	/*
 	 * Don't allow a zfs_zget() while were trying to release this znode.
+	 *
+	 * Linux allows direct memory reclaim which means that any KM_SLEEP
+	 * allocation may trigger inode eviction.  This can lead to a deadlock
+	 * through the ->shrink_icache_memory()->evict()->zfs_inactive()->
+	 * zfs_zinactive() call path.  To avoid this deadlock the process
+	 * must not reacquire the mutex when it is already holding it.
 	 */
-	ZFS_OBJ_HOLD_ENTER(zsb, z_id);
+	if (!ZFS_OBJ_HOLD_OWNED(zsb, z_id)) {
+		ZFS_OBJ_HOLD_ENTER(zsb, z_id);
+		drop_mutex = 1;
+	}
 
 	mutex_enter(&zp->z_lock);
 
@@ -1116,7 +1126,8 @@ zfs_zinactive(znode_t *zp)
 	if (zp->z_unlinked) {
 		mutex_exit(&zp->z_lock);
 
-		ZFS_OBJ_HOLD_EXIT(zsb, z_id);
+		if (drop_mutex)
+			ZFS_OBJ_HOLD_EXIT(zsb, z_id);
 
 		zfs_rmnode(zp);
 		return;
@@ -1125,7 +1136,8 @@ zfs_zinactive(znode_t *zp)
 	mutex_exit(&zp->z_lock);
 	zfs_znode_dmu_fini(zp);
 
-	ZFS_OBJ_HOLD_EXIT(zsb, z_id);
+	if (drop_mutex)
+		ZFS_OBJ_HOLD_EXIT(zsb, z_id);
 }
 
 static inline int