Index: sys/fs/autofs/autofs_vnops.c
===================================================================
--- sys/fs/autofs/autofs_vnops.c
+++ sys/fs/autofs/autofs_vnops.c
@@ -214,7 +214,7 @@
 	struct autofs_mount *amp;
 	struct autofs_node *anp, *child;
 	struct componentname *cnp;
-	int error, lock_flags;
+	int error;
 
 	dvp = ap->a_dvp;
 	vpp = ap->a_vpp;
@@ -257,23 +257,17 @@
 			return (error);
 
 		if (newvp != NULL) {
-			error = VOP_LOOKUP(newvp, ap->a_vpp, ap->a_cnp);
-
 			/*
-			 * Instead of figuring out whether our vnode should
-			 * be locked or not given the error and cnp flags,
-			 * just "copy" the lock status from vnode returned
-			 * by mounted filesystem's VOP_LOOKUP().  Get rid
-			 * of that new vnode afterwards.
+			 * The target filesystem got automounted.  Make
+			 * the lookup(9) go around with a new parent vnode
+			 * and the same path component.
 			 */
-			lock_flags = VOP_ISLOCKED(newvp);
-			if (lock_flags == 0) {
-				VOP_UNLOCK(dvp, 0);
-				vrele(newvp);
-			} else {
-				vput(newvp);
-			}
-			return (error);
+			cnp->cn_consume = -(cnp->cn_namelen + 1);
+			vput(newvp);
+			vref(dvp);
+			*vpp = dvp;
+
+			return (0);
 		}
 	}
 
Index: sys/kern/vfs_lookup.c
===================================================================
--- sys/kern/vfs_lookup.c
+++ sys/kern/vfs_lookup.c
@@ -784,7 +784,7 @@
 	 * Take into account any additional components consumed by
 	 * the underlying filesystem.
 	 */
-	if (cnp->cn_consume > 0) {
+	if (cnp->cn_consume != 0) {
 		cnp->cn_nameptr += cnp->cn_consume;
 		ndp->ni_next += cnp->cn_consume;
 		ndp->ni_pathlen -= cnp->cn_consume;