In the words of the submitter:

---------
Make callers of namei() responsible for releasing references or locks
instead of having the underlying filesystems do it.  This eliminates
redundancy in all terminal filesystems and makes it possible for stacked
transport layers such as umapfs or nullfs to operate correctly.

Quality testing was done with testvn, and lat_fs from the lmbench suite.

Some NFS client testing courtesy of Patrik Kudo.

vop_mknod and vop_symlink still release the returned vpp.  vop_rename
still releases 4 vnode arguments before it returns.  These remaining cases
will be corrected in the next set of patches.
---------

Submitted by:	Michael Hancock <michaelh@cet.co.jp>

1 files changed, 11 insertions, 26 deletions

diff --git a/sys/miscfs/devfs/devfs_vnops.c b/sys/miscfs/devfs/devfs_vnops.c
index 6c9bae7..c35d63f 100644
--- a/sys/miscfs/devfs/devfs_vnops.c
+++ b/sys/miscfs/devfs/devfs_vnops.c
@@ -24,7 +24,7 @@
  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  * SUCH DAMAGE.
  * 
- *	$Id: devfs_vnops.c,v 1.53 1998/03/26 20:52:12 phk Exp $
+ *	$Id: devfs_vnops.c,v 1.54 1998/04/19 23:32:17 julian Exp $
  */
 
 
@@ -320,7 +320,6 @@ devfs_create(struct vop_mknod_args  *ap)
         } */
 {
 DBPRINT(("create\n"));
-	vput(ap->a_dvp);
         return EINVAL;
 }
 
@@ -773,18 +772,12 @@ DBPRINT(("remove\n"));
 	 * are the end of the path. Get pointers to all our
 	 * devfs structures.
 	 */
-	if ( error = devfs_vntodn(dvp,&tdp)) {
+	if (error = devfs_vntodn(dvp, &tdp)) {
 abortit:
 		VOP_ABORTOP(dvp, cnp); 
-		if (dvp == vp) /* eh? */
-			vrele(dvp);
-		else
-			vput(dvp);
-		if (vp)
-			vput(vp);
 		return (error);
 	}
-	if ( error = devfs_vntodn(vp,&tp)) goto abortit;
+	if (error = devfs_vntodn(vp, &tp)) goto abortit;
 	/*
 	 * Assuming we are atomic, dev_lookup left this for us
 	 */
@@ -846,8 +839,6 @@ abortit:
 	}
 	dev_free_name(tnp);
 	tp = NULL;
-	vput(vp);
-	vput(dvp);
 	return (error);
 }
 
@@ -915,7 +906,6 @@ abortit:
 			fp,
 			&tnp);
 out:
-	vput(tdvp);
 	return (error);
 
 }
@@ -1186,9 +1176,7 @@ devfs_rmdir(struct vop_rmdir_args *ap)
         } */ 
 {
 DBPRINT(("rmdir\n"));
-	vput(ap->a_dvp);
-	vput(ap->a_vp);
-	return 0;
+	return (0);
 }
 #endif
 
@@ -1202,30 +1190,27 @@ devfs_symlink(struct vop_symlink_args *ap)
                 char *a_target;
         } */
 {
-	int err;
+	struct vnode *vp;
+	int error;
 	dn_p dnp;
 	union typeinfo by;
 	devnm_p nm_p;
-	struct vnode *vp;
 
 DBPRINT(("symlink\n"));
-	if(err = devfs_vntodn(ap->a_dvp,&dnp)) {
-		vput(ap->a_dvp);
-		return err;
+	if(error = devfs_vntodn(ap->a_dvp, &dnp)) {
+		return (error);
 	}
 		
 	by.Slnk.name = ap->a_target;
 	by.Slnk.namelen = strlen(ap->a_target);
-	dev_add_entry(	ap->a_cnp->cn_nameptr, dnp, DEV_SLNK, &by,
+	dev_add_entry(ap->a_cnp->cn_nameptr, dnp, DEV_SLNK, &by,
 		NULL, NULL, &nm_p);
-	if(err = devfs_dntovn(nm_p->dnp,&vp) ) {
-		vput(ap->a_dvp);
-		return err;
+	if(error = devfs_dntovn(nm_p->dnp, &vp)) {
+		return (error);
 	}
 	VOP_SETATTR(vp, ap->a_vap, ap->a_cnp->cn_cred, ap->a_cnp->cn_proc);
 	*ap->a_vpp = NULL;
 	vput(vp);
-	vput(ap->a_dvp);
 	return 0;
 }