Add new flag PDIRUNLOCK to the component.cn_flags which should be set by

filesystem lookup() routine if it unlocks parent directory. This flag should
be carefully tracked by filesystems if they want to work properly with nullfs
and other stacked filesystems.

VFS takes advantage of this flag to perform symantically correct usage
of vrele() instead of vput() if parent directory already unlocked.

If filesystem fails to track this flag then previous codepath in VFS left
unchanged.

Convert UFS code to set PDIRUNLOCK flag if necessary. Other filesystmes will
be changed after some period of testing.

Reviewed in general by:	mckusick, dillon, adrian
Obtained from:	NetBSD

2 files changed, 25 insertions, 9 deletions

diff --git a/sys/kern/vfs_cache.c b/sys/kern/vfs_cache.c
index 5cc4f1c..976f85e 100644
--- a/sys/kern/vfs_cache.c
+++ b/sys/kern/vfs_cache.c
@@ -458,18 +458,24 @@ vfs_cache_lookup(ap)
 
 	vp = *vpp;
 	vpid = vp->v_id;
+	cnp->cn_flags &= ~PDIRUNLOCK;
 	if (dvp == vp) {   /* lookup on "." */
 		VREF(vp);
 		error = 0;
 	} else if (flags & ISDOTDOT) {
 		VOP_UNLOCK(dvp, 0, p);
+		cnp->cn_flags |= PDIRUNLOCK;
 		error = vget(vp, LK_EXCLUSIVE, p);
-		if (!error && lockparent && (flags & ISLASTCN))
-			error = vn_lock(dvp, LK_EXCLUSIVE, p);
+		if (!error && lockparent && (flags & ISLASTCN)) {
+			if ((error = vn_lock(dvp, LK_EXCLUSIVE, p)) == 0)
+				cnp->cn_flags &= ~PDIRUNLOCK;
+		}
 	} else {
 		error = vget(vp, LK_EXCLUSIVE, p);
-		if (!lockparent || error || !(flags & ISLASTCN))
+		if (!lockparent || error || !(flags & ISLASTCN)) {
 			VOP_UNLOCK(dvp, 0, p);
+			cnp->cn_flags |= PDIRUNLOCK;
+		}
 	}
 	/*
 	 * Check that the capability number did not change
@@ -479,12 +485,17 @@ vfs_cache_lookup(ap)
 		if (vpid == vp->v_id)
 			return (0);
 		vput(vp);
-		if (lockparent && dvp != vp && (flags & ISLASTCN))
+		if (lockparent && dvp != vp && (flags & ISLASTCN)) {
 			VOP_UNLOCK(dvp, 0, p);
+			cnp->cn_flags |= PDIRUNLOCK;
+		}
+	}
+	if (cnp->cn_flags & PDIRUNLOCK) {
+		error = vn_lock(dvp, LK_EXCLUSIVE, p);
+		if (error)
+			return (error);
+		cnp->cn_flags &= ~PDIRUNLOCK;
 	}
-	error = vn_lock(dvp, LK_EXCLUSIVE, p);
-	if (error)
-		return (error);
 	return (VOP_CACHEDLOOKUP(dvp, vpp, cnp));
 }
 
diff --git a/sys/kern/vfs_lookup.c b/sys/kern/vfs_lookup.c
index c9f43d4..4cd4b01 100644
--- a/sys/kern/vfs_lookup.c
+++ b/sys/kern/vfs_lookup.c
@@ -418,6 +418,7 @@ dirloop:
 unionlookup:
 	ndp->ni_dvp = dp;
 	ndp->ni_vp = NULL;
+	cnp->cn_flags &= ~PDIRUNLOCK;
 	ASSERT_VOP_LOCKED(dp, "lookup");
 	if ((error = VOP_LOOKUP(dp, &ndp->ni_vp, cnp)) != 0) {
 		KASSERT(ndp->ni_vp == NULL, ("leaf should be empty"));
@@ -429,7 +430,10 @@ unionlookup:
 		    (dp->v_mount->mnt_flag & MNT_UNION)) {
 			tdp = dp;
 			dp = dp->v_mount->mnt_vnodecovered;
-			vput(tdp);
+			if (cnp->cn_flags & PDIRUNLOCK)
+				vrele(tdp);
+			else
+				vput(tdp);
 			VREF(dp);
 			vn_lock(dp, LK_EXCLUSIVE | LK_RETRY, p);
 			goto unionlookup;
@@ -557,7 +561,8 @@ nextname:
 	return (0);
 
 bad2:
-	if ((cnp->cn_flags & LOCKPARENT) && *ndp->ni_next == '\0')
+	if ((cnp->cn_flags & (LOCKPARENT | PDIRUNLOCK)) == LOCKPARENT &&
+	    *ndp->ni_next == '\0')
 		VOP_UNLOCK(ndp->ni_dvp, 0, p);
 	vrele(ndp->ni_dvp);
 bad: