summaryrefslogtreecommitdiff
path: root/sys/kern
diff options
context:
space:
mode:
authorConstantine Sapuntzakis <csapuntz@cvs.openbsd.org>1997-11-06 22:46:20 +0000
committerConstantine Sapuntzakis <csapuntz@cvs.openbsd.org>1997-11-06 22:46:20 +0000
commit28ad34a5066792793751e691e5f730e5c6e46d06 (patch)
tree5faabe3b6625e1e67c444627ca1647d816398a10 /sys/kern
parentf6f720773d9c569b71220d37d8945ba6c462352b (diff)
DEBUG now compiles
Diffstat (limited to 'sys/kern')
-rw-r--r--sys/kern/vfs_subr.c19
1 files changed, 5 insertions, 14 deletions
diff --git a/sys/kern/vfs_subr.c b/sys/kern/vfs_subr.c
index 2cf1e492d0b..120b33a76e7 100644
--- a/sys/kern/vfs_subr.c
+++ b/sys/kern/vfs_subr.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: vfs_subr.c,v 1.13 1997/11/06 05:58:28 csapuntz Exp $ */
+/* $OpenBSD: vfs_subr.c,v 1.14 1997/11/06 22:46:19 csapuntz Exp $ */
/* $NetBSD: vfs_subr.c,v 1.53 1996/04/22 01:39:13 christos Exp $ */
/*
@@ -1063,7 +1063,6 @@ vop_nolock(v)
struct proc *a_p;
} */ *ap = v;
-#ifdef notyet
/*
* This code cannot be used until all the non-locking filesystems
* (notably NFS) are converted to properly lock and release nodes.
@@ -1107,15 +1106,6 @@ vop_nolock(v)
if (flags & LK_INTERLOCK)
vnflags |= LK_INTERLOCK;
return(lockmgr(vp->v_vnlock, vnflags, &vp->v_interlock, ap->a_p));
-#else /* for now */
- /*
- * Since we are not using the lock manager, we must clear
- * the interlock here.
- */
- if (ap->a_flags & LK_INTERLOCK)
- simple_unlock(&ap->a_vp->v_interlock);
- return (0);
-#endif
}
/*
@@ -1805,10 +1795,10 @@ printlockedvnodes()
printf("Locked vnodes\n");
simple_lock(&mountlist_slock);
- for (mp = mountlist.cqh_first; mp != (void *)&mountlist;
+ for (mp = CIRCLEQ_FIRST(&mountlist); mp != (void *)&mountlist;
mp = nmp) {
if (vfs_busy(mp, LK_NOWAIT, &mountlist_slock, p)) {
- nmp = mp->mnt_list.cque_next;
+ nmp = CIRCLEQ_NEXT(mp, mnt_list);
continue;
}
for (vp = mp->mnt_vnodelist.lh_first;
@@ -1816,8 +1806,9 @@ printlockedvnodes()
vp = vp->v_mntvnodes.le_next) {
if (VOP_ISLOCKED(vp))
vprint((char *)0, vp);
+ }
simple_lock(&mountlist_slock);
- nmp = mp->mnt_list.cqe_next;
+ nmp = CIRCLEQ_NEXT(mp, mnt_list);
vfs_unbusy(mp, p);
}
simple_unlock(&mountlist_slock);