X-Git-Url: https://git.subgeniuskitty.com/unix-history/.git/blobdiff_plain/1d2d7c6d645a327a2088eb0716a6df6e0ed7c8e0..99afd20ce5420eed51b50d2ea4806d864724cf2c:/usr/src/sys/kern/vfs_subr.c

diff --git a/usr/src/sys/kern/vfs_subr.c b/usr/src/sys/kern/vfs_subr.c
index 176db6b0d0..cc599a5ed1 100644
--- a/usr/src/sys/kern/vfs_subr.c
+++ b/usr/src/sys/kern/vfs_subr.c
@@ -9,7 +9,7 @@
  *
  * %sccs.include.redist.c%
  *
- *	@(#)vfs_subr.c	8.12 (Berkeley) %G%
+ *	@(#)vfs_subr.c	8.25 (Berkeley) %G%
  */
 
 /*
@@ -49,35 +49,44 @@ int	vttoif_tab[9] = {
  * Insq/Remq for the vnode usage lists.
  */
 #define	bufinsvn(bp, dp)	LIST_INSERT_HEAD(dp, bp, b_vnbufs)
-#define	bufremvn(bp) {  \
-	LIST_REMOVE(bp, b_vnbufs); \
-	(bp)->b_vnbufs.le_next = NOLIST; \
+#define	bufremvn(bp) {							\
+	LIST_REMOVE(bp, b_vnbufs);					\
+	(bp)->b_vnbufs.le_next = NOLIST;				\
 }
-
 TAILQ_HEAD(freelst, vnode) vnode_free_list;	/* vnode free list */
 struct mntlist mountlist;			/* mounted filesystem list */
+static struct simplelock mntid_slock;
+struct simplelock mntvnode_slock;
+static struct simplelock spechash_slock;
+static struct simplelock vnode_free_list_slock;
 
 /*
  * Initialize the vnode management data structures.
  */
+void
 vntblinit()
 {
 
+	simple_lock_init(&mntvnode_slock);
+	simple_lock_init(&mntid_slock);
+	simple_lock_init(&spechash_slock);
 	TAILQ_INIT(&vnode_free_list);
-	TAILQ_INIT(&mountlist);
+	simple_lock_init(&vnode_free_list_slock);
+	CIRCLEQ_INIT(&mountlist);
 }
 
 /*
  * Lock a filesystem.
  * Used to prevent access to it while mounting and unmounting.
  */
+int
 vfs_lock(mp)
 	register struct mount *mp;
 {
 
-	while(mp->mnt_flag & MNT_MLOCK) {
+	while (mp->mnt_flag & MNT_MLOCK) {
 		mp->mnt_flag |= MNT_MWAIT;
-		sleep((caddr_t)mp, PVFS);
+		tsleep((caddr_t)mp, PVFS, "vfslock", 0);
 	}
 	mp->mnt_flag |= MNT_MLOCK;
 	return (0);
@@ -105,13 +114,14 @@ vfs_unlock(mp)
  * Mark a mount point as busy.
  * Used to synchronize access and to delay unmounting.
  */
+int
 vfs_busy(mp)
 	register struct mount *mp;
 {
 
-	while(mp->mnt_flag & MNT_MPBUSY) {
+	while (mp->mnt_flag & MNT_MPBUSY) {
 		mp->mnt_flag |= MNT_MPWANT;
-		sleep((caddr_t)&mp->mnt_flag, PVFS);
+		tsleep((caddr_t)&mp->mnt_flag, PVFS, "vfsbusy", 0);
 	}
 	if (mp->mnt_flag & MNT_UNMOUNT)
 		return (1);
@@ -123,6 +133,7 @@ vfs_busy(mp)
  * Free a busy filesystem.
  * Panic if filesystem is not busy.
  */
+void
 vfs_unbusy(mp)
 	register struct mount *mp;
 {
@@ -136,16 +147,79 @@ vfs_unbusy(mp)
 	}
 }
 
+/*
+ * Lookup a filesystem type, and if found allocate and initialize
+ * a mount structure for it.
+ *
+ * Devname is usually updated by mount(8) after booting.
+ */
+int
+vfs_rootmountalloc(fstypename, devname, mpp)
+	char *fstypename;
+	char *devname;
+	struct mount **mpp;
+{
+	struct vfsconf *vfsp;
+	struct mount *mp;
+
+	for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
+		if (!strcmp(vfsp->vfc_name, fstypename))
+			break;
+	if (vfsp == NULL)
+		return (ENODEV);
+	mp = malloc((u_long)sizeof(struct mount), M_MOUNT, M_WAITOK);
+	bzero((char *)mp, (u_long)sizeof(struct mount));
+	LIST_INIT(&mp->mnt_vnodelist);
+	mp->mnt_vfc = vfsp;
+	mp->mnt_op = vfsp->vfc_vfsops;
+	mp->mnt_flag = MNT_RDONLY;
+	mp->mnt_vnodecovered = NULLVP;
+	vfsp->vfc_refcount++;
+	mp->mnt_stat.f_type = vfsp->vfc_typenum;
+	mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
+	strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
+	mp->mnt_stat.f_mntonname[0] = '/';
+	(void) copystr(devname, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, 0);
+	*mpp = mp;
+	return (0);
+}
+
+/*
+ * Find an appropriate filesystem to use for the root. If a filesystem
+ * has not been preselected, walk through the list of known filesystems
+ * trying those that have mountroot routines, and try them until one
+ * works or we have tried them all.
+ */
+int
+vfs_mountroot()
+{
+	struct vfsconf *vfsp;
+	extern int (*mountroot)(void);
+	int error;
+
+	if (mountroot != NULL)
+		return ((*vfsp->vfc_mountroot)());
+	for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next) {
+		if (vfsp->vfc_mountroot == NULL)
+			continue;
+		if ((error = (*vfsp->vfc_mountroot)()) == 0)
+			return (0);
+		printf("%s_mountroot failed: %d\n", vfsp->vfc_name, error);
+	}
+	return (ENODEV);
+}
+
 /*
  * Lookup a mount point by filesystem identifier.
  */
 struct mount *
-getvfs(fsid)
+vfs_getvfs(fsid)
 	fsid_t *fsid;
 {
 	register struct mount *mp;
 
-	for (mp = mountlist.tqh_first; mp != NULL; mp = mp->mnt_list.tqe_next) {
+	for (mp = mountlist.cqh_first; mp != (void *)&mountlist;
+	     mp = mp->mnt_list.cqe_next) {
 		if (mp->mnt_stat.f_fsid.val[0] == fsid->val[0] &&
 		    mp->mnt_stat.f_fsid.val[1] == fsid->val[1])
 			return (mp);
@@ -157,33 +231,37 @@ getvfs(fsid)
  * Get a new unique fsid
  */
 void
-getnewfsid(mp, mtype)
+vfs_getnewfsid(mp)
 	struct mount *mp;
-	int mtype;
 {
 static u_short xxxfs_mntid;
 
 	fsid_t tfsid;
+	int mtype;
 
+	simple_lock(&mntid_slock);
+	mtype = mp->mnt_vfc->vfc_typenum;
 	mp->mnt_stat.f_fsid.val[0] = makedev(nblkdev + mtype, 0);
 	mp->mnt_stat.f_fsid.val[1] = mtype;
 	if (xxxfs_mntid == 0)
 		++xxxfs_mntid;
 	tfsid.val[0] = makedev(nblkdev + mtype, xxxfs_mntid);
 	tfsid.val[1] = mtype;
-	if (mountlist.tqh_first != NULL) {
-		while (getvfs(&tfsid)) {
+	if (mountlist.cqh_first != (void *)&mountlist) {
+		while (vfs_getvfs(&tfsid)) {
 			tfsid.val[0]++;
 			xxxfs_mntid++;
 		}
 	}
 	mp->mnt_stat.f_fsid.val[0] = tfsid.val[0];
+	simple_unlock(&mntid_slock);
 }
 
 /*
  * Set vnode attributes to VNOVAL
  */
-void vattr_null(vap)
+void
+vattr_null(vap)
 	register struct vattr *vap;
 {
 
@@ -203,7 +281,8 @@ void vattr_null(vap)
  * Routines having to do with the management of the vnode table.
  */
 extern int (**dead_vnodeop_p)();
-extern void vclean();
+static void vclean __P((struct vnode *vp, int flag, struct proc *p));
+extern void vgonel __P((struct vnode *vp, struct proc *p));
 long numvnodes;
 extern struct vattr va_null;
 int newnodes = 0;
@@ -212,19 +291,25 @@ int printcnt = 0;
 /*
  * Return the next vnode from the free list.
  */
+int
 getnewvnode(tag, mp, vops, vpp)
 	enum vtagtype tag;
 	struct mount *mp;
 	int (**vops)();
 	struct vnode **vpp;
 {
-	register struct vnode *vp;
+	struct proc *p = curproc;	/* XXX */
+	struct vnode *vp;
 	int s;
+	int cnt;
 
+top:
+	simple_lock(&vnode_free_list_slock);
 newnodes++;
 	if ((vnode_free_list.tqh_first == NULL &&
 	     numvnodes < 2 * desiredvnodes) ||
 	    numvnodes < desiredvnodes) {
+		simple_unlock(&vnode_free_list_slock);
 		vp = (struct vnode *)malloc((u_long)sizeof *vp,
 		    M_VNODE, M_WAITOK);
 		bzero((char *)vp, sizeof *vp);
@@ -235,7 +320,18 @@ newnodes++;
 		numvnodes++;
 		vp->v_spare[0] = numvnodes;
 	} else {
-		if ((vp = vnode_free_list.tqh_first) == NULL) {
+		for (vp = vnode_free_list.tqh_first;
+				vp != NULLVP; vp = vp->v_freelist.tqe_next) {
+			if (simple_lock_try(&vp->v_interlock))
+				break;
+		}
+		/*
+		 * Unless this is a bad time of the month, at most
+		 * the first NCPUS items on the free list are
+		 * locked, so this is close enough to being empty.
+		 */
+		if (vp == NULLVP) {
+			simple_unlock(&vnode_free_list_slock);
 			tablefull("vnode");
 			*vpp = 0;
 			return (ENFILE);
@@ -249,9 +345,12 @@ newnodes++;
 		vp->v_freelist.tqe_next = (struct vnode *)0xdeadf;
 		/* see comment on why 0xdeadb is set at end of vgone (below) */
 		vp->v_freelist.tqe_prev = (struct vnode **)0xdeadb;
+		simple_unlock(&vnode_free_list_slock);
 		vp->v_lease = NULL;
 		if (vp->v_type != VBAD)
-			vgone(vp);
+			vgonel(vp, p);
+		else
+			simple_unlock(&vp->v_interlock);
 #ifdef DIAGNOSTIC
 		if (vp->v_data)
 			panic("cleaned vnode isn't");
@@ -285,11 +384,13 @@ newnodes++;
 /*
  * Move a vnode from one mount queue to another.
  */
+void
 insmntque(vp, mp)
-	register struct vnode *vp;
-	register struct mount *mp;
+	struct vnode *vp;
+	struct mount *mp;
 {
 
+	simple_lock(&mntvnode_slock);
 	/*
 	 * Delete from old mount point vnode list, if on one.
 	 */
@@ -304,17 +405,15 @@ insmntque(vp, mp)
 	/*
 	 * Insert into list of vnodes for the new mount point, if available.
 	 */
-	if ((vp->v_mount = mp) == NULL)
-		return;
-	if (vp->v_mntvnodes.le_next != (struct vnode *)0xdeadf ||
-	    vp->v_mntvnodes.le_prev != (struct vnode **)0xdeadb)
-		panic("insmntque: already on queue");
-	LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes);
+	if ((vp->v_mount = mp) != NULL)
+		LIST_INSERT_HEAD(&mp->mnt_vnodelist, vp, v_mntvnodes);
+	simple_unlock(&mntvnode_slock);
 }
 
 /*
  * Update outstanding I/O count and do wakeup if requested.
  */
+void
 vwakeup(bp)
 	register struct buf *bp;
 {
@@ -322,12 +421,11 @@ vwakeup(bp)
 
 	bp->b_flags &= ~B_WRITEINPROG;
 	if (vp = bp->b_vp) {
-		vp->v_numoutput--;
-		if (vp->v_numoutput < 0)
+		if (--vp->v_numoutput < 0)
 			panic("vwakeup: neg numoutput");
 		if ((vp->v_flag & VBWAIT) && vp->v_numoutput <= 0) {
 			if (vp->v_numoutput < 0)
-				panic("vwakeup: neg numoutput");
+				panic("vwakeup: neg numoutput 2");
 			vp->v_flag &= ~VBWAIT;
 			wakeup((caddr_t)&vp->v_numoutput);
 		}
@@ -360,7 +458,7 @@ vinvalbuf(vp, flags, cred, p, slpflag, slptimeo)
 		if ((blist = vp->v_cleanblkhd.lh_first) && flags & V_SAVEMETA)
 			while (blist && blist->b_lblkno < 0)
 				blist = blist->b_vnbufs.le_next;
-		if (!blist && (blist = vp->v_dirtyblkhd.lh_first) && 
+		if (!blist && (blist = vp->v_dirtyblkhd.lh_first) &&
 		    (flags & V_SAVEMETA))
 			while (blist && blist->b_lblkno < 0)
 				blist = blist->b_vnbufs.le_next;
@@ -407,6 +505,7 @@ vinvalbuf(vp, flags, cred, p, slpflag, slptimeo)
 /*
  * Associate a buffer with a vnode.
  */
+void
 bgetvp(vp, bp)
 	register struct vnode *vp;
 	register struct buf *bp;
@@ -429,6 +528,7 @@ bgetvp(vp, bp)
 /*
  * Disassociate a buffer from a vnode.
  */
+void
 brelvp(bp)
 	register struct buf *bp;
 {
@@ -451,6 +551,7 @@ brelvp(bp)
  * Used to assign file specific control information
  * (indirect blocks) to the vnode to which they belong.
  */
+void
 reassignbuf(bp, newvp)
 	register struct buf *bp;
 	register struct vnode *newvp;
@@ -482,6 +583,7 @@ reassignbuf(bp, newvp)
  * Used for root filesystem, argdev, and swap areas.
  * Also used for memory file system special devices.
  */
+int
 bdevvp(dev, vpp)
 	dev_t dev;
 	struct vnode **vpp;
@@ -490,11 +592,13 @@ bdevvp(dev, vpp)
 	struct vnode *nvp;
 	int error;
 
-	if (dev == NODEV)
-		return (0);
+	if (dev == NODEV) {
+		*vpp = NULLVP;
+		return (ENODEV);
+	}
 	error = getnewvnode(VT_NON, (struct mount *)0, spec_vnodeop_p, &nvp);
 	if (error) {
-		*vpp = 0;
+		*vpp = NULLVP;
 		return (error);
 	}
 	vp = nvp;
@@ -521,7 +625,8 @@ checkalias(nvp, nvp_rdev, mp)
 	dev_t nvp_rdev;
 	struct mount *mp;
 {
-	register struct vnode *vp;
+	struct proc *p = curproc;	/* XXX */
+	struct vnode *vp;
 	struct vnode **vpp;
 
 	if (nvp->v_type != VBLK && nvp->v_type != VCHR)
@@ -529,18 +634,23 @@ checkalias(nvp, nvp_rdev, mp)
 
 	vpp = &speclisth[SPECHASH(nvp_rdev)];
 loop:
+	simple_lock(&spechash_slock);
 	for (vp = *vpp; vp; vp = vp->v_specnext) {
 		if (nvp_rdev != vp->v_rdev || nvp->v_type != vp->v_type)
 			continue;
 		/*
 		 * Alias, but not in use, so flush it out.
 		 */
+		simple_lock(&vp->v_interlock);
 		if (vp->v_usecount == 0) {
-			vgone(vp);
+			simple_unlock(&spechash_slock);
+			vgonel(vp, p);
 			goto loop;
 		}
-		if (vget(vp, 1))
+		if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, p)) {
+			simple_unlock(&spechash_slock);
 			goto loop;
+		}
 		break;
 	}
 	if (vp == NULL || vp->v_tag != VT_NON) {
@@ -550,16 +660,19 @@ loop:
 		nvp->v_hashchain = vpp;
 		nvp->v_specnext = *vpp;
 		nvp->v_specflags = 0;
+		simple_unlock(&spechash_slock);
 		*vpp = nvp;
-		if (vp != NULL) {
+		if (vp != NULLVP) {
 			nvp->v_flag |= VALIASED;
 			vp->v_flag |= VALIASED;
 			vput(vp);
 		}
 		return (NULLVP);
 	}
-	VOP_UNLOCK(vp);
-	vclean(vp, 0);
+	simple_unlock(&spechash_slock);
+	VOP_UNLOCK(vp, 0, p);
+	simple_lock(&vp->v_interlock);
+	vclean(vp, 0, p);
 	vp->v_op = nvp->v_op;
 	vp->v_tag = nvp->v_tag;
 	nvp->v_type = VNON;
@@ -575,28 +688,41 @@ loop:
  * indicate that the vnode is no longer usable (possibly having
  * been changed to a new file system type).
  */
-vget(vp, lockflag)
-	register struct vnode *vp;
-	int lockflag;
+int
+vget(vp, flags, p)
+	struct vnode *vp;
+	int flags;
+	struct proc *p;
 {
 
+	/*
+	 * If the vnode is in the process of being cleaned out for
+	 * another use, we wait for the cleaning to finish and then
+	 * return failure. Cleaning is determined by checking that
+	 * the VXLOCK flag is set.
+	 */
+	if ((flags & LK_INTERLOCK) == 0)
+		simple_lock(&vp->v_interlock);
 	if (vp->v_flag & VXLOCK) {
 		vp->v_flag |= VXWANT;
-		sleep((caddr_t)vp, PINOD);
-		return (1);
+		simple_unlock(&vp->v_interlock);
+		tsleep((caddr_t)vp, PINOD, "vget", 0);
+		return (ENOENT);
 	}
 	if (vp->v_usecount == 0) {
 		if (vp->v_freelist.tqe_next == (struct vnode *)0xdeadf ||
 		    vp->v_freelist.tqe_prev == (struct vnode **)0xdeadb)
 			panic("vget: not on queue");
 		TAILQ_REMOVE(&vnode_free_list, vp, v_freelist);
+		simple_unlock(&vnode_free_list_slock);
 	}
 		vp->v_freelist.tqe_next = (struct vnode *)0xdeadf;
 		vp->v_freelist.tqe_prev = (struct vnode **)0xdeadb;
 	}
 	vp->v_usecount++;
-	if (lockflag)
-		VOP_LOCK(vp);
+	if (flags & LK_TYPE_MASK)
+		return (vn_lock(vp, flags | LK_INTERLOCK, p));
+	simple_unlock(&vp->v_interlock);
 	if (printcnt-- > 0) vprint("vget got", vp);
 	return (0);
 }
@@ -604,18 +730,73 @@ vget(vp, lockflag)
 int bug_refs = 0;
 
 /*
- * Vnode reference, just increment the count
+ * Stubs to use when there is no locking to be done on the underlying object.
+ *
+ * Getting a lock just clears the interlock if necessary.
+ */
+int
+vop_nolock(ap)
+	struct vop_lock_args /* {
+		struct vnode *a_vp;
+		int a_flags;
+		struct proc *a_p;
+	} */ *ap;
+{
+	struct vnode *vp = ap->a_vp;
+
+	/*
+	 * Since we are not using the lock manager, we must clear
+	 * the interlock here.
+	 */
+	if (ap->a_flags & LK_INTERLOCK)
+		simple_unlock(&vp->v_interlock);
+	return (0);
+}
+
+/*
+ * Unlock has nothing to do.
+ */
+int
+vop_nounlock(ap)
+	struct vop_unlock_args /* {
+		struct vnode *a_vp;
+		int a_flags;
+		struct proc *a_p;
+	} */ *ap;
+{
+
+	return (0);
+}
+
+/*
+ * Nothing is ever locked.
+ */
+int
+vop_noislocked(ap)
+	struct vop_islocked_args /* {
+		struct vnode *a_vp;
+	} */ *ap;
+{
+
+	return (0);
+}
+
+/*
+ * Vnode reference.
  */
-void vref(vp)
+void
+vref(vp)
 	struct vnode *vp;
 {
 
+	simple_lock(&vp->v_interlock);
 	if (vp->v_usecount <= 0)
 		panic("vref used where vget required");
 	if (vp->v_freelist.tqe_next != (struct vnode *)0xdeadf ||
 	    vp->v_freelist.tqe_prev != (struct vnode **)0xdeadb)
 		panic("vref: not free");
 	vp->v_usecount++;
+	simple_unlock(&vp->v_interlock);
 	if (printcnt-- > 0) vprint("vref get", vp);
 	if (vp->v_type != VBLK && curproc)
 		curproc->p_spare[0]++;
@@ -626,36 +807,66 @@ void vref(vp)
 /*
  * vput(), just unlock and vrele()
  */
-void vput(vp)
-	register struct vnode *vp;
+void
+vput(vp)
+	struct vnode *vp;
 {
+	struct proc *p = curproc;	/* XXX */
 
-	VOP_UNLOCK(vp);
-	vrele(vp);
+#ifdef DIGANOSTIC
+	if (vp == NULL)
+		panic("vput: null vp");
+#endif
+	simple_lock(&vp->v_interlock);
+	vp->v_usecount--;
+	if (vp->v_usecount > 0) {
+		simple_unlock(&vp->v_interlock);
+		VOP_UNLOCK(vp, 0, p);
+		return;
+	}
+#ifdef DIAGNOSTIC
+	if (vp->v_usecount < 0 || vp->v_writecount != 0) {
+		vprint("vput: bad ref count", vp);
+		panic("vput: ref cnt");
+	}
+#endif
+	/*
+	 * insert at tail of LRU list
+	 */
+	simple_lock(&vnode_free_list_slock);
+	TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist);
+	simple_unlock(&vnode_free_list_slock);
+	simple_unlock(&vp->v_interlock);
+	VOP_INACTIVE(vp, p);
 }
 
 /*
  * Vnode release.
  * If count drops to zero, call inactive routine and return to freelist.
  */
-void vrele(vp)
-	register struct vnode *vp;
+void
+vrele(vp)
+	struct vnode *vp;
 {
+	struct proc *p = curproc;	/* XXX */
 
 #ifdef DIAGNOSTIC
 	if (vp == NULL)
 		panic("vrele: null vp");
 #endif
+	simple_lock(&vp->v_interlock);
 	vp->v_usecount--;
 	if (printcnt-- > 0) vprint("vrele put", vp);
 	if (vp->v_type != VBLK && curproc)
 		curproc->p_spare[0]--;
 	if (bug_refs)
 		vprint("vref: ");
-	if (vp->v_usecount > 0)
+	if (vp->v_usecount > 0) {
+		simple_unlock(&vp->v_interlock);
 		return;
+	}
 #ifdef DIAGNOSTIC
-	if (vp->v_usecount != 0 || vp->v_writecount != 0) {
+	if (vp->v_usecount < 0 || vp->v_writecount != 0) {
 		vprint("vrele: bad ref count", vp);
 		panic("vrele: ref cnt");
 	}
@@ -663,34 +874,45 @@ void vrele(vp)
 	/*
 	 * insert at tail of LRU list
 	 */
+	simple_lock(&vnode_free_list_slock);
 	if (vp->v_freelist.tqe_next != (struct vnode *)0xdeadf ||
 	    vp->v_freelist.tqe_prev != (struct vnode **)0xdeadb)
 		panic("vrele: not free");
 	TAILQ_INSERT_TAIL(&vnode_free_list, vp, v_freelist);
-	VOP_INACTIVE(vp);
+	simple_unlock(&vnode_free_list_slock);
+	if (vn_lock(vp, LK_EXCLUSIVE | LK_INTERLOCK, p) == 0)
+		VOP_INACTIVE(vp, p);
 }
 
+#ifdef DIAGNOSTIC
 /*
  * Page or buffer structure gets a reference.
  */
-void vhold(vp)
+void
+vhold(vp)
 	register struct vnode *vp;
 {
 
+	simple_lock(&vp->v_interlock);
 	vp->v_holdcnt++;
+	simple_unlock(&vp->v_interlock);
 }
 
 /*
  * Page or buffer structure frees a reference.
  */
-void holdrele(vp)
+void
+holdrele(vp)
 	register struct vnode *vp;
 {
 
+	simple_lock(&vp->v_interlock);
 	if (vp->v_holdcnt <= 0)
 		panic("holdrele: holdcnt");
 	vp->v_holdcnt--;
+	simple_unlock(&vp->v_interlock);
 }
+#endif /* DIAGNOSTIC */
 
 /*
  * Remove any vnodes in the vnode table belonging to mount point mp.
@@ -705,16 +927,22 @@ int busyprt = 0;	/* print out busy vnodes */
 struct ctldebug debug1 = { "busyprt", &busyprt };
 #endif
 
+int
 vflush(mp, skipvp, flags)
 	struct mount *mp;
 	struct vnode *skipvp;
 	int flags;
 {
-	register struct vnode *vp, *nvp;
+	struct proc *p = curproc;	/* XXX */
+	struct vnode *vp, *nvp;
 	int busy = 0;
 
+#ifdef DIAGNOSTIC
 	if ((mp->mnt_flag & MNT_MPBUSY) == 0)
 		panic("vflush: not busy");
+#endif
+
+	simple_lock(&mntvnode_slock);
 loop:
 	for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) {
 		if (vp->v_mount != mp)
@@ -725,24 +953,32 @@ loop:
 		 */
 		if (vp == skipvp)
 			continue;
+
+		simple_lock(&vp->v_interlock);
 		/*
 		 * Skip over a vnodes marked VSYSTEM.
 		 */
-		if ((flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM))
+		if ((flags & SKIPSYSTEM) && (vp->v_flag & VSYSTEM)) {
+			simple_unlock(&vp->v_interlock);
 			continue;
+		}
 		/*
 		 * If WRITECLOSE is set, only flush out regular file
 		 * vnodes open for writing.
 		 */
 		if ((flags & WRITECLOSE) &&
-		    (vp->v_writecount == 0 || vp->v_type != VREG))
+		    (vp->v_writecount == 0 || vp->v_type != VREG)) {
+			simple_unlock(&vp->v_interlock);
 			continue;
+		}
 		/*
 		 * With v_usecount == 0, all we need to do is clear
 		 * out the vnode data structures and we are done.
 		 */
 		if (vp->v_usecount == 0) {
-			vgone(vp);
+			simple_unlock(&mntvnode_slock);
+			vgonel(vp, p);
+			simple_lock(&mntvnode_slock);
 			continue;
 		}
 		/*
@@ -751,21 +987,25 @@ loop:
 		 * anonymous device. For all other files, just kill them.
 		 */
 		if (flags & FORCECLOSE) {
+			simple_unlock(&mntvnode_slock);
 			if (vp->v_type != VBLK && vp->v_type != VCHR) {
-				vgone(vp);
+				vgonel(vp, p);
 			} else {
-				vclean(vp, 0);
+				vclean(vp, 0, p);
 				vp->v_op = spec_vnodeop_p;
 				insmntque(vp, (struct mount *)0);
 			}
+			simple_lock(&mntvnode_slock);
 			continue;
 		}
 #ifdef DIAGNOSTIC
 		if (busyprt)
 			vprint("vflush: busy vnode", vp);
 #endif
+		simple_unlock(&vp->v_interlock);
 		busy++;
 	}
+	simple_unlock(&mntvnode_slock);
 	if (busy)
 		return (EBUSY);
 	return (0);
@@ -773,11 +1013,13 @@ loop:
 
 /*
  * Disassociate the underlying file system from a vnode.
+ * The vnode interlock is held on entry.
  */
-void
-vclean(vp, flags)
-	register struct vnode *vp;
+static void
+vclean(vp, flags, p)
+	struct vnode *vp;
 	int flags;
+	struct proc *p;
 {
 	int active;
 
@@ -788,15 +1030,7 @@ vclean(vp, flags)
 	 * race against ourselves to recycle it.
 	 */
 	if (active = vp->v_usecount)
-		VREF(vp);
-	/*
-	 * Even if the count is zero, the VOP_INACTIVE routine may still
-	 * have the object locked while it cleans it out. The VOP_LOCK
-	 * ensures that the VOP_INACTIVE routine is done with its work.
-	 * For active vnodes, it ensures that no other activity can
-	 * occur while the underlying object is being cleaned out.
-	 */
-	VOP_LOCK(vp);
+		vp->v_usecount++;
 	/*
 	 * Prevent the vnode from being recycled or
 	 * brought into use while we clean it out.
@@ -804,32 +1038,43 @@ vclean(vp, flags)
 	if (vp->v_flag & VXLOCK)
 		panic("vclean: deadlock");
 	vp->v_flag |= VXLOCK;
+	/*
+	 * Even if the count is zero, the VOP_INACTIVE routine may still
+	 * have the object locked while it cleans it out. The VOP_LOCK
+	 * ensures that the VOP_INACTIVE routine is done with its work.
+	 * For active vnodes, it ensures that no other activity can
+	 * occur while the underlying object is being cleaned out.
+	 */
+	VOP_LOCK(vp, LK_DRAIN | LK_INTERLOCK, p);
 	/*
 	 * Clean out any buffers associated with the vnode.
 	 */
 	if (flags & DOCLOSE)
 		vinvalbuf(vp, V_SAVE, NOCRED, NULL, 0, 0);
-	/*
-	 * Any other processes trying to obtain this lock must first
-	 * wait for VXLOCK to clear, then call the new lock operation.
-	 */
-	VOP_UNLOCK(vp);
 	/*
 	 * If purging an active vnode, it must be closed and
-	 * deactivated before being reclaimed.
+	 * deactivated before being reclaimed. Note that the
+	 * VOP_INACTIVE will unlock the vnode.
 	 */
 	if (active) {
 		if (flags & DOCLOSE)
-			VOP_CLOSE(vp, IO_NDELAY, NOCRED, NULL);
-		VOP_INACTIVE(vp);
+			VOP_CLOSE(vp, IO_NDELAY, NOCRED, p);
+		VOP_INACTIVE(vp, p);
+	} else {
+		/*
+		 * Any other processes trying to obtain this lock must first
+		 * wait for VXLOCK to clear, then call the new lock operation.
+		 */
+		VOP_UNLOCK(vp, 0, p);
 	}
 	/*
 	 * Reclaim the vnode.
 	 */
-	if (VOP_RECLAIM(vp))
+	if (VOP_RECLAIM(vp, p))
 		panic("vclean: cannot reclaim");
 	if (active)
 		vrele(vp);
+	cache_purge(vp);
 
 	/*
 	 * Done with purge, notify sleepers of the grim news.
@@ -847,10 +1092,23 @@ vclean(vp, flags)
  * Eliminate all activity associated with  the requested vnode
  * and with all vnodes aliased to the requested vnode.
  */
-void vgoneall(vp)
-	register struct vnode *vp;
+int
+vop_revoke(ap)
+	struct vop_revoke_args /* {
+		struct vnode *a_vp;
+		int a_flags;
+	} */ *ap;
 {
-	register struct vnode *vq;
+	struct vnode *vp, *vq;
+	struct proc *p = curproc;	/* XXX */
+
+#ifdef DIAGNOSTIC
+	if ((ap->a_flags & REVOKEALL) == 0)
+		panic("vop_revoke");
+#endif
+
+	vp = ap->a_vp;
+	simple_lock(&vp->v_interlock);
 
 	if (vp->v_flag & VALIASED) {
 		/*
@@ -859,41 +1117,86 @@ void vgoneall(vp)
 		 */
 		if (vp->v_flag & VXLOCK) {
 			vp->v_flag |= VXWANT;
-			sleep((caddr_t)vp, PINOD);
-			return;
+			simple_unlock(&vp->v_interlock);
+			tsleep((caddr_t)vp, PINOD, "vop_revokeall", 0);
+			return (0);
 		}
 		/*
 		 * Ensure that vp will not be vgone'd while we
 		 * are eliminating its aliases.
 		 */
 		vp->v_flag |= VXLOCK;
+		simple_unlock(&vp->v_interlock);
 		while (vp->v_flag & VALIASED) {
+			simple_lock(&spechash_slock);
 			for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) {
 				if (vq->v_rdev != vp->v_rdev ||
 				    vq->v_type != vp->v_type || vp == vq)
 					continue;
+				simple_unlock(&spechash_slock);
 				vgone(vq);
 				break;
 			}
+			if (vq == NULLVP)
+				simple_unlock(&spechash_slock);
 		}
 		/*
 		 * Remove the lock so that vgone below will
 		 * really eliminate the vnode after which time
 		 * vgone will awaken any sleepers.
 		 */
+		simple_lock(&vp->v_interlock);
 		vp->v_flag &= ~VXLOCK;
 	}
-	vgone(vp);
+	vgonel(vp, p);
+	return (0);
+}
+
+/*
+ * Recycle an unused vnode to the front of the free list.
+ * Release the passed interlock if the vnode will be recycled.
+ */
+int
+vrecycle(vp, inter_lkp, p)
+	struct vnode *vp;
+	struct simplelock *inter_lkp;
+	struct proc *p;
+{
+
+	simple_lock(&vp->v_interlock);
+	if (vp->v_usecount == 0) {
+		if (inter_lkp)
+			simple_unlock(inter_lkp);
+		vgonel(vp, p);
+		return (1);
+	}
+	simple_unlock(&vp->v_interlock);
+	return (0);
 }
 
 /*
  * Eliminate all activity associated with a vnode
  * in preparation for reuse.
  */
-void vgone(vp)
-	register struct vnode *vp;
+void
+vgone(vp)
+	struct vnode *vp;
+{
+	struct proc *p = curproc;	/* XXX */
+
+	simple_lock(&vp->v_interlock);
+	vgonel(vp, p);
+}
+
+/*
+ * vgone, with the vp interlock held.
+ */
+void
+vgonel(vp, p)
+	struct vnode *vp;
+	struct proc *p;
 {
-	register struct vnode *vq;
+	struct vnode *vq;
 	struct vnode *vx;
 
 	/*
@@ -902,29 +1205,25 @@ void vgone(vp)
 	 */
 	if (vp->v_flag & VXLOCK) {
 		vp->v_flag |= VXWANT;
-		sleep((caddr_t)vp, PINOD);
+		simple_unlock(&vp->v_interlock);
+		tsleep((caddr_t)vp, PINOD, "vgone", 0);
 		return;
 	}
 	/*
 	 * Clean out the filesystem specific data.
 	 */
-	vclean(vp, DOCLOSE);
+	vclean(vp, DOCLOSE, p);
 	/*
 	 * Delete from old mount point vnode list, if on one.
 	 */
-	if (vp->v_mount != NULL) {
-		if (vp->v_mntvnodes.le_next == (struct vnode *)0xdeadf ||
-		    vp->v_mntvnodes.le_prev == (struct vnode **)0xdeadb)
-			panic("vgone: not on queue");
-		LIST_REMOVE(vp, v_mntvnodes);
-		vp->v_mntvnodes.le_next = (struct vnode *)0xdeadf;
-		vp->v_mntvnodes.le_prev = (struct vnode **)0xdeadb;
-		vp->v_mount = NULL;
-	}
+	if (vp->v_mount != NULL)
+		insmntque(vp, (struct mount *)0);
 	/*
-	 * If special device, remove it from special device alias list.
+	 * If special device, remove it from special device alias list
+	 * if it is on one.
 	 */
-	if (vp->v_type == VBLK || vp->v_type == VCHR) {
+	if ((vp->v_type == VBLK || vp->v_type == VCHR) && vp->v_specinfo != 0) {
+		simple_lock(&spechash_slock);
 		if (*vp->v_hashchain == vp) {
 			*vp->v_hashchain = vp->v_specnext;
 		} else {
@@ -953,6 +1252,7 @@ void vgone(vp)
 				vx->v_flag &= ~VALIASED;
 			vp->v_flag &= ~VALIASED;
 		}
+		simple_unlock(&spechash_slock);
 		FREE(vp->v_specinfo, M_VNODE);
 		vp->v_specinfo = NULL;
 	}
@@ -969,13 +1269,14 @@ void vgone(vp)
 	 * getnewvnode after removing it from the freelist to ensure
 	 * that we do not try to move it here.
 	 */
-	if (vp->v_usecount == 0 &&
-	    vp->v_freelist.tqe_prev != (struct vnode **)0xdeadb &&
-	    vnode_free_list.tqh_first != vp) {
-		if (vp->v_freelist.tqe_next == (struct vnode *)0xdeadf)
-			panic("vgone: use 0, not free");
-		TAILQ_REMOVE(&vnode_free_list, vp, v_freelist);
-		TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist);
+	if (vp->v_usecount == 0) {
+		simple_lock(&vnode_free_list_slock);
+		if ((vp->v_freelist.tqe_prev != (struct vnode **)0xdeadb) &&
+		    vnode_free_list.tqh_first != vp) {
+			TAILQ_REMOVE(&vnode_free_list, vp, v_freelist);
+			TAILQ_INSERT_HEAD(&vnode_free_list, vp, v_freelist);
+		}
+		simple_unlock(&vnode_free_list_slock);
 	}
 	vp->v_type = VBAD;
 }
@@ -983,34 +1284,41 @@ void vgone(vp)
 /*
  * Lookup a vnode by device number.
  */
+int
 vfinddev(dev, type, vpp)
 	dev_t dev;
 	enum vtype type;
 	struct vnode **vpp;
 {
-	register struct vnode *vp;
+	struct vnode *vp;
+	int rc = 0;
 
+	simple_lock(&spechash_slock);
 	for (vp = speclisth[SPECHASH(dev)]; vp; vp = vp->v_specnext) {
 		if (dev != vp->v_rdev || type != vp->v_type)
 			continue;
 		*vpp = vp;
-		return (1);
+		rc = 1;
+		break;
 	}
-	return (0);
+	simple_unlock(&spechash_slock);
+	return (rc);
 }
 
 /*
  * Calculate the total number of references to a special device.
  */
+int
 vcount(vp)
-	register struct vnode *vp;
+	struct vnode *vp;
 {
-	register struct vnode *vq, *vnext;
+	struct vnode *vq, *vnext;
 	int count;
 
 loop:
 	if ((vp->v_flag & VALIASED) == 0)
 		return (vp->v_usecount);
+	simple_lock(&spechash_slock);
 	for (count = 0, vq = *vp->v_hashchain; vq; vq = vnext) {
 		vnext = vq->v_specnext;
 		if (vq->v_rdev != vp->v_rdev || vq->v_type != vp->v_type)
@@ -1019,11 +1327,13 @@ loop:
 		 * Alias, but not in use, so flush it out.
 		 */
 		if (vq->v_usecount == 0 && vq != vp) {
+			simple_unlock(&spechash_slock);
 			vgone(vq);
 			goto loop;
 		}
 		count += vq->v_usecount;
 	}
+	simple_unlock(&spechash_slock);
 	return (count);
 }
 
@@ -1033,6 +1343,7 @@ loop:
 static char *typename[] =
    { "VNON", "VREG", "VDIR", "VBLK", "VCHR", "VLNK", "VSOCK", "VFIFO", "VBAD" };
 
+void
 vprint(label, vp)
 	char *label;
 	register struct vnode *vp;
@@ -1075,22 +1386,70 @@ vprint(label, vp)
  * List all of the locked vnodes in the system.
  * Called when debugging the kernel.
  */
+void
 printlockedvnodes()
 {
 	register struct mount *mp;
 	register struct vnode *vp;
 
 	printf("Locked vnodes\n");
-	for (mp = mountlist.tqh_first; mp != NULL; mp = mp->mnt_list.tqe_next) {
+	for (mp = mountlist.cqh_first; mp != (void *)&mountlist;
+	     mp = mp->mnt_list.cqe_next) {
 		for (vp = mp->mnt_vnodelist.lh_first;
 		     vp != NULL;
-		     vp = vp->v_mntvnodes.le_next)
+		     vp = vp->v_mntvnodes.le_next) {
 			if (VOP_ISLOCKED(vp))
 				vprint((char *)0, vp);
+		}
 	}
 }
 #endif
 
+/*
+ * Top level filesystem related information gathering.
+ */
+int
+vfs_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
+	int *name;
+	u_int namelen;
+	void *oldp;
+	size_t *oldlenp;
+	void *newp;
+	size_t newlen;
+	struct proc *p;
+{
+	struct ctldebug *cdp;
+	struct vfsconf *vfsp;
+
+	/* all sysctl names at this level are at least name and field */
+	if (namelen < 2)
+		return (ENOTDIR);		/* overloaded */
+	if (name[0] != VFS_GENERIC) {
+		for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
+			if (vfsp->vfc_typenum == name[0])
+				break;
+		if (vfsp == NULL)
+			return (EOPNOTSUPP);
+		return ((*vfsp->vfc_vfsops->vfs_sysctl)(&name[1], namelen - 1,
+		    oldp, oldlenp, newp, newlen, p));
+	}
+	switch (name[1]) {
+	case VFS_MAXTYPENUM:
+		return (sysctl_rdint(oldp, oldlenp, newp, maxvfsconf));
+	case VFS_CONF:
+		if (namelen < 3)
+			return (ENOTDIR);	/* overloaded */
+		for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
+			if (vfsp->vfc_typenum == name[2])
+				break;
+		if (vfsp == NULL)
+			return (EOPNOTSUPP);
+		return (sysctl_rdstruct(oldp, oldlenp, newp, vfsp,
+		    sizeof(struct vfsconf)));
+	}
+	return (EOPNOTSUPP);
+}
+
 int kinfo_vdebug = 1;
 int kinfo_vgetfailed;
 #define KINFO_VNODESLOP	10
@@ -1099,12 +1458,13 @@ int kinfo_vgetfailed;
  * Copyout address of vnode followed by vnode.
  */
 /* ARGSUSED */
+int
 sysctl_vnode(where, sizep)
 	char *where;
 	size_t *sizep;
 {
 	register struct mount *mp, *nmp;
-	struct vnode *vp;
+	struct vnode *nvp, *vp;
 	register char *bp = where, *savebp;
 	char *ewhere;
 	int error;
@@ -1117,35 +1477,42 @@ sysctl_vnode(where, sizep)
 	}
 	ewhere = where + *sizep;
 		
-	for (mp = mountlist.tqh_first; mp != NULL; mp = nmp) {
-		nmp = mp->mnt_list.tqe_next;
+	for (mp = mountlist.cqh_first; mp != (void *)&mountlist; mp = nmp) {
+		nmp = mp->mnt_list.cqe_next;
 		if (vfs_busy(mp))
 			continue;
 		savebp = bp;
 again:
+		simple_lock(&mntvnode_slock);
 		for (vp = mp->mnt_vnodelist.lh_first;
 		     vp != NULL;
-		     vp = vp->v_mntvnodes.le_next) {
+		     vp = nvp) {
 			/*
 			 * Check that the vp is still associated with
 			 * this filesystem.  RACE: could have been
 			 * recycled onto the same filesystem.
 			 */
 			if (vp->v_mount != mp) {
+				simple_unlock(&mntvnode_slock);
 				if (kinfo_vdebug)
 					printf("kinfo: vp changed\n");
 				bp = savebp;
 				goto again;
 			}
+			nvp = vp->v_mntvnodes.le_next;
 			if (bp + VPTRSZ + VNODESZ > ewhere) {
+				simple_unlock(&mntvnode_slock);
 				*sizep = bp - where;
 				return (ENOMEM);
 			}
+			simple_unlock(&mntvnode_slock);
 			if ((error = copyout((caddr_t)&vp, bp, VPTRSZ)) ||
 			   (error = copyout((caddr_t)vp, bp + VPTRSZ, VNODESZ)))
 				return (error);
 			bp += VPTRSZ + VNODESZ;
+			simple_lock(&mntvnode_slock);
 		}
+		simple_unlock(&mntvnode_slock);
 		vfs_unbusy(mp);
 	}
 
@@ -1158,22 +1525,42 @@ again:
  */
 int
 vfs_mountedon(vp)
-	register struct vnode *vp;
+	struct vnode *vp;
 {
-	register struct vnode *vq;
+	struct vnode *vq;
+	int error = 0;
 
 	if (vp->v_specflags & SI_MOUNTEDON)
 		return (EBUSY);
 	if (vp->v_flag & VALIASED) {
+		simple_lock(&spechash_slock);
 		for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) {
 			if (vq->v_rdev != vp->v_rdev ||
 			    vq->v_type != vp->v_type)
 				continue;
-			if (vq->v_specflags & SI_MOUNTEDON)
-				return (EBUSY);
+			if (vq->v_specflags & SI_MOUNTEDON) {
+				error = EBUSY;
+				break;
+			}
 		}
+		simple_unlock(&spechash_slock);
+	}
+	return (error);
+}
+
+/*
+ * Unmount all filesystems. The list is traversed in reverse order
+ * of mounting to avoid dependencies.
+ */
+void
+vfs_unmountall()
+{
+	struct mount *mp, *nmp;
+
+	for (mp = mountlist.cqh_last; mp != (void *)&mountlist; mp = nmp) {
+		nmp = mp->mnt_list.cqe_prev;
+		(void) dounmount(mp, MNT_FORCE, &proc0);
 	}
-	return (0);
 }
 
 /*
@@ -1239,9 +1626,21 @@ vfs_hang_addrlist(mp, nep, argp)
 	}
 	rn = (*rnh->rnh_addaddr)((caddr_t)saddr, (caddr_t)smask, rnh,
 		np->netc_rnodes);
-	if (rn == 0 || np != (struct netcred *)rn) { /* already exists */
-		error = EPERM;
-		goto out;
+	if (rn == 0) {
+		/*
+		 * One of the reasons that rnh_addaddr may fail is that
+		 * the entry already exists. To check for this case, we
+		 * look up the entry to see if it is there. If so, we
+		 * do not need to make a new entry but do return success.
+		 */
+		free(np, M_NETADDR);
+		rn = (*rnh->rnh_matchaddr)((caddr_t)saddr, rnh);
+		if (rn != 0 && (rn->rn_flags & RNF_ROOT) == 0 &&
+		    ((struct netcred *)rn)->netc_exflags == argp->ex_flags &&
+		    !bcmp((caddr_t)&((struct netcred *)rn)->netc_anon,
+			    (caddr_t)&argp->ex_anon, sizeof(struct ucred)))
+			return (0);
+		return (EPERM);
 	}
 	np->netc_exflags = argp->ex_flags;
 	np->netc_anon = argp->ex_anon;
@@ -1264,7 +1663,7 @@ vfs_free_netcred(rn, w)
 	free((caddr_t)rn, M_NETADDR);
 	return (0);
 }
-	
+
 /*
  * Free the net address hash lists that are hanging off the mount points.
  */