Add snapshots to the fast filesystem. Most of the changes support

the gating of system calls that cause modifications to the underlying
filesystem. The gating can be enabled by any filesystem that needs
to consistently suspend operations by adding the vop_stdgetwritemount
to their set of vnops. Once gating is enabled, the function
vfs_write_suspend stops all new write operations to a filesystem,
allows any filesystem modifying system calls already in progress
to complete, then sync's the filesystem to disk and returns. The
function vfs_write_resume allows the suspended write operations to
begin again. Gating is not added by default for all filesystems as
for SMP systems it adds two extra locks to such critical kernel
paths as the write system call. Thus, gating should only be added
as needed.

Details on the use and current status of snapshots in FFS can be
found in /sys/ufs/ffs/README.snapshot so for brevity and timelyness
is not included here. Unless and until you create a snapshot file,
these changes should have no effect on your system (famous last words).

5 files changed, 66 insertions, 12 deletions

diff --git a/sys/miscfs/fdesc/fdesc_vnops.c b/sys/miscfs/fdesc/fdesc_vnops.c
index 772a94c..72c7cae 100644
--- a/sys/miscfs/fdesc/fdesc_vnops.c
+++ b/sys/miscfs/fdesc/fdesc_vnops.c
@@ -383,6 +383,8 @@ fdesc_setattr(ap)
 {
 	struct filedesc *fdp = ap->a_p->p_fd;
 	struct vattr *vap = ap->a_vap;
+	struct vnode *vp;
+	struct mount *mp;
 	struct file *fp;
 	unsigned fd;
 	int error;
@@ -403,8 +405,11 @@ fdesc_setattr(ap)
 	switch (fp->f_type) {
 	case DTYPE_FIFO:
 	case DTYPE_VNODE:
-		error = VOP_SETATTR((struct vnode *) fp->f_data, ap->a_vap,
-		    ap->a_cred, ap->a_p);
+		vp = (struct vnode *)fp->f_data;
+		if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
+			return (error);
+		error = VOP_SETATTR(vp, ap->a_vap, ap->a_cred, ap->a_p);
+		vn_finished_write(mp);
 		break;
 
 	default:
diff --git a/sys/miscfs/fifofs/fifo_vnops.c b/sys/miscfs/fifofs/fifo_vnops.c
index 5bd13a7..03e3e37 100644
--- a/sys/miscfs/fifofs/fifo_vnops.c
+++ b/sys/miscfs/fifofs/fifo_vnops.c
@@ -107,6 +107,7 @@ static struct vnodeopv_entry_desc fifo_vnodeop_entries[] = {
 	{ &vop_open_desc,		(vop_t *) fifo_open },
 	{ &vop_pathconf_desc,		(vop_t *) fifo_pathconf },
 	{ &vop_poll_desc,		(vop_t *) fifo_poll },
+	{ &vop_getwritemount_desc, 	(vop_t *) vop_stdgetwritemount },
 	{ &vop_print_desc,		(vop_t *) fifo_print },
 	{ &vop_read_desc,		(vop_t *) fifo_read },
 	{ &vop_readdir_desc,		(vop_t *) fifo_badop },
diff --git a/sys/miscfs/specfs/spec_vnops.c b/sys/miscfs/specfs/spec_vnops.c
index cbe52f4..baf40c3 100644
--- a/sys/miscfs/specfs/spec_vnops.c
+++ b/sys/miscfs/specfs/spec_vnops.c
@@ -88,6 +88,7 @@ static struct vnodeopv_entry_desc spec_vnodeop_entries[] = {
 	{ &vop_open_desc,		(vop_t *) spec_open },
 	{ &vop_pathconf_desc,		(vop_t *) vop_stdpathconf },
 	{ &vop_poll_desc,		(vop_t *) spec_poll },
+	{ &vop_getwritemount_desc, 	(vop_t *) vop_stdgetwritemount },
 	{ &vop_print_desc,		(vop_t *) spec_print },
 	{ &vop_read_desc,		(vop_t *) spec_read },
 	{ &vop_readdir_desc,		(vop_t *) vop_panic },
@@ -415,16 +416,29 @@ spec_strategy(ap)
 	struct buf *bp;
 	struct vnode *vp;
 	struct mount *mp;
+	int error;
 
 	bp = ap->a_bp;
-	if ((bp->b_iocmd == BIO_WRITE) && (LIST_FIRST(&bp->b_dep)) != NULL)
-		buf_start(bp);
-
+	vp = ap->a_vp;
+	if ((bp->b_iocmd == BIO_WRITE)) {
+		if (vp->v_mount != NULL &&
+		    (vp->v_mount->mnt_kern_flag & MNTK_SUSPENDED) != 0)
+			panic("spec_strategy: bad I/O");
+		if (LIST_FIRST(&bp->b_dep) != NULL)
+			buf_start(bp);
+		if ((vp->v_flag & VCOPYONWRITE) &&
+		    (error = VOP_COPYONWRITE(vp, bp)) != 0 &&
+		    error != EOPNOTSUPP) {
+			bp->b_io.bio_error = error;
+			bp->b_io.bio_flags |= BIO_ERROR;
+			biodone(&bp->b_io);
+			return (0);
+		}
+	}
 	/*
 	 * Collect statistics on synchronous and asynchronous read
 	 * and write counts for disks that have associated filesystems.
 	 */
-	vp = ap->a_vp;
 	if (vn_isdisk(vp, NULL) && (mp = vp->v_specmountpoint) != NULL) {
 		if (bp->b_iocmd == BIO_WRITE) {
 			if (bp->b_lock.lk_lockholder == LK_KERNPROC)
diff --git a/sys/miscfs/union/union_subr.c b/sys/miscfs/union/union_subr.c
index 6b88bef..d1d6e31 100644
--- a/sys/miscfs/union/union_subr.c
+++ b/sys/miscfs/union/union_subr.c
@@ -747,6 +747,7 @@ union_copyup(un, docopy, cred, p)
 	struct proc *p;
 {
 	int error;
+	struct mount *mp;
 	struct vnode *lvp, *uvp;
 
 	/*
@@ -759,9 +760,12 @@ union_copyup(un, docopy, cred, p)
 	if (error)
 		return (error);
 
-	error = union_vn_create(&uvp, un, p);
-	if (error)
+	if ((error = vn_start_write(un->un_dirvp, &mp, V_WAIT | PCATCH)) != 0)
 		return (error);
+	if ((error = union_vn_create(&uvp, un, p)) != 0) {
+		vn_finished_write(mp);
+		return (error);
+	}
 
 	lvp = un->un_lowervp;
 
@@ -785,6 +789,7 @@ union_copyup(un, docopy, cred, p)
 
 	}
 	VOP_UNLOCK(uvp, 0, p);
+	vn_finished_write(mp);
 	union_newupper(un, uvp);
 	KASSERT(uvp->v_usecount > 0, ("copy: uvp refcount 0: %d", uvp->v_usecount));
 	union_vn_close(uvp, FWRITE, cred, p);
@@ -910,11 +915,15 @@ union_mkshadow(um, dvp, cnp, vpp)
 	struct vattr va;
 	struct proc *p = cnp->cn_proc;
 	struct componentname cn;
+	struct mount *mp;
 
-	error = union_relookup(um, dvp, vpp, cnp, &cn,
-			cnp->cn_nameptr, cnp->cn_namelen);
-	if (error)
+	if ((error = vn_start_write(dvp, &mp, V_WAIT | PCATCH)) != 0)
+		return (error);
+	if ((error = union_relookup(um, dvp, vpp, cnp, &cn,
+			cnp->cn_nameptr, cnp->cn_namelen)) != 0) {
+		vn_finished_write(mp);
 		return (error);
+	}
 
 	if (*vpp) {
 		if (cn.cn_flags & HASBUF) {
@@ -925,6 +934,7 @@ union_mkshadow(um, dvp, cnp, vpp)
 			vrele(*vpp);
 		else
 			vput(*vpp);
+		vn_finished_write(mp);
 		*vpp = NULLVP;
 		return (EEXIST);
 	}
@@ -950,6 +960,7 @@ union_mkshadow(um, dvp, cnp, vpp)
 		cn.cn_flags &= ~HASBUF;
 	}
 	/*vput(dvp);*/
+	vn_finished_write(mp);
 	return (error);
 }
 
@@ -973,10 +984,15 @@ union_mkwhiteout(um, dvp, cnp, path)
 	struct proc *p = cnp->cn_proc;
 	struct vnode *wvp;
 	struct componentname cn;
+	struct mount *mp;
 
+	if ((error = vn_start_write(dvp, &mp, V_WAIT | PCATCH)) != 0)
+		return (error);
 	error = union_relookup(um, dvp, &wvp, cnp, &cn, path, strlen(path));
-	if (error)
+	if (error) {
+		vn_finished_write(mp);
 		return (error);
+	}
 
 	if (wvp) {
 		if (cn.cn_flags & HASBUF) {
@@ -987,6 +1003,7 @@ union_mkwhiteout(um, dvp, cnp, path)
 			vrele(wvp);
 		else
 			vput(wvp);
+		vn_finished_write(mp);
 		return (EEXIST);
 	}
 
@@ -998,6 +1015,7 @@ union_mkwhiteout(um, dvp, cnp, path)
 		zfree(namei_zone, cn.cn_pnbuf);
 		cn.cn_flags &= ~HASBUF;
 	}
+	vn_finished_write(mp);
 	return (error);
 }
 
diff --git a/sys/miscfs/union/union_vnops.c b/sys/miscfs/union/union_vnops.c
index 1c5ed5d..d7b95f3 100644
--- a/sys/miscfs/union/union_vnops.c
+++ b/sys/miscfs/union/union_vnops.c
@@ -93,6 +93,7 @@ static int	union_print __P((struct vop_print_args *ap));
 static int	union_read __P((struct vop_read_args *ap));
 static int	union_readdir __P((struct vop_readdir_args *ap));
 static int	union_readlink __P((struct vop_readlink_args *ap));
+static int	union_getwritemount __P((struct vop_getwritemount_args *ap));
 static int	union_reclaim __P((struct vop_reclaim_args *ap));
 static int	union_remove __P((struct vop_remove_args *ap));
 static int	union_rename __P((struct vop_rename_args *ap));
@@ -1681,6 +1682,20 @@ union_readlink(ap)
 	return (error);
 }
 
+static int
+union_getwritemount(ap)
+	struct vop_getwritemount_args /* {
+		struct vnode *a_vp;
+		struct mount **a_mpp;
+	} */ *ap;
+{
+	struct vnode *vp = UPPERVP(ap->a_vp);
+
+	if (vp == NULL)
+		panic("union: missing upper layer in getwritemount");
+	return(VOP_GETWRITEMOUNT(vp, ap->a_mpp));
+}
+
 /*
  *	union_inactive:
  *
@@ -1963,6 +1978,7 @@ static struct vnodeopv_entry_desc union_vnodeop_entries[] = {
 	{ &vop_read_desc,		(vop_t *) union_read },
 	{ &vop_readdir_desc,		(vop_t *) union_readdir },
 	{ &vop_readlink_desc,		(vop_t *) union_readlink },
+	{ &vop_getwritemount_desc,	(vop_t *) union_getwritemount },
 	{ &vop_reclaim_desc,		(vop_t *) union_reclaim },
 	{ &vop_remove_desc,		(vop_t *) union_remove },
 	{ &vop_rename_desc,		(vop_t *) union_rename },