2 * Copyright (c) 1994, 1995 The Regents of the University of California.
3 * Copyright (c) 1994, 1995 Jan-Simon Pendry.
6 * This code is derived from software donated to Berkeley by
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. All advertising materials mentioning features or use of this software
18 * must display the following acknowledgement:
19 * This product includes software developed by the University of
20 * California, Berkeley and its contributors.
21 * 4. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 * @(#)union_vfsops.c 8.20 (Berkeley) 5/20/95
38 * $FreeBSD: src/sys/miscfs/union/union_vfsops.c,v 1.39.2.2 2001/10/25 19:18:53 dillon Exp $
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/kernel.h>
49 #include <sys/vnode.h>
50 #include <sys/mount.h>
51 #include <sys/nlookup.h>
52 #include <sys/namei.h>
53 #include <sys/malloc.h>
54 #include <sys/filedesc.h>
56 #include <vm/vm_zone.h>
58 extern struct vop_ops union_vnode_vops;
60 static MALLOC_DEFINE(M_UNIONFSMNT, "UNION mount", "UNION mount structure");
62 extern int union_init (struct vfsconf *);
63 static int union_mount (struct mount *mp, char *path, caddr_t data,
65 static int union_root (struct mount *mp, struct vnode **vpp);
66 static int union_statfs (struct mount *mp, struct statfs *sbp,
68 static int union_unmount (struct mount *mp, int mntflags);
71 * Mount union filesystem
74 union_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
77 struct union_args args;
78 struct vnode *lowerrootvp = NULLVP;
79 struct vnode *upperrootvp = NULLVP;
80 struct union_mount *um = NULL;
81 struct ucred *cred = NULL;
82 struct nlookupdata nd;
87 UDEBUG(("union_mount(mp = %p)\n", (void *)mp));
90 * Disable clustered write, otherwise system becomes unstable.
92 mp->mnt_flag |= MNT_NOCLUSTERW;
97 if (mp->mnt_flag & MNT_UPDATE) {
100 * 1. a way to convert between rdonly and rdwr mounts.
101 * 2. support for nfs exports.
110 error = copyin(data, (caddr_t)&args, sizeof(struct union_args));
115 * Obtain lower vnode.
118 lowerrootvp = mp->mnt_vnodecovered;
123 * Unlock lower node to avoid deadlock.
125 if (lowerrootvp->v_tag == VT_UNION)
126 vn_unlock(lowerrootvp);
130 * Obtain upper vnode by calling nlookup() on the path. The
131 * upperrootvp will be turned referenced but not locked.
133 error = nlookup_init(&nd, args.target, UIO_USERSPACE, NLC_FOLLOW);
135 error = nlookup(&nd);
137 error = cache_vref(&nd.nl_nch, nd.nl_cred, &upperrootvp);
142 UDEBUG(("mount_root UPPERVP %p locked = %d\n", upperrootvp,
143 vn_islocked(upperrootvp)));
146 * Check multi union mount to avoid `lock myself again' panic.
147 * Also require that it be a directory.
149 if (upperrootvp == VTOUNION(lowerrootvp)->un_uppervp) {
151 kprintf("union_mount: multi union mount?\n");
157 if (upperrootvp->v_type != VDIR) {
163 * Allocate our union_mount structure and populate the fields.
164 * The vnode references are stored in the union_mount as held,
165 * unlocked references. Depending on the _BELOW flag, the
166 * filesystems are viewed in a different order. In effect this
167 * is the same as providing a mount-under option to the mount
171 um = (struct union_mount *) malloc(sizeof(struct union_mount),
172 M_UNIONFSMNT, M_WAITOK);
174 bzero(um, sizeof(struct union_mount));
176 um->um_op = args.mntflags & UNMNT_OPMASK;
180 um->um_lowervp = lowerrootvp;
181 um->um_uppervp = upperrootvp;
187 um->um_lowervp = upperrootvp;
188 um->um_uppervp = lowerrootvp;
196 um->um_uppervp = upperrootvp;
197 um->um_lowervp = lowerrootvp;
207 * Unless the mount is readonly, ensure that the top layer
208 * supports whiteout operations
210 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
211 error = VOP_WHITEOUT(um->um_uppervp, NULL, NAMEI_LOOKUP);
217 * File creds and modes for shadowed files are based on the user
218 * that did the mount.
220 um->um_cred = crhold(cred);
221 um->um_cmode = UN_DIRMODE;
223 um->um_cmode &= ~curproc->p_fd->fd_cmask;
226 * Depending on what you think the MNT_LOCAL flag might mean,
227 * you may want the && to be || on the conditional below.
228 * At the moment it has been defined that the filesystem is
229 * only local if it is all local, ie the MNT_LOCAL flag implies
230 * that the entire namespace is local. If you think the MNT_LOCAL
231 * flag implies that some of the files might be stored locally
232 * then you will want to change the conditional.
234 if (um->um_op == UNMNT_ABOVE) {
235 if (((um->um_lowervp == NULLVP) ||
236 (um->um_lowervp->v_mount->mnt_flag & MNT_LOCAL)) &&
237 (um->um_uppervp->v_mount->mnt_flag & MNT_LOCAL))
238 mp->mnt_flag |= MNT_LOCAL;
242 * Copy in the upper layer's RDONLY flag. This is for the benefit
243 * of lookup() which explicitly checks the flag, rather than asking
244 * the filesystem for its own opinion. This means, that an update
245 * mount of the underlying filesystem to go from rdonly to rdwr
246 * will leave the unioned view as read-only.
248 mp->mnt_flag |= (um->um_uppervp->v_mount->mnt_flag & MNT_RDONLY);
250 mp->mnt_data = (qaddr_t) um;
265 bcopy(cp, mp->mnt_stat.f_mntfromname, len);
267 cp = mp->mnt_stat.f_mntfromname + len;
268 len = MNAMELEN - len;
270 (void) copyinstr(args.target, cp, len - 1, &size);
271 bzero(cp + size, len - size);
273 vfs_add_vnodeops(mp, &union_vnode_vops, &mp->mnt_vn_norm_ops);
275 (void)union_statfs(mp, &mp->mnt_stat, cred);
282 vrele(um->um_uppervp);
284 vrele(um->um_lowervp);
285 /* XXX other fields */
286 kfree(um, M_UNIONFSMNT);
298 * Free reference to union layer
301 union_unmount(struct mount *mp, int mntflags)
303 struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
308 UDEBUG(("union_unmount(mp = %p)\n", (void *)mp));
310 if (mntflags & MNT_FORCE)
314 * Keep flushing vnodes from the mount list.
315 * This is needed because of the un_pvp held
316 * reference to the parent vnode.
317 * If more vnodes have been freed on a given pass,
318 * the try again. The loop will iterate at most
319 * (d) times, where (d) is the maximum tree depth
322 for (freeing = 0; (error = vflush(mp, 0, flags)) != 0;) {
323 int n = mp->mnt_nvnodelistsize;
325 /* if this is unchanged then stop */
329 /* otherwise try once more time */
333 /* If the most recent vflush failed, the filesystem is still busy. */
338 * Discard references to upper and lower target vnodes.
341 vrele(um->um_lowervp);
342 vrele(um->um_uppervp);
345 * Finally, throw away the union_mount structure
347 kfree(mp->mnt_data, M_UNIONFSMNT); /* XXX */
353 union_root(struct mount *mp, struct vnode **vpp)
355 struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
359 * Supply an unlocked reference to um_uppervp and to um_lowervp. It
360 * is possible for um_uppervp to be locked without the associated
361 * root union_node being locked. We let union_allocvp() deal with
364 UDEBUG(("union_root UPPERVP %p locked = %d\n", um->um_uppervp,
365 vn_islocked(um->um_uppervp)));
367 vref(um->um_uppervp);
369 vref(um->um_lowervp);
371 error = union_allocvp(vpp, mp, NULLVP, NULLVP, NULL,
372 um->um_uppervp, um->um_lowervp, 1);
373 UDEBUG(("error %d\n", error));
374 UDEBUG(("union_root2 UPPERVP %p locked = %d\n", um->um_uppervp,
375 vn_islocked(um->um_uppervp)));
381 union_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
384 struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
388 UDEBUG(("union_statfs(mp = %p, lvp = %p, uvp = %p)\n",
389 (void *)mp, (void *)um->um_lowervp, (void *)um->um_uppervp));
391 bzero(&mstat, sizeof(mstat));
393 if (um->um_lowervp) {
394 error = VFS_STATFS(um->um_lowervp->v_mount, &mstat, cred);
399 /* now copy across the "interesting" information and fake the rest */
401 sbp->f_type = mstat.f_type;
402 sbp->f_flags = mstat.f_flags;
403 sbp->f_bsize = mstat.f_bsize;
404 sbp->f_iosize = mstat.f_iosize;
406 lbsize = mstat.f_bsize;
407 sbp->f_blocks = mstat.f_blocks;
408 sbp->f_bfree = mstat.f_bfree;
409 sbp->f_bavail = mstat.f_bavail;
410 sbp->f_files = mstat.f_files;
411 sbp->f_ffree = mstat.f_ffree;
413 error = VFS_STATFS(um->um_uppervp->v_mount, &mstat, cred);
417 sbp->f_flags = mstat.f_flags;
418 sbp->f_bsize = mstat.f_bsize;
419 sbp->f_iosize = mstat.f_iosize;
422 * if the lower and upper blocksizes differ, then frig the
423 * block counts so that the sizes reported by df make some
424 * kind of sense. none of this makes sense though.
427 if (mstat.f_bsize != lbsize)
428 sbp->f_blocks = ((off_t) sbp->f_blocks * lbsize) / mstat.f_bsize;
431 * The "total" fields count total resources in all layers,
432 * the "free" fields count only those resources which are
433 * free in the upper layer (since only the upper layer
436 sbp->f_blocks += mstat.f_blocks;
437 sbp->f_bfree = mstat.f_bfree;
438 sbp->f_bavail = mstat.f_bavail;
439 sbp->f_files += mstat.f_files;
440 sbp->f_ffree = mstat.f_ffree;
442 if (sbp != &mp->mnt_stat) {
443 sbp->f_type = mp->mnt_vfc->vfc_typenum;
444 bcopy(&mp->mnt_stat.f_fsid, &sbp->f_fsid, sizeof(sbp->f_fsid));
445 bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
450 static struct vfsops union_vfsops = {
451 .vfs_mount = union_mount,
452 .vfs_unmount = union_unmount,
453 .vfs_root = union_root,
454 .vfs_statfs = union_statfs,
455 .vfs_sync = vfs_stdsync,
456 .vfs_init = union_init
459 VFS_SET(union_vfsops, union, VFCF_LOOPBACK);