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 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)union_vfsops.c 8.20 (Berkeley) 5/20/95
34 * $FreeBSD: src/sys/miscfs/union/union_vfsops.c,v 1.39.2.2 2001/10/25 19:18:53 dillon Exp $
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kernel.h>
45 #include <sys/vnode.h>
46 #include <sys/mount.h>
47 #include <sys/nlookup.h>
48 #include <sys/namei.h>
49 #include <sys/malloc.h>
50 #include <sys/filedesc.h>
52 #include <vm/vm_zone.h>
54 extern struct vop_ops union_vnode_vops;
56 static MALLOC_DEFINE(M_UNIONFSMNT, "UNION mount", "UNION mount structure");
58 extern int union_init (struct vfsconf *);
59 static int union_mount (struct mount *mp, char *path, caddr_t data,
61 static int union_root (struct mount *mp, struct vnode **vpp);
62 static int union_statfs (struct mount *mp, struct statfs *sbp,
64 static int union_unmount (struct mount *mp, int mntflags);
67 * Mount union filesystem
70 union_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
73 struct union_args args;
74 struct vnode *lowerrootvp = NULLVP;
75 struct vnode *upperrootvp = NULLVP;
76 struct union_mount *um = NULL;
77 struct ucred *cred = NULL;
78 struct nlookupdata nd;
83 UDEBUG(("union_mount(mp = %p)\n", (void *)mp));
86 * Disable clustered write, otherwise system becomes unstable.
88 mp->mnt_flag |= MNT_NOCLUSTERW;
93 if (mp->mnt_flag & MNT_UPDATE) {
96 * 1. a way to convert between rdonly and rdwr mounts.
97 * 2. support for nfs exports.
106 error = copyin(data, (caddr_t)&args, sizeof(struct union_args));
111 * Obtain lower vnode.
114 lowerrootvp = mp->mnt_vnodecovered;
119 * Unlock lower node to avoid deadlock.
121 if (lowerrootvp->v_tag == VT_UNION)
122 vn_unlock(lowerrootvp);
126 * Obtain upper vnode by calling nlookup() on the path. The
127 * upperrootvp will be turned referenced but not locked.
129 error = nlookup_init(&nd, args.target, UIO_USERSPACE, NLC_FOLLOW);
131 error = nlookup(&nd);
133 error = cache_vref(&nd.nl_nch, nd.nl_cred, &upperrootvp);
138 UDEBUG(("mount_root UPPERVP %p locked = %d\n", upperrootvp,
139 vn_islocked(upperrootvp)));
142 * Check multi union mount to avoid `lock myself again' panic.
143 * Also require that it be a directory.
145 if (upperrootvp == VTOUNION(lowerrootvp)->un_uppervp) {
147 kprintf("union_mount: multi union mount?\n");
153 if (upperrootvp->v_type != VDIR) {
159 * Allocate our union_mount structure and populate the fields.
160 * The vnode references are stored in the union_mount as held,
161 * unlocked references. Depending on the _BELOW flag, the
162 * filesystems are viewed in a different order. In effect this
163 * is the same as providing a mount-under option to the mount
167 um = (struct union_mount *) malloc(sizeof(struct union_mount),
168 M_UNIONFSMNT, M_WAITOK);
170 bzero(um, sizeof(struct union_mount));
172 um->um_op = args.mntflags & UNMNT_OPMASK;
176 um->um_lowervp = lowerrootvp;
177 um->um_uppervp = upperrootvp;
183 um->um_lowervp = upperrootvp;
184 um->um_uppervp = lowerrootvp;
192 um->um_uppervp = upperrootvp;
193 um->um_lowervp = lowerrootvp;
203 * Unless the mount is readonly, ensure that the top layer
204 * supports whiteout operations
206 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
207 error = VOP_WHITEOUT(um->um_uppervp, NULL, NAMEI_LOOKUP);
213 * File creds and modes for shadowed files are based on the user
214 * that did the mount.
216 um->um_cred = crhold(cred);
217 um->um_cmode = UN_DIRMODE;
219 um->um_cmode &= ~curproc->p_fd->fd_cmask;
222 * Depending on what you think the MNT_LOCAL flag might mean,
223 * you may want the && to be || on the conditional below.
224 * At the moment it has been defined that the filesystem is
225 * only local if it is all local, ie the MNT_LOCAL flag implies
226 * that the entire namespace is local. If you think the MNT_LOCAL
227 * flag implies that some of the files might be stored locally
228 * then you will want to change the conditional.
230 if (um->um_op == UNMNT_ABOVE) {
231 if (((um->um_lowervp == NULLVP) ||
232 (um->um_lowervp->v_mount->mnt_flag & MNT_LOCAL)) &&
233 (um->um_uppervp->v_mount->mnt_flag & MNT_LOCAL))
234 mp->mnt_flag |= MNT_LOCAL;
238 * Copy in the upper layer's RDONLY flag. This is for the benefit
239 * of lookup() which explicitly checks the flag, rather than asking
240 * the filesystem for its own opinion. This means, that an update
241 * mount of the underlying filesystem to go from rdonly to rdwr
242 * will leave the unioned view as read-only.
244 mp->mnt_flag |= (um->um_uppervp->v_mount->mnt_flag & MNT_RDONLY);
246 mp->mnt_data = (qaddr_t) um;
261 bcopy(cp, mp->mnt_stat.f_mntfromname, len);
263 cp = mp->mnt_stat.f_mntfromname + len;
264 len = MNAMELEN - len;
266 (void) copyinstr(args.target, cp, len - 1, &size);
267 bzero(cp + size, len - size);
269 vfs_add_vnodeops(mp, &union_vnode_vops, &mp->mnt_vn_norm_ops);
271 (void)union_statfs(mp, &mp->mnt_stat, cred);
278 vrele(um->um_uppervp);
280 vrele(um->um_lowervp);
281 /* XXX other fields */
282 kfree(um, M_UNIONFSMNT);
294 * Free reference to union layer
297 union_unmount(struct mount *mp, int mntflags)
299 struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
304 UDEBUG(("union_unmount(mp = %p)\n", (void *)mp));
306 if (mntflags & MNT_FORCE)
310 * Keep flushing vnodes from the mount list.
311 * This is needed because of the un_pvp held
312 * reference to the parent vnode.
313 * If more vnodes have been freed on a given pass,
314 * the try again. The loop will iterate at most
315 * (d) times, where (d) is the maximum tree depth
318 for (freeing = 0; (error = vflush(mp, 0, flags)) != 0;) {
319 int n = mp->mnt_nvnodelistsize;
321 /* if this is unchanged then stop */
325 /* otherwise try once more time */
329 /* If the most recent vflush failed, the filesystem is still busy. */
334 * Discard references to upper and lower target vnodes.
337 vrele(um->um_lowervp);
338 vrele(um->um_uppervp);
341 * Finally, throw away the union_mount structure
343 kfree(mp->mnt_data, M_UNIONFSMNT); /* XXX */
349 union_root(struct mount *mp, struct vnode **vpp)
351 struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
355 * Supply an unlocked reference to um_uppervp and to um_lowervp. It
356 * is possible for um_uppervp to be locked without the associated
357 * root union_node being locked. We let union_allocvp() deal with
360 UDEBUG(("union_root UPPERVP %p locked = %d\n", um->um_uppervp,
361 vn_islocked(um->um_uppervp)));
363 vref(um->um_uppervp);
365 vref(um->um_lowervp);
367 error = union_allocvp(vpp, mp, NULLVP, NULLVP, NULL,
368 um->um_uppervp, um->um_lowervp, 1);
369 UDEBUG(("error %d\n", error));
370 UDEBUG(("union_root2 UPPERVP %p locked = %d\n", um->um_uppervp,
371 vn_islocked(um->um_uppervp)));
377 union_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
380 struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
384 UDEBUG(("union_statfs(mp = %p, lvp = %p, uvp = %p)\n",
385 (void *)mp, (void *)um->um_lowervp, (void *)um->um_uppervp));
387 bzero(&mstat, sizeof(mstat));
389 if (um->um_lowervp) {
390 error = VFS_STATFS(um->um_lowervp->v_mount, &mstat, cred);
395 /* now copy across the "interesting" information and fake the rest */
397 sbp->f_type = mstat.f_type;
398 sbp->f_flags = mstat.f_flags;
399 sbp->f_bsize = mstat.f_bsize;
400 sbp->f_iosize = mstat.f_iosize;
402 lbsize = mstat.f_bsize;
403 sbp->f_blocks = mstat.f_blocks;
404 sbp->f_bfree = mstat.f_bfree;
405 sbp->f_bavail = mstat.f_bavail;
406 sbp->f_files = mstat.f_files;
407 sbp->f_ffree = mstat.f_ffree;
409 error = VFS_STATFS(um->um_uppervp->v_mount, &mstat, cred);
413 sbp->f_flags = mstat.f_flags;
414 sbp->f_bsize = mstat.f_bsize;
415 sbp->f_iosize = mstat.f_iosize;
418 * if the lower and upper blocksizes differ, then frig the
419 * block counts so that the sizes reported by df make some
420 * kind of sense. none of this makes sense though.
423 if (mstat.f_bsize != lbsize)
424 sbp->f_blocks = ((off_t) sbp->f_blocks * lbsize) / mstat.f_bsize;
427 * The "total" fields count total resources in all layers,
428 * the "free" fields count only those resources which are
429 * free in the upper layer (since only the upper layer
432 sbp->f_blocks += mstat.f_blocks;
433 sbp->f_bfree = mstat.f_bfree;
434 sbp->f_bavail = mstat.f_bavail;
435 sbp->f_files += mstat.f_files;
436 sbp->f_ffree = mstat.f_ffree;
438 if (sbp != &mp->mnt_stat) {
439 sbp->f_type = mp->mnt_vfc->vfc_typenum;
440 bcopy(&mp->mnt_stat.f_fsid, &sbp->f_fsid, sizeof(sbp->f_fsid));
441 bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
446 static struct vfsops union_vfsops = {
447 .vfs_mount = union_mount,
448 .vfs_unmount = union_unmount,
449 .vfs_root = union_root,
450 .vfs_statfs = union_statfs,
451 .vfs_sync = vfs_stdsync,
452 .vfs_init = union_init
455 VFS_SET(union_vfsops, union, VFCF_LOOPBACK);