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 $
39 * $DragonFly: src/sys/vfs/union/union_vfsops.c,v 1.19 2005/02/02 21:34:19 joerg Exp $
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
50 #include <sys/vnode.h>
51 #include <sys/mount.h>
52 #include <sys/nlookup.h>
53 #include <sys/namei.h>
54 #include <sys/malloc.h>
55 #include <sys/filedesc.h>
57 #include <vm/vm_zone.h>
59 extern struct vnodeopv_entry_desc union_vnodeop_entries[];
61 static MALLOC_DEFINE(M_UNIONFSMNT, "UNION mount", "UNION mount structure");
63 extern int union_init (struct vfsconf *);
64 static int union_mount (struct mount *mp, char *path, caddr_t data,
66 static int union_root (struct mount *mp, struct vnode **vpp);
67 static int union_statfs (struct mount *mp, struct statfs *sbp,
69 static int union_unmount (struct mount *mp, int mntflags,
73 * Mount union filesystem
76 union_mount(struct mount *mp, char *path, caddr_t data, struct thread *td)
79 struct union_args args;
80 struct vnode *lowerrootvp = NULLVP;
81 struct vnode *upperrootvp = NULLVP;
82 struct union_mount *um = 0;
83 struct ucred *cred = 0;
84 struct nlookupdata nd;
89 UDEBUG(("union_mount(mp = %p)\n", (void *)mp));
91 KKASSERT(td->td_proc);
94 * Disable clustered write, otherwise system becomes unstable.
96 mp->mnt_flag |= MNT_NOCLUSTERW;
101 if (mp->mnt_flag & MNT_UPDATE) {
104 * 1. a way to convert between rdonly and rdwr mounts.
105 * 2. support for nfs exports.
114 error = copyin(data, (caddr_t)&args, sizeof(struct union_args));
119 * Obtain lower vnode. Vnode is stored in mp->mnt_vnodecovered.
120 * We need to reference it but not lock it.
123 lowerrootvp = mp->mnt_vnodecovered;
128 * Unlock lower node to avoid deadlock.
130 if (lowerrootvp->v_tag == VT_UNION)
131 VOP_UNLOCK(lowerrootvp, 0, td);
135 * Obtain upper vnode by calling nlookup() on the path. The
136 * upperrootvp will be turned referenced but not locked.
138 error = nlookup_init(&nd, args.target, UIO_USERSPACE, NLC_FOLLOW);
140 error = nlookup(&nd);
142 error = cache_vref(nd.nl_ncp, nd.nl_cred, &upperrootvp);
147 UDEBUG(("mount_root UPPERVP %p locked = %d\n", upperrootvp,
148 VOP_ISLOCKED(upperrootvp, NULL)));
151 * Check multi union mount to avoid `lock myself again' panic.
152 * Also require that it be a directory.
154 if (upperrootvp == VTOUNION(lowerrootvp)->un_uppervp) {
156 printf("union_mount: multi union mount?\n");
162 if (upperrootvp->v_type != VDIR) {
168 * Allocate our union_mount structure and populate the fields.
169 * The vnode references are stored in the union_mount as held,
170 * unlocked references. Depending on the _BELOW flag, the
171 * filesystems are viewed in a different order. In effect this
172 * is the same as providing a mount-under option to the mount
176 um = (struct union_mount *) malloc(sizeof(struct union_mount),
177 M_UNIONFSMNT, M_WAITOK);
179 bzero(um, sizeof(struct union_mount));
181 um->um_op = args.mntflags & UNMNT_OPMASK;
185 um->um_lowervp = lowerrootvp;
186 um->um_uppervp = upperrootvp;
192 um->um_lowervp = upperrootvp;
193 um->um_uppervp = lowerrootvp;
201 um->um_uppervp = upperrootvp;
202 um->um_lowervp = lowerrootvp;
212 * Unless the mount is readonly, ensure that the top layer
213 * supports whiteout operations
215 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
216 error = VOP_WHITEOUT(um->um_uppervp, NULL, NAMEI_LOOKUP);
221 um->um_cred = crhold(td->td_proc->p_ucred);
222 um->um_cmode = UN_DIRMODE & ~td->td_proc->p_fd->fd_cmask;
225 * Depending on what you think the MNT_LOCAL flag might mean,
226 * you may want the && to be || on the conditional below.
227 * At the moment it has been defined that the filesystem is
228 * only local if it is all local, ie the MNT_LOCAL flag implies
229 * that the entire namespace is local. If you think the MNT_LOCAL
230 * flag implies that some of the files might be stored locally
231 * then you will want to change the conditional.
233 if (um->um_op == UNMNT_ABOVE) {
234 if (((um->um_lowervp == NULLVP) ||
235 (um->um_lowervp->v_mount->mnt_flag & MNT_LOCAL)) &&
236 (um->um_uppervp->v_mount->mnt_flag & MNT_LOCAL))
237 mp->mnt_flag |= MNT_LOCAL;
241 * Copy in the upper layer's RDONLY flag. This is for the benefit
242 * of lookup() which explicitly checks the flag, rather than asking
243 * the filesystem for its own opinion. This means, that an update
244 * mount of the underlying filesystem to go from rdonly to rdwr
245 * will leave the unioned view as read-only.
247 mp->mnt_flag |= (um->um_uppervp->v_mount->mnt_flag & MNT_RDONLY);
249 mp->mnt_data = (qaddr_t) um;
264 bcopy(cp, mp->mnt_stat.f_mntfromname, len);
266 cp = mp->mnt_stat.f_mntfromname + len;
267 len = MNAMELEN - len;
269 (void) copyinstr(args.target, cp, len - 1, &size);
270 bzero(cp + size, len - size);
272 vfs_add_vnodeops(mp, &mp->mnt_vn_norm_ops, union_vnodeop_entries);
274 (void)union_statfs(mp, &mp->mnt_stat, td);
281 vrele(um->um_uppervp);
283 vrele(um->um_lowervp);
284 /* XXX other fields */
285 free(um, M_UNIONFSMNT);
297 * Free reference to union layer
300 union_unmount(struct mount *mp, int mntflags, struct thread *td)
302 struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
307 UDEBUG(("union_unmount(mp = %p)\n", (void *)mp));
309 if (mntflags & MNT_FORCE)
313 * Keep flushing vnodes from the mount list.
314 * This is needed because of the un_pvp held
315 * reference to the parent vnode.
316 * If more vnodes have been freed on a given pass,
317 * the try again. The loop will iterate at most
318 * (d) times, where (d) is the maximum tree depth
321 for (freeing = 0; (error = vflush(mp, 0, flags)) != 0;) {
322 int n = mp->mnt_nvnodelistsize;
324 /* if this is unchanged then stop */
328 /* otherwise try once more time */
332 /* If the most recent vflush failed, the filesystem is still busy. */
337 * Discard references to upper and lower target vnodes.
340 vrele(um->um_lowervp);
341 vrele(um->um_uppervp);
344 * Finally, throw away the union_mount structure
346 free(mp->mnt_data, M_UNIONFSMNT); /* XXX */
352 union_root(struct mount *mp, struct vnode **vpp)
354 struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
358 * Supply an unlocked reference to um_uppervp and to um_lowervp. It
359 * is possible for um_uppervp to be locked without the associated
360 * root union_node being locked. We let union_allocvp() deal with
363 UDEBUG(("union_root UPPERVP %p locked = %d\n", um->um_uppervp,
364 VOP_ISLOCKED(um->um_uppervp, NULL)));
366 vref(um->um_uppervp);
368 vref(um->um_lowervp);
370 error = union_allocvp(vpp, mp, NULLVP, NULLVP, NULL,
371 um->um_uppervp, um->um_lowervp, 1);
372 UDEBUG(("error %d\n", error));
373 UDEBUG(("union_root2 UPPERVP %p locked = %d\n", um->um_uppervp,
374 VOP_ISLOCKED(um->um_uppervp, NULL)));
380 union_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
383 struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
387 UDEBUG(("union_statfs(mp = %p, lvp = %p, uvp = %p)\n",
388 (void *)mp, (void *)um->um_lowervp, (void *)um->um_uppervp));
390 bzero(&mstat, sizeof(mstat));
392 if (um->um_lowervp) {
393 error = VFS_STATFS(um->um_lowervp->v_mount, &mstat, td);
398 /* now copy across the "interesting" information and fake the rest */
400 sbp->f_type = mstat.f_type;
401 sbp->f_flags = mstat.f_flags;
402 sbp->f_bsize = mstat.f_bsize;
403 sbp->f_iosize = mstat.f_iosize;
405 lbsize = mstat.f_bsize;
406 sbp->f_blocks = mstat.f_blocks;
407 sbp->f_bfree = mstat.f_bfree;
408 sbp->f_bavail = mstat.f_bavail;
409 sbp->f_files = mstat.f_files;
410 sbp->f_ffree = mstat.f_ffree;
412 error = VFS_STATFS(um->um_uppervp->v_mount, &mstat, td);
416 sbp->f_flags = mstat.f_flags;
417 sbp->f_bsize = mstat.f_bsize;
418 sbp->f_iosize = mstat.f_iosize;
421 * if the lower and upper blocksizes differ, then frig the
422 * block counts so that the sizes reported by df make some
423 * kind of sense. none of this makes sense though.
426 if (mstat.f_bsize != lbsize)
427 sbp->f_blocks = ((off_t) sbp->f_blocks * lbsize) / mstat.f_bsize;
430 * The "total" fields count total resources in all layers,
431 * the "free" fields count only those resources which are
432 * free in the upper layer (since only the upper layer
435 sbp->f_blocks += mstat.f_blocks;
436 sbp->f_bfree = mstat.f_bfree;
437 sbp->f_bavail = mstat.f_bavail;
438 sbp->f_files += mstat.f_files;
439 sbp->f_ffree = mstat.f_ffree;
441 if (sbp != &mp->mnt_stat) {
442 sbp->f_type = mp->mnt_vfc->vfc_typenum;
443 bcopy(&mp->mnt_stat.f_fsid, &sbp->f_fsid, sizeof(sbp->f_fsid));
444 bcopy(mp->mnt_stat.f_mntfromname, sbp->f_mntfromname, MNAMELEN);
449 static struct vfsops union_vfsops = {
451 vfs_stdstart, /* underlying start already done */
456 vfs_stdsync, /* XXX assumes no cached data on union level */
466 VFS_SET(union_vfsops, union, VFCF_LOOPBACK);