union_vfsops.c revision 1.8.2.9 1 /* $NetBSD: union_vfsops.c,v 1.8.2.9 2005/11/10 14:09:44 skrll Exp $ */
2
3 /*
4 * Copyright (c) 1994 The Regents of the University of California.
5 * All rights reserved.
6 *
7 * This code is derived from software donated to Berkeley by
8 * Jan-Simon Pendry.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 *
34 * @(#)union_vfsops.c 8.20 (Berkeley) 5/20/95
35 */
36
37 /*
38 * Copyright (c) 1994 Jan-Simon Pendry.
39 * All rights reserved.
40 *
41 * This code is derived from software donated to Berkeley by
42 * Jan-Simon Pendry.
43 *
44 * Redistribution and use in source and binary forms, with or without
45 * modification, are permitted provided that the following conditions
46 * are met:
47 * 1. Redistributions of source code must retain the above copyright
48 * notice, this list of conditions and the following disclaimer.
49 * 2. Redistributions in binary form must reproduce the above copyright
50 * notice, this list of conditions and the following disclaimer in the
51 * documentation and/or other materials provided with the distribution.
52 * 3. All advertising materials mentioning features or use of this software
53 * must display the following acknowledgement:
54 * This product includes software developed by the University of
55 * California, Berkeley and its contributors.
56 * 4. Neither the name of the University nor the names of its contributors
57 * may be used to endorse or promote products derived from this software
58 * without specific prior written permission.
59 *
60 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
61 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
62 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
63 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
64 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
65 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
66 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
67 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
68 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
69 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
70 * SUCH DAMAGE.
71 *
72 * @(#)union_vfsops.c 8.20 (Berkeley) 5/20/95
73 */
74
75 /*
76 * Union Layer
77 */
78
79 #include <sys/cdefs.h>
80 __KERNEL_RCSID(0, "$NetBSD: union_vfsops.c,v 1.8.2.9 2005/11/10 14:09:44 skrll Exp $");
81
82 #include <sys/param.h>
83 #include <sys/systm.h>
84 #include <sys/sysctl.h>
85 #include <sys/time.h>
86 #include <sys/proc.h>
87 #include <sys/vnode.h>
88 #include <sys/mount.h>
89 #include <sys/namei.h>
90 #include <sys/malloc.h>
91 #include <sys/filedesc.h>
92 #include <sys/queue.h>
93 #include <sys/stat.h>
94
95 #include <fs/union/union.h>
96
97 int union_mount(struct mount *, const char *, void *, struct nameidata *,
98 struct lwp *);
99 int union_start(struct mount *, int, struct lwp *);
100 int union_unmount(struct mount *, int, struct lwp *);
101 int union_root(struct mount *, struct vnode **);
102 int union_quotactl(struct mount *, int, uid_t, void *, struct lwp *);
103 int union_statvfs(struct mount *, struct statvfs *, struct lwp *);
104 int union_sync(struct mount *, int, struct ucred *, struct lwp *);
105 int union_vget(struct mount *, ino_t, struct vnode **);
106
107 /*
108 * Mount union filesystem
109 */
110 int
111 union_mount(mp, path, data, ndp, l)
112 struct mount *mp;
113 const char *path;
114 void *data;
115 struct nameidata *ndp;
116 struct lwp *l;
117 {
118 int error = 0;
119 struct union_args args;
120 struct vnode *lowerrootvp = NULLVP;
121 struct vnode *upperrootvp = NULLVP;
122 struct union_mount *um = 0;
123 struct ucred *cred = 0;
124 const char *cp;
125 char *xp;
126 int len;
127 size_t size;
128
129 #ifdef UNION_DIAGNOSTIC
130 printf("union_mount(mp = %p)\n", mp);
131 #endif
132
133 if (mp->mnt_flag & MNT_GETARGS) {
134 um = MOUNTTOUNIONMOUNT(mp);
135 if (um == NULL)
136 return EIO;
137 args.target = NULL;
138 args.mntflags = um->um_op;
139 return copyout(&args, data, sizeof(args));
140 }
141 /*
142 * Update is a no-op
143 */
144 if (mp->mnt_flag & MNT_UPDATE) {
145 /*
146 * Need to provide.
147 * 1. a way to convert between rdonly and rdwr mounts.
148 * 2. support for nfs exports.
149 */
150 error = EOPNOTSUPP;
151 goto bad;
152 }
153
154 /*
155 * Get argument
156 */
157 error = copyin(data, &args, sizeof(struct union_args));
158 if (error)
159 goto bad;
160
161 lowerrootvp = mp->mnt_vnodecovered;
162 VREF(lowerrootvp);
163
164 /*
165 * Find upper node.
166 */
167 NDINIT(ndp, LOOKUP, FOLLOW,
168 UIO_USERSPACE, args.target, l);
169
170 if ((error = namei(ndp)) != 0)
171 goto bad;
172
173 upperrootvp = ndp->ni_vp;
174
175 if (upperrootvp->v_type != VDIR) {
176 error = EINVAL;
177 goto bad;
178 }
179
180 um = (struct union_mount *) malloc(sizeof(struct union_mount),
181 M_UFSMNT, M_WAITOK); /* XXX */
182
183 /*
184 * Keep a held reference to the target vnodes.
185 * They are vrele'd in union_unmount.
186 *
187 * Depending on the _BELOW flag, the filesystems are
188 * viewed in a different order. In effect, this is the
189 * same as providing a mount under option to the mount syscall.
190 */
191
192 um->um_op = args.mntflags & UNMNT_OPMASK;
193 switch (um->um_op) {
194 case UNMNT_ABOVE:
195 um->um_lowervp = lowerrootvp;
196 um->um_uppervp = upperrootvp;
197 break;
198
199 case UNMNT_BELOW:
200 um->um_lowervp = upperrootvp;
201 um->um_uppervp = lowerrootvp;
202 break;
203
204 case UNMNT_REPLACE:
205 vrele(lowerrootvp);
206 lowerrootvp = NULLVP;
207 um->um_uppervp = upperrootvp;
208 um->um_lowervp = lowerrootvp;
209 break;
210
211 default:
212 error = EINVAL;
213 goto bad;
214 }
215
216 /*
217 * Unless the mount is readonly, ensure that the top layer
218 * supports whiteout operations
219 */
220 if ((mp->mnt_flag & MNT_RDONLY) == 0) {
221 error = VOP_WHITEOUT(um->um_uppervp, (struct componentname *) 0, LOOKUP);
222 if (error)
223 goto bad;
224 }
225
226 um->um_cred = l->l_proc->p_ucred;
227 crhold(um->um_cred);
228 um->um_cmode = UN_DIRMODE &~ l->l_proc->p_cwdi->cwdi_cmask;
229
230 /*
231 * Depending on what you think the MNT_LOCAL flag might mean,
232 * you may want the && to be || on the conditional below.
233 * At the moment it has been defined that the filesystem is
234 * only local if it is all local, ie the MNT_LOCAL flag implies
235 * that the entire namespace is local. If you think the MNT_LOCAL
236 * flag implies that some of the files might be stored locally
237 * then you will want to change the conditional.
238 */
239 if (um->um_op == UNMNT_ABOVE) {
240 if (((um->um_lowervp == NULLVP) ||
241 (um->um_lowervp->v_mount->mnt_flag & MNT_LOCAL)) &&
242 (um->um_uppervp->v_mount->mnt_flag & MNT_LOCAL))
243 mp->mnt_flag |= MNT_LOCAL;
244 }
245
246 /*
247 * Copy in the upper layer's RDONLY flag. This is for the benefit
248 * of lookup() which explicitly checks the flag, rather than asking
249 * the filesystem for it's own opinion. This means, that an update
250 * mount of the underlying filesystem to go from rdonly to rdwr
251 * will leave the unioned view as read-only.
252 */
253 mp->mnt_flag |= (um->um_uppervp->v_mount->mnt_flag & MNT_RDONLY);
254
255 mp->mnt_data = um;
256 mp->mnt_leaf = um->um_uppervp->v_mount->mnt_leaf;
257 vfs_getnewfsid(mp);
258
259 error = set_statvfs_info( path, UIO_USERSPACE, NULL, UIO_USERSPACE,
260 mp, l);
261 if (error)
262 goto bad;
263
264 switch (um->um_op) {
265 case UNMNT_ABOVE:
266 cp = "<above>:";
267 break;
268 case UNMNT_BELOW:
269 cp = "<below>:";
270 break;
271 case UNMNT_REPLACE:
272 cp = "";
273 break;
274 default:
275 cp = "<invalid>:";
276 #ifdef DIAGNOSTIC
277 panic("union_mount: bad um_op");
278 #endif
279 break;
280 }
281 len = strlen(cp);
282 memcpy(mp->mnt_stat.f_mntfromname, cp, len);
283
284 xp = mp->mnt_stat.f_mntfromname + len;
285 len = MNAMELEN - len;
286
287 (void) copyinstr(args.target, xp, len - 1, &size);
288 memset(xp + size, 0, len - size);
289
290 #ifdef UNION_DIAGNOSTIC
291 printf("union_mount: from %s, on %s\n",
292 mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname);
293 #endif
294
295 /* Setup the readdir hook if it's not set already */
296 if (!vn_union_readdir_hook)
297 vn_union_readdir_hook = union_readdirhook;
298
299 return (0);
300
301 bad:
302 if (um)
303 free(um, M_UFSMNT);
304 if (cred)
305 crfree(cred);
306 if (upperrootvp)
307 vrele(upperrootvp);
308 if (lowerrootvp)
309 vrele(lowerrootvp);
310 return (error);
311 }
312
313 /*
314 * VFS start. Nothing needed here - the start routine
315 * on the underlying filesystem(s) will have been called
316 * when that filesystem was mounted.
317 */
318 /*ARGSUSED*/
319 int
320 union_start(mp, flags, l)
321 struct mount *mp;
322 int flags;
323 struct lwp *l;
324 {
325
326 return (0);
327 }
328
329 /*
330 * Free reference to union layer
331 */
332 int
333 union_unmount(mp, mntflags, l)
334 struct mount *mp;
335 int mntflags;
336 struct lwp *l;
337 {
338 struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
339 struct vnode *um_rootvp;
340 int error;
341 int freeing;
342
343 #ifdef UNION_DIAGNOSTIC
344 printf("union_unmount(mp = %p)\n", mp);
345 #endif
346
347 if ((error = union_root(mp, &um_rootvp)) != 0)
348 return (error);
349
350 /*
351 * Keep flushing vnodes from the mount list.
352 * This is needed because of the un_pvp held
353 * reference to the parent vnode.
354 * If more vnodes have been freed on a given pass,
355 * the try again. The loop will iterate at most
356 * (d) times, where (d) is the maximum tree depth
357 * in the filesystem.
358 */
359 for (freeing = 0; vflush(mp, um_rootvp, 0) != 0;) {
360 struct vnode *vp;
361 int n;
362
363 /* count #vnodes held on mount list */
364 for (n = 0, vp = mp->mnt_vnodelist.lh_first;
365 vp != NULLVP;
366 vp = vp->v_mntvnodes.le_next)
367 n++;
368
369 /* if this is unchanged then stop */
370 if (n == freeing)
371 break;
372
373 /* otherwise try once more time */
374 freeing = n;
375 }
376
377 /*
378 * Ok, now that we've tried doing it gently, get out the hammer.
379 */
380
381 if (mntflags & MNT_FORCE)
382 vflush(mp, um_rootvp, FORCECLOSE);
383
384
385 /* At this point the root vnode should have a single reference */
386 if (um_rootvp->v_usecount > 1) {
387 vput(um_rootvp);
388 return (EBUSY);
389 }
390
391 #ifdef UNION_DIAGNOSTIC
392 vprint("union root", um_rootvp);
393 #endif
394 /*
395 * Discard references to upper and lower target vnodes.
396 */
397 if (um->um_lowervp)
398 vrele(um->um_lowervp);
399 vrele(um->um_uppervp);
400 crfree(um->um_cred);
401 /*
402 * Release reference on underlying root vnode
403 */
404 vput(um_rootvp);
405 /*
406 * And blow it away for future re-use
407 */
408 vgone(um_rootvp);
409 /*
410 * Finally, throw away the union_mount structure
411 */
412 free(mp->mnt_data, M_UFSMNT); /* XXX */
413 mp->mnt_data = 0;
414 return (0);
415 }
416
417 int
418 union_root(mp, vpp)
419 struct mount *mp;
420 struct vnode **vpp;
421 {
422 struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
423 int error;
424 int loselock;
425
426 /*
427 * Return locked reference to root.
428 */
429 VREF(um->um_uppervp);
430 if ((um->um_op == UNMNT_BELOW) &&
431 VOP_ISLOCKED(um->um_uppervp)) {
432 loselock = 1;
433 } else {
434 vn_lock(um->um_uppervp, LK_EXCLUSIVE | LK_RETRY);
435 loselock = 0;
436 }
437 if (um->um_lowervp)
438 VREF(um->um_lowervp);
439 error = union_allocvp(vpp, mp,
440 (struct vnode *) 0,
441 (struct vnode *) 0,
442 (struct componentname *) 0,
443 um->um_uppervp,
444 um->um_lowervp,
445 1);
446
447 if (error) {
448 if (!loselock)
449 VOP_UNLOCK(um->um_uppervp, 0);
450 vrele(um->um_uppervp);
451 if (um->um_lowervp)
452 vrele(um->um_lowervp);
453 } else {
454 if (loselock)
455 VTOUNION(*vpp)->un_flags &= ~UN_ULOCK;
456 }
457
458 return (error);
459 }
460
461 /*ARGSUSED*/
462 int
463 union_quotactl(mp, cmd, uid, arg, l)
464 struct mount *mp;
465 int cmd;
466 uid_t uid;
467 void *arg;
468 struct lwp *l;
469 {
470
471 return (EOPNOTSUPP);
472 }
473
474 int
475 union_statvfs(mp, sbp, l)
476 struct mount *mp;
477 struct statvfs *sbp;
478 struct lwp *l;
479 {
480 int error;
481 struct union_mount *um = MOUNTTOUNIONMOUNT(mp);
482 struct statvfs *sbuf = malloc(sizeof(*sbuf), M_TEMP, M_WAITOK | M_ZERO);
483 unsigned long lbsize;
484
485 #ifdef UNION_DIAGNOSTIC
486 printf("union_statvfs(mp = %p, lvp = %p, uvp = %p)\n", mp,
487 um->um_lowervp, um->um_uppervp);
488 #endif
489
490 if (um->um_lowervp) {
491 error = VFS_STATVFS(um->um_lowervp->v_mount, sbuf, l);
492 if (error)
493 goto done;
494 }
495
496 /* now copy across the "interesting" information and fake the rest */
497 lbsize = sbuf->f_bsize;
498 sbp->f_blocks = sbuf->f_blocks - sbuf->f_bfree;
499 sbp->f_files = sbuf->f_files - sbuf->f_ffree;
500
501 error = VFS_STATVFS(um->um_uppervp->v_mount, sbuf, l);
502 if (error)
503 goto done;
504
505 sbp->f_flag = sbuf->f_flag;
506 sbp->f_bsize = sbuf->f_bsize;
507 sbp->f_frsize = sbuf->f_frsize;
508 sbp->f_iosize = sbuf->f_iosize;
509
510 /*
511 * The "total" fields count total resources in all layers,
512 * the "free" fields count only those resources which are
513 * free in the upper layer (since only the upper layer
514 * is writable).
515 */
516
517 if (sbuf->f_bsize != lbsize)
518 sbp->f_blocks = sbp->f_blocks * lbsize / sbuf->f_bsize;
519 sbp->f_blocks += sbuf->f_blocks;
520 sbp->f_bfree = sbuf->f_bfree;
521 sbp->f_bavail = sbuf->f_bavail;
522 sbp->f_bresvd = sbuf->f_bresvd;
523 sbp->f_files += sbuf->f_files;
524 sbp->f_ffree = sbuf->f_ffree;
525 sbp->f_favail = sbuf->f_favail;
526 sbp->f_fresvd = sbuf->f_fresvd;
527
528 copy_statvfs_info(sbp, mp);
529 done:
530 free(sbuf, M_TEMP);
531 return error;
532 }
533
534 /*ARGSUSED*/
535 int
536 union_sync(mp, waitfor, cred, l)
537 struct mount *mp;
538 int waitfor;
539 struct ucred *cred;
540 struct lwp *l;
541 {
542
543 /*
544 * XXX - Assumes no data cached at union layer.
545 */
546 return (0);
547 }
548
549 /*ARGSUSED*/
550 int
551 union_vget(mp, ino, vpp)
552 struct mount *mp;
553 ino_t ino;
554 struct vnode **vpp;
555 {
556
557 return (EOPNOTSUPP);
558 }
559
560 SYSCTL_SETUP(sysctl_vfs_union_setup, "sysctl vfs.union subtree setup")
561 {
562
563 sysctl_createv(clog, 0, NULL, NULL,
564 CTLFLAG_PERMANENT,
565 CTLTYPE_NODE, "vfs", NULL,
566 NULL, 0, NULL, 0,
567 CTL_VFS, CTL_EOL);
568 sysctl_createv(clog, 0, NULL, NULL,
569 CTLFLAG_PERMANENT,
570 CTLTYPE_NODE, "union",
571 SYSCTL_DESCR("Union file system"),
572 NULL, 0, NULL, 0,
573 CTL_VFS, 15, CTL_EOL);
574 /*
575 * XXX the "15" above could be dynamic, thereby eliminating
576 * one more instance of the "number to vfs" mapping problem,
577 * but "15" is the order as taken from sys/mount.h
578 */
579 }
580
581 extern const struct vnodeopv_desc union_vnodeop_opv_desc;
582
583 const struct vnodeopv_desc * const union_vnodeopv_descs[] = {
584 &union_vnodeop_opv_desc,
585 NULL,
586 };
587
588 struct vfsops union_vfsops = {
589 MOUNT_UNION,
590 union_mount,
591 union_start,
592 union_unmount,
593 union_root,
594 union_quotactl,
595 union_statvfs,
596 union_sync,
597 union_vget,
598 NULL, /* vfs_fhtovp */
599 NULL, /* vfs_vptofh */
600 union_init,
601 NULL, /* vfs_reinit */
602 union_done,
603 NULL, /* vfs_mountroot */
604 (int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp,
605 vfs_stdextattrctl,
606 union_vnodeopv_descs,
607 };
608 VFS_ATTACH(union_vfsops);
609