xref: /netbsd-src/sys/fs/unionfs/unionfs_vfsops.c (revision c2f76ff004a2cb67efe5b12d97bd3ef7fe89e18d)
1 /*-
2  * Copyright (c) 1994, 1995 The Regents of the University of California.
3  * Copyright (c) 1994, 1995 Jan-Simon Pendry.
4  * Copyright (c) 2005, 2006 Masanori Ozawa <ozawa@ongs.co.jp>, ONGS Inc.
5  * Copyright (c) 2006 Daichi Goto <daichi@freebsd.org>
6  * All rights reserved.
7  *
8  * This code is derived from software donated to Berkeley by
9  * Jan-Simon Pendry.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)union_vfsops.c	8.20 (Berkeley) 5/20/95
36  * $FreeBSD: src/sys/fs/unionfs/union_vfsops.c,v 1.89 2008/01/13 14:44:06 attilio Exp $
37  */
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/kernel.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/mount.h>
45 #include <sys/namei.h>
46 #include <sys/proc.h>
47 #include <sys/vnode.h>
48 #include <sys/stat.h>
49 #include <sys/sysctl.h>
50 #include <sys/module.h>
51 
52 #include <fs/unionfs/unionfs.h>
53 
54 MODULE(MODULE_CLASS_VFS, unionfs, "layerfs");
55 
56 MALLOC_DEFINE(M_UNIONFSMNT, "UNIONFS mount", "UNIONFS mount structure");
57 
58 struct vfsops unionfs_vfsops;
59 
60 VFS_PROTOS(unionfs);
61 
62 static struct sysctllog *unionfs_sysctl_log;
63 
64 /*
65  * Mount unionfs layer.
66  */
67 int
68 unionfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
69 {
70 	int		error;
71 	struct vnode   *lowerrootvp;
72 	struct vnode   *upperrootvp;
73 	struct unionfs_mount *ump;
74 	int		below;
75 	uid_t		uid;
76 	gid_t		gid;
77 	u_short		udir;
78 	u_short		ufile;
79 	unionfs_copymode copymode;
80 	unionfs_whitemode whitemode;
81 	struct pathbuf *pb;
82 	struct componentname fakecn;
83 	struct nameidata nd, *ndp;
84 	struct vattr	va;
85 	struct union_args *args = data;
86 	kauth_cred_t	cred;
87 	size_t		size;
88 	size_t		len;
89 	const char	*cp;
90 	char		*xp;
91 
92 	if (*data_len < sizeof *args)
93 		return EINVAL;
94 
95 	UNIONFSDEBUG("unionfs_mount(mp = %p)\n", (void *)mp);
96 
97 	error = 0;
98 	below = 0;
99 	uid = 0;
100 	gid = 0;
101 	udir = 0;
102 	ufile = 0;
103 	copymode = UNIONFS_TRANSPARENT;	/* default */
104 	whitemode = UNIONFS_WHITE_ALWAYS;
105 	ndp = &nd;
106 	cred = kauth_cred_get();
107 
108 	if (mp->mnt_flag & MNT_ROOTFS) {
109 		printf("union_mount: cannot union mount root filesystem\n");
110 		return (EOPNOTSUPP);
111 	}
112 
113 	if (mp->mnt_flag & MNT_GETARGS) {
114 		ump = MOUNTTOUNIONFSMOUNT(mp);
115 		if (ump == NULL)
116 			return EIO;
117 		args->target = NULL;
118 		args->mntflags = ump->um_op;
119 		*data_len = sizeof *args;
120 		return 0;
121 	}
122 
123 	/*
124 	 * Update is a no operation.
125 	 */
126 	if (mp->mnt_flag & MNT_UPDATE) {
127 		printf("union_mount: cannot update union mount\n");
128 		return (EOPNOTSUPP);
129 	}
130 
131 	vn_lock(mp->mnt_vnodecovered, LK_EXCLUSIVE | LK_RETRY);
132 	error = VOP_GETATTR(mp->mnt_vnodecovered, &va, cred);
133 	if (!error) {
134 		if (udir == 0)
135 			udir = va.va_mode;
136 		if (ufile == 0)
137 			ufile = va.va_mode;
138 		uid = va.va_uid;
139 		gid = va.va_gid;
140 	}
141 	VOP_UNLOCK(mp->mnt_vnodecovered);
142 	if (error)
143 		return (error);
144 
145 	switch (args->mntflags & UNMNT_OPMASK) {
146 	case UNMNT_ABOVE:
147 		below = 0;
148 		break;
149 
150 	case UNMNT_BELOW:
151 		below = 1;
152 		break;
153 
154 	case UNMNT_REPLACE:
155 	default:
156 		return EINVAL;
157 	}
158 
159 	/* If copymode is UNIONFS_TRADITIONAL, uid/gid is mounted user. */
160 	if (copymode == UNIONFS_TRADITIONAL) {
161 		uid = kauth_cred_getuid(cred);
162 		gid = kauth_cred_getgid(cred);
163 	}
164 
165 	UNIONFSDEBUG("unionfs_mount: uid=%d, gid=%d\n", uid, gid);
166 	UNIONFSDEBUG("unionfs_mount: udir=0%03o, ufile=0%03o\n", udir, ufile);
167 	UNIONFSDEBUG("unionfs_mount: copymode=%d\n", copymode);
168 
169 	/*
170 	 * Find upper node
171 	 */
172 	error = pathbuf_copyin(args->target, &pb);
173 	if (error) {
174 		return error;
175 	}
176 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, pb);
177 	if ((error = namei(ndp))) {
178 		pathbuf_destroy(pb);
179 		return error;
180 	}
181 
182 	/* get root vnodes */
183 	lowerrootvp = mp->mnt_vnodecovered;
184 	upperrootvp = ndp->ni_vp;
185 
186 	vrele(ndp->ni_dvp);
187 	ndp->ni_dvp = NULLVP;
188 	pathbuf_destroy(pb);
189 
190 	/* create unionfs_mount */
191 	ump = (struct unionfs_mount *)malloc(sizeof(struct unionfs_mount),
192 	    M_UNIONFSMNT, M_WAITOK | M_ZERO);
193 
194 	/*
195 	 * Save reference
196 	 */
197 	if (below) {
198 		VOP_UNLOCK(upperrootvp);
199 		vn_lock(lowerrootvp, LK_EXCLUSIVE | LK_RETRY);
200 		ump->um_lowervp = upperrootvp;
201 		ump->um_uppervp = lowerrootvp;
202 	} else {
203 		ump->um_lowervp = lowerrootvp;
204 		ump->um_uppervp = upperrootvp;
205 	}
206 	ump->um_rootvp = NULLVP;
207 	ump->um_uid = uid;
208 	ump->um_gid = gid;
209 	ump->um_udir = udir;
210 	ump->um_ufile = ufile;
211 	ump->um_copymode = copymode;
212 	ump->um_whitemode = whitemode;
213 
214 	if ((lowerrootvp->v_mount->mnt_iflag & IMNT_MPSAFE) &&
215 	    (upperrootvp->v_mount->mnt_flag & IMNT_MPSAFE))
216 		mp->mnt_iflag |= IMNT_MPSAFE;
217 	mp->mnt_data = ump;
218 
219 	/*
220 	 * Copy upper layer's RDONLY flag.
221 	 */
222 	mp->mnt_flag |= ump->um_uppervp->v_mount->mnt_flag & MNT_RDONLY;
223 
224 	/*
225 	 * Check whiteout
226 	 */
227 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
228 		memset(&fakecn, 0, sizeof(fakecn));
229 		fakecn.cn_nameiop = LOOKUP;
230 		error = VOP_WHITEOUT(ump->um_uppervp, &fakecn, LOOKUP);
231 		if (error) {
232 			if (below) {
233 				VOP_UNLOCK(ump->um_uppervp);
234 				vrele(upperrootvp);
235 			} else
236 				vput(ump->um_uppervp);
237 			free(ump, M_UNIONFSMNT);
238 			mp->mnt_data = NULL;
239 			return (error);
240 		}
241 	}
242 
243 	/*
244 	 * Unlock the node
245 	 */
246 	VOP_UNLOCK(ump->um_uppervp);
247 
248 	ump->um_op = args->mntflags & UNMNT_OPMASK;
249 
250 	/*
251 	 * Get the unionfs root vnode.
252 	 */
253 	error = unionfs_nodeget(mp, ump->um_uppervp, ump->um_lowervp,
254 	    NULLVP, &(ump->um_rootvp), NULL);
255 	vrele(upperrootvp);
256 	if (error) {
257 		free(ump, M_UNIONFSMNT);
258 		mp->mnt_data = NULL;
259 		return (error);
260 	}
261 
262 	/*
263 	 * Check mnt_flag
264 	 */
265 	if ((ump->um_lowervp->v_mount->mnt_flag & MNT_LOCAL) &&
266 	    (ump->um_uppervp->v_mount->mnt_flag & MNT_LOCAL))
267 		mp->mnt_flag |= MNT_LOCAL;
268 
269 	/*
270 	 * Get new fsid
271 	 */
272 	vfs_getnewfsid(mp);
273 
274 	error = set_statvfs_info(path, UIO_USERSPACE, NULL, UIO_USERSPACE,
275 	    mp->mnt_op->vfs_name, mp, curlwp);
276 	if (error) {
277 		unionfs_noderem(ump->um_rootvp);
278 		free(ump, M_UNIONFSMNT);
279 		mp->mnt_data = NULL;
280 		return (error);
281 	}
282 
283 	switch (ump->um_op) {
284 	case UNMNT_ABOVE:
285 		cp = "<above>:";
286 		break;
287 	case UNMNT_BELOW:
288 		cp = "<below>:";
289 		break;
290 	default:
291 		panic("union_mount: bad um_op");
292 		break;
293 	}
294 	len = strlen(cp);
295 	memcpy(mp->mnt_stat.f_mntfromname, cp, len);
296 	xp = mp->mnt_stat.f_mntfromname + len;
297 	len = MNAMELEN - len;
298 	(void) copyinstr(args->target, xp, len - 1, &size);
299 	memset(xp + size, 0, len - size);
300 
301 	UNIONFSDEBUG("unionfs_mount: from %s, on %s\n",
302 	    mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname);
303 
304 	return (0);
305 }
306 
307 /*
308  * Free reference to unionfs layer
309  */
310 int
311 unionfs_unmount(struct mount *mp, int mntflags)
312 {
313 	struct unionfs_mount *ump;
314 	int		error;
315 	int		freeing;
316 	int		flags;
317 
318 	UNIONFSDEBUG("unionfs_unmount: mp = %p\n", (void *)mp);
319 
320 	ump = MOUNTTOUNIONFSMOUNT(mp);
321 	flags = 0;
322 
323 	if (mntflags & MNT_FORCE)
324 		flags |= FORCECLOSE;
325 
326 	/* vflush (no need to call vrele) */
327 	for (freeing = 0; (error = vflush(mp, NULL, flags)) != 0;) {
328 		struct vnode *vp;
329 		int n;
330 
331 		/* count #vnodes held on mount list */
332 		n = 0;
333 		TAILQ_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes)
334 			n++;
335 
336 		/* if this is unchanged then stop */
337 		if (n == freeing)
338 			break;
339 
340 		/* otherwise try once more time */
341 		freeing = n;
342 	}
343 
344 	if (error)
345 		return (error);
346 
347 	free(ump, M_UNIONFSMNT);
348 	mp->mnt_data = NULL;
349 
350 	return (0);
351 }
352 
353 int
354 unionfs_root(struct mount *mp, struct vnode **vpp)
355 {
356 	struct unionfs_mount *ump;
357 	struct vnode   *vp;
358 
359 	ump = MOUNTTOUNIONFSMOUNT(mp);
360 	vp = ump->um_rootvp;
361 
362 	UNIONFSDEBUG("unionfs_root: rootvp=%p locked=%x\n",
363 	    vp, VOP_ISLOCKED(vp));
364 
365 	vref(vp);
366 	vn_lock(vp, LK_EXCLUSIVE);
367 
368 	*vpp = vp;
369 
370 	return (0);
371 }
372 
373 int
374 unionfs_quotactl(struct mount *mp, int cmd, uid_t uid, void *arg)
375 {
376 	struct unionfs_mount *ump;
377 
378 	ump = MOUNTTOUNIONFSMOUNT(mp);
379 
380 	/*
381 	 * Writing is always performed to upper vnode.
382 	 */
383 	return (VFS_QUOTACTL(ump->um_uppervp->v_mount, cmd, uid, arg));
384 }
385 
386 int
387 unionfs_statvfs(struct mount *mp, struct statvfs *sbp)
388 {
389 	struct unionfs_mount *ump;
390 	int		error;
391 	uint64_t	lbsize;
392 	struct statvfs *sbuf = malloc(sizeof(*sbuf), M_TEMP, M_WAITOK | M_ZERO);
393 
394 	ump = MOUNTTOUNIONFSMOUNT(mp);
395 
396 	UNIONFSDEBUG("unionfs_statvfs(mp = %p, lvp = %p, uvp = %p)\n",
397 	    (void *)mp, (void *)ump->um_lowervp, (void *)ump->um_uppervp);
398 
399 	error = VFS_STATVFS(ump->um_lowervp->v_mount, sbuf);
400 	if (error)
401 		goto done;
402 
403 	/* now copy across the "interesting" information and fake the rest */
404 	sbp->f_blocks = sbuf->f_blocks;
405 	sbp->f_files = sbuf->f_files;
406 
407 	lbsize = sbuf->f_bsize;
408 
409 	error = VFS_STATVFS(ump->um_uppervp->v_mount, sbuf);
410 	if (error)
411 		goto done;
412 
413 	/*
414 	 * The FS type etc is copy from upper vfs.
415 	 * (write able vfs have priority)
416 	 */
417 	sbp->f_flag = sbuf->f_flag;
418 	sbp->f_bsize = sbuf->f_bsize;
419 	sbp->f_iosize = sbuf->f_iosize;
420 
421 	if (sbuf->f_bsize != lbsize)
422 		sbp->f_blocks = ((off_t)sbp->f_blocks * lbsize) / sbuf->f_bsize;
423 
424 	sbp->f_blocks += sbuf->f_blocks;
425 	sbp->f_bfree = sbuf->f_bfree;
426 	sbp->f_bavail = sbuf->f_bavail;
427 	sbp->f_files += sbuf->f_files;
428 	sbp->f_ffree = sbuf->f_ffree;
429 
430  done:
431 	free(sbuf, M_TEMP);
432 	return (error);
433 }
434 
435 int
436 unionfs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
437 {
438 	/* nothing to do */
439 	return (0);
440 }
441 
442 int
443 unionfs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
444 {
445 	return (EOPNOTSUPP);
446 }
447 
448 int
449 unionfs_fhtovp(struct mount *mp, struct fid *fidp, struct vnode **vpp)
450 {
451 	return (EOPNOTSUPP);
452 }
453 
454 int
455 unionfs_extattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
456     int namespace, const char *attrname)
457 {
458 	struct unionfs_mount *ump;
459 	struct unionfs_node *unp;
460 
461 	ump = MOUNTTOUNIONFSMOUNT(mp);
462 	unp = VTOUNIONFS(filename_vp);
463 
464 	if (unp->un_uppervp != NULLVP) {
465 		return (VFS_EXTATTRCTL(ump->um_uppervp->v_mount, cmd,
466 		    unp->un_uppervp, namespace, attrname));
467 	} else {
468 		return (VFS_EXTATTRCTL(ump->um_lowervp->v_mount, cmd,
469 		    unp->un_lowervp, namespace, attrname));
470 	}
471 }
472 
473 /*
474  * Initialize
475  */
476 void
477 unionfs_init(void)
478 {
479 	UNIONFSDEBUG("unionfs_init\n");	/* printed during system boot */
480 }
481 
482 static int
483 unionfs_renamelock_enter(struct mount *mp)
484 {
485 	struct unionfs_mount *um = MOUNTTOUNIONFSMOUNT(mp);
486 
487 	/* Lock just the upper fs, where the action happens. */
488 	return VFS_RENAMELOCK_ENTER(um->um_uppervp->v_mount);
489 }
490 
491 static void
492 unionfs_renamelock_exit(struct mount *mp)
493 {
494 	struct unionfs_mount *um = MOUNTTOUNIONFSMOUNT(mp);
495 
496 	VFS_RENAMELOCK_EXIT(um->um_uppervp->v_mount);
497 }
498 
499 int
500 unionfs_start(struct mount *mp, int flags)
501 {
502 
503 	return (0);
504 }
505 
506 void
507 unionfs_done(void)
508 {
509 
510 	/* Make sure to unset the readdir hook. */
511 	vn_union_readdir_hook = NULL;
512 }
513 
514 extern const struct vnodeopv_desc unionfs_vnodeop_opv_desc;
515 
516 const struct vnodeopv_desc * const unionfs_vnodeopv_descs[] = {
517 	&unionfs_vnodeop_opv_desc,
518 	NULL,
519 };
520 
521 struct vfsops unionfs_vfsops = {
522 	MOUNT_UNION,
523 	sizeof (struct unionfs_args),
524 	unionfs_mount,
525 	unionfs_start,
526 	unionfs_unmount,
527 	unionfs_root,
528 	(void *)eopnotsupp,		/* vfs_quotactl */
529 	unionfs_statvfs,
530 	unionfs_sync,
531 	unionfs_vget,
532 	(void *)eopnotsupp,		/* vfs_fhtovp */
533 	(void *)eopnotsupp,		/* vfs_vptofh */
534 	unionfs_init,
535 	NULL,				/* vfs_reinit */
536 	unionfs_done,
537 	NULL,				/* vfs_mountroot */
538 	(int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp,
539 	vfs_stdextattrctl,
540 	(void *)eopnotsupp,		/* vfs_suspendctl */
541 	unionfs_renamelock_enter,
542 	unionfs_renamelock_exit,
543 	(void *)eopnotsupp,
544 	unionfs_vnodeopv_descs,
545 	0,				/* vfs_refcount */
546 	{ NULL, NULL },
547 };
548 
549 static int
550 unionfs_modcmd(modcmd_t cmd, void *arg)
551 {
552 	int error;
553 
554 	switch (cmd) {
555 	case MODULE_CMD_INIT:
556 		error = vfs_attach(&unionfs_vfsops);
557 		if (error != 0)
558 			break;
559 		sysctl_createv(&unionfs_sysctl_log, 0, NULL, NULL,
560 			       CTLFLAG_PERMANENT,
561 			       CTLTYPE_NODE, "vfs", NULL,
562 			       NULL, 0, NULL, 0,
563 			       CTL_VFS, CTL_EOL);
564 		sysctl_createv(&unionfs_sysctl_log, 0, NULL, NULL,
565 			       CTLFLAG_PERMANENT,
566 			       CTLTYPE_NODE, "union",
567 			       SYSCTL_DESCR("Union file system"),
568 			       NULL, 0, NULL, 0,
569 			       CTL_VFS, 15, CTL_EOL);
570 		/*
571 		 * XXX the "15" above could be dynamic, thereby eliminating
572 		 * one more instance of the "number to vfs" mapping problem,
573 		 * but "15" is the order as taken from sys/mount.h
574 		 */
575 		break;
576 	case MODULE_CMD_FINI:
577 		error = vfs_detach(&unionfs_vfsops);
578 		if (error != 0)
579 			break;
580 		sysctl_teardown(&unionfs_sysctl_log);
581 		break;
582 	default:
583 		error = ENOTTY;
584 		break;
585 	}
586 
587 	return (error);
588 }
589