xref: /netbsd-src/sys/fs/unionfs/unionfs_vfsops.c (revision cd22f25e6f6d1cc1f197fe8c5468a80f51d1c4e1)
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 
51 #include <fs/unionfs/unionfs.h>
52 
53 MALLOC_DEFINE(M_UNIONFSMNT, "UNIONFS mount", "UNIONFS mount structure");
54 
55 struct vfsops unionfs_vfsops;
56 
57 VFS_PROTOS(unionfs);
58 
59 /*
60  * Mount unionfs layer.
61  */
62 int
63 unionfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
64 {
65 	int		error;
66 	struct vnode   *lowerrootvp;
67 	struct vnode   *upperrootvp;
68 	struct unionfs_mount *ump;
69 	int		below;
70 	uid_t		uid;
71 	gid_t		gid;
72 	u_short		udir;
73 	u_short		ufile;
74 	unionfs_copymode copymode;
75 	unionfs_whitemode whitemode;
76 	struct componentname fakecn;
77 	struct nameidata nd, *ndp;
78 	struct vattr	va;
79 	struct union_args *args = data;
80 	kauth_cred_t	cred;
81 	size_t		size;
82 	size_t		len;
83 	const char	*cp;
84 	char		*xp;
85 
86 	if (*data_len < sizeof *args)
87 		return EINVAL;
88 
89 	UNIONFSDEBUG("unionfs_mount(mp = %p)\n", (void *)mp);
90 
91 	error = 0;
92 	below = 0;
93 	uid = 0;
94 	gid = 0;
95 	udir = 0;
96 	ufile = 0;
97 	copymode = UNIONFS_TRANSPARENT;	/* default */
98 	whitemode = UNIONFS_WHITE_ALWAYS;
99 	ndp = &nd;
100 	cred = kauth_cred_get();
101 
102 	if (mp->mnt_flag & MNT_ROOTFS) {
103 		printf("union_mount: cannot union mount root filesystem\n");
104 		return (EOPNOTSUPP);
105 	}
106 
107 	if (mp->mnt_flag & MNT_GETARGS) {
108 		ump = MOUNTTOUNIONFSMOUNT(mp);
109 		if (ump == NULL)
110 			return EIO;
111 		args->target = NULL;
112 		args->mntflags = ump->um_op;
113 		*data_len = sizeof *args;
114 		return 0;
115 	}
116 
117 	/*
118 	 * Update is a no operation.
119 	 */
120 	if (mp->mnt_flag & MNT_UPDATE) {
121 		printf("union_mount: cannot update union mount\n");
122 		return (EOPNOTSUPP);
123 	}
124 
125 	vn_lock(mp->mnt_vnodecovered, LK_EXCLUSIVE | LK_RETRY);
126 	error = VOP_GETATTR(mp->mnt_vnodecovered, &va, cred);
127 	if (!error) {
128 		if (udir == 0)
129 			udir = va.va_mode;
130 		if (ufile == 0)
131 			ufile = va.va_mode;
132 		uid = va.va_uid;
133 		gid = va.va_gid;
134 	}
135 	VOP_UNLOCK(mp->mnt_vnodecovered, 0);
136 	if (error)
137 		return (error);
138 
139 	switch (args->mntflags & UNMNT_OPMASK) {
140 	case UNMNT_ABOVE:
141 		below = 0;
142 		break;
143 
144 	case UNMNT_BELOW:
145 		below = 1;
146 		break;
147 
148 	case UNMNT_REPLACE:
149 	default:
150 		return EINVAL;
151 	}
152 
153 	/* If copymode is UNIONFS_TRADITIONAL, uid/gid is mounted user. */
154 	if (copymode == UNIONFS_TRADITIONAL) {
155 		uid = kauth_cred_getuid(cred);
156 		gid = kauth_cred_getgid(cred);
157 	}
158 
159 	UNIONFSDEBUG("unionfs_mount: uid=%d, gid=%d\n", uid, gid);
160 	UNIONFSDEBUG("unionfs_mount: udir=0%03o, ufile=0%03o\n", udir, ufile);
161 	UNIONFSDEBUG("unionfs_mount: copymode=%d\n", copymode);
162 
163 	/*
164 	 * Find upper node
165 	 */
166 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, args->target);
167 	if ((error = namei(ndp)))
168 		return (error);
169 
170 	/* get root vnodes */
171 	lowerrootvp = mp->mnt_vnodecovered;
172 	upperrootvp = ndp->ni_vp;
173 
174 	vrele(ndp->ni_dvp);
175 	ndp->ni_dvp = NULLVP;
176 
177 	/* create unionfs_mount */
178 	ump = (struct unionfs_mount *)malloc(sizeof(struct unionfs_mount),
179 	    M_UNIONFSMNT, M_WAITOK | M_ZERO);
180 
181 	/*
182 	 * Save reference
183 	 */
184 	if (below) {
185 		VOP_UNLOCK(upperrootvp, 0);
186 		vn_lock(lowerrootvp, LK_EXCLUSIVE | LK_RETRY);
187 		ump->um_lowervp = upperrootvp;
188 		ump->um_uppervp = lowerrootvp;
189 	} else {
190 		ump->um_lowervp = lowerrootvp;
191 		ump->um_uppervp = upperrootvp;
192 	}
193 	ump->um_rootvp = NULLVP;
194 	ump->um_uid = uid;
195 	ump->um_gid = gid;
196 	ump->um_udir = udir;
197 	ump->um_ufile = ufile;
198 	ump->um_copymode = copymode;
199 	ump->um_whitemode = whitemode;
200 
201 	if ((lowerrootvp->v_mount->mnt_iflag & IMNT_MPSAFE) &&
202 	    (upperrootvp->v_mount->mnt_flag & IMNT_MPSAFE))
203 		mp->mnt_iflag |= IMNT_MPSAFE;
204 	mp->mnt_data = ump;
205 
206 	/*
207 	 * Copy upper layer's RDONLY flag.
208 	 */
209 	mp->mnt_flag |= ump->um_uppervp->v_mount->mnt_flag & MNT_RDONLY;
210 
211 	/*
212 	 * Check whiteout
213 	 */
214 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
215 		memset(&fakecn, 0, sizeof(fakecn));
216 		fakecn.cn_nameiop = LOOKUP;
217 		error = VOP_WHITEOUT(ump->um_uppervp, &fakecn, LOOKUP);
218 		if (error) {
219 			if (below) {
220 				VOP_UNLOCK(ump->um_uppervp, 0);
221 				vrele(upperrootvp);
222 			} else
223 				vput(ump->um_uppervp);
224 			free(ump, M_UNIONFSMNT);
225 			mp->mnt_data = NULL;
226 			return (error);
227 		}
228 	}
229 
230 	/*
231 	 * Unlock the node
232 	 */
233 	VOP_UNLOCK(ump->um_uppervp, 0);
234 
235 	ump->um_op = args->mntflags & UNMNT_OPMASK;
236 
237 	/*
238 	 * Get the unionfs root vnode.
239 	 */
240 	error = unionfs_nodeget(mp, ump->um_uppervp, ump->um_lowervp,
241 	    NULLVP, &(ump->um_rootvp), NULL);
242 	vrele(upperrootvp);
243 	if (error) {
244 		free(ump, M_UNIONFSMNT);
245 		mp->mnt_data = NULL;
246 		return (error);
247 	}
248 
249 	/*
250 	 * Check mnt_flag
251 	 */
252 	if ((ump->um_lowervp->v_mount->mnt_flag & MNT_LOCAL) &&
253 	    (ump->um_uppervp->v_mount->mnt_flag & MNT_LOCAL))
254 		mp->mnt_flag |= MNT_LOCAL;
255 
256 	/*
257 	 * Get new fsid
258 	 */
259 	vfs_getnewfsid(mp);
260 
261 	error = set_statvfs_info(path, UIO_USERSPACE, NULL, UIO_USERSPACE,
262 	    mp->mnt_op->vfs_name, mp, curlwp);
263 	if (error) {
264 		unionfs_noderem(ump->um_rootvp);
265 		free(ump, M_UNIONFSMNT);
266 		mp->mnt_data = NULL;
267 		return (error);
268 	}
269 
270 	switch (ump->um_op) {
271 	case UNMNT_ABOVE:
272 		cp = "<above>:";
273 		break;
274 	case UNMNT_BELOW:
275 		cp = "<below>:";
276 		break;
277 	default:
278 		panic("union_mount: bad um_op");
279 		break;
280 	}
281 	len = strlen(cp);
282 	memcpy(mp->mnt_stat.f_mntfromname, cp, len);
283 	xp = mp->mnt_stat.f_mntfromname + len;
284 	len = MNAMELEN - len;
285 	(void) copyinstr(args->target, xp, len - 1, &size);
286 	memset(xp + size, 0, len - size);
287 
288 	UNIONFSDEBUG("unionfs_mount: from %s, on %s\n",
289 	    mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname);
290 
291 	return (0);
292 }
293 
294 /*
295  * Free reference to unionfs layer
296  */
297 int
298 unionfs_unmount(struct mount *mp, int mntflags)
299 {
300 	struct unionfs_mount *ump;
301 	int		error;
302 	int		freeing;
303 	int		flags;
304 
305 	UNIONFSDEBUG("unionfs_unmount: mp = %p\n", (void *)mp);
306 
307 	ump = MOUNTTOUNIONFSMOUNT(mp);
308 	flags = 0;
309 
310 	if (mntflags & MNT_FORCE)
311 		flags |= FORCECLOSE;
312 
313 	/* vflush (no need to call vrele) */
314 	for (freeing = 0; (error = vflush(mp, NULL, flags)) != 0;) {
315 		struct vnode *vp;
316 		int n;
317 
318 		/* count #vnodes held on mount list */
319 		n = 0;
320 		TAILQ_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes)
321 			n++;
322 
323 		/* if this is unchanged then stop */
324 		if (n == freeing)
325 			break;
326 
327 		/* otherwise try once more time */
328 		freeing = n;
329 	}
330 
331 	if (error)
332 		return (error);
333 
334 	free(ump, M_UNIONFSMNT);
335 	mp->mnt_data = 0;
336 
337 	return (0);
338 }
339 
340 int
341 unionfs_root(struct mount *mp, struct vnode **vpp)
342 {
343 	struct unionfs_mount *ump;
344 	struct vnode   *vp;
345 
346 	ump = MOUNTTOUNIONFSMOUNT(mp);
347 	vp = ump->um_rootvp;
348 
349 	UNIONFSDEBUG("unionfs_root: rootvp=%p locked=%x\n",
350 	    vp, VOP_ISLOCKED(vp));
351 
352 	vref(vp);
353 	vn_lock(vp, LK_EXCLUSIVE);
354 
355 	*vpp = vp;
356 
357 	return (0);
358 }
359 
360 int
361 unionfs_quotactl(struct mount *mp, int cmd, uid_t uid, void *arg)
362 {
363 	struct unionfs_mount *ump;
364 
365 	ump = MOUNTTOUNIONFSMOUNT(mp);
366 
367 	/*
368 	 * Writing is always performed to upper vnode.
369 	 */
370 	return (VFS_QUOTACTL(ump->um_uppervp->v_mount, cmd, uid, arg));
371 }
372 
373 int
374 unionfs_statvfs(struct mount *mp, struct statvfs *sbp)
375 {
376 	struct unionfs_mount *ump;
377 	int		error;
378 	uint64_t	lbsize;
379 	struct statvfs *sbuf = malloc(sizeof(*sbuf), M_TEMP, M_WAITOK | M_ZERO);
380 
381 	ump = MOUNTTOUNIONFSMOUNT(mp);
382 
383 	UNIONFSDEBUG("unionfs_statvfs(mp = %p, lvp = %p, uvp = %p)\n",
384 	    (void *)mp, (void *)ump->um_lowervp, (void *)ump->um_uppervp);
385 
386 	error = VFS_STATVFS(ump->um_lowervp->v_mount, sbuf);
387 	if (error)
388 		goto done;
389 
390 	/* now copy across the "interesting" information and fake the rest */
391 	sbp->f_blocks = sbuf->f_blocks;
392 	sbp->f_files = sbuf->f_files;
393 
394 	lbsize = sbuf->f_bsize;
395 
396 	error = VFS_STATVFS(ump->um_uppervp->v_mount, sbuf);
397 	if (error)
398 		goto done;
399 
400 	/*
401 	 * The FS type etc is copy from upper vfs.
402 	 * (write able vfs have priority)
403 	 */
404 	sbp->f_flag = sbuf->f_flag;
405 	sbp->f_bsize = sbuf->f_bsize;
406 	sbp->f_iosize = sbuf->f_iosize;
407 
408 	if (sbuf->f_bsize != lbsize)
409 		sbp->f_blocks = ((off_t)sbp->f_blocks * lbsize) / sbuf->f_bsize;
410 
411 	sbp->f_blocks += sbuf->f_blocks;
412 	sbp->f_bfree = sbuf->f_bfree;
413 	sbp->f_bavail = sbuf->f_bavail;
414 	sbp->f_files += sbuf->f_files;
415 	sbp->f_ffree = sbuf->f_ffree;
416 
417  done:
418 	free(sbuf, M_TEMP);
419 	return (error);
420 }
421 
422 int
423 unionfs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
424 {
425 	/* nothing to do */
426 	return (0);
427 }
428 
429 int
430 unionfs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
431 {
432 	return (EOPNOTSUPP);
433 }
434 
435 int
436 unionfs_fhtovp(struct mount *mp, struct fid *fidp, struct vnode **vpp)
437 {
438 	return (EOPNOTSUPP);
439 }
440 
441 int
442 unionfs_extattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
443     int namespace, const char *attrname)
444 {
445 	struct unionfs_mount *ump;
446 	struct unionfs_node *unp;
447 
448 	ump = MOUNTTOUNIONFSMOUNT(mp);
449 	unp = VTOUNIONFS(filename_vp);
450 
451 	if (unp->un_uppervp != NULLVP) {
452 		return (VFS_EXTATTRCTL(ump->um_uppervp->v_mount, cmd,
453 		    unp->un_uppervp, namespace, attrname));
454 	} else {
455 		return (VFS_EXTATTRCTL(ump->um_lowervp->v_mount, cmd,
456 		    unp->un_lowervp, namespace, attrname));
457 	}
458 }
459 
460 /*
461  * Initialize
462  */
463 void
464 unionfs_init(void)
465 {
466 	UNIONFSDEBUG("unionfs_init\n");	/* printed during system boot */
467 }
468 
469 static int
470 unionfs_renamelock_enter(struct mount *mp)
471 {
472 	struct unionfs_mount *um = MOUNTTOUNIONFSMOUNT(mp);
473 
474 	/* Lock just the upper fs, where the action happens. */
475 	return VFS_RENAMELOCK_ENTER(um->um_uppervp->v_mount);
476 }
477 
478 static void
479 unionfs_renamelock_exit(struct mount *mp)
480 {
481 	struct unionfs_mount *um = MOUNTTOUNIONFSMOUNT(mp);
482 
483 	VFS_RENAMELOCK_EXIT(um->um_uppervp->v_mount);
484 }
485 
486 int
487 unionfs_start(struct mount *mp, int flags)
488 {
489 
490 	return (0);
491 }
492 
493 void
494 unionfs_done(void)
495 {
496 
497 	/* Make sure to unset the readdir hook. */
498 	vn_union_readdir_hook = NULL;
499 }
500 
501 SYSCTL_SETUP(sysctl_vfs_unionfs_setup, "sysctl vfs.union subtree setup")
502 {
503 
504 	sysctl_createv(clog, 0, NULL, NULL,
505 		       CTLFLAG_PERMANENT,
506 		       CTLTYPE_NODE, "vfs", NULL,
507 		       NULL, 0, NULL, 0,
508 		       CTL_VFS, CTL_EOL);
509 	sysctl_createv(clog, 0, NULL, NULL,
510 		       CTLFLAG_PERMANENT,
511 		       CTLTYPE_NODE, "union",
512 		       SYSCTL_DESCR("Union file system"),
513 		       NULL, 0, NULL, 0,
514 		       CTL_VFS, 15, CTL_EOL);
515 	/*
516 	 * XXX the "15" above could be dynamic, thereby eliminating
517 	 * one more instance of the "number to vfs" mapping problem,
518 	 * but "15" is the order as taken from sys/mount.h
519 	 */
520 }
521 
522 extern const struct vnodeopv_desc unionfs_vnodeop_opv_desc;
523 
524 const struct vnodeopv_desc * const unionfs_vnodeopv_descs[] = {
525 	&unionfs_vnodeop_opv_desc,
526 	NULL,
527 };
528 
529 struct vfsops unionfs_vfsops = {
530 	MOUNT_UNION,
531 	sizeof (struct unionfs_args),
532 	unionfs_mount,
533 	unionfs_start,
534 	unionfs_unmount,
535 	unionfs_root,
536 	(void *)eopnotsupp,		/* vfs_quotactl */
537 	unionfs_statvfs,
538 	unionfs_sync,
539 	unionfs_vget,
540 	(void *)eopnotsupp,		/* vfs_fhtovp */
541 	(void *)eopnotsupp,		/* vfs_vptofh */
542 	unionfs_init,
543 	NULL,				/* vfs_reinit */
544 	unionfs_done,
545 	NULL,				/* vfs_mountroot */
546 	(int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp,
547 	vfs_stdextattrctl,
548 	(void *)eopnotsupp,		/* vfs_suspendctl */
549 	unionfs_renamelock_enter,
550 	unionfs_renamelock_exit,
551 	(void *)eopnotsupp,
552 	unionfs_vnodeopv_descs,
553 	0,				/* vfs_refcount */
554 	{ NULL, NULL },
555 };
556 VFS_ATTACH(unionfs_vfsops);
557