xref: /netbsd-src/sys/fs/unionfs/unionfs_vfsops.c (revision b1c86f5f087524e68db12794ee9c3e3da1ab17a0)
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 componentname fakecn;
82 	struct nameidata nd, *ndp;
83 	struct vattr	va;
84 	struct union_args *args = data;
85 	kauth_cred_t	cred;
86 	size_t		size;
87 	size_t		len;
88 	const char	*cp;
89 	char		*xp;
90 
91 	if (*data_len < sizeof *args)
92 		return EINVAL;
93 
94 	UNIONFSDEBUG("unionfs_mount(mp = %p)\n", (void *)mp);
95 
96 	error = 0;
97 	below = 0;
98 	uid = 0;
99 	gid = 0;
100 	udir = 0;
101 	ufile = 0;
102 	copymode = UNIONFS_TRANSPARENT;	/* default */
103 	whitemode = UNIONFS_WHITE_ALWAYS;
104 	ndp = &nd;
105 	cred = kauth_cred_get();
106 
107 	if (mp->mnt_flag & MNT_ROOTFS) {
108 		printf("union_mount: cannot union mount root filesystem\n");
109 		return (EOPNOTSUPP);
110 	}
111 
112 	if (mp->mnt_flag & MNT_GETARGS) {
113 		ump = MOUNTTOUNIONFSMOUNT(mp);
114 		if (ump == NULL)
115 			return EIO;
116 		args->target = NULL;
117 		args->mntflags = ump->um_op;
118 		*data_len = sizeof *args;
119 		return 0;
120 	}
121 
122 	/*
123 	 * Update is a no operation.
124 	 */
125 	if (mp->mnt_flag & MNT_UPDATE) {
126 		printf("union_mount: cannot update union mount\n");
127 		return (EOPNOTSUPP);
128 	}
129 
130 	vn_lock(mp->mnt_vnodecovered, LK_EXCLUSIVE | LK_RETRY);
131 	error = VOP_GETATTR(mp->mnt_vnodecovered, &va, cred);
132 	if (!error) {
133 		if (udir == 0)
134 			udir = va.va_mode;
135 		if (ufile == 0)
136 			ufile = va.va_mode;
137 		uid = va.va_uid;
138 		gid = va.va_gid;
139 	}
140 	VOP_UNLOCK(mp->mnt_vnodecovered);
141 	if (error)
142 		return (error);
143 
144 	switch (args->mntflags & UNMNT_OPMASK) {
145 	case UNMNT_ABOVE:
146 		below = 0;
147 		break;
148 
149 	case UNMNT_BELOW:
150 		below = 1;
151 		break;
152 
153 	case UNMNT_REPLACE:
154 	default:
155 		return EINVAL;
156 	}
157 
158 	/* If copymode is UNIONFS_TRADITIONAL, uid/gid is mounted user. */
159 	if (copymode == UNIONFS_TRADITIONAL) {
160 		uid = kauth_cred_getuid(cred);
161 		gid = kauth_cred_getgid(cred);
162 	}
163 
164 	UNIONFSDEBUG("unionfs_mount: uid=%d, gid=%d\n", uid, gid);
165 	UNIONFSDEBUG("unionfs_mount: udir=0%03o, ufile=0%03o\n", udir, ufile);
166 	UNIONFSDEBUG("unionfs_mount: copymode=%d\n", copymode);
167 
168 	/*
169 	 * Find upper node
170 	 */
171 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_USERSPACE, args->target);
172 	if ((error = namei(ndp)))
173 		return (error);
174 
175 	/* get root vnodes */
176 	lowerrootvp = mp->mnt_vnodecovered;
177 	upperrootvp = ndp->ni_vp;
178 
179 	vrele(ndp->ni_dvp);
180 	ndp->ni_dvp = NULLVP;
181 
182 	/* create unionfs_mount */
183 	ump = (struct unionfs_mount *)malloc(sizeof(struct unionfs_mount),
184 	    M_UNIONFSMNT, M_WAITOK | M_ZERO);
185 
186 	/*
187 	 * Save reference
188 	 */
189 	if (below) {
190 		VOP_UNLOCK(upperrootvp);
191 		vn_lock(lowerrootvp, LK_EXCLUSIVE | LK_RETRY);
192 		ump->um_lowervp = upperrootvp;
193 		ump->um_uppervp = lowerrootvp;
194 	} else {
195 		ump->um_lowervp = lowerrootvp;
196 		ump->um_uppervp = upperrootvp;
197 	}
198 	ump->um_rootvp = NULLVP;
199 	ump->um_uid = uid;
200 	ump->um_gid = gid;
201 	ump->um_udir = udir;
202 	ump->um_ufile = ufile;
203 	ump->um_copymode = copymode;
204 	ump->um_whitemode = whitemode;
205 
206 	if ((lowerrootvp->v_mount->mnt_iflag & IMNT_MPSAFE) &&
207 	    (upperrootvp->v_mount->mnt_flag & IMNT_MPSAFE))
208 		mp->mnt_iflag |= IMNT_MPSAFE;
209 	mp->mnt_data = ump;
210 
211 	/*
212 	 * Copy upper layer's RDONLY flag.
213 	 */
214 	mp->mnt_flag |= ump->um_uppervp->v_mount->mnt_flag & MNT_RDONLY;
215 
216 	/*
217 	 * Check whiteout
218 	 */
219 	if ((mp->mnt_flag & MNT_RDONLY) == 0) {
220 		memset(&fakecn, 0, sizeof(fakecn));
221 		fakecn.cn_nameiop = LOOKUP;
222 		error = VOP_WHITEOUT(ump->um_uppervp, &fakecn, LOOKUP);
223 		if (error) {
224 			if (below) {
225 				VOP_UNLOCK(ump->um_uppervp);
226 				vrele(upperrootvp);
227 			} else
228 				vput(ump->um_uppervp);
229 			free(ump, M_UNIONFSMNT);
230 			mp->mnt_data = NULL;
231 			return (error);
232 		}
233 	}
234 
235 	/*
236 	 * Unlock the node
237 	 */
238 	VOP_UNLOCK(ump->um_uppervp);
239 
240 	ump->um_op = args->mntflags & UNMNT_OPMASK;
241 
242 	/*
243 	 * Get the unionfs root vnode.
244 	 */
245 	error = unionfs_nodeget(mp, ump->um_uppervp, ump->um_lowervp,
246 	    NULLVP, &(ump->um_rootvp), NULL);
247 	vrele(upperrootvp);
248 	if (error) {
249 		free(ump, M_UNIONFSMNT);
250 		mp->mnt_data = NULL;
251 		return (error);
252 	}
253 
254 	/*
255 	 * Check mnt_flag
256 	 */
257 	if ((ump->um_lowervp->v_mount->mnt_flag & MNT_LOCAL) &&
258 	    (ump->um_uppervp->v_mount->mnt_flag & MNT_LOCAL))
259 		mp->mnt_flag |= MNT_LOCAL;
260 
261 	/*
262 	 * Get new fsid
263 	 */
264 	vfs_getnewfsid(mp);
265 
266 	error = set_statvfs_info(path, UIO_USERSPACE, NULL, UIO_USERSPACE,
267 	    mp->mnt_op->vfs_name, mp, curlwp);
268 	if (error) {
269 		unionfs_noderem(ump->um_rootvp);
270 		free(ump, M_UNIONFSMNT);
271 		mp->mnt_data = NULL;
272 		return (error);
273 	}
274 
275 	switch (ump->um_op) {
276 	case UNMNT_ABOVE:
277 		cp = "<above>:";
278 		break;
279 	case UNMNT_BELOW:
280 		cp = "<below>:";
281 		break;
282 	default:
283 		panic("union_mount: bad um_op");
284 		break;
285 	}
286 	len = strlen(cp);
287 	memcpy(mp->mnt_stat.f_mntfromname, cp, len);
288 	xp = mp->mnt_stat.f_mntfromname + len;
289 	len = MNAMELEN - len;
290 	(void) copyinstr(args->target, xp, len - 1, &size);
291 	memset(xp + size, 0, len - size);
292 
293 	UNIONFSDEBUG("unionfs_mount: from %s, on %s\n",
294 	    mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname);
295 
296 	return (0);
297 }
298 
299 /*
300  * Free reference to unionfs layer
301  */
302 int
303 unionfs_unmount(struct mount *mp, int mntflags)
304 {
305 	struct unionfs_mount *ump;
306 	int		error;
307 	int		freeing;
308 	int		flags;
309 
310 	UNIONFSDEBUG("unionfs_unmount: mp = %p\n", (void *)mp);
311 
312 	ump = MOUNTTOUNIONFSMOUNT(mp);
313 	flags = 0;
314 
315 	if (mntflags & MNT_FORCE)
316 		flags |= FORCECLOSE;
317 
318 	/* vflush (no need to call vrele) */
319 	for (freeing = 0; (error = vflush(mp, NULL, flags)) != 0;) {
320 		struct vnode *vp;
321 		int n;
322 
323 		/* count #vnodes held on mount list */
324 		n = 0;
325 		TAILQ_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes)
326 			n++;
327 
328 		/* if this is unchanged then stop */
329 		if (n == freeing)
330 			break;
331 
332 		/* otherwise try once more time */
333 		freeing = n;
334 	}
335 
336 	if (error)
337 		return (error);
338 
339 	free(ump, M_UNIONFSMNT);
340 	mp->mnt_data = NULL;
341 
342 	return (0);
343 }
344 
345 int
346 unionfs_root(struct mount *mp, struct vnode **vpp)
347 {
348 	struct unionfs_mount *ump;
349 	struct vnode   *vp;
350 
351 	ump = MOUNTTOUNIONFSMOUNT(mp);
352 	vp = ump->um_rootvp;
353 
354 	UNIONFSDEBUG("unionfs_root: rootvp=%p locked=%x\n",
355 	    vp, VOP_ISLOCKED(vp));
356 
357 	vref(vp);
358 	vn_lock(vp, LK_EXCLUSIVE);
359 
360 	*vpp = vp;
361 
362 	return (0);
363 }
364 
365 int
366 unionfs_quotactl(struct mount *mp, int cmd, uid_t uid, void *arg)
367 {
368 	struct unionfs_mount *ump;
369 
370 	ump = MOUNTTOUNIONFSMOUNT(mp);
371 
372 	/*
373 	 * Writing is always performed to upper vnode.
374 	 */
375 	return (VFS_QUOTACTL(ump->um_uppervp->v_mount, cmd, uid, arg));
376 }
377 
378 int
379 unionfs_statvfs(struct mount *mp, struct statvfs *sbp)
380 {
381 	struct unionfs_mount *ump;
382 	int		error;
383 	uint64_t	lbsize;
384 	struct statvfs *sbuf = malloc(sizeof(*sbuf), M_TEMP, M_WAITOK | M_ZERO);
385 
386 	ump = MOUNTTOUNIONFSMOUNT(mp);
387 
388 	UNIONFSDEBUG("unionfs_statvfs(mp = %p, lvp = %p, uvp = %p)\n",
389 	    (void *)mp, (void *)ump->um_lowervp, (void *)ump->um_uppervp);
390 
391 	error = VFS_STATVFS(ump->um_lowervp->v_mount, sbuf);
392 	if (error)
393 		goto done;
394 
395 	/* now copy across the "interesting" information and fake the rest */
396 	sbp->f_blocks = sbuf->f_blocks;
397 	sbp->f_files = sbuf->f_files;
398 
399 	lbsize = sbuf->f_bsize;
400 
401 	error = VFS_STATVFS(ump->um_uppervp->v_mount, sbuf);
402 	if (error)
403 		goto done;
404 
405 	/*
406 	 * The FS type etc is copy from upper vfs.
407 	 * (write able vfs have priority)
408 	 */
409 	sbp->f_flag = sbuf->f_flag;
410 	sbp->f_bsize = sbuf->f_bsize;
411 	sbp->f_iosize = sbuf->f_iosize;
412 
413 	if (sbuf->f_bsize != lbsize)
414 		sbp->f_blocks = ((off_t)sbp->f_blocks * lbsize) / sbuf->f_bsize;
415 
416 	sbp->f_blocks += sbuf->f_blocks;
417 	sbp->f_bfree = sbuf->f_bfree;
418 	sbp->f_bavail = sbuf->f_bavail;
419 	sbp->f_files += sbuf->f_files;
420 	sbp->f_ffree = sbuf->f_ffree;
421 
422  done:
423 	free(sbuf, M_TEMP);
424 	return (error);
425 }
426 
427 int
428 unionfs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
429 {
430 	/* nothing to do */
431 	return (0);
432 }
433 
434 int
435 unionfs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
436 {
437 	return (EOPNOTSUPP);
438 }
439 
440 int
441 unionfs_fhtovp(struct mount *mp, struct fid *fidp, struct vnode **vpp)
442 {
443 	return (EOPNOTSUPP);
444 }
445 
446 int
447 unionfs_extattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
448     int namespace, const char *attrname)
449 {
450 	struct unionfs_mount *ump;
451 	struct unionfs_node *unp;
452 
453 	ump = MOUNTTOUNIONFSMOUNT(mp);
454 	unp = VTOUNIONFS(filename_vp);
455 
456 	if (unp->un_uppervp != NULLVP) {
457 		return (VFS_EXTATTRCTL(ump->um_uppervp->v_mount, cmd,
458 		    unp->un_uppervp, namespace, attrname));
459 	} else {
460 		return (VFS_EXTATTRCTL(ump->um_lowervp->v_mount, cmd,
461 		    unp->un_lowervp, namespace, attrname));
462 	}
463 }
464 
465 /*
466  * Initialize
467  */
468 void
469 unionfs_init(void)
470 {
471 	UNIONFSDEBUG("unionfs_init\n");	/* printed during system boot */
472 }
473 
474 static int
475 unionfs_renamelock_enter(struct mount *mp)
476 {
477 	struct unionfs_mount *um = MOUNTTOUNIONFSMOUNT(mp);
478 
479 	/* Lock just the upper fs, where the action happens. */
480 	return VFS_RENAMELOCK_ENTER(um->um_uppervp->v_mount);
481 }
482 
483 static void
484 unionfs_renamelock_exit(struct mount *mp)
485 {
486 	struct unionfs_mount *um = MOUNTTOUNIONFSMOUNT(mp);
487 
488 	VFS_RENAMELOCK_EXIT(um->um_uppervp->v_mount);
489 }
490 
491 int
492 unionfs_start(struct mount *mp, int flags)
493 {
494 
495 	return (0);
496 }
497 
498 void
499 unionfs_done(void)
500 {
501 
502 	/* Make sure to unset the readdir hook. */
503 	vn_union_readdir_hook = NULL;
504 }
505 
506 extern const struct vnodeopv_desc unionfs_vnodeop_opv_desc;
507 
508 const struct vnodeopv_desc * const unionfs_vnodeopv_descs[] = {
509 	&unionfs_vnodeop_opv_desc,
510 	NULL,
511 };
512 
513 struct vfsops unionfs_vfsops = {
514 	MOUNT_UNION,
515 	sizeof (struct unionfs_args),
516 	unionfs_mount,
517 	unionfs_start,
518 	unionfs_unmount,
519 	unionfs_root,
520 	(void *)eopnotsupp,		/* vfs_quotactl */
521 	unionfs_statvfs,
522 	unionfs_sync,
523 	unionfs_vget,
524 	(void *)eopnotsupp,		/* vfs_fhtovp */
525 	(void *)eopnotsupp,		/* vfs_vptofh */
526 	unionfs_init,
527 	NULL,				/* vfs_reinit */
528 	unionfs_done,
529 	NULL,				/* vfs_mountroot */
530 	(int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp,
531 	vfs_stdextattrctl,
532 	(void *)eopnotsupp,		/* vfs_suspendctl */
533 	unionfs_renamelock_enter,
534 	unionfs_renamelock_exit,
535 	(void *)eopnotsupp,
536 	unionfs_vnodeopv_descs,
537 	0,				/* vfs_refcount */
538 	{ NULL, NULL },
539 };
540 
541 static int
542 unionfs_modcmd(modcmd_t cmd, void *arg)
543 {
544 	int error;
545 
546 	switch (cmd) {
547 	case MODULE_CMD_INIT:
548 		error = vfs_attach(&unionfs_vfsops);
549 		if (error != 0)
550 			break;
551 		sysctl_createv(&unionfs_sysctl_log, 0, NULL, NULL,
552 			       CTLFLAG_PERMANENT,
553 			       CTLTYPE_NODE, "vfs", NULL,
554 			       NULL, 0, NULL, 0,
555 			       CTL_VFS, CTL_EOL);
556 		sysctl_createv(&unionfs_sysctl_log, 0, NULL, NULL,
557 			       CTLFLAG_PERMANENT,
558 			       CTLTYPE_NODE, "union",
559 			       SYSCTL_DESCR("Union file system"),
560 			       NULL, 0, NULL, 0,
561 			       CTL_VFS, 15, CTL_EOL);
562 		/*
563 		 * XXX the "15" above could be dynamic, thereby eliminating
564 		 * one more instance of the "number to vfs" mapping problem,
565 		 * but "15" is the order as taken from sys/mount.h
566 		 */
567 		break;
568 	case MODULE_CMD_FINI:
569 		error = vfs_detach(&unionfs_vfsops);
570 		if (error != 0)
571 			break;
572 		sysctl_teardown(&unionfs_sysctl_log);
573 		break;
574 	default:
575 		error = ENOTTY;
576 		break;
577 	}
578 
579 	return (error);
580 }
581