xref: /freebsd/sys/fs/unionfs/union_vfsops.c (revision 8881d206f4e68b564c2c5f50fc717086fc3e827a)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1994, 1995 The Regents of the University of California.
5  * Copyright (c) 1994, 1995 Jan-Simon Pendry.
6  * Copyright (c) 2005, 2006, 2012 Masanori Ozawa <ozawa@ongs.co.jp>, ONGS Inc.
7  * Copyright (c) 2006, 2012 Daichi Goto <daichi@freebsd.org>
8  * All rights reserved.
9  *
10  * This code is derived from software donated to Berkeley by
11  * Jan-Simon Pendry.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1. Redistributions of source code must retain the above copyright
17  *    notice, this list of conditions and the following disclaimer.
18  * 2. Redistributions in binary form must reproduce the above copyright
19  *    notice, this list of conditions and the following disclaimer in the
20  *    documentation and/or other materials provided with the distribution.
21  * 3. 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.
24  *
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
35  * SUCH DAMAGE.
36  *
37  *	@(#)union_vfsops.c	8.20 (Berkeley) 5/20/95
38  * $FreeBSD$
39  */
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kdb.h>
44 #include <sys/fcntl.h>
45 #include <sys/kernel.h>
46 #include <sys/lock.h>
47 #include <sys/malloc.h>
48 #include <sys/mount.h>
49 #include <sys/namei.h>
50 #include <sys/proc.h>
51 #include <sys/vnode.h>
52 #include <sys/stat.h>
53 
54 #include <fs/unionfs/union.h>
55 
56 static MALLOC_DEFINE(M_UNIONFSMNT, "UNIONFS mount", "UNIONFS mount structure");
57 
58 static vfs_fhtovp_t	unionfs_fhtovp;
59 static vfs_checkexp_t	unionfs_checkexp;
60 static vfs_mount_t	unionfs_domount;
61 static vfs_quotactl_t	unionfs_quotactl;
62 static vfs_root_t	unionfs_root;
63 static vfs_sync_t	unionfs_sync;
64 static vfs_statfs_t	unionfs_statfs;
65 static vfs_unmount_t	unionfs_unmount;
66 static vfs_vget_t	unionfs_vget;
67 static vfs_extattrctl_t	unionfs_extattrctl;
68 
69 static struct vfsops unionfs_vfsops;
70 
71 /*
72  * Mount unionfs layer.
73  */
74 static int
75 unionfs_domount(struct mount *mp)
76 {
77 	struct mount   *lowermp, *uppermp;
78 	struct vnode   *lowerrootvp;
79 	struct vnode   *upperrootvp;
80 	struct unionfs_mount *ump;
81 	char           *target;
82 	char           *tmp;
83 	char           *ep;
84 	struct nameidata nd, *ndp;
85 	struct vattr	va;
86 	unionfs_copymode copymode;
87 	unionfs_whitemode whitemode;
88 	int		below;
89 	int		error;
90 	int		len;
91 	uid_t		uid;
92 	gid_t		gid;
93 	u_short		udir;
94 	u_short		ufile;
95 
96 	UNIONFSDEBUG("unionfs_mount(mp = %p)\n", mp);
97 
98 	error = 0;
99 	below = 0;
100 	uid = 0;
101 	gid = 0;
102 	udir = 0;
103 	ufile = 0;
104 	copymode = UNIONFS_TRANSPARENT;	/* default */
105 	whitemode = UNIONFS_WHITE_ALWAYS;
106 	ndp = &nd;
107 
108 	if (mp->mnt_flag & MNT_ROOTFS) {
109 		vfs_mount_error(mp, "Cannot union mount root filesystem");
110 		return (EOPNOTSUPP);
111 	}
112 
113 	/*
114 	 * Update is a no operation.
115 	 */
116 	if (mp->mnt_flag & MNT_UPDATE) {
117 		vfs_mount_error(mp, "unionfs does not support mount update");
118 		return (EOPNOTSUPP);
119 	}
120 
121 	/*
122 	 * Get argument
123 	 */
124 	error = vfs_getopt(mp->mnt_optnew, "target", (void **)&target, &len);
125 	if (error)
126 		error = vfs_getopt(mp->mnt_optnew, "from", (void **)&target,
127 		    &len);
128 	if (error || target[len - 1] != '\0') {
129 		vfs_mount_error(mp, "Invalid target");
130 		return (EINVAL);
131 	}
132 	if (vfs_getopt(mp->mnt_optnew, "below", NULL, NULL) == 0)
133 		below = 1;
134 	if (vfs_getopt(mp->mnt_optnew, "udir", (void **)&tmp, NULL) == 0) {
135 		if (tmp != NULL)
136 			udir = (mode_t)strtol(tmp, &ep, 8);
137 		if (tmp == NULL || *ep) {
138 			vfs_mount_error(mp, "Invalid udir");
139 			return (EINVAL);
140 		}
141 		udir &= S_IRWXU | S_IRWXG | S_IRWXO;
142 	}
143 	if (vfs_getopt(mp->mnt_optnew, "ufile", (void **)&tmp, NULL) == 0) {
144 		if (tmp != NULL)
145 			ufile = (mode_t)strtol(tmp, &ep, 8);
146 		if (tmp == NULL || *ep) {
147 			vfs_mount_error(mp, "Invalid ufile");
148 			return (EINVAL);
149 		}
150 		ufile &= S_IRWXU | S_IRWXG | S_IRWXO;
151 	}
152 	/* check umask, uid and gid */
153 	if (udir == 0 && ufile != 0)
154 		udir = ufile;
155 	if (ufile == 0 && udir != 0)
156 		ufile = udir;
157 
158 	vn_lock(mp->mnt_vnodecovered, LK_SHARED | LK_RETRY);
159 	error = VOP_GETATTR(mp->mnt_vnodecovered, &va, mp->mnt_cred);
160 	if (!error) {
161 		if (udir == 0)
162 			udir = va.va_mode;
163 		if (ufile == 0)
164 			ufile = va.va_mode;
165 		uid = va.va_uid;
166 		gid = va.va_gid;
167 	}
168 	VOP_UNLOCK(mp->mnt_vnodecovered);
169 	if (error)
170 		return (error);
171 
172 	if (mp->mnt_cred->cr_ruid == 0) {	/* root only */
173 		if (vfs_getopt(mp->mnt_optnew, "uid", (void **)&tmp,
174 		    NULL) == 0) {
175 			if (tmp != NULL)
176 				uid = (uid_t)strtol(tmp, &ep, 10);
177 			if (tmp == NULL || *ep) {
178 				vfs_mount_error(mp, "Invalid uid");
179 				return (EINVAL);
180 			}
181 		}
182 		if (vfs_getopt(mp->mnt_optnew, "gid", (void **)&tmp,
183 		    NULL) == 0) {
184 			if (tmp != NULL)
185 				gid = (gid_t)strtol(tmp, &ep, 10);
186 			if (tmp == NULL || *ep) {
187 				vfs_mount_error(mp, "Invalid gid");
188 				return (EINVAL);
189 			}
190 		}
191 		if (vfs_getopt(mp->mnt_optnew, "copymode", (void **)&tmp,
192 		    NULL) == 0) {
193 			if (tmp == NULL) {
194 				vfs_mount_error(mp, "Invalid copymode");
195 				return (EINVAL);
196 			} else if (strcasecmp(tmp, "traditional") == 0)
197 				copymode = UNIONFS_TRADITIONAL;
198 			else if (strcasecmp(tmp, "transparent") == 0)
199 				copymode = UNIONFS_TRANSPARENT;
200 			else if (strcasecmp(tmp, "masquerade") == 0)
201 				copymode = UNIONFS_MASQUERADE;
202 			else {
203 				vfs_mount_error(mp, "Invalid copymode");
204 				return (EINVAL);
205 			}
206 		}
207 		if (vfs_getopt(mp->mnt_optnew, "whiteout", (void **)&tmp,
208 		    NULL) == 0) {
209 			if (tmp == NULL) {
210 				vfs_mount_error(mp, "Invalid whiteout mode");
211 				return (EINVAL);
212 			} else if (strcasecmp(tmp, "always") == 0)
213 				whitemode = UNIONFS_WHITE_ALWAYS;
214 			else if (strcasecmp(tmp, "whenneeded") == 0)
215 				whitemode = UNIONFS_WHITE_WHENNEEDED;
216 			else {
217 				vfs_mount_error(mp, "Invalid whiteout mode");
218 				return (EINVAL);
219 			}
220 		}
221 	}
222 	/* If copymode is UNIONFS_TRADITIONAL, uid/gid is mounted user. */
223 	if (copymode == UNIONFS_TRADITIONAL) {
224 		uid = mp->mnt_cred->cr_ruid;
225 		gid = mp->mnt_cred->cr_rgid;
226 	}
227 
228 	UNIONFSDEBUG("unionfs_mount: uid=%d, gid=%d\n", uid, gid);
229 	UNIONFSDEBUG("unionfs_mount: udir=0%03o, ufile=0%03o\n", udir, ufile);
230 	UNIONFSDEBUG("unionfs_mount: copymode=%d\n", copymode);
231 
232 	/*
233 	 * Find upper node
234 	 */
235 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, target);
236 	if ((error = namei(ndp)))
237 		return (error);
238 
239 	NDFREE(ndp, NDF_ONLY_PNBUF);
240 
241 	/* get root vnodes */
242 	lowerrootvp = mp->mnt_vnodecovered;
243 	upperrootvp = ndp->ni_vp;
244 	KASSERT(lowerrootvp != NULL, ("%s: NULL lower root vp", __func__));
245 	KASSERT(upperrootvp != NULL, ("%s: NULL upper root vp", __func__));
246 
247 	/* create unionfs_mount */
248 	ump = malloc(sizeof(struct unionfs_mount), M_UNIONFSMNT,
249 	    M_WAITOK | M_ZERO);
250 
251 	/*
252 	 * Save reference
253 	 */
254 	if (below) {
255 		VOP_UNLOCK(upperrootvp);
256 		vn_lock(lowerrootvp, LK_EXCLUSIVE | LK_RETRY);
257 		ump->um_lowervp = upperrootvp;
258 		ump->um_uppervp = lowerrootvp;
259 	} else {
260 		ump->um_lowervp = lowerrootvp;
261 		ump->um_uppervp = upperrootvp;
262 	}
263 	ump->um_rootvp = NULLVP;
264 	ump->um_uid = uid;
265 	ump->um_gid = gid;
266 	ump->um_udir = udir;
267 	ump->um_ufile = ufile;
268 	ump->um_copymode = copymode;
269 	ump->um_whitemode = whitemode;
270 
271 	mp->mnt_data = ump;
272 
273 	/*
274 	 * Copy upper layer's RDONLY flag.
275 	 */
276 	mp->mnt_flag |= ump->um_uppervp->v_mount->mnt_flag & MNT_RDONLY;
277 
278 	/*
279 	 * Unlock the node
280 	 */
281 	VOP_UNLOCK(ump->um_uppervp);
282 
283 	/*
284 	 * Get the unionfs root vnode.
285 	 */
286 	error = unionfs_nodeget(mp, ump->um_uppervp, ump->um_lowervp,
287 	    NULLVP, &(ump->um_rootvp), NULL);
288 	vrele(upperrootvp);
289 	if (error != 0) {
290 		free(ump, M_UNIONFSMNT);
291 		mp->mnt_data = NULL;
292 		return (error);
293 	}
294 	KASSERT(ump->um_rootvp != NULL, ("rootvp cannot be NULL"));
295 	KASSERT((ump->um_rootvp->v_vflag & VV_ROOT) != 0,
296 	    ("%s: rootvp without VV_ROOT", __func__));
297 
298 	lowermp = vfs_register_upper_from_vp(ump->um_lowervp, mp,
299 	    &ump->um_lower_link);
300 	uppermp = vfs_register_upper_from_vp(ump->um_uppervp, mp,
301 	    &ump->um_upper_link);
302 
303 	if (lowermp == NULL || uppermp == NULL) {
304 		if (lowermp != NULL)
305 			vfs_unregister_upper(lowermp, &ump->um_lower_link);
306 		if (uppermp != NULL)
307 			vfs_unregister_upper(uppermp, &ump->um_upper_link);
308 		free(ump, M_UNIONFSMNT);
309 		mp->mnt_data = NULL;
310 		return (ENOENT);
311 	}
312 
313 	MNT_ILOCK(mp);
314 	if ((lowermp->mnt_flag & MNT_LOCAL) != 0 &&
315 	    (uppermp->mnt_flag & MNT_LOCAL) != 0)
316 		mp->mnt_flag |= MNT_LOCAL;
317 	mp->mnt_kern_flag |= MNTK_NOMSYNC | MNTK_UNIONFS;
318 	MNT_IUNLOCK(mp);
319 
320 	/*
321 	 * Get new fsid
322 	 */
323 	vfs_getnewfsid(mp);
324 
325 	snprintf(mp->mnt_stat.f_mntfromname, MNAMELEN, "<%s>:%s",
326 	    below ? "below" : "above", target);
327 
328 	UNIONFSDEBUG("unionfs_mount: from %s, on %s\n",
329 	    mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname);
330 
331 	return (0);
332 }
333 
334 /*
335  * Free reference to unionfs layer
336  */
337 static int
338 unionfs_unmount(struct mount *mp, int mntflags)
339 {
340 	struct unionfs_mount *ump;
341 	int		error;
342 	int		num;
343 	int		freeing;
344 	int		flags;
345 
346 	UNIONFSDEBUG("unionfs_unmount: mp = %p\n", mp);
347 
348 	ump = MOUNTTOUNIONFSMOUNT(mp);
349 	flags = 0;
350 
351 	if (mntflags & MNT_FORCE)
352 		flags |= FORCECLOSE;
353 
354 	/* vflush (no need to call vrele) */
355 	for (freeing = 0; (error = vflush(mp, 1, flags, curthread)) != 0;) {
356 		num = mp->mnt_nvnodelistsize;
357 		if (num == freeing)
358 			break;
359 		freeing = num;
360 	}
361 
362 	if (error)
363 		return (error);
364 
365 	vfs_unregister_upper(ump->um_lowervp->v_mount, &ump->um_lower_link);
366 	vfs_unregister_upper(ump->um_uppervp->v_mount, &ump->um_upper_link);
367 	free(ump, M_UNIONFSMNT);
368 	mp->mnt_data = NULL;
369 
370 	return (0);
371 }
372 
373 static int
374 unionfs_root(struct mount *mp, int flags, struct vnode **vpp)
375 {
376 	struct unionfs_mount *ump;
377 	struct vnode *vp;
378 
379 	ump = MOUNTTOUNIONFSMOUNT(mp);
380 	vp = ump->um_rootvp;
381 
382 	UNIONFSDEBUG("unionfs_root: rootvp=%p locked=%x\n",
383 	    vp, VOP_ISLOCKED(vp));
384 
385 	vref(vp);
386 	if (flags & LK_TYPE_MASK)
387 		vn_lock(vp, flags);
388 
389 	*vpp = vp;
390 
391 	return (0);
392 }
393 
394 static int
395 unionfs_quotactl(struct mount *mp, int cmd, uid_t uid, void *arg,
396     bool *mp_busy)
397 {
398 	struct mount *uppermp;
399 	struct unionfs_mount *ump;
400 	int error;
401 	bool unbusy;
402 
403 	ump = MOUNTTOUNIONFSMOUNT(mp);
404 	uppermp = atomic_load_ptr(&ump->um_uppervp->v_mount);
405 	KASSERT(*mp_busy == true, ("upper mount not busy"));
406 	/*
407 	 * See comment in sys_quotactl() for an explanation of why the
408 	 * lower mount needs to be busied by the caller of VFS_QUOTACTL()
409 	 * but may be unbusied by the implementation.  We must unbusy
410 	 * the upper mount for the same reason; otherwise a namei lookup
411 	 * issued by the VFS_QUOTACTL() implementation could traverse the
412 	 * upper mount and deadlock.
413 	 */
414 	vfs_unbusy(mp);
415 	*mp_busy = false;
416 	unbusy = true;
417 	error = vfs_busy(uppermp, 0);
418 	/*
419 	 * Writing is always performed to upper vnode.
420 	 */
421 	if (error == 0)
422 		error = VFS_QUOTACTL(uppermp, cmd, uid, arg, &unbusy);
423 	if (unbusy)
424 		vfs_unbusy(uppermp);
425 
426 	return (error);
427 }
428 
429 static int
430 unionfs_statfs(struct mount *mp, struct statfs *sbp)
431 {
432 	struct unionfs_mount *ump;
433 	struct statfs	*mstat;
434 	uint64_t	lbsize;
435 	int		error;
436 
437 	ump = MOUNTTOUNIONFSMOUNT(mp);
438 
439 	UNIONFSDEBUG("unionfs_statfs(mp = %p, lvp = %p, uvp = %p)\n",
440 	    mp, ump->um_lowervp, ump->um_uppervp);
441 
442 	mstat = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK | M_ZERO);
443 
444 	error = VFS_STATFS(ump->um_lowervp->v_mount, mstat);
445 	if (error) {
446 		free(mstat, M_STATFS);
447 		return (error);
448 	}
449 
450 	/* now copy across the "interesting" information and fake the rest */
451 	sbp->f_blocks = mstat->f_blocks;
452 	sbp->f_files = mstat->f_files;
453 
454 	lbsize = mstat->f_bsize;
455 
456 	error = VFS_STATFS(ump->um_uppervp->v_mount, mstat);
457 	if (error) {
458 		free(mstat, M_STATFS);
459 		return (error);
460 	}
461 
462 	/*
463 	 * The FS type etc is copy from upper vfs.
464 	 * (write able vfs have priority)
465 	 */
466 	sbp->f_type = mstat->f_type;
467 	sbp->f_flags = mstat->f_flags;
468 	sbp->f_bsize = mstat->f_bsize;
469 	sbp->f_iosize = mstat->f_iosize;
470 
471 	if (mstat->f_bsize != lbsize)
472 		sbp->f_blocks = ((off_t)sbp->f_blocks * lbsize) /
473 		    mstat->f_bsize;
474 
475 	sbp->f_blocks += mstat->f_blocks;
476 	sbp->f_bfree = mstat->f_bfree;
477 	sbp->f_bavail = mstat->f_bavail;
478 	sbp->f_files += mstat->f_files;
479 	sbp->f_ffree = mstat->f_ffree;
480 
481 	free(mstat, M_STATFS);
482 	return (0);
483 }
484 
485 static int
486 unionfs_sync(struct mount *mp, int waitfor)
487 {
488 	/* nothing to do */
489 	return (0);
490 }
491 
492 static int
493 unionfs_vget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
494 {
495 	return (EOPNOTSUPP);
496 }
497 
498 static int
499 unionfs_fhtovp(struct mount *mp, struct fid *fidp, int flags,
500     struct vnode **vpp)
501 {
502 	return (EOPNOTSUPP);
503 }
504 
505 static int
506 unionfs_checkexp(struct mount *mp, struct sockaddr *nam, uint64_t *extflagsp,
507     struct ucred **credanonp, int *numsecflavors, int *secflavors)
508 {
509 	return (EOPNOTSUPP);
510 }
511 
512 static int
513 unionfs_extattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
514     int namespace, const char *attrname)
515 {
516 	struct unionfs_mount *ump;
517 	struct unionfs_node *unp;
518 
519 	ump = MOUNTTOUNIONFSMOUNT(mp);
520 	unp = VTOUNIONFS(filename_vp);
521 
522 	if (unp->un_uppervp != NULLVP) {
523 		return (VFS_EXTATTRCTL(ump->um_uppervp->v_mount, cmd,
524 		    unp->un_uppervp, namespace, attrname));
525 	} else {
526 		return (VFS_EXTATTRCTL(ump->um_lowervp->v_mount, cmd,
527 		    unp->un_lowervp, namespace, attrname));
528 	}
529 }
530 
531 static struct vfsops unionfs_vfsops = {
532 	.vfs_checkexp =		unionfs_checkexp,
533 	.vfs_extattrctl =	unionfs_extattrctl,
534 	.vfs_fhtovp =		unionfs_fhtovp,
535 	.vfs_init =		unionfs_init,
536 	.vfs_mount =		unionfs_domount,
537 	.vfs_quotactl =		unionfs_quotactl,
538 	.vfs_root =		unionfs_root,
539 	.vfs_statfs =		unionfs_statfs,
540 	.vfs_sync =		unionfs_sync,
541 	.vfs_uninit =		unionfs_uninit,
542 	.vfs_unmount =		unionfs_unmount,
543 	.vfs_vget =		unionfs_vget,
544 };
545 
546 VFS_SET(unionfs_vfsops, unionfs, VFCF_LOOPBACK);
547