xref: /freebsd/sys/fs/unionfs/union_vfsops.c (revision dd41de95a84d979615a2ef11df6850622bf6184e)
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 	int		error;
78 	struct vnode   *lowerrootvp;
79 	struct vnode   *upperrootvp;
80 	struct unionfs_mount *ump;
81 	struct thread *td;
82 	char           *target;
83 	char           *tmp;
84 	char           *ep;
85 	int		len;
86 	int		below;
87 	uid_t		uid;
88 	gid_t		gid;
89 	u_short		udir;
90 	u_short		ufile;
91 	unionfs_copymode copymode;
92 	unionfs_whitemode whitemode;
93 	struct nameidata nd, *ndp;
94 	struct vattr	va;
95 
96 	UNIONFSDEBUG("unionfs_mount(mp = %p)\n", (void *)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 	td = curthread;
108 
109 	if (mp->mnt_flag & MNT_ROOTFS) {
110 		vfs_mount_error(mp, "Cannot union mount root filesystem");
111 		return (EOPNOTSUPP);
112 	}
113 
114 	/*
115 	 * Update is a no operation.
116 	 */
117 	if (mp->mnt_flag & MNT_UPDATE) {
118 		vfs_mount_error(mp, "unionfs does not support mount update");
119 		return (EOPNOTSUPP);
120 	}
121 
122 	/*
123 	 * Get argument
124 	 */
125 	error = vfs_getopt(mp->mnt_optnew, "target", (void **)&target, &len);
126 	if (error)
127 		error = vfs_getopt(mp->mnt_optnew, "from", (void **)&target,
128 		    &len);
129 	if (error || target[len - 1] != '\0') {
130 		vfs_mount_error(mp, "Invalid target");
131 		return (EINVAL);
132 	}
133 	if (vfs_getopt(mp->mnt_optnew, "below", NULL, NULL) == 0)
134 		below = 1;
135 	if (vfs_getopt(mp->mnt_optnew, "udir", (void **)&tmp, NULL) == 0) {
136 		if (tmp != NULL)
137 			udir = (mode_t)strtol(tmp, &ep, 8);
138 		if (tmp == NULL || *ep) {
139 			vfs_mount_error(mp, "Invalid udir");
140 			return (EINVAL);
141 		}
142 		udir &= S_IRWXU | S_IRWXG | S_IRWXO;
143 	}
144 	if (vfs_getopt(mp->mnt_optnew, "ufile", (void **)&tmp, NULL) == 0) {
145 		if (tmp != NULL)
146 			ufile = (mode_t)strtol(tmp, &ep, 8);
147 		if (tmp == NULL || *ep) {
148 			vfs_mount_error(mp, "Invalid ufile");
149 			return (EINVAL);
150 		}
151 		ufile &= S_IRWXU | S_IRWXG | S_IRWXO;
152 	}
153 	/* check umask, uid and gid */
154 	if (udir == 0 && ufile != 0)
155 		udir = ufile;
156 	if (ufile == 0 && udir != 0)
157 		ufile = udir;
158 
159 	vn_lock(mp->mnt_vnodecovered, LK_SHARED | LK_RETRY);
160 	error = VOP_GETATTR(mp->mnt_vnodecovered, &va, mp->mnt_cred);
161 	if (!error) {
162 		if (udir == 0)
163 			udir = va.va_mode;
164 		if (ufile == 0)
165 			ufile = va.va_mode;
166 		uid = va.va_uid;
167 		gid = va.va_gid;
168 	}
169 	VOP_UNLOCK(mp->mnt_vnodecovered);
170 	if (error)
171 		return (error);
172 
173 	if (mp->mnt_cred->cr_ruid == 0) {	/* root only */
174 		if (vfs_getopt(mp->mnt_optnew, "uid", (void **)&tmp,
175 		    NULL) == 0) {
176 			if (tmp != NULL)
177 				uid = (uid_t)strtol(tmp, &ep, 10);
178 			if (tmp == NULL || *ep) {
179 				vfs_mount_error(mp, "Invalid uid");
180 				return (EINVAL);
181 			}
182 		}
183 		if (vfs_getopt(mp->mnt_optnew, "gid", (void **)&tmp,
184 		    NULL) == 0) {
185 			if (tmp != NULL)
186 				gid = (gid_t)strtol(tmp, &ep, 10);
187 			if (tmp == NULL || *ep) {
188 				vfs_mount_error(mp, "Invalid gid");
189 				return (EINVAL);
190 			}
191 		}
192 		if (vfs_getopt(mp->mnt_optnew, "copymode", (void **)&tmp,
193 		    NULL) == 0) {
194 			if (tmp == NULL) {
195 				vfs_mount_error(mp, "Invalid copymode");
196 				return (EINVAL);
197 			} else if (strcasecmp(tmp, "traditional") == 0)
198 				copymode = UNIONFS_TRADITIONAL;
199 			else if (strcasecmp(tmp, "transparent") == 0)
200 				copymode = UNIONFS_TRANSPARENT;
201 			else if (strcasecmp(tmp, "masquerade") == 0)
202 				copymode = UNIONFS_MASQUERADE;
203 			else {
204 				vfs_mount_error(mp, "Invalid copymode");
205 				return (EINVAL);
206 			}
207 		}
208 		if (vfs_getopt(mp->mnt_optnew, "whiteout", (void **)&tmp,
209 		    NULL) == 0) {
210 			if (tmp == NULL) {
211 				vfs_mount_error(mp, "Invalid whiteout mode");
212 				return (EINVAL);
213 			} else if (strcasecmp(tmp, "always") == 0)
214 				whitemode = UNIONFS_WHITE_ALWAYS;
215 			else if (strcasecmp(tmp, "whenneeded") == 0)
216 				whitemode = UNIONFS_WHITE_WHENNEEDED;
217 			else {
218 				vfs_mount_error(mp, "Invalid whiteout mode");
219 				return (EINVAL);
220 			}
221 		}
222 	}
223 	/* If copymode is UNIONFS_TRADITIONAL, uid/gid is mounted user. */
224 	if (copymode == UNIONFS_TRADITIONAL) {
225 		uid = mp->mnt_cred->cr_ruid;
226 		gid = mp->mnt_cred->cr_rgid;
227 	}
228 
229 	UNIONFSDEBUG("unionfs_mount: uid=%d, gid=%d\n", uid, gid);
230 	UNIONFSDEBUG("unionfs_mount: udir=0%03o, ufile=0%03o\n", udir, ufile);
231 	UNIONFSDEBUG("unionfs_mount: copymode=%d\n", copymode);
232 
233 	/*
234 	 * Find upper node
235 	 */
236 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, target, td);
237 	if ((error = namei(ndp)))
238 		return (error);
239 
240 	NDFREE(ndp, NDF_ONLY_PNBUF);
241 
242 	/* get root vnodes */
243 	lowerrootvp = mp->mnt_vnodecovered;
244 	upperrootvp = ndp->ni_vp;
245 
246 	/* create unionfs_mount */
247 	ump = (struct unionfs_mount *)malloc(sizeof(struct unionfs_mount),
248 	    M_UNIONFSMNT, M_WAITOK | M_ZERO);
249 
250 	/*
251 	 * Save reference
252 	 */
253 	if (below) {
254 		VOP_UNLOCK(upperrootvp);
255 		vn_lock(lowerrootvp, LK_EXCLUSIVE | LK_RETRY);
256 		ump->um_lowervp = upperrootvp;
257 		ump->um_uppervp = lowerrootvp;
258 	} else {
259 		ump->um_lowervp = lowerrootvp;
260 		ump->um_uppervp = upperrootvp;
261 	}
262 	ump->um_rootvp = NULLVP;
263 	ump->um_uid = uid;
264 	ump->um_gid = gid;
265 	ump->um_udir = udir;
266 	ump->um_ufile = ufile;
267 	ump->um_copymode = copymode;
268 	ump->um_whitemode = whitemode;
269 
270 	mp->mnt_data = ump;
271 
272 	/*
273 	 * Copy upper layer's RDONLY flag.
274 	 */
275 	mp->mnt_flag |= ump->um_uppervp->v_mount->mnt_flag & MNT_RDONLY;
276 
277 	/*
278 	 * Unlock the node
279 	 */
280 	VOP_UNLOCK(ump->um_uppervp);
281 
282 	/*
283 	 * Get the unionfs root vnode.
284 	 */
285 	error = unionfs_nodeget(mp, ump->um_uppervp, ump->um_lowervp,
286 	    NULLVP, &(ump->um_rootvp), NULL, td);
287 	vrele(upperrootvp);
288 	if (error) {
289 		free(ump, M_UNIONFSMNT);
290 		mp->mnt_data = NULL;
291 		return (error);
292 	}
293 
294 	MNT_ILOCK(mp);
295 	if ((ump->um_lowervp->v_mount->mnt_flag & MNT_LOCAL) &&
296 	    (ump->um_uppervp->v_mount->mnt_flag & MNT_LOCAL))
297 		mp->mnt_flag |= MNT_LOCAL;
298 	mp->mnt_kern_flag |= MNTK_NOMSYNC | MNTK_UNIONFS;
299 	MNT_IUNLOCK(mp);
300 
301 	/*
302 	 * Get new fsid
303 	 */
304 	vfs_getnewfsid(mp);
305 
306 	snprintf(mp->mnt_stat.f_mntfromname, MNAMELEN, "<%s>:%s",
307 	    below ? "below" : "above", target);
308 
309 	UNIONFSDEBUG("unionfs_mount: from %s, on %s\n",
310 	    mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname);
311 
312 	return (0);
313 }
314 
315 /*
316  * Free reference to unionfs layer
317  */
318 static int
319 unionfs_unmount(struct mount *mp, int mntflags)
320 {
321 	struct unionfs_mount *ump;
322 	int		error;
323 	int		num;
324 	int		freeing;
325 	int		flags;
326 
327 	UNIONFSDEBUG("unionfs_unmount: mp = %p\n", (void *)mp);
328 
329 	ump = MOUNTTOUNIONFSMOUNT(mp);
330 	flags = 0;
331 
332 	if (mntflags & MNT_FORCE)
333 		flags |= FORCECLOSE;
334 
335 	/* vflush (no need to call vrele) */
336 	for (freeing = 0; (error = vflush(mp, 1, flags, curthread)) != 0;) {
337 		num = mp->mnt_nvnodelistsize;
338 		if (num == freeing)
339 			break;
340 		freeing = num;
341 	}
342 
343 	if (error)
344 		return (error);
345 
346 	free(ump, M_UNIONFSMNT);
347 	mp->mnt_data = NULL;
348 
349 	return (0);
350 }
351 
352 static int
353 unionfs_root(struct mount *mp, int flags, struct vnode **vpp)
354 {
355 	struct unionfs_mount *ump;
356 	struct vnode   *vp;
357 
358 	ump = MOUNTTOUNIONFSMOUNT(mp);
359 	vp = ump->um_rootvp;
360 
361 	UNIONFSDEBUG("unionfs_root: rootvp=%p locked=%x\n",
362 	    vp, VOP_ISLOCKED(vp));
363 
364 	vref(vp);
365 	if (flags & LK_TYPE_MASK)
366 		vn_lock(vp, flags);
367 
368 	*vpp = vp;
369 
370 	return (0);
371 }
372 
373 static int
374 unionfs_quotactl(struct mount *mp, int cmd, uid_t uid, void *arg,
375     bool *mp_busy)
376 {
377 	struct mount *uppermp;
378 	struct unionfs_mount *ump;
379 	int error;
380 	bool unbusy;
381 
382 	ump = MOUNTTOUNIONFSMOUNT(mp);
383 	uppermp = atomic_load_ptr(&ump->um_uppervp->v_mount);
384 	KASSERT(*mp_busy == true, ("upper mount not busy"));
385 	/*
386 	 * See comment in sys_quotactl() for an explanation of why the
387 	 * lower mount needs to be busied by the caller of VFS_QUOTACTL()
388 	 * but may be unbusied by the implementation.  We must unbusy
389 	 * the upper mount for the same reason; otherwise a namei lookup
390 	 * issued by the VFS_QUOTACTL() implementation could traverse the
391 	 * upper mount and deadlock.
392 	 */
393 	vfs_unbusy(mp);
394 	*mp_busy = false;
395 	unbusy = true;
396 	error = vfs_busy(uppermp, 0);
397 	/*
398 	 * Writing is always performed to upper vnode.
399 	 */
400 	if (error == 0)
401 		error = VFS_QUOTACTL(uppermp, cmd, uid, arg, &unbusy);
402 	if (unbusy)
403 		vfs_unbusy(uppermp);
404 
405 	return (error);
406 }
407 
408 static int
409 unionfs_statfs(struct mount *mp, struct statfs *sbp)
410 {
411 	struct unionfs_mount *ump;
412 	int		error;
413 	struct statfs	*mstat;
414 	uint64_t	lbsize;
415 
416 	ump = MOUNTTOUNIONFSMOUNT(mp);
417 
418 	UNIONFSDEBUG("unionfs_statfs(mp = %p, lvp = %p, uvp = %p)\n",
419 	    (void *)mp, (void *)ump->um_lowervp, (void *)ump->um_uppervp);
420 
421 	mstat = malloc(sizeof(struct statfs), M_STATFS, M_WAITOK | M_ZERO);
422 
423 	error = VFS_STATFS(ump->um_lowervp->v_mount, mstat);
424 	if (error) {
425 		free(mstat, M_STATFS);
426 		return (error);
427 	}
428 
429 	/* now copy across the "interesting" information and fake the rest */
430 	sbp->f_blocks = mstat->f_blocks;
431 	sbp->f_files = mstat->f_files;
432 
433 	lbsize = mstat->f_bsize;
434 
435 	error = VFS_STATFS(ump->um_uppervp->v_mount, mstat);
436 	if (error) {
437 		free(mstat, M_STATFS);
438 		return (error);
439 	}
440 
441 	/*
442 	 * The FS type etc is copy from upper vfs.
443 	 * (write able vfs have priority)
444 	 */
445 	sbp->f_type = mstat->f_type;
446 	sbp->f_flags = mstat->f_flags;
447 	sbp->f_bsize = mstat->f_bsize;
448 	sbp->f_iosize = mstat->f_iosize;
449 
450 	if (mstat->f_bsize != lbsize)
451 		sbp->f_blocks = ((off_t)sbp->f_blocks * lbsize) /
452 		    mstat->f_bsize;
453 
454 	sbp->f_blocks += mstat->f_blocks;
455 	sbp->f_bfree = mstat->f_bfree;
456 	sbp->f_bavail = mstat->f_bavail;
457 	sbp->f_files += mstat->f_files;
458 	sbp->f_ffree = mstat->f_ffree;
459 
460 	free(mstat, M_STATFS);
461 	return (0);
462 }
463 
464 static int
465 unionfs_sync(struct mount *mp, int waitfor)
466 {
467 	/* nothing to do */
468 	return (0);
469 }
470 
471 static int
472 unionfs_vget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
473 {
474 	return (EOPNOTSUPP);
475 }
476 
477 static int
478 unionfs_fhtovp(struct mount *mp, struct fid *fidp, int flags,
479     struct vnode **vpp)
480 {
481 	return (EOPNOTSUPP);
482 }
483 
484 static int
485 unionfs_checkexp(struct mount *mp, struct sockaddr *nam, uint64_t *extflagsp,
486     struct ucred **credanonp, int *numsecflavors, int *secflavors)
487 {
488 	return (EOPNOTSUPP);
489 }
490 
491 static int
492 unionfs_extattrctl(struct mount *mp, int cmd, struct vnode *filename_vp,
493     int namespace, const char *attrname)
494 {
495 	struct unionfs_mount *ump;
496 	struct unionfs_node *unp;
497 
498 	ump = MOUNTTOUNIONFSMOUNT(mp);
499 	unp = VTOUNIONFS(filename_vp);
500 
501 	if (unp->un_uppervp != NULLVP) {
502 		return (VFS_EXTATTRCTL(ump->um_uppervp->v_mount, cmd,
503 		    unp->un_uppervp, namespace, attrname));
504 	} else {
505 		return (VFS_EXTATTRCTL(ump->um_lowervp->v_mount, cmd,
506 		    unp->un_lowervp, namespace, attrname));
507 	}
508 }
509 
510 static struct vfsops unionfs_vfsops = {
511 	.vfs_checkexp =		unionfs_checkexp,
512 	.vfs_extattrctl =	unionfs_extattrctl,
513 	.vfs_fhtovp =		unionfs_fhtovp,
514 	.vfs_init =		unionfs_init,
515 	.vfs_mount =		unionfs_domount,
516 	.vfs_quotactl =		unionfs_quotactl,
517 	.vfs_root =		unionfs_root,
518 	.vfs_statfs =		unionfs_statfs,
519 	.vfs_sync =		unionfs_sync,
520 	.vfs_uninit =		unionfs_uninit,
521 	.vfs_unmount =		unionfs_unmount,
522 	.vfs_vget =		unionfs_vget,
523 };
524 
525 VFS_SET(unionfs_vfsops, unionfs, VFCF_LOOPBACK);
526