xref: /freebsd/sys/kern/vfs_mount.c (revision 8a4217aacf57330755501a349d0ea662d4880386)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1999-2004 Poul-Henning Kamp
5  * Copyright (c) 1999 Michael Smith
6  * Copyright (c) 1989, 1993
7  *	The Regents of the University of California.  All rights reserved.
8  * (c) UNIX System Laboratories, Inc.
9  * All or some portions of this file are derived from material licensed
10  * to the University of California by American Telephone and Telegraph
11  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
12  * the permission of UNIX System Laboratories, Inc.
13  *
14  * Redistribution and use in source and binary forms, with or without
15  * modification, are permitted provided that the following conditions
16  * are met:
17  * 1. Redistributions of source code must retain the above copyright
18  *    notice, this list of conditions and the following disclaimer.
19  * 2. Redistributions in binary form must reproduce the above copyright
20  *    notice, this list of conditions and the following disclaimer in the
21  *    documentation and/or other materials provided with the distribution.
22  * 3. Neither the name of the University nor the names of its contributors
23  *    may be used to endorse or promote products derived from this software
24  *    without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36  * SUCH DAMAGE.
37  */
38 
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41 
42 #include <sys/param.h>
43 #include <sys/conf.h>
44 #include <sys/fcntl.h>
45 #include <sys/jail.h>
46 #include <sys/kernel.h>
47 #include <sys/libkern.h>
48 #include <sys/malloc.h>
49 #include <sys/mount.h>
50 #include <sys/mutex.h>
51 #include <sys/namei.h>
52 #include <sys/priv.h>
53 #include <sys/proc.h>
54 #include <sys/filedesc.h>
55 #include <sys/reboot.h>
56 #include <sys/sbuf.h>
57 #include <sys/syscallsubr.h>
58 #include <sys/sysproto.h>
59 #include <sys/sx.h>
60 #include <sys/sysctl.h>
61 #include <sys/sysent.h>
62 #include <sys/systm.h>
63 #include <sys/vnode.h>
64 #include <vm/uma.h>
65 
66 #include <geom/geom.h>
67 
68 #include <machine/stdarg.h>
69 
70 #include <security/audit/audit.h>
71 #include <security/mac/mac_framework.h>
72 
73 #define	VFS_MOUNTARG_SIZE_MAX	(1024 * 64)
74 
75 static int	vfs_domount(struct thread *td, const char *fstype, char *fspath,
76 		    uint64_t fsflags, struct vfsoptlist **optlist);
77 static void	free_mntarg(struct mntarg *ma);
78 
79 static int	usermount = 0;
80 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0,
81     "Unprivileged users may mount and unmount file systems");
82 
83 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
84 MALLOC_DEFINE(M_STATFS, "statfs", "statfs structure");
85 static uma_zone_t mount_zone;
86 
87 /* List of mounted filesystems. */
88 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
89 
90 /* For any iteration/modification of mountlist */
91 struct mtx mountlist_mtx;
92 MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF);
93 
94 /*
95  * Global opts, taken by all filesystems
96  */
97 static const char *global_opts[] = {
98 	"errmsg",
99 	"fstype",
100 	"fspath",
101 	"ro",
102 	"rw",
103 	"nosuid",
104 	"noexec",
105 	NULL
106 };
107 
108 static int
109 mount_init(void *mem, int size, int flags)
110 {
111 	struct mount *mp;
112 
113 	mp = (struct mount *)mem;
114 	mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF);
115 	mtx_init(&mp->mnt_listmtx, "struct mount vlist mtx", NULL, MTX_DEF);
116 	lockinit(&mp->mnt_explock, PVFS, "explock", 0, 0);
117 	return (0);
118 }
119 
120 static void
121 mount_fini(void *mem, int size)
122 {
123 	struct mount *mp;
124 
125 	mp = (struct mount *)mem;
126 	lockdestroy(&mp->mnt_explock);
127 	mtx_destroy(&mp->mnt_listmtx);
128 	mtx_destroy(&mp->mnt_mtx);
129 }
130 
131 static void
132 vfs_mount_init(void *dummy __unused)
133 {
134 
135 	mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount), NULL,
136 	    NULL, mount_init, mount_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
137 }
138 SYSINIT(vfs_mount, SI_SUB_VFS, SI_ORDER_ANY, vfs_mount_init, NULL);
139 
140 /*
141  * ---------------------------------------------------------------------
142  * Functions for building and sanitizing the mount options
143  */
144 
145 /* Remove one mount option. */
146 static void
147 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
148 {
149 
150 	TAILQ_REMOVE(opts, opt, link);
151 	free(opt->name, M_MOUNT);
152 	if (opt->value != NULL)
153 		free(opt->value, M_MOUNT);
154 	free(opt, M_MOUNT);
155 }
156 
157 /* Release all resources related to the mount options. */
158 void
159 vfs_freeopts(struct vfsoptlist *opts)
160 {
161 	struct vfsopt *opt;
162 
163 	while (!TAILQ_EMPTY(opts)) {
164 		opt = TAILQ_FIRST(opts);
165 		vfs_freeopt(opts, opt);
166 	}
167 	free(opts, M_MOUNT);
168 }
169 
170 void
171 vfs_deleteopt(struct vfsoptlist *opts, const char *name)
172 {
173 	struct vfsopt *opt, *temp;
174 
175 	if (opts == NULL)
176 		return;
177 	TAILQ_FOREACH_SAFE(opt, opts, link, temp)  {
178 		if (strcmp(opt->name, name) == 0)
179 			vfs_freeopt(opts, opt);
180 	}
181 }
182 
183 static int
184 vfs_isopt_ro(const char *opt)
185 {
186 
187 	if (strcmp(opt, "ro") == 0 || strcmp(opt, "rdonly") == 0 ||
188 	    strcmp(opt, "norw") == 0)
189 		return (1);
190 	return (0);
191 }
192 
193 static int
194 vfs_isopt_rw(const char *opt)
195 {
196 
197 	if (strcmp(opt, "rw") == 0 || strcmp(opt, "noro") == 0)
198 		return (1);
199 	return (0);
200 }
201 
202 /*
203  * Check if options are equal (with or without the "no" prefix).
204  */
205 static int
206 vfs_equalopts(const char *opt1, const char *opt2)
207 {
208 	char *p;
209 
210 	/* "opt" vs. "opt" or "noopt" vs. "noopt" */
211 	if (strcmp(opt1, opt2) == 0)
212 		return (1);
213 	/* "noopt" vs. "opt" */
214 	if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
215 		return (1);
216 	/* "opt" vs. "noopt" */
217 	if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
218 		return (1);
219 	while ((p = strchr(opt1, '.')) != NULL &&
220 	    !strncmp(opt1, opt2, ++p - opt1)) {
221 		opt2 += p - opt1;
222 		opt1 = p;
223 		/* "foo.noopt" vs. "foo.opt" */
224 		if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
225 			return (1);
226 		/* "foo.opt" vs. "foo.noopt" */
227 		if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
228 			return (1);
229 	}
230 	/* "ro" / "rdonly" / "norw" / "rw" / "noro" */
231 	if ((vfs_isopt_ro(opt1) || vfs_isopt_rw(opt1)) &&
232 	    (vfs_isopt_ro(opt2) || vfs_isopt_rw(opt2)))
233 		return (1);
234 	return (0);
235 }
236 
237 /*
238  * If a mount option is specified several times,
239  * (with or without the "no" prefix) only keep
240  * the last occurrence of it.
241  */
242 static void
243 vfs_sanitizeopts(struct vfsoptlist *opts)
244 {
245 	struct vfsopt *opt, *opt2, *tmp;
246 
247 	TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
248 		opt2 = TAILQ_PREV(opt, vfsoptlist, link);
249 		while (opt2 != NULL) {
250 			if (vfs_equalopts(opt->name, opt2->name)) {
251 				tmp = TAILQ_PREV(opt2, vfsoptlist, link);
252 				vfs_freeopt(opts, opt2);
253 				opt2 = tmp;
254 			} else {
255 				opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
256 			}
257 		}
258 	}
259 }
260 
261 /*
262  * Build a linked list of mount options from a struct uio.
263  */
264 int
265 vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
266 {
267 	struct vfsoptlist *opts;
268 	struct vfsopt *opt;
269 	size_t memused, namelen, optlen;
270 	unsigned int i, iovcnt;
271 	int error;
272 
273 	opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
274 	TAILQ_INIT(opts);
275 	memused = 0;
276 	iovcnt = auio->uio_iovcnt;
277 	for (i = 0; i < iovcnt; i += 2) {
278 		namelen = auio->uio_iov[i].iov_len;
279 		optlen = auio->uio_iov[i + 1].iov_len;
280 		memused += sizeof(struct vfsopt) + optlen + namelen;
281 		/*
282 		 * Avoid consuming too much memory, and attempts to overflow
283 		 * memused.
284 		 */
285 		if (memused > VFS_MOUNTARG_SIZE_MAX ||
286 		    optlen > VFS_MOUNTARG_SIZE_MAX ||
287 		    namelen > VFS_MOUNTARG_SIZE_MAX) {
288 			error = EINVAL;
289 			goto bad;
290 		}
291 
292 		opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
293 		opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
294 		opt->value = NULL;
295 		opt->len = 0;
296 		opt->pos = i / 2;
297 		opt->seen = 0;
298 
299 		/*
300 		 * Do this early, so jumps to "bad" will free the current
301 		 * option.
302 		 */
303 		TAILQ_INSERT_TAIL(opts, opt, link);
304 
305 		if (auio->uio_segflg == UIO_SYSSPACE) {
306 			bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
307 		} else {
308 			error = copyin(auio->uio_iov[i].iov_base, opt->name,
309 			    namelen);
310 			if (error)
311 				goto bad;
312 		}
313 		/* Ensure names are null-terminated strings. */
314 		if (namelen == 0 || opt->name[namelen - 1] != '\0') {
315 			error = EINVAL;
316 			goto bad;
317 		}
318 		if (optlen != 0) {
319 			opt->len = optlen;
320 			opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
321 			if (auio->uio_segflg == UIO_SYSSPACE) {
322 				bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
323 				    optlen);
324 			} else {
325 				error = copyin(auio->uio_iov[i + 1].iov_base,
326 				    opt->value, optlen);
327 				if (error)
328 					goto bad;
329 			}
330 		}
331 	}
332 	vfs_sanitizeopts(opts);
333 	*options = opts;
334 	return (0);
335 bad:
336 	vfs_freeopts(opts);
337 	return (error);
338 }
339 
340 /*
341  * Merge the old mount options with the new ones passed
342  * in the MNT_UPDATE case.
343  *
344  * XXX: This function will keep a "nofoo" option in the new
345  * options.  E.g, if the option's canonical name is "foo",
346  * "nofoo" ends up in the mount point's active options.
347  */
348 static void
349 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *oldopts)
350 {
351 	struct vfsopt *opt, *new;
352 
353 	TAILQ_FOREACH(opt, oldopts, link) {
354 		new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
355 		new->name = strdup(opt->name, M_MOUNT);
356 		if (opt->len != 0) {
357 			new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
358 			bcopy(opt->value, new->value, opt->len);
359 		} else
360 			new->value = NULL;
361 		new->len = opt->len;
362 		new->seen = opt->seen;
363 		TAILQ_INSERT_HEAD(toopts, new, link);
364 	}
365 	vfs_sanitizeopts(toopts);
366 }
367 
368 /*
369  * Mount a filesystem.
370  */
371 #ifndef _SYS_SYSPROTO_H_
372 struct nmount_args {
373 	struct iovec *iovp;
374 	unsigned int iovcnt;
375 	int flags;
376 };
377 #endif
378 int
379 sys_nmount(struct thread *td, struct nmount_args *uap)
380 {
381 	struct uio *auio;
382 	int error;
383 	u_int iovcnt;
384 	uint64_t flags;
385 
386 	/*
387 	 * Mount flags are now 64-bits. On 32-bit archtectures only
388 	 * 32-bits are passed in, but from here on everything handles
389 	 * 64-bit flags correctly.
390 	 */
391 	flags = uap->flags;
392 
393 	AUDIT_ARG_FFLAGS(flags);
394 	CTR4(KTR_VFS, "%s: iovp %p with iovcnt %d and flags %d", __func__,
395 	    uap->iovp, uap->iovcnt, flags);
396 
397 	/*
398 	 * Filter out MNT_ROOTFS.  We do not want clients of nmount() in
399 	 * userspace to set this flag, but we must filter it out if we want
400 	 * MNT_UPDATE on the root file system to work.
401 	 * MNT_ROOTFS should only be set by the kernel when mounting its
402 	 * root file system.
403 	 */
404 	flags &= ~MNT_ROOTFS;
405 
406 	iovcnt = uap->iovcnt;
407 	/*
408 	 * Check that we have an even number of iovec's
409 	 * and that we have at least two options.
410 	 */
411 	if ((iovcnt & 1) || (iovcnt < 4)) {
412 		CTR2(KTR_VFS, "%s: failed for invalid iovcnt %d", __func__,
413 		    uap->iovcnt);
414 		return (EINVAL);
415 	}
416 
417 	error = copyinuio(uap->iovp, iovcnt, &auio);
418 	if (error) {
419 		CTR2(KTR_VFS, "%s: failed for invalid uio op with %d errno",
420 		    __func__, error);
421 		return (error);
422 	}
423 	error = vfs_donmount(td, flags, auio);
424 
425 	free(auio, M_IOV);
426 	return (error);
427 }
428 
429 /*
430  * ---------------------------------------------------------------------
431  * Various utility functions
432  */
433 
434 void
435 vfs_ref(struct mount *mp)
436 {
437 
438 	CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
439 	MNT_ILOCK(mp);
440 	MNT_REF(mp);
441 	MNT_IUNLOCK(mp);
442 }
443 
444 void
445 vfs_rel(struct mount *mp)
446 {
447 
448 	CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
449 	MNT_ILOCK(mp);
450 	MNT_REL(mp);
451 	MNT_IUNLOCK(mp);
452 }
453 
454 /*
455  * Allocate and initialize the mount point struct.
456  */
457 struct mount *
458 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, const char *fspath,
459     struct ucred *cred)
460 {
461 	struct mount *mp;
462 
463 	mp = uma_zalloc(mount_zone, M_WAITOK);
464 	bzero(&mp->mnt_startzero,
465 	    __rangeof(struct mount, mnt_startzero, mnt_endzero));
466 	TAILQ_INIT(&mp->mnt_nvnodelist);
467 	mp->mnt_nvnodelistsize = 0;
468 	TAILQ_INIT(&mp->mnt_activevnodelist);
469 	mp->mnt_activevnodelistsize = 0;
470 	TAILQ_INIT(&mp->mnt_tmpfreevnodelist);
471 	mp->mnt_tmpfreevnodelistsize = 0;
472 	mp->mnt_ref = 0;
473 	(void) vfs_busy(mp, MBF_NOWAIT);
474 	atomic_add_acq_int(&vfsp->vfc_refcount, 1);
475 	mp->mnt_op = vfsp->vfc_vfsops;
476 	mp->mnt_vfc = vfsp;
477 	mp->mnt_stat.f_type = vfsp->vfc_typenum;
478 	mp->mnt_gen++;
479 	strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
480 	mp->mnt_vnodecovered = vp;
481 	mp->mnt_cred = crdup(cred);
482 	mp->mnt_stat.f_owner = cred->cr_uid;
483 	strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
484 	mp->mnt_iosize_max = DFLTPHYS;
485 #ifdef MAC
486 	mac_mount_init(mp);
487 	mac_mount_create(cred, mp);
488 #endif
489 	arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0);
490 	TAILQ_INIT(&mp->mnt_uppers);
491 	return (mp);
492 }
493 
494 /*
495  * Destroy the mount struct previously allocated by vfs_mount_alloc().
496  */
497 void
498 vfs_mount_destroy(struct mount *mp)
499 {
500 
501 	MNT_ILOCK(mp);
502 	mp->mnt_kern_flag |= MNTK_REFEXPIRE;
503 	if (mp->mnt_kern_flag & MNTK_MWAIT) {
504 		mp->mnt_kern_flag &= ~MNTK_MWAIT;
505 		wakeup(mp);
506 	}
507 	while (mp->mnt_ref)
508 		msleep(mp, MNT_MTX(mp), PVFS, "mntref", 0);
509 	KASSERT(mp->mnt_ref == 0,
510 	    ("%s: invalid refcount in the drain path @ %s:%d", __func__,
511 	    __FILE__, __LINE__));
512 	if (mp->mnt_writeopcount != 0)
513 		panic("vfs_mount_destroy: nonzero writeopcount");
514 	if (mp->mnt_secondary_writes != 0)
515 		panic("vfs_mount_destroy: nonzero secondary_writes");
516 	atomic_subtract_rel_int(&mp->mnt_vfc->vfc_refcount, 1);
517 	if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
518 		struct vnode *vp;
519 
520 		TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
521 			vn_printf(vp, "dangling vnode ");
522 		panic("unmount: dangling vnode");
523 	}
524 	KASSERT(TAILQ_EMPTY(&mp->mnt_uppers), ("mnt_uppers"));
525 	if (mp->mnt_nvnodelistsize != 0)
526 		panic("vfs_mount_destroy: nonzero nvnodelistsize");
527 	if (mp->mnt_activevnodelistsize != 0)
528 		panic("vfs_mount_destroy: nonzero activevnodelistsize");
529 	if (mp->mnt_tmpfreevnodelistsize != 0)
530 		panic("vfs_mount_destroy: nonzero tmpfreevnodelistsize");
531 	if (mp->mnt_lockref != 0)
532 		panic("vfs_mount_destroy: nonzero lock refcount");
533 	MNT_IUNLOCK(mp);
534 	if (mp->mnt_vnodecovered != NULL)
535 		vrele(mp->mnt_vnodecovered);
536 #ifdef MAC
537 	mac_mount_destroy(mp);
538 #endif
539 	if (mp->mnt_opt != NULL)
540 		vfs_freeopts(mp->mnt_opt);
541 	crfree(mp->mnt_cred);
542 	uma_zfree(mount_zone, mp);
543 }
544 
545 int
546 vfs_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions)
547 {
548 	struct vfsoptlist *optlist;
549 	struct vfsopt *opt, *tmp_opt;
550 	char *fstype, *fspath, *errmsg;
551 	int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
552 
553 	errmsg = fspath = NULL;
554 	errmsg_len = fspathlen = 0;
555 	errmsg_pos = -1;
556 
557 	error = vfs_buildopts(fsoptions, &optlist);
558 	if (error)
559 		return (error);
560 
561 	if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
562 		errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
563 
564 	/*
565 	 * We need these two options before the others,
566 	 * and they are mandatory for any filesystem.
567 	 * Ensure they are NUL terminated as well.
568 	 */
569 	fstypelen = 0;
570 	error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
571 	if (error || fstype[fstypelen - 1] != '\0') {
572 		error = EINVAL;
573 		if (errmsg != NULL)
574 			strncpy(errmsg, "Invalid fstype", errmsg_len);
575 		goto bail;
576 	}
577 	fspathlen = 0;
578 	error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
579 	if (error || fspath[fspathlen - 1] != '\0') {
580 		error = EINVAL;
581 		if (errmsg != NULL)
582 			strncpy(errmsg, "Invalid fspath", errmsg_len);
583 		goto bail;
584 	}
585 
586 	/*
587 	 * We need to see if we have the "update" option
588 	 * before we call vfs_domount(), since vfs_domount() has special
589 	 * logic based on MNT_UPDATE.  This is very important
590 	 * when we want to update the root filesystem.
591 	 */
592 	TAILQ_FOREACH_SAFE(opt, optlist, link, tmp_opt) {
593 		if (strcmp(opt->name, "update") == 0) {
594 			fsflags |= MNT_UPDATE;
595 			vfs_freeopt(optlist, opt);
596 		}
597 		else if (strcmp(opt->name, "async") == 0)
598 			fsflags |= MNT_ASYNC;
599 		else if (strcmp(opt->name, "force") == 0) {
600 			fsflags |= MNT_FORCE;
601 			vfs_freeopt(optlist, opt);
602 		}
603 		else if (strcmp(opt->name, "reload") == 0) {
604 			fsflags |= MNT_RELOAD;
605 			vfs_freeopt(optlist, opt);
606 		}
607 		else if (strcmp(opt->name, "multilabel") == 0)
608 			fsflags |= MNT_MULTILABEL;
609 		else if (strcmp(opt->name, "noasync") == 0)
610 			fsflags &= ~MNT_ASYNC;
611 		else if (strcmp(opt->name, "noatime") == 0)
612 			fsflags |= MNT_NOATIME;
613 		else if (strcmp(opt->name, "atime") == 0) {
614 			free(opt->name, M_MOUNT);
615 			opt->name = strdup("nonoatime", M_MOUNT);
616 		}
617 		else if (strcmp(opt->name, "noclusterr") == 0)
618 			fsflags |= MNT_NOCLUSTERR;
619 		else if (strcmp(opt->name, "clusterr") == 0) {
620 			free(opt->name, M_MOUNT);
621 			opt->name = strdup("nonoclusterr", M_MOUNT);
622 		}
623 		else if (strcmp(opt->name, "noclusterw") == 0)
624 			fsflags |= MNT_NOCLUSTERW;
625 		else if (strcmp(opt->name, "clusterw") == 0) {
626 			free(opt->name, M_MOUNT);
627 			opt->name = strdup("nonoclusterw", M_MOUNT);
628 		}
629 		else if (strcmp(opt->name, "noexec") == 0)
630 			fsflags |= MNT_NOEXEC;
631 		else if (strcmp(opt->name, "exec") == 0) {
632 			free(opt->name, M_MOUNT);
633 			opt->name = strdup("nonoexec", M_MOUNT);
634 		}
635 		else if (strcmp(opt->name, "nosuid") == 0)
636 			fsflags |= MNT_NOSUID;
637 		else if (strcmp(opt->name, "suid") == 0) {
638 			free(opt->name, M_MOUNT);
639 			opt->name = strdup("nonosuid", M_MOUNT);
640 		}
641 		else if (strcmp(opt->name, "nosymfollow") == 0)
642 			fsflags |= MNT_NOSYMFOLLOW;
643 		else if (strcmp(opt->name, "symfollow") == 0) {
644 			free(opt->name, M_MOUNT);
645 			opt->name = strdup("nonosymfollow", M_MOUNT);
646 		}
647 		else if (strcmp(opt->name, "noro") == 0)
648 			fsflags &= ~MNT_RDONLY;
649 		else if (strcmp(opt->name, "rw") == 0)
650 			fsflags &= ~MNT_RDONLY;
651 		else if (strcmp(opt->name, "ro") == 0)
652 			fsflags |= MNT_RDONLY;
653 		else if (strcmp(opt->name, "rdonly") == 0) {
654 			free(opt->name, M_MOUNT);
655 			opt->name = strdup("ro", M_MOUNT);
656 			fsflags |= MNT_RDONLY;
657 		}
658 		else if (strcmp(opt->name, "suiddir") == 0)
659 			fsflags |= MNT_SUIDDIR;
660 		else if (strcmp(opt->name, "sync") == 0)
661 			fsflags |= MNT_SYNCHRONOUS;
662 		else if (strcmp(opt->name, "union") == 0)
663 			fsflags |= MNT_UNION;
664 		else if (strcmp(opt->name, "automounted") == 0) {
665 			fsflags |= MNT_AUTOMOUNTED;
666 			vfs_freeopt(optlist, opt);
667 		}
668 	}
669 
670 	/*
671 	 * Be ultra-paranoid about making sure the type and fspath
672 	 * variables will fit in our mp buffers, including the
673 	 * terminating NUL.
674 	 */
675 	if (fstypelen > MFSNAMELEN || fspathlen > MNAMELEN) {
676 		error = ENAMETOOLONG;
677 		goto bail;
678 	}
679 
680 	error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
681 bail:
682 	/* copyout the errmsg */
683 	if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
684 	    && errmsg_len > 0 && errmsg != NULL) {
685 		if (fsoptions->uio_segflg == UIO_SYSSPACE) {
686 			bcopy(errmsg,
687 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
688 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
689 		} else {
690 			copyout(errmsg,
691 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
692 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
693 		}
694 	}
695 
696 	if (optlist != NULL)
697 		vfs_freeopts(optlist);
698 	return (error);
699 }
700 
701 /*
702  * Old mount API.
703  */
704 #ifndef _SYS_SYSPROTO_H_
705 struct mount_args {
706 	char	*type;
707 	char	*path;
708 	int	flags;
709 	caddr_t	data;
710 };
711 #endif
712 /* ARGSUSED */
713 int
714 sys_mount(struct thread *td, struct mount_args *uap)
715 {
716 	char *fstype;
717 	struct vfsconf *vfsp = NULL;
718 	struct mntarg *ma = NULL;
719 	uint64_t flags;
720 	int error;
721 
722 	/*
723 	 * Mount flags are now 64-bits. On 32-bit architectures only
724 	 * 32-bits are passed in, but from here on everything handles
725 	 * 64-bit flags correctly.
726 	 */
727 	flags = uap->flags;
728 
729 	AUDIT_ARG_FFLAGS(flags);
730 
731 	/*
732 	 * Filter out MNT_ROOTFS.  We do not want clients of mount() in
733 	 * userspace to set this flag, but we must filter it out if we want
734 	 * MNT_UPDATE on the root file system to work.
735 	 * MNT_ROOTFS should only be set by the kernel when mounting its
736 	 * root file system.
737 	 */
738 	flags &= ~MNT_ROOTFS;
739 
740 	fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
741 	error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
742 	if (error) {
743 		free(fstype, M_TEMP);
744 		return (error);
745 	}
746 
747 	AUDIT_ARG_TEXT(fstype);
748 	vfsp = vfs_byname_kld(fstype, td, &error);
749 	free(fstype, M_TEMP);
750 	if (vfsp == NULL)
751 		return (ENOENT);
752 	if (vfsp->vfc_vfsops->vfs_cmount == NULL)
753 		return (EOPNOTSUPP);
754 
755 	ma = mount_argsu(ma, "fstype", uap->type, MFSNAMELEN);
756 	ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
757 	ma = mount_argb(ma, flags & MNT_RDONLY, "noro");
758 	ma = mount_argb(ma, !(flags & MNT_NOSUID), "nosuid");
759 	ma = mount_argb(ma, !(flags & MNT_NOEXEC), "noexec");
760 
761 	error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags);
762 	return (error);
763 }
764 
765 /*
766  * vfs_domount_first(): first file system mount (not update)
767  */
768 static int
769 vfs_domount_first(
770 	struct thread *td,		/* Calling thread. */
771 	struct vfsconf *vfsp,		/* File system type. */
772 	char *fspath,			/* Mount path. */
773 	struct vnode *vp,		/* Vnode to be covered. */
774 	uint64_t fsflags,		/* Flags common to all filesystems. */
775 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
776 	)
777 {
778 	struct vattr va;
779 	struct mount *mp;
780 	struct vnode *newdp;
781 	int error;
782 
783 	ASSERT_VOP_ELOCKED(vp, __func__);
784 	KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
785 
786 	/*
787 	 * If the user is not root, ensure that they own the directory
788 	 * onto which we are attempting to mount.
789 	 */
790 	error = VOP_GETATTR(vp, &va, td->td_ucred);
791 	if (error == 0 && va.va_uid != td->td_ucred->cr_uid)
792 		error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN, 0);
793 	if (error == 0)
794 		error = vinvalbuf(vp, V_SAVE, 0, 0);
795 	if (error == 0 && vp->v_type != VDIR)
796 		error = ENOTDIR;
797 	if (error == 0) {
798 		VI_LOCK(vp);
799 		if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
800 			vp->v_iflag |= VI_MOUNT;
801 		else
802 			error = EBUSY;
803 		VI_UNLOCK(vp);
804 	}
805 	if (error != 0) {
806 		vput(vp);
807 		return (error);
808 	}
809 	VOP_UNLOCK(vp, 0);
810 
811 	/* Allocate and initialize the filesystem. */
812 	mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred);
813 	/* XXXMAC: pass to vfs_mount_alloc? */
814 	mp->mnt_optnew = *optlist;
815 	/* Set the mount level flags. */
816 	mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY));
817 
818 	/*
819 	 * Mount the filesystem.
820 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
821 	 * get.  No freeing of cn_pnbuf.
822 	 */
823 	error = VFS_MOUNT(mp);
824 	if (error != 0) {
825 		vfs_unbusy(mp);
826 		mp->mnt_vnodecovered = NULL;
827 		vfs_mount_destroy(mp);
828 		VI_LOCK(vp);
829 		vp->v_iflag &= ~VI_MOUNT;
830 		VI_UNLOCK(vp);
831 		vrele(vp);
832 		return (error);
833 	}
834 
835 	if (mp->mnt_opt != NULL)
836 		vfs_freeopts(mp->mnt_opt);
837 	mp->mnt_opt = mp->mnt_optnew;
838 	*optlist = NULL;
839 	(void)VFS_STATFS(mp, &mp->mnt_stat);
840 
841 	/*
842 	 * Prevent external consumers of mount options from reading mnt_optnew.
843 	 */
844 	mp->mnt_optnew = NULL;
845 
846 	MNT_ILOCK(mp);
847 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
848 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
849 		mp->mnt_kern_flag |= MNTK_ASYNC;
850 	else
851 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
852 	MNT_IUNLOCK(mp);
853 
854 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
855 	cache_purge(vp);
856 	VI_LOCK(vp);
857 	vp->v_iflag &= ~VI_MOUNT;
858 	VI_UNLOCK(vp);
859 	vp->v_mountedhere = mp;
860 	/* Place the new filesystem at the end of the mount list. */
861 	mtx_lock(&mountlist_mtx);
862 	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
863 	mtx_unlock(&mountlist_mtx);
864 	vfs_event_signal(NULL, VQ_MOUNT, 0);
865 	if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp))
866 		panic("mount: lost mount");
867 	VOP_UNLOCK(vp, 0);
868 	EVENTHANDLER_INVOKE(vfs_mounted, mp, newdp, td);
869 	VOP_UNLOCK(newdp, 0);
870 	mountcheckdirs(vp, newdp);
871 	vrele(newdp);
872 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
873 		vfs_allocate_syncvnode(mp);
874 	vfs_unbusy(mp);
875 	return (0);
876 }
877 
878 /*
879  * vfs_domount_update(): update of mounted file system
880  */
881 static int
882 vfs_domount_update(
883 	struct thread *td,		/* Calling thread. */
884 	struct vnode *vp,		/* Mount point vnode. */
885 	uint64_t fsflags,		/* Flags common to all filesystems. */
886 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
887 	)
888 {
889 	struct export_args export;
890 	void *bufp;
891 	struct mount *mp;
892 	int error, export_error, len;
893 	uint64_t flag;
894 
895 	ASSERT_VOP_ELOCKED(vp, __func__);
896 	KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
897 	mp = vp->v_mount;
898 
899 	if ((vp->v_vflag & VV_ROOT) == 0) {
900 		if (vfs_copyopt(*optlist, "export", &export, sizeof(export))
901 		    == 0)
902 			error = EXDEV;
903 		else
904 			error = EINVAL;
905 		vput(vp);
906 		return (error);
907 	}
908 
909 	/*
910 	 * We only allow the filesystem to be reloaded if it
911 	 * is currently mounted read-only.
912 	 */
913 	flag = mp->mnt_flag;
914 	if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
915 		vput(vp);
916 		return (EOPNOTSUPP);	/* Needs translation */
917 	}
918 	/*
919 	 * Only privileged root, or (if MNT_USER is set) the user that
920 	 * did the original mount is permitted to update it.
921 	 */
922 	error = vfs_suser(mp, td);
923 	if (error != 0) {
924 		vput(vp);
925 		return (error);
926 	}
927 	if (vfs_busy(mp, MBF_NOWAIT)) {
928 		vput(vp);
929 		return (EBUSY);
930 	}
931 	VI_LOCK(vp);
932 	if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
933 		VI_UNLOCK(vp);
934 		vfs_unbusy(mp);
935 		vput(vp);
936 		return (EBUSY);
937 	}
938 	vp->v_iflag |= VI_MOUNT;
939 	VI_UNLOCK(vp);
940 	VOP_UNLOCK(vp, 0);
941 
942 	MNT_ILOCK(mp);
943 	if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0) {
944 		MNT_IUNLOCK(mp);
945 		error = EBUSY;
946 		goto end;
947 	}
948 	mp->mnt_flag &= ~MNT_UPDATEMASK;
949 	mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE |
950 	    MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY);
951 	if ((mp->mnt_flag & MNT_ASYNC) == 0)
952 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
953 	MNT_IUNLOCK(mp);
954 	mp->mnt_optnew = *optlist;
955 	vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
956 
957 	/*
958 	 * Mount the filesystem.
959 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
960 	 * get.  No freeing of cn_pnbuf.
961 	 */
962 	error = VFS_MOUNT(mp);
963 
964 	export_error = 0;
965 	/* Process the export option. */
966 	if (error == 0 && vfs_getopt(mp->mnt_optnew, "export", &bufp,
967 	    &len) == 0) {
968 		/* Assume that there is only 1 ABI for each length. */
969 		switch (len) {
970 		case (sizeof(struct oexport_args)):
971 			bzero(&export, sizeof(export));
972 			/* FALLTHROUGH */
973 		case (sizeof(export)):
974 			bcopy(bufp, &export, len);
975 			export_error = vfs_export(mp, &export);
976 			break;
977 		default:
978 			export_error = EINVAL;
979 			break;
980 		}
981 	}
982 
983 	MNT_ILOCK(mp);
984 	if (error == 0) {
985 		mp->mnt_flag &=	~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
986 		    MNT_SNAPSHOT);
987 	} else {
988 		/*
989 		 * If we fail, restore old mount flags. MNT_QUOTA is special,
990 		 * because it is not part of MNT_UPDATEMASK, but it could have
991 		 * changed in the meantime if quotactl(2) was called.
992 		 * All in all we want current value of MNT_QUOTA, not the old
993 		 * one.
994 		 */
995 		mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
996 	}
997 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
998 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
999 		mp->mnt_kern_flag |= MNTK_ASYNC;
1000 	else
1001 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
1002 	MNT_IUNLOCK(mp);
1003 
1004 	if (error != 0)
1005 		goto end;
1006 
1007 	if (mp->mnt_opt != NULL)
1008 		vfs_freeopts(mp->mnt_opt);
1009 	mp->mnt_opt = mp->mnt_optnew;
1010 	*optlist = NULL;
1011 	(void)VFS_STATFS(mp, &mp->mnt_stat);
1012 	/*
1013 	 * Prevent external consumers of mount options from reading
1014 	 * mnt_optnew.
1015 	 */
1016 	mp->mnt_optnew = NULL;
1017 
1018 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
1019 		vfs_allocate_syncvnode(mp);
1020 	else
1021 		vfs_deallocate_syncvnode(mp);
1022 end:
1023 	vfs_unbusy(mp);
1024 	VI_LOCK(vp);
1025 	vp->v_iflag &= ~VI_MOUNT;
1026 	VI_UNLOCK(vp);
1027 	vrele(vp);
1028 	return (error != 0 ? error : export_error);
1029 }
1030 
1031 /*
1032  * vfs_domount(): actually attempt a filesystem mount.
1033  */
1034 static int
1035 vfs_domount(
1036 	struct thread *td,		/* Calling thread. */
1037 	const char *fstype,		/* Filesystem type. */
1038 	char *fspath,			/* Mount path. */
1039 	uint64_t fsflags,		/* Flags common to all filesystems. */
1040 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
1041 	)
1042 {
1043 	struct vfsconf *vfsp;
1044 	struct nameidata nd;
1045 	struct vnode *vp;
1046 	char *pathbuf;
1047 	int error;
1048 
1049 	/*
1050 	 * Be ultra-paranoid about making sure the type and fspath
1051 	 * variables will fit in our mp buffers, including the
1052 	 * terminating NUL.
1053 	 */
1054 	if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
1055 		return (ENAMETOOLONG);
1056 
1057 	if (jailed(td->td_ucred) || usermount == 0) {
1058 		if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
1059 			return (error);
1060 	}
1061 
1062 	/*
1063 	 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
1064 	 */
1065 	if (fsflags & MNT_EXPORTED) {
1066 		error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
1067 		if (error)
1068 			return (error);
1069 	}
1070 	if (fsflags & MNT_SUIDDIR) {
1071 		error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
1072 		if (error)
1073 			return (error);
1074 	}
1075 	/*
1076 	 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
1077 	 */
1078 	if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
1079 		if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
1080 			fsflags |= MNT_NOSUID | MNT_USER;
1081 	}
1082 
1083 	/* Load KLDs before we lock the covered vnode to avoid reversals. */
1084 	vfsp = NULL;
1085 	if ((fsflags & MNT_UPDATE) == 0) {
1086 		/* Don't try to load KLDs if we're mounting the root. */
1087 		if (fsflags & MNT_ROOTFS)
1088 			vfsp = vfs_byname(fstype);
1089 		else
1090 			vfsp = vfs_byname_kld(fstype, td, &error);
1091 		if (vfsp == NULL)
1092 			return (ENODEV);
1093 		if (jailed(td->td_ucred) && !(vfsp->vfc_flags & VFCF_JAIL))
1094 			return (EPERM);
1095 	}
1096 
1097 	/*
1098 	 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
1099 	 */
1100 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1101 	    UIO_SYSSPACE, fspath, td);
1102 	error = namei(&nd);
1103 	if (error != 0)
1104 		return (error);
1105 	NDFREE(&nd, NDF_ONLY_PNBUF);
1106 	vp = nd.ni_vp;
1107 	if ((fsflags & MNT_UPDATE) == 0) {
1108 		pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1109 		strcpy(pathbuf, fspath);
1110 		error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN);
1111 		/* debug.disablefullpath == 1 results in ENODEV */
1112 		if (error == 0 || error == ENODEV) {
1113 			error = vfs_domount_first(td, vfsp, pathbuf, vp,
1114 			    fsflags, optlist);
1115 		}
1116 		free(pathbuf, M_TEMP);
1117 	} else
1118 		error = vfs_domount_update(td, vp, fsflags, optlist);
1119 
1120 	return (error);
1121 }
1122 
1123 /*
1124  * Unmount a filesystem.
1125  *
1126  * Note: unmount takes a path to the vnode mounted on as argument, not
1127  * special file (as before).
1128  */
1129 #ifndef _SYS_SYSPROTO_H_
1130 struct unmount_args {
1131 	char	*path;
1132 	int	flags;
1133 };
1134 #endif
1135 /* ARGSUSED */
1136 int
1137 sys_unmount(struct thread *td, struct unmount_args *uap)
1138 {
1139 	struct nameidata nd;
1140 	struct mount *mp;
1141 	char *pathbuf;
1142 	int error, id0, id1;
1143 
1144 	AUDIT_ARG_VALUE(uap->flags);
1145 	if (jailed(td->td_ucred) || usermount == 0) {
1146 		error = priv_check(td, PRIV_VFS_UNMOUNT);
1147 		if (error)
1148 			return (error);
1149 	}
1150 
1151 	pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1152 	error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1153 	if (error) {
1154 		free(pathbuf, M_TEMP);
1155 		return (error);
1156 	}
1157 	if (uap->flags & MNT_BYFSID) {
1158 		AUDIT_ARG_TEXT(pathbuf);
1159 		/* Decode the filesystem ID. */
1160 		if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
1161 			free(pathbuf, M_TEMP);
1162 			return (EINVAL);
1163 		}
1164 
1165 		mtx_lock(&mountlist_mtx);
1166 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1167 			if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1168 			    mp->mnt_stat.f_fsid.val[1] == id1) {
1169 				vfs_ref(mp);
1170 				break;
1171 			}
1172 		}
1173 		mtx_unlock(&mountlist_mtx);
1174 	} else {
1175 		/*
1176 		 * Try to find global path for path argument.
1177 		 */
1178 		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1179 		    UIO_SYSSPACE, pathbuf, td);
1180 		if (namei(&nd) == 0) {
1181 			NDFREE(&nd, NDF_ONLY_PNBUF);
1182 			error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
1183 			    MNAMELEN);
1184 			if (error == 0 || error == ENODEV)
1185 				vput(nd.ni_vp);
1186 		}
1187 		mtx_lock(&mountlist_mtx);
1188 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1189 			if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) {
1190 				vfs_ref(mp);
1191 				break;
1192 			}
1193 		}
1194 		mtx_unlock(&mountlist_mtx);
1195 	}
1196 	free(pathbuf, M_TEMP);
1197 	if (mp == NULL) {
1198 		/*
1199 		 * Previously we returned ENOENT for a nonexistent path and
1200 		 * EINVAL for a non-mountpoint.  We cannot tell these apart
1201 		 * now, so in the !MNT_BYFSID case return the more likely
1202 		 * EINVAL for compatibility.
1203 		 */
1204 		return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1205 	}
1206 
1207 	/*
1208 	 * Don't allow unmounting the root filesystem.
1209 	 */
1210 	if (mp->mnt_flag & MNT_ROOTFS) {
1211 		vfs_rel(mp);
1212 		return (EINVAL);
1213 	}
1214 	error = dounmount(mp, uap->flags, td);
1215 	return (error);
1216 }
1217 
1218 /*
1219  * Return error if any of the vnodes, ignoring the root vnode
1220  * and the syncer vnode, have non-zero usecount.
1221  *
1222  * This function is purely advisory - it can return false positives
1223  * and negatives.
1224  */
1225 static int
1226 vfs_check_usecounts(struct mount *mp)
1227 {
1228 	struct vnode *vp, *mvp;
1229 
1230 	MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1231 		if ((vp->v_vflag & VV_ROOT) == 0 && vp->v_type != VNON &&
1232 		    vp->v_usecount != 0) {
1233 			VI_UNLOCK(vp);
1234 			MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1235 			return (EBUSY);
1236 		}
1237 		VI_UNLOCK(vp);
1238 	}
1239 
1240 	return (0);
1241 }
1242 
1243 static void
1244 dounmount_cleanup(struct mount *mp, struct vnode *coveredvp, int mntkflags)
1245 {
1246 
1247 	mtx_assert(MNT_MTX(mp), MA_OWNED);
1248 	mp->mnt_kern_flag &= ~mntkflags;
1249 	if ((mp->mnt_kern_flag & MNTK_MWAIT) != 0) {
1250 		mp->mnt_kern_flag &= ~MNTK_MWAIT;
1251 		wakeup(mp);
1252 	}
1253 	MNT_IUNLOCK(mp);
1254 	if (coveredvp != NULL) {
1255 		VOP_UNLOCK(coveredvp, 0);
1256 		vdrop(coveredvp);
1257 	}
1258 	vn_finished_write(mp);
1259 }
1260 
1261 /*
1262  * Do the actual filesystem unmount.
1263  */
1264 int
1265 dounmount(struct mount *mp, int flags, struct thread *td)
1266 {
1267 	struct vnode *coveredvp;
1268 	int error;
1269 	uint64_t async_flag;
1270 	int mnt_gen_r;
1271 
1272 	if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
1273 		mnt_gen_r = mp->mnt_gen;
1274 		VI_LOCK(coveredvp);
1275 		vholdl(coveredvp);
1276 		vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
1277 		/*
1278 		 * Check for mp being unmounted while waiting for the
1279 		 * covered vnode lock.
1280 		 */
1281 		if (coveredvp->v_mountedhere != mp ||
1282 		    coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
1283 			VOP_UNLOCK(coveredvp, 0);
1284 			vdrop(coveredvp);
1285 			vfs_rel(mp);
1286 			return (EBUSY);
1287 		}
1288 	}
1289 
1290 	/*
1291 	 * Only privileged root, or (if MNT_USER is set) the user that did the
1292 	 * original mount is permitted to unmount this filesystem.
1293 	 */
1294 	error = vfs_suser(mp, td);
1295 	if (error != 0) {
1296 		if (coveredvp != NULL) {
1297 			VOP_UNLOCK(coveredvp, 0);
1298 			vdrop(coveredvp);
1299 		}
1300 		vfs_rel(mp);
1301 		return (error);
1302 	}
1303 
1304 	vn_start_write(NULL, &mp, V_WAIT | V_MNTREF);
1305 	MNT_ILOCK(mp);
1306 	if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 ||
1307 	    (mp->mnt_flag & MNT_UPDATE) != 0 ||
1308 	    !TAILQ_EMPTY(&mp->mnt_uppers)) {
1309 		dounmount_cleanup(mp, coveredvp, 0);
1310 		return (EBUSY);
1311 	}
1312 	mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ;
1313 	if (flags & MNT_NONBUSY) {
1314 		MNT_IUNLOCK(mp);
1315 		error = vfs_check_usecounts(mp);
1316 		MNT_ILOCK(mp);
1317 		if (error != 0) {
1318 			dounmount_cleanup(mp, coveredvp, MNTK_UNMOUNT |
1319 			    MNTK_NOINSMNTQ);
1320 			return (error);
1321 		}
1322 	}
1323 	/* Allow filesystems to detect that a forced unmount is in progress. */
1324 	if (flags & MNT_FORCE) {
1325 		mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1326 		MNT_IUNLOCK(mp);
1327 		/*
1328 		 * Must be done after setting MNTK_UNMOUNTF and before
1329 		 * waiting for mnt_lockref to become 0.
1330 		 */
1331 		VFS_PURGE(mp);
1332 		MNT_ILOCK(mp);
1333 	}
1334 	error = 0;
1335 	if (mp->mnt_lockref) {
1336 		mp->mnt_kern_flag |= MNTK_DRAINING;
1337 		error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
1338 		    "mount drain", 0);
1339 	}
1340 	MNT_IUNLOCK(mp);
1341 	KASSERT(mp->mnt_lockref == 0,
1342 	    ("%s: invalid lock refcount in the drain path @ %s:%d",
1343 	    __func__, __FILE__, __LINE__));
1344 	KASSERT(error == 0,
1345 	    ("%s: invalid return value for msleep in the drain path @ %s:%d",
1346 	    __func__, __FILE__, __LINE__));
1347 
1348 	if (mp->mnt_flag & MNT_EXPUBLIC)
1349 		vfs_setpublicfs(NULL, NULL, NULL);
1350 
1351 	/*
1352 	 * From now, we can claim that the use reference on the
1353 	 * coveredvp is ours, and the ref can be released only by
1354 	 * successfull unmount by us, or left for later unmount
1355 	 * attempt.  The previously acquired hold reference is no
1356 	 * longer needed to protect the vnode from reuse.
1357 	 */
1358 	if (coveredvp != NULL)
1359 		vdrop(coveredvp);
1360 
1361 	vfs_msync(mp, MNT_WAIT);
1362 	MNT_ILOCK(mp);
1363 	async_flag = mp->mnt_flag & MNT_ASYNC;
1364 	mp->mnt_flag &= ~MNT_ASYNC;
1365 	mp->mnt_kern_flag &= ~MNTK_ASYNC;
1366 	MNT_IUNLOCK(mp);
1367 	cache_purgevfs(mp, false); /* remove cache entries for this file sys */
1368 	vfs_deallocate_syncvnode(mp);
1369 	if ((mp->mnt_flag & MNT_RDONLY) != 0 || (flags & MNT_FORCE) != 0 ||
1370 	    (error = VFS_SYNC(mp, MNT_WAIT)) == 0)
1371 		error = VFS_UNMOUNT(mp, flags);
1372 	vn_finished_write(mp);
1373 	/*
1374 	 * If we failed to flush the dirty blocks for this mount point,
1375 	 * undo all the cdir/rdir and rootvnode changes we made above.
1376 	 * Unless we failed to do so because the device is reporting that
1377 	 * it doesn't exist anymore.
1378 	 */
1379 	if (error && error != ENXIO) {
1380 		MNT_ILOCK(mp);
1381 		mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ;
1382 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
1383 			MNT_IUNLOCK(mp);
1384 			vfs_allocate_syncvnode(mp);
1385 			MNT_ILOCK(mp);
1386 		}
1387 		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1388 		mp->mnt_flag |= async_flag;
1389 		if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1390 		    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1391 			mp->mnt_kern_flag |= MNTK_ASYNC;
1392 		if (mp->mnt_kern_flag & MNTK_MWAIT) {
1393 			mp->mnt_kern_flag &= ~MNTK_MWAIT;
1394 			wakeup(mp);
1395 		}
1396 		MNT_IUNLOCK(mp);
1397 		if (coveredvp)
1398 			VOP_UNLOCK(coveredvp, 0);
1399 		return (error);
1400 	}
1401 	mtx_lock(&mountlist_mtx);
1402 	TAILQ_REMOVE(&mountlist, mp, mnt_list);
1403 	mtx_unlock(&mountlist_mtx);
1404 	EVENTHANDLER_INVOKE(vfs_unmounted, mp, td);
1405 	if (coveredvp != NULL) {
1406 		coveredvp->v_mountedhere = NULL;
1407 		VOP_UNLOCK(coveredvp, 0);
1408 	}
1409 	vfs_event_signal(NULL, VQ_UNMOUNT, 0);
1410 	if (rootvnode != NULL && mp == rootvnode->v_mount) {
1411 		vrele(rootvnode);
1412 		rootvnode = NULL;
1413 	}
1414 	if (mp == rootdevmp)
1415 		rootdevmp = NULL;
1416 	vfs_mount_destroy(mp);
1417 	return (0);
1418 }
1419 
1420 /*
1421  * Report errors during filesystem mounting.
1422  */
1423 void
1424 vfs_mount_error(struct mount *mp, const char *fmt, ...)
1425 {
1426 	struct vfsoptlist *moptlist = mp->mnt_optnew;
1427 	va_list ap;
1428 	int error, len;
1429 	char *errmsg;
1430 
1431 	error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
1432 	if (error || errmsg == NULL || len <= 0)
1433 		return;
1434 
1435 	va_start(ap, fmt);
1436 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1437 	va_end(ap);
1438 }
1439 
1440 void
1441 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
1442 {
1443 	va_list ap;
1444 	int error, len;
1445 	char *errmsg;
1446 
1447 	error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
1448 	if (error || errmsg == NULL || len <= 0)
1449 		return;
1450 
1451 	va_start(ap, fmt);
1452 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1453 	va_end(ap);
1454 }
1455 
1456 /*
1457  * ---------------------------------------------------------------------
1458  * Functions for querying mount options/arguments from filesystems.
1459  */
1460 
1461 /*
1462  * Check that no unknown options are given
1463  */
1464 int
1465 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1466 {
1467 	struct vfsopt *opt;
1468 	char errmsg[255];
1469 	const char **t, *p, *q;
1470 	int ret = 0;
1471 
1472 	TAILQ_FOREACH(opt, opts, link) {
1473 		p = opt->name;
1474 		q = NULL;
1475 		if (p[0] == 'n' && p[1] == 'o')
1476 			q = p + 2;
1477 		for(t = global_opts; *t != NULL; t++) {
1478 			if (strcmp(*t, p) == 0)
1479 				break;
1480 			if (q != NULL) {
1481 				if (strcmp(*t, q) == 0)
1482 					break;
1483 			}
1484 		}
1485 		if (*t != NULL)
1486 			continue;
1487 		for(t = legal; *t != NULL; t++) {
1488 			if (strcmp(*t, p) == 0)
1489 				break;
1490 			if (q != NULL) {
1491 				if (strcmp(*t, q) == 0)
1492 					break;
1493 			}
1494 		}
1495 		if (*t != NULL)
1496 			continue;
1497 		snprintf(errmsg, sizeof(errmsg),
1498 		    "mount option <%s> is unknown", p);
1499 		ret = EINVAL;
1500 	}
1501 	if (ret != 0) {
1502 		TAILQ_FOREACH(opt, opts, link) {
1503 			if (strcmp(opt->name, "errmsg") == 0) {
1504 				strncpy((char *)opt->value, errmsg, opt->len);
1505 				break;
1506 			}
1507 		}
1508 		if (opt == NULL)
1509 			printf("%s\n", errmsg);
1510 	}
1511 	return (ret);
1512 }
1513 
1514 /*
1515  * Get a mount option by its name.
1516  *
1517  * Return 0 if the option was found, ENOENT otherwise.
1518  * If len is non-NULL it will be filled with the length
1519  * of the option. If buf is non-NULL, it will be filled
1520  * with the address of the option.
1521  */
1522 int
1523 vfs_getopt(struct vfsoptlist *opts, const char *name, void **buf, int *len)
1524 {
1525 	struct vfsopt *opt;
1526 
1527 	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1528 
1529 	TAILQ_FOREACH(opt, opts, link) {
1530 		if (strcmp(name, opt->name) == 0) {
1531 			opt->seen = 1;
1532 			if (len != NULL)
1533 				*len = opt->len;
1534 			if (buf != NULL)
1535 				*buf = opt->value;
1536 			return (0);
1537 		}
1538 	}
1539 	return (ENOENT);
1540 }
1541 
1542 int
1543 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
1544 {
1545 	struct vfsopt *opt;
1546 
1547 	if (opts == NULL)
1548 		return (-1);
1549 
1550 	TAILQ_FOREACH(opt, opts, link) {
1551 		if (strcmp(name, opt->name) == 0) {
1552 			opt->seen = 1;
1553 			return (opt->pos);
1554 		}
1555 	}
1556 	return (-1);
1557 }
1558 
1559 int
1560 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value)
1561 {
1562 	char *opt_value, *vtp;
1563 	quad_t iv;
1564 	int error, opt_len;
1565 
1566 	error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len);
1567 	if (error != 0)
1568 		return (error);
1569 	if (opt_len == 0 || opt_value == NULL)
1570 		return (EINVAL);
1571 	if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0')
1572 		return (EINVAL);
1573 	iv = strtoq(opt_value, &vtp, 0);
1574 	if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0'))
1575 		return (EINVAL);
1576 	if (iv < 0)
1577 		return (EINVAL);
1578 	switch (vtp[0]) {
1579 	case 't':
1580 	case 'T':
1581 		iv *= 1024;
1582 	case 'g':
1583 	case 'G':
1584 		iv *= 1024;
1585 	case 'm':
1586 	case 'M':
1587 		iv *= 1024;
1588 	case 'k':
1589 	case 'K':
1590 		iv *= 1024;
1591 	case '\0':
1592 		break;
1593 	default:
1594 		return (EINVAL);
1595 	}
1596 	*value = iv;
1597 
1598 	return (0);
1599 }
1600 
1601 char *
1602 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
1603 {
1604 	struct vfsopt *opt;
1605 
1606 	*error = 0;
1607 	TAILQ_FOREACH(opt, opts, link) {
1608 		if (strcmp(name, opt->name) != 0)
1609 			continue;
1610 		opt->seen = 1;
1611 		if (opt->len == 0 ||
1612 		    ((char *)opt->value)[opt->len - 1] != '\0') {
1613 			*error = EINVAL;
1614 			return (NULL);
1615 		}
1616 		return (opt->value);
1617 	}
1618 	*error = ENOENT;
1619 	return (NULL);
1620 }
1621 
1622 int
1623 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
1624 	uint64_t val)
1625 {
1626 	struct vfsopt *opt;
1627 
1628 	TAILQ_FOREACH(opt, opts, link) {
1629 		if (strcmp(name, opt->name) == 0) {
1630 			opt->seen = 1;
1631 			if (w != NULL)
1632 				*w |= val;
1633 			return (1);
1634 		}
1635 	}
1636 	if (w != NULL)
1637 		*w &= ~val;
1638 	return (0);
1639 }
1640 
1641 int
1642 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
1643 {
1644 	va_list ap;
1645 	struct vfsopt *opt;
1646 	int ret;
1647 
1648 	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1649 
1650 	TAILQ_FOREACH(opt, opts, link) {
1651 		if (strcmp(name, opt->name) != 0)
1652 			continue;
1653 		opt->seen = 1;
1654 		if (opt->len == 0 || opt->value == NULL)
1655 			return (0);
1656 		if (((char *)opt->value)[opt->len - 1] != '\0')
1657 			return (0);
1658 		va_start(ap, fmt);
1659 		ret = vsscanf(opt->value, fmt, ap);
1660 		va_end(ap);
1661 		return (ret);
1662 	}
1663 	return (0);
1664 }
1665 
1666 int
1667 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len)
1668 {
1669 	struct vfsopt *opt;
1670 
1671 	TAILQ_FOREACH(opt, opts, link) {
1672 		if (strcmp(name, opt->name) != 0)
1673 			continue;
1674 		opt->seen = 1;
1675 		if (opt->value == NULL)
1676 			opt->len = len;
1677 		else {
1678 			if (opt->len != len)
1679 				return (EINVAL);
1680 			bcopy(value, opt->value, len);
1681 		}
1682 		return (0);
1683 	}
1684 	return (ENOENT);
1685 }
1686 
1687 int
1688 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len)
1689 {
1690 	struct vfsopt *opt;
1691 
1692 	TAILQ_FOREACH(opt, opts, link) {
1693 		if (strcmp(name, opt->name) != 0)
1694 			continue;
1695 		opt->seen = 1;
1696 		if (opt->value == NULL)
1697 			opt->len = len;
1698 		else {
1699 			if (opt->len < len)
1700 				return (EINVAL);
1701 			opt->len = len;
1702 			bcopy(value, opt->value, len);
1703 		}
1704 		return (0);
1705 	}
1706 	return (ENOENT);
1707 }
1708 
1709 int
1710 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value)
1711 {
1712 	struct vfsopt *opt;
1713 
1714 	TAILQ_FOREACH(opt, opts, link) {
1715 		if (strcmp(name, opt->name) != 0)
1716 			continue;
1717 		opt->seen = 1;
1718 		if (opt->value == NULL)
1719 			opt->len = strlen(value) + 1;
1720 		else if (strlcpy(opt->value, value, opt->len) >= opt->len)
1721 			return (EINVAL);
1722 		return (0);
1723 	}
1724 	return (ENOENT);
1725 }
1726 
1727 /*
1728  * Find and copy a mount option.
1729  *
1730  * The size of the buffer has to be specified
1731  * in len, if it is not the same length as the
1732  * mount option, EINVAL is returned.
1733  * Returns ENOENT if the option is not found.
1734  */
1735 int
1736 vfs_copyopt(struct vfsoptlist *opts, const char *name, void *dest, int len)
1737 {
1738 	struct vfsopt *opt;
1739 
1740 	KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1741 
1742 	TAILQ_FOREACH(opt, opts, link) {
1743 		if (strcmp(name, opt->name) == 0) {
1744 			opt->seen = 1;
1745 			if (len != opt->len)
1746 				return (EINVAL);
1747 			bcopy(opt->value, dest, opt->len);
1748 			return (0);
1749 		}
1750 	}
1751 	return (ENOENT);
1752 }
1753 
1754 int
1755 __vfs_statfs(struct mount *mp, struct statfs *sbp)
1756 {
1757 	int error;
1758 
1759 	error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat);
1760 	if (sbp != &mp->mnt_stat)
1761 		*sbp = mp->mnt_stat;
1762 	return (error);
1763 }
1764 
1765 void
1766 vfs_mountedfrom(struct mount *mp, const char *from)
1767 {
1768 
1769 	bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
1770 	strlcpy(mp->mnt_stat.f_mntfromname, from,
1771 	    sizeof mp->mnt_stat.f_mntfromname);
1772 }
1773 
1774 /*
1775  * ---------------------------------------------------------------------
1776  * This is the api for building mount args and mounting filesystems from
1777  * inside the kernel.
1778  *
1779  * The API works by accumulation of individual args.  First error is
1780  * latched.
1781  *
1782  * XXX: should be documented in new manpage kernel_mount(9)
1783  */
1784 
1785 /* A memory allocation which must be freed when we are done */
1786 struct mntaarg {
1787 	SLIST_ENTRY(mntaarg)	next;
1788 };
1789 
1790 /* The header for the mount arguments */
1791 struct mntarg {
1792 	struct iovec *v;
1793 	int len;
1794 	int error;
1795 	SLIST_HEAD(, mntaarg)	list;
1796 };
1797 
1798 /*
1799  * Add a boolean argument.
1800  *
1801  * flag is the boolean value.
1802  * name must start with "no".
1803  */
1804 struct mntarg *
1805 mount_argb(struct mntarg *ma, int flag, const char *name)
1806 {
1807 
1808 	KASSERT(name[0] == 'n' && name[1] == 'o',
1809 	    ("mount_argb(...,%s): name must start with 'no'", name));
1810 
1811 	return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
1812 }
1813 
1814 /*
1815  * Add an argument printf style
1816  */
1817 struct mntarg *
1818 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
1819 {
1820 	va_list ap;
1821 	struct mntaarg *maa;
1822 	struct sbuf *sb;
1823 	int len;
1824 
1825 	if (ma == NULL) {
1826 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1827 		SLIST_INIT(&ma->list);
1828 	}
1829 	if (ma->error)
1830 		return (ma);
1831 
1832 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1833 	    M_MOUNT, M_WAITOK);
1834 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1835 	ma->v[ma->len].iov_len = strlen(name) + 1;
1836 	ma->len++;
1837 
1838 	sb = sbuf_new_auto();
1839 	va_start(ap, fmt);
1840 	sbuf_vprintf(sb, fmt, ap);
1841 	va_end(ap);
1842 	sbuf_finish(sb);
1843 	len = sbuf_len(sb) + 1;
1844 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1845 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1846 	bcopy(sbuf_data(sb), maa + 1, len);
1847 	sbuf_delete(sb);
1848 
1849 	ma->v[ma->len].iov_base = maa + 1;
1850 	ma->v[ma->len].iov_len = len;
1851 	ma->len++;
1852 
1853 	return (ma);
1854 }
1855 
1856 /*
1857  * Add an argument which is a userland string.
1858  */
1859 struct mntarg *
1860 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
1861 {
1862 	struct mntaarg *maa;
1863 	char *tbuf;
1864 
1865 	if (val == NULL)
1866 		return (ma);
1867 	if (ma == NULL) {
1868 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1869 		SLIST_INIT(&ma->list);
1870 	}
1871 	if (ma->error)
1872 		return (ma);
1873 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1874 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1875 	tbuf = (void *)(maa + 1);
1876 	ma->error = copyinstr(val, tbuf, len, NULL);
1877 	return (mount_arg(ma, name, tbuf, -1));
1878 }
1879 
1880 /*
1881  * Plain argument.
1882  *
1883  * If length is -1, treat value as a C string.
1884  */
1885 struct mntarg *
1886 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
1887 {
1888 
1889 	if (ma == NULL) {
1890 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1891 		SLIST_INIT(&ma->list);
1892 	}
1893 	if (ma->error)
1894 		return (ma);
1895 
1896 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1897 	    M_MOUNT, M_WAITOK);
1898 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1899 	ma->v[ma->len].iov_len = strlen(name) + 1;
1900 	ma->len++;
1901 
1902 	ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
1903 	if (len < 0)
1904 		ma->v[ma->len].iov_len = strlen(val) + 1;
1905 	else
1906 		ma->v[ma->len].iov_len = len;
1907 	ma->len++;
1908 	return (ma);
1909 }
1910 
1911 /*
1912  * Free a mntarg structure
1913  */
1914 static void
1915 free_mntarg(struct mntarg *ma)
1916 {
1917 	struct mntaarg *maa;
1918 
1919 	while (!SLIST_EMPTY(&ma->list)) {
1920 		maa = SLIST_FIRST(&ma->list);
1921 		SLIST_REMOVE_HEAD(&ma->list, next);
1922 		free(maa, M_MOUNT);
1923 	}
1924 	free(ma->v, M_MOUNT);
1925 	free(ma, M_MOUNT);
1926 }
1927 
1928 /*
1929  * Mount a filesystem
1930  */
1931 int
1932 kernel_mount(struct mntarg *ma, uint64_t flags)
1933 {
1934 	struct uio auio;
1935 	int error;
1936 
1937 	KASSERT(ma != NULL, ("kernel_mount NULL ma"));
1938 	KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
1939 	KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
1940 
1941 	auio.uio_iov = ma->v;
1942 	auio.uio_iovcnt = ma->len;
1943 	auio.uio_segflg = UIO_SYSSPACE;
1944 
1945 	error = ma->error;
1946 	if (!error)
1947 		error = vfs_donmount(curthread, flags, &auio);
1948 	free_mntarg(ma);
1949 	return (error);
1950 }
1951 
1952 /*
1953  * A printflike function to mount a filesystem.
1954  */
1955 int
1956 kernel_vmount(int flags, ...)
1957 {
1958 	struct mntarg *ma = NULL;
1959 	va_list ap;
1960 	const char *cp;
1961 	const void *vp;
1962 	int error;
1963 
1964 	va_start(ap, flags);
1965 	for (;;) {
1966 		cp = va_arg(ap, const char *);
1967 		if (cp == NULL)
1968 			break;
1969 		vp = va_arg(ap, const void *);
1970 		ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0));
1971 	}
1972 	va_end(ap);
1973 
1974 	error = kernel_mount(ma, flags);
1975 	return (error);
1976 }
1977 
1978 void
1979 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp)
1980 {
1981 
1982 	bcopy(oexp, exp, sizeof(*oexp));
1983 	exp->ex_numsecflavors = 0;
1984 }
1985