xref: /freebsd/sys/kern/vfs_mount.c (revision 63a938566d524836885917d95bd491aa4400b181)
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/eventhandler.h>
45 #include <sys/fcntl.h>
46 #include <sys/jail.h>
47 #include <sys/kernel.h>
48 #include <sys/libkern.h>
49 #include <sys/malloc.h>
50 #include <sys/mount.h>
51 #include <sys/mutex.h>
52 #include <sys/namei.h>
53 #include <sys/priv.h>
54 #include <sys/proc.h>
55 #include <sys/filedesc.h>
56 #include <sys/reboot.h>
57 #include <sys/sbuf.h>
58 #include <sys/syscallsubr.h>
59 #include <sys/sysproto.h>
60 #include <sys/sx.h>
61 #include <sys/sysctl.h>
62 #include <sys/sysent.h>
63 #include <sys/systm.h>
64 #include <sys/vnode.h>
65 #include <vm/uma.h>
66 
67 #include <geom/geom.h>
68 
69 #include <machine/stdarg.h>
70 
71 #include <security/audit/audit.h>
72 #include <security/mac/mac_framework.h>
73 
74 #define	VFS_MOUNTARG_SIZE_MAX	(1024 * 64)
75 
76 static int	vfs_domount(struct thread *td, const char *fstype, char *fspath,
77 		    uint64_t fsflags, struct vfsoptlist **optlist);
78 static void	free_mntarg(struct mntarg *ma);
79 
80 static int	usermount = 0;
81 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0,
82     "Unprivileged users may mount and unmount file systems");
83 
84 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
85 MALLOC_DEFINE(M_STATFS, "statfs", "statfs structure");
86 static uma_zone_t mount_zone;
87 
88 /* List of mounted filesystems. */
89 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
90 
91 /* For any iteration/modification of mountlist */
92 struct mtx mountlist_mtx;
93 MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF);
94 
95 EVENTHANDLER_LIST_DEFINE(vfs_mounted);
96 EVENTHANDLER_LIST_DEFINE(vfs_unmounted);
97 
98 /*
99  * Global opts, taken by all filesystems
100  */
101 static const char *global_opts[] = {
102 	"errmsg",
103 	"fstype",
104 	"fspath",
105 	"ro",
106 	"rw",
107 	"nosuid",
108 	"noexec",
109 	NULL
110 };
111 
112 static int
113 mount_init(void *mem, int size, int flags)
114 {
115 	struct mount *mp;
116 
117 	mp = (struct mount *)mem;
118 	mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF);
119 	mtx_init(&mp->mnt_listmtx, "struct mount vlist mtx", NULL, MTX_DEF);
120 	lockinit(&mp->mnt_explock, PVFS, "explock", 0, 0);
121 	return (0);
122 }
123 
124 static void
125 mount_fini(void *mem, int size)
126 {
127 	struct mount *mp;
128 
129 	mp = (struct mount *)mem;
130 	lockdestroy(&mp->mnt_explock);
131 	mtx_destroy(&mp->mnt_listmtx);
132 	mtx_destroy(&mp->mnt_mtx);
133 }
134 
135 static void
136 vfs_mount_init(void *dummy __unused)
137 {
138 
139 	mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount), NULL,
140 	    NULL, mount_init, mount_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
141 }
142 SYSINIT(vfs_mount, SI_SUB_VFS, SI_ORDER_ANY, vfs_mount_init, NULL);
143 
144 /*
145  * ---------------------------------------------------------------------
146  * Functions for building and sanitizing the mount options
147  */
148 
149 /* Remove one mount option. */
150 static void
151 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
152 {
153 
154 	TAILQ_REMOVE(opts, opt, link);
155 	free(opt->name, M_MOUNT);
156 	if (opt->value != NULL)
157 		free(opt->value, M_MOUNT);
158 	free(opt, M_MOUNT);
159 }
160 
161 /* Release all resources related to the mount options. */
162 void
163 vfs_freeopts(struct vfsoptlist *opts)
164 {
165 	struct vfsopt *opt;
166 
167 	while (!TAILQ_EMPTY(opts)) {
168 		opt = TAILQ_FIRST(opts);
169 		vfs_freeopt(opts, opt);
170 	}
171 	free(opts, M_MOUNT);
172 }
173 
174 void
175 vfs_deleteopt(struct vfsoptlist *opts, const char *name)
176 {
177 	struct vfsopt *opt, *temp;
178 
179 	if (opts == NULL)
180 		return;
181 	TAILQ_FOREACH_SAFE(opt, opts, link, temp)  {
182 		if (strcmp(opt->name, name) == 0)
183 			vfs_freeopt(opts, opt);
184 	}
185 }
186 
187 static int
188 vfs_isopt_ro(const char *opt)
189 {
190 
191 	if (strcmp(opt, "ro") == 0 || strcmp(opt, "rdonly") == 0 ||
192 	    strcmp(opt, "norw") == 0)
193 		return (1);
194 	return (0);
195 }
196 
197 static int
198 vfs_isopt_rw(const char *opt)
199 {
200 
201 	if (strcmp(opt, "rw") == 0 || strcmp(opt, "noro") == 0)
202 		return (1);
203 	return (0);
204 }
205 
206 /*
207  * Check if options are equal (with or without the "no" prefix).
208  */
209 static int
210 vfs_equalopts(const char *opt1, const char *opt2)
211 {
212 	char *p;
213 
214 	/* "opt" vs. "opt" or "noopt" vs. "noopt" */
215 	if (strcmp(opt1, opt2) == 0)
216 		return (1);
217 	/* "noopt" vs. "opt" */
218 	if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
219 		return (1);
220 	/* "opt" vs. "noopt" */
221 	if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
222 		return (1);
223 	while ((p = strchr(opt1, '.')) != NULL &&
224 	    !strncmp(opt1, opt2, ++p - opt1)) {
225 		opt2 += p - opt1;
226 		opt1 = p;
227 		/* "foo.noopt" vs. "foo.opt" */
228 		if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
229 			return (1);
230 		/* "foo.opt" vs. "foo.noopt" */
231 		if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
232 			return (1);
233 	}
234 	/* "ro" / "rdonly" / "norw" / "rw" / "noro" */
235 	if ((vfs_isopt_ro(opt1) || vfs_isopt_rw(opt1)) &&
236 	    (vfs_isopt_ro(opt2) || vfs_isopt_rw(opt2)))
237 		return (1);
238 	return (0);
239 }
240 
241 /*
242  * If a mount option is specified several times,
243  * (with or without the "no" prefix) only keep
244  * the last occurrence of it.
245  */
246 static void
247 vfs_sanitizeopts(struct vfsoptlist *opts)
248 {
249 	struct vfsopt *opt, *opt2, *tmp;
250 
251 	TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
252 		opt2 = TAILQ_PREV(opt, vfsoptlist, link);
253 		while (opt2 != NULL) {
254 			if (vfs_equalopts(opt->name, opt2->name)) {
255 				tmp = TAILQ_PREV(opt2, vfsoptlist, link);
256 				vfs_freeopt(opts, opt2);
257 				opt2 = tmp;
258 			} else {
259 				opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
260 			}
261 		}
262 	}
263 }
264 
265 /*
266  * Build a linked list of mount options from a struct uio.
267  */
268 int
269 vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
270 {
271 	struct vfsoptlist *opts;
272 	struct vfsopt *opt;
273 	size_t memused, namelen, optlen;
274 	unsigned int i, iovcnt;
275 	int error;
276 
277 	opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
278 	TAILQ_INIT(opts);
279 	memused = 0;
280 	iovcnt = auio->uio_iovcnt;
281 	for (i = 0; i < iovcnt; i += 2) {
282 		namelen = auio->uio_iov[i].iov_len;
283 		optlen = auio->uio_iov[i + 1].iov_len;
284 		memused += sizeof(struct vfsopt) + optlen + namelen;
285 		/*
286 		 * Avoid consuming too much memory, and attempts to overflow
287 		 * memused.
288 		 */
289 		if (memused > VFS_MOUNTARG_SIZE_MAX ||
290 		    optlen > VFS_MOUNTARG_SIZE_MAX ||
291 		    namelen > VFS_MOUNTARG_SIZE_MAX) {
292 			error = EINVAL;
293 			goto bad;
294 		}
295 
296 		opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
297 		opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
298 		opt->value = NULL;
299 		opt->len = 0;
300 		opt->pos = i / 2;
301 		opt->seen = 0;
302 
303 		/*
304 		 * Do this early, so jumps to "bad" will free the current
305 		 * option.
306 		 */
307 		TAILQ_INSERT_TAIL(opts, opt, link);
308 
309 		if (auio->uio_segflg == UIO_SYSSPACE) {
310 			bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
311 		} else {
312 			error = copyin(auio->uio_iov[i].iov_base, opt->name,
313 			    namelen);
314 			if (error)
315 				goto bad;
316 		}
317 		/* Ensure names are null-terminated strings. */
318 		if (namelen == 0 || opt->name[namelen - 1] != '\0') {
319 			error = EINVAL;
320 			goto bad;
321 		}
322 		if (optlen != 0) {
323 			opt->len = optlen;
324 			opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
325 			if (auio->uio_segflg == UIO_SYSSPACE) {
326 				bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
327 				    optlen);
328 			} else {
329 				error = copyin(auio->uio_iov[i + 1].iov_base,
330 				    opt->value, optlen);
331 				if (error)
332 					goto bad;
333 			}
334 		}
335 	}
336 	vfs_sanitizeopts(opts);
337 	*options = opts;
338 	return (0);
339 bad:
340 	vfs_freeopts(opts);
341 	return (error);
342 }
343 
344 /*
345  * Merge the old mount options with the new ones passed
346  * in the MNT_UPDATE case.
347  *
348  * XXX: This function will keep a "nofoo" option in the new
349  * options.  E.g, if the option's canonical name is "foo",
350  * "nofoo" ends up in the mount point's active options.
351  */
352 static void
353 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *oldopts)
354 {
355 	struct vfsopt *opt, *new;
356 
357 	TAILQ_FOREACH(opt, oldopts, link) {
358 		new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
359 		new->name = strdup(opt->name, M_MOUNT);
360 		if (opt->len != 0) {
361 			new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
362 			bcopy(opt->value, new->value, opt->len);
363 		} else
364 			new->value = NULL;
365 		new->len = opt->len;
366 		new->seen = opt->seen;
367 		TAILQ_INSERT_HEAD(toopts, new, link);
368 	}
369 	vfs_sanitizeopts(toopts);
370 }
371 
372 /*
373  * Mount a filesystem.
374  */
375 #ifndef _SYS_SYSPROTO_H_
376 struct nmount_args {
377 	struct iovec *iovp;
378 	unsigned int iovcnt;
379 	int flags;
380 };
381 #endif
382 int
383 sys_nmount(struct thread *td, struct nmount_args *uap)
384 {
385 	struct uio *auio;
386 	int error;
387 	u_int iovcnt;
388 	uint64_t flags;
389 
390 	/*
391 	 * Mount flags are now 64-bits. On 32-bit archtectures only
392 	 * 32-bits are passed in, but from here on everything handles
393 	 * 64-bit flags correctly.
394 	 */
395 	flags = uap->flags;
396 
397 	AUDIT_ARG_FFLAGS(flags);
398 	CTR4(KTR_VFS, "%s: iovp %p with iovcnt %d and flags %d", __func__,
399 	    uap->iovp, uap->iovcnt, flags);
400 
401 	/*
402 	 * Filter out MNT_ROOTFS.  We do not want clients of nmount() in
403 	 * userspace to set this flag, but we must filter it out if we want
404 	 * MNT_UPDATE on the root file system to work.
405 	 * MNT_ROOTFS should only be set by the kernel when mounting its
406 	 * root file system.
407 	 */
408 	flags &= ~MNT_ROOTFS;
409 
410 	iovcnt = uap->iovcnt;
411 	/*
412 	 * Check that we have an even number of iovec's
413 	 * and that we have at least two options.
414 	 */
415 	if ((iovcnt & 1) || (iovcnt < 4)) {
416 		CTR2(KTR_VFS, "%s: failed for invalid iovcnt %d", __func__,
417 		    uap->iovcnt);
418 		return (EINVAL);
419 	}
420 
421 	error = copyinuio(uap->iovp, iovcnt, &auio);
422 	if (error) {
423 		CTR2(KTR_VFS, "%s: failed for invalid uio op with %d errno",
424 		    __func__, error);
425 		return (error);
426 	}
427 	error = vfs_donmount(td, flags, auio);
428 
429 	free(auio, M_IOV);
430 	return (error);
431 }
432 
433 /*
434  * ---------------------------------------------------------------------
435  * Various utility functions
436  */
437 
438 void
439 vfs_ref(struct mount *mp)
440 {
441 
442 	CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
443 	MNT_ILOCK(mp);
444 	MNT_REF(mp);
445 	MNT_IUNLOCK(mp);
446 }
447 
448 void
449 vfs_rel(struct mount *mp)
450 {
451 
452 	CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
453 	MNT_ILOCK(mp);
454 	MNT_REL(mp);
455 	MNT_IUNLOCK(mp);
456 }
457 
458 /*
459  * Allocate and initialize the mount point struct.
460  */
461 struct mount *
462 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, const char *fspath,
463     struct ucred *cred)
464 {
465 	struct mount *mp;
466 
467 	mp = uma_zalloc(mount_zone, M_WAITOK);
468 	bzero(&mp->mnt_startzero,
469 	    __rangeof(struct mount, mnt_startzero, mnt_endzero));
470 	TAILQ_INIT(&mp->mnt_nvnodelist);
471 	mp->mnt_nvnodelistsize = 0;
472 	TAILQ_INIT(&mp->mnt_activevnodelist);
473 	mp->mnt_activevnodelistsize = 0;
474 	TAILQ_INIT(&mp->mnt_tmpfreevnodelist);
475 	mp->mnt_tmpfreevnodelistsize = 0;
476 	mp->mnt_ref = 0;
477 	(void) vfs_busy(mp, MBF_NOWAIT);
478 	atomic_add_acq_int(&vfsp->vfc_refcount, 1);
479 	mp->mnt_op = vfsp->vfc_vfsops;
480 	mp->mnt_vfc = vfsp;
481 	mp->mnt_stat.f_type = vfsp->vfc_typenum;
482 	mp->mnt_gen++;
483 	strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
484 	mp->mnt_vnodecovered = vp;
485 	mp->mnt_cred = crdup(cred);
486 	mp->mnt_stat.f_owner = cred->cr_uid;
487 	strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
488 	mp->mnt_iosize_max = DFLTPHYS;
489 #ifdef MAC
490 	mac_mount_init(mp);
491 	mac_mount_create(cred, mp);
492 #endif
493 	arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0);
494 	TAILQ_INIT(&mp->mnt_uppers);
495 	return (mp);
496 }
497 
498 /*
499  * Destroy the mount struct previously allocated by vfs_mount_alloc().
500  */
501 void
502 vfs_mount_destroy(struct mount *mp)
503 {
504 
505 	MNT_ILOCK(mp);
506 	mp->mnt_kern_flag |= MNTK_REFEXPIRE;
507 	if (mp->mnt_kern_flag & MNTK_MWAIT) {
508 		mp->mnt_kern_flag &= ~MNTK_MWAIT;
509 		wakeup(mp);
510 	}
511 	while (mp->mnt_ref)
512 		msleep(mp, MNT_MTX(mp), PVFS, "mntref", 0);
513 	KASSERT(mp->mnt_ref == 0,
514 	    ("%s: invalid refcount in the drain path @ %s:%d", __func__,
515 	    __FILE__, __LINE__));
516 	if (mp->mnt_writeopcount != 0)
517 		panic("vfs_mount_destroy: nonzero writeopcount");
518 	if (mp->mnt_secondary_writes != 0)
519 		panic("vfs_mount_destroy: nonzero secondary_writes");
520 	atomic_subtract_rel_int(&mp->mnt_vfc->vfc_refcount, 1);
521 	if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
522 		struct vnode *vp;
523 
524 		TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
525 			vn_printf(vp, "dangling vnode ");
526 		panic("unmount: dangling vnode");
527 	}
528 	KASSERT(TAILQ_EMPTY(&mp->mnt_uppers), ("mnt_uppers"));
529 	if (mp->mnt_nvnodelistsize != 0)
530 		panic("vfs_mount_destroy: nonzero nvnodelistsize");
531 	if (mp->mnt_activevnodelistsize != 0)
532 		panic("vfs_mount_destroy: nonzero activevnodelistsize");
533 	if (mp->mnt_tmpfreevnodelistsize != 0)
534 		panic("vfs_mount_destroy: nonzero tmpfreevnodelistsize");
535 	if (mp->mnt_lockref != 0)
536 		panic("vfs_mount_destroy: nonzero lock refcount");
537 	MNT_IUNLOCK(mp);
538 	if (mp->mnt_vnodecovered != NULL)
539 		vrele(mp->mnt_vnodecovered);
540 #ifdef MAC
541 	mac_mount_destroy(mp);
542 #endif
543 	if (mp->mnt_opt != NULL)
544 		vfs_freeopts(mp->mnt_opt);
545 	crfree(mp->mnt_cred);
546 	uma_zfree(mount_zone, mp);
547 }
548 
549 int
550 vfs_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions)
551 {
552 	struct vfsoptlist *optlist;
553 	struct vfsopt *opt, *tmp_opt;
554 	char *fstype, *fspath, *errmsg;
555 	int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
556 
557 	errmsg = fspath = NULL;
558 	errmsg_len = fspathlen = 0;
559 	errmsg_pos = -1;
560 
561 	error = vfs_buildopts(fsoptions, &optlist);
562 	if (error)
563 		return (error);
564 
565 	if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
566 		errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
567 
568 	/*
569 	 * We need these two options before the others,
570 	 * and they are mandatory for any filesystem.
571 	 * Ensure they are NUL terminated as well.
572 	 */
573 	fstypelen = 0;
574 	error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
575 	if (error || fstype[fstypelen - 1] != '\0') {
576 		error = EINVAL;
577 		if (errmsg != NULL)
578 			strncpy(errmsg, "Invalid fstype", errmsg_len);
579 		goto bail;
580 	}
581 	fspathlen = 0;
582 	error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
583 	if (error || fspath[fspathlen - 1] != '\0') {
584 		error = EINVAL;
585 		if (errmsg != NULL)
586 			strncpy(errmsg, "Invalid fspath", errmsg_len);
587 		goto bail;
588 	}
589 
590 	/*
591 	 * We need to see if we have the "update" option
592 	 * before we call vfs_domount(), since vfs_domount() has special
593 	 * logic based on MNT_UPDATE.  This is very important
594 	 * when we want to update the root filesystem.
595 	 */
596 	TAILQ_FOREACH_SAFE(opt, optlist, link, tmp_opt) {
597 		if (strcmp(opt->name, "update") == 0) {
598 			fsflags |= MNT_UPDATE;
599 			vfs_freeopt(optlist, opt);
600 		}
601 		else if (strcmp(opt->name, "async") == 0)
602 			fsflags |= MNT_ASYNC;
603 		else if (strcmp(opt->name, "force") == 0) {
604 			fsflags |= MNT_FORCE;
605 			vfs_freeopt(optlist, opt);
606 		}
607 		else if (strcmp(opt->name, "reload") == 0) {
608 			fsflags |= MNT_RELOAD;
609 			vfs_freeopt(optlist, opt);
610 		}
611 		else if (strcmp(opt->name, "multilabel") == 0)
612 			fsflags |= MNT_MULTILABEL;
613 		else if (strcmp(opt->name, "noasync") == 0)
614 			fsflags &= ~MNT_ASYNC;
615 		else if (strcmp(opt->name, "noatime") == 0)
616 			fsflags |= MNT_NOATIME;
617 		else if (strcmp(opt->name, "atime") == 0) {
618 			free(opt->name, M_MOUNT);
619 			opt->name = strdup("nonoatime", M_MOUNT);
620 		}
621 		else if (strcmp(opt->name, "noclusterr") == 0)
622 			fsflags |= MNT_NOCLUSTERR;
623 		else if (strcmp(opt->name, "clusterr") == 0) {
624 			free(opt->name, M_MOUNT);
625 			opt->name = strdup("nonoclusterr", M_MOUNT);
626 		}
627 		else if (strcmp(opt->name, "noclusterw") == 0)
628 			fsflags |= MNT_NOCLUSTERW;
629 		else if (strcmp(opt->name, "clusterw") == 0) {
630 			free(opt->name, M_MOUNT);
631 			opt->name = strdup("nonoclusterw", M_MOUNT);
632 		}
633 		else if (strcmp(opt->name, "noexec") == 0)
634 			fsflags |= MNT_NOEXEC;
635 		else if (strcmp(opt->name, "exec") == 0) {
636 			free(opt->name, M_MOUNT);
637 			opt->name = strdup("nonoexec", M_MOUNT);
638 		}
639 		else if (strcmp(opt->name, "nosuid") == 0)
640 			fsflags |= MNT_NOSUID;
641 		else if (strcmp(opt->name, "suid") == 0) {
642 			free(opt->name, M_MOUNT);
643 			opt->name = strdup("nonosuid", M_MOUNT);
644 		}
645 		else if (strcmp(opt->name, "nosymfollow") == 0)
646 			fsflags |= MNT_NOSYMFOLLOW;
647 		else if (strcmp(opt->name, "symfollow") == 0) {
648 			free(opt->name, M_MOUNT);
649 			opt->name = strdup("nonosymfollow", M_MOUNT);
650 		}
651 		else if (strcmp(opt->name, "noro") == 0)
652 			fsflags &= ~MNT_RDONLY;
653 		else if (strcmp(opt->name, "rw") == 0)
654 			fsflags &= ~MNT_RDONLY;
655 		else if (strcmp(opt->name, "ro") == 0)
656 			fsflags |= MNT_RDONLY;
657 		else if (strcmp(opt->name, "rdonly") == 0) {
658 			free(opt->name, M_MOUNT);
659 			opt->name = strdup("ro", M_MOUNT);
660 			fsflags |= MNT_RDONLY;
661 		}
662 		else if (strcmp(opt->name, "suiddir") == 0)
663 			fsflags |= MNT_SUIDDIR;
664 		else if (strcmp(opt->name, "sync") == 0)
665 			fsflags |= MNT_SYNCHRONOUS;
666 		else if (strcmp(opt->name, "union") == 0)
667 			fsflags |= MNT_UNION;
668 		else if (strcmp(opt->name, "automounted") == 0) {
669 			fsflags |= MNT_AUTOMOUNTED;
670 			vfs_freeopt(optlist, opt);
671 		}
672 	}
673 
674 	/*
675 	 * Be ultra-paranoid about making sure the type and fspath
676 	 * variables will fit in our mp buffers, including the
677 	 * terminating NUL.
678 	 */
679 	if (fstypelen > MFSNAMELEN || fspathlen > MNAMELEN) {
680 		error = ENAMETOOLONG;
681 		goto bail;
682 	}
683 
684 	error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
685 bail:
686 	/* copyout the errmsg */
687 	if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
688 	    && errmsg_len > 0 && errmsg != NULL) {
689 		if (fsoptions->uio_segflg == UIO_SYSSPACE) {
690 			bcopy(errmsg,
691 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
692 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
693 		} else {
694 			copyout(errmsg,
695 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
696 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
697 		}
698 	}
699 
700 	if (optlist != NULL)
701 		vfs_freeopts(optlist);
702 	return (error);
703 }
704 
705 /*
706  * Old mount API.
707  */
708 #ifndef _SYS_SYSPROTO_H_
709 struct mount_args {
710 	char	*type;
711 	char	*path;
712 	int	flags;
713 	caddr_t	data;
714 };
715 #endif
716 /* ARGSUSED */
717 int
718 sys_mount(struct thread *td, struct mount_args *uap)
719 {
720 	char *fstype;
721 	struct vfsconf *vfsp = NULL;
722 	struct mntarg *ma = NULL;
723 	uint64_t flags;
724 	int error;
725 
726 	/*
727 	 * Mount flags are now 64-bits. On 32-bit architectures only
728 	 * 32-bits are passed in, but from here on everything handles
729 	 * 64-bit flags correctly.
730 	 */
731 	flags = uap->flags;
732 
733 	AUDIT_ARG_FFLAGS(flags);
734 
735 	/*
736 	 * Filter out MNT_ROOTFS.  We do not want clients of mount() in
737 	 * userspace to set this flag, but we must filter it out if we want
738 	 * MNT_UPDATE on the root file system to work.
739 	 * MNT_ROOTFS should only be set by the kernel when mounting its
740 	 * root file system.
741 	 */
742 	flags &= ~MNT_ROOTFS;
743 
744 	fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
745 	error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
746 	if (error) {
747 		free(fstype, M_TEMP);
748 		return (error);
749 	}
750 
751 	AUDIT_ARG_TEXT(fstype);
752 	vfsp = vfs_byname_kld(fstype, td, &error);
753 	free(fstype, M_TEMP);
754 	if (vfsp == NULL)
755 		return (ENOENT);
756 	if (vfsp->vfc_vfsops->vfs_cmount == NULL)
757 		return (EOPNOTSUPP);
758 
759 	ma = mount_argsu(ma, "fstype", uap->type, MFSNAMELEN);
760 	ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
761 	ma = mount_argb(ma, flags & MNT_RDONLY, "noro");
762 	ma = mount_argb(ma, !(flags & MNT_NOSUID), "nosuid");
763 	ma = mount_argb(ma, !(flags & MNT_NOEXEC), "noexec");
764 
765 	error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags);
766 	return (error);
767 }
768 
769 /*
770  * vfs_domount_first(): first file system mount (not update)
771  */
772 static int
773 vfs_domount_first(
774 	struct thread *td,		/* Calling thread. */
775 	struct vfsconf *vfsp,		/* File system type. */
776 	char *fspath,			/* Mount path. */
777 	struct vnode *vp,		/* Vnode to be covered. */
778 	uint64_t fsflags,		/* Flags common to all filesystems. */
779 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
780 	)
781 {
782 	struct vattr va;
783 	struct mount *mp;
784 	struct vnode *newdp;
785 	int error;
786 
787 	ASSERT_VOP_ELOCKED(vp, __func__);
788 	KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
789 
790 	/*
791 	 * If the user is not root, ensure that they own the directory
792 	 * onto which we are attempting to mount.
793 	 */
794 	error = VOP_GETATTR(vp, &va, td->td_ucred);
795 	if (error == 0 && va.va_uid != td->td_ucred->cr_uid)
796 		error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN, 0);
797 	if (error == 0)
798 		error = vinvalbuf(vp, V_SAVE, 0, 0);
799 	if (error == 0 && vp->v_type != VDIR)
800 		error = ENOTDIR;
801 	if (error == 0) {
802 		VI_LOCK(vp);
803 		if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
804 			vp->v_iflag |= VI_MOUNT;
805 		else
806 			error = EBUSY;
807 		VI_UNLOCK(vp);
808 	}
809 	if (error != 0) {
810 		vput(vp);
811 		return (error);
812 	}
813 	VOP_UNLOCK(vp, 0);
814 
815 	/* Allocate and initialize the filesystem. */
816 	mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred);
817 	/* XXXMAC: pass to vfs_mount_alloc? */
818 	mp->mnt_optnew = *optlist;
819 	/* Set the mount level flags. */
820 	mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY));
821 
822 	/*
823 	 * Mount the filesystem.
824 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
825 	 * get.  No freeing of cn_pnbuf.
826 	 */
827 	error = VFS_MOUNT(mp);
828 	if (error != 0) {
829 		vfs_unbusy(mp);
830 		mp->mnt_vnodecovered = NULL;
831 		vfs_mount_destroy(mp);
832 		VI_LOCK(vp);
833 		vp->v_iflag &= ~VI_MOUNT;
834 		VI_UNLOCK(vp);
835 		vrele(vp);
836 		return (error);
837 	}
838 
839 	if (mp->mnt_opt != NULL)
840 		vfs_freeopts(mp->mnt_opt);
841 	mp->mnt_opt = mp->mnt_optnew;
842 	*optlist = NULL;
843 	(void)VFS_STATFS(mp, &mp->mnt_stat);
844 
845 	/*
846 	 * Prevent external consumers of mount options from reading mnt_optnew.
847 	 */
848 	mp->mnt_optnew = NULL;
849 
850 	MNT_ILOCK(mp);
851 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
852 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
853 		mp->mnt_kern_flag |= MNTK_ASYNC;
854 	else
855 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
856 	MNT_IUNLOCK(mp);
857 
858 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
859 	cache_purge(vp);
860 	VI_LOCK(vp);
861 	vp->v_iflag &= ~VI_MOUNT;
862 	VI_UNLOCK(vp);
863 	vp->v_mountedhere = mp;
864 	/* Place the new filesystem at the end of the mount list. */
865 	mtx_lock(&mountlist_mtx);
866 	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
867 	mtx_unlock(&mountlist_mtx);
868 	vfs_event_signal(NULL, VQ_MOUNT, 0);
869 	if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp))
870 		panic("mount: lost mount");
871 	VOP_UNLOCK(vp, 0);
872 	EVENTHANDLER_DIRECT_INVOKE(vfs_mounted, mp, newdp, td);
873 	VOP_UNLOCK(newdp, 0);
874 	mountcheckdirs(vp, newdp);
875 	vrele(newdp);
876 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
877 		vfs_allocate_syncvnode(mp);
878 	vfs_unbusy(mp);
879 	return (0);
880 }
881 
882 /*
883  * vfs_domount_update(): update of mounted file system
884  */
885 static int
886 vfs_domount_update(
887 	struct thread *td,		/* Calling thread. */
888 	struct vnode *vp,		/* Mount point vnode. */
889 	uint64_t fsflags,		/* Flags common to all filesystems. */
890 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
891 	)
892 {
893 	struct export_args export;
894 	void *bufp;
895 	struct mount *mp;
896 	int error, export_error, len;
897 	uint64_t flag;
898 
899 	ASSERT_VOP_ELOCKED(vp, __func__);
900 	KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
901 	mp = vp->v_mount;
902 
903 	if ((vp->v_vflag & VV_ROOT) == 0) {
904 		if (vfs_copyopt(*optlist, "export", &export, sizeof(export))
905 		    == 0)
906 			error = EXDEV;
907 		else
908 			error = EINVAL;
909 		vput(vp);
910 		return (error);
911 	}
912 
913 	/*
914 	 * We only allow the filesystem to be reloaded if it
915 	 * is currently mounted read-only.
916 	 */
917 	flag = mp->mnt_flag;
918 	if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
919 		vput(vp);
920 		return (EOPNOTSUPP);	/* Needs translation */
921 	}
922 	/*
923 	 * Only privileged root, or (if MNT_USER is set) the user that
924 	 * did the original mount is permitted to update it.
925 	 */
926 	error = vfs_suser(mp, td);
927 	if (error != 0) {
928 		vput(vp);
929 		return (error);
930 	}
931 	if (vfs_busy(mp, MBF_NOWAIT)) {
932 		vput(vp);
933 		return (EBUSY);
934 	}
935 	VI_LOCK(vp);
936 	if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
937 		VI_UNLOCK(vp);
938 		vfs_unbusy(mp);
939 		vput(vp);
940 		return (EBUSY);
941 	}
942 	vp->v_iflag |= VI_MOUNT;
943 	VI_UNLOCK(vp);
944 	VOP_UNLOCK(vp, 0);
945 
946 	MNT_ILOCK(mp);
947 	if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0) {
948 		MNT_IUNLOCK(mp);
949 		error = EBUSY;
950 		goto end;
951 	}
952 	mp->mnt_flag &= ~MNT_UPDATEMASK;
953 	mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE |
954 	    MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY);
955 	if ((mp->mnt_flag & MNT_ASYNC) == 0)
956 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
957 	MNT_IUNLOCK(mp);
958 	mp->mnt_optnew = *optlist;
959 	vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
960 
961 	/*
962 	 * Mount the filesystem.
963 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
964 	 * get.  No freeing of cn_pnbuf.
965 	 */
966 	error = VFS_MOUNT(mp);
967 
968 	export_error = 0;
969 	/* Process the export option. */
970 	if (error == 0 && vfs_getopt(mp->mnt_optnew, "export", &bufp,
971 	    &len) == 0) {
972 		/* Assume that there is only 1 ABI for each length. */
973 		switch (len) {
974 		case (sizeof(struct oexport_args)):
975 			bzero(&export, sizeof(export));
976 			/* FALLTHROUGH */
977 		case (sizeof(export)):
978 			bcopy(bufp, &export, len);
979 			export_error = vfs_export(mp, &export);
980 			break;
981 		default:
982 			export_error = EINVAL;
983 			break;
984 		}
985 	}
986 
987 	MNT_ILOCK(mp);
988 	if (error == 0) {
989 		mp->mnt_flag &=	~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
990 		    MNT_SNAPSHOT);
991 	} else {
992 		/*
993 		 * If we fail, restore old mount flags. MNT_QUOTA is special,
994 		 * because it is not part of MNT_UPDATEMASK, but it could have
995 		 * changed in the meantime if quotactl(2) was called.
996 		 * All in all we want current value of MNT_QUOTA, not the old
997 		 * one.
998 		 */
999 		mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
1000 	}
1001 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1002 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1003 		mp->mnt_kern_flag |= MNTK_ASYNC;
1004 	else
1005 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
1006 	MNT_IUNLOCK(mp);
1007 
1008 	if (error != 0)
1009 		goto end;
1010 
1011 	if (mp->mnt_opt != NULL)
1012 		vfs_freeopts(mp->mnt_opt);
1013 	mp->mnt_opt = mp->mnt_optnew;
1014 	*optlist = NULL;
1015 	(void)VFS_STATFS(mp, &mp->mnt_stat);
1016 	/*
1017 	 * Prevent external consumers of mount options from reading
1018 	 * mnt_optnew.
1019 	 */
1020 	mp->mnt_optnew = NULL;
1021 
1022 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
1023 		vfs_allocate_syncvnode(mp);
1024 	else
1025 		vfs_deallocate_syncvnode(mp);
1026 end:
1027 	vfs_unbusy(mp);
1028 	VI_LOCK(vp);
1029 	vp->v_iflag &= ~VI_MOUNT;
1030 	VI_UNLOCK(vp);
1031 	vrele(vp);
1032 	return (error != 0 ? error : export_error);
1033 }
1034 
1035 /*
1036  * vfs_domount(): actually attempt a filesystem mount.
1037  */
1038 static int
1039 vfs_domount(
1040 	struct thread *td,		/* Calling thread. */
1041 	const char *fstype,		/* Filesystem type. */
1042 	char *fspath,			/* Mount path. */
1043 	uint64_t fsflags,		/* Flags common to all filesystems. */
1044 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
1045 	)
1046 {
1047 	struct vfsconf *vfsp;
1048 	struct nameidata nd;
1049 	struct vnode *vp;
1050 	char *pathbuf;
1051 	int error;
1052 
1053 	/*
1054 	 * Be ultra-paranoid about making sure the type and fspath
1055 	 * variables will fit in our mp buffers, including the
1056 	 * terminating NUL.
1057 	 */
1058 	if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
1059 		return (ENAMETOOLONG);
1060 
1061 	if (jailed(td->td_ucred) || usermount == 0) {
1062 		if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
1063 			return (error);
1064 	}
1065 
1066 	/*
1067 	 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
1068 	 */
1069 	if (fsflags & MNT_EXPORTED) {
1070 		error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
1071 		if (error)
1072 			return (error);
1073 	}
1074 	if (fsflags & MNT_SUIDDIR) {
1075 		error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
1076 		if (error)
1077 			return (error);
1078 	}
1079 	/*
1080 	 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
1081 	 */
1082 	if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
1083 		if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
1084 			fsflags |= MNT_NOSUID | MNT_USER;
1085 	}
1086 
1087 	/* Load KLDs before we lock the covered vnode to avoid reversals. */
1088 	vfsp = NULL;
1089 	if ((fsflags & MNT_UPDATE) == 0) {
1090 		/* Don't try to load KLDs if we're mounting the root. */
1091 		if (fsflags & MNT_ROOTFS)
1092 			vfsp = vfs_byname(fstype);
1093 		else
1094 			vfsp = vfs_byname_kld(fstype, td, &error);
1095 		if (vfsp == NULL)
1096 			return (ENODEV);
1097 		if (jailed(td->td_ucred) && !(vfsp->vfc_flags & VFCF_JAIL))
1098 			return (EPERM);
1099 	}
1100 
1101 	/*
1102 	 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
1103 	 */
1104 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1105 	    UIO_SYSSPACE, fspath, td);
1106 	error = namei(&nd);
1107 	if (error != 0)
1108 		return (error);
1109 	NDFREE(&nd, NDF_ONLY_PNBUF);
1110 	vp = nd.ni_vp;
1111 	if ((fsflags & MNT_UPDATE) == 0) {
1112 		pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1113 		strcpy(pathbuf, fspath);
1114 		error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN);
1115 		/* debug.disablefullpath == 1 results in ENODEV */
1116 		if (error == 0 || error == ENODEV) {
1117 			error = vfs_domount_first(td, vfsp, pathbuf, vp,
1118 			    fsflags, optlist);
1119 		}
1120 		free(pathbuf, M_TEMP);
1121 	} else
1122 		error = vfs_domount_update(td, vp, fsflags, optlist);
1123 
1124 	return (error);
1125 }
1126 
1127 /*
1128  * Unmount a filesystem.
1129  *
1130  * Note: unmount takes a path to the vnode mounted on as argument, not
1131  * special file (as before).
1132  */
1133 #ifndef _SYS_SYSPROTO_H_
1134 struct unmount_args {
1135 	char	*path;
1136 	int	flags;
1137 };
1138 #endif
1139 /* ARGSUSED */
1140 int
1141 sys_unmount(struct thread *td, struct unmount_args *uap)
1142 {
1143 	struct nameidata nd;
1144 	struct mount *mp;
1145 	char *pathbuf;
1146 	int error, id0, id1;
1147 
1148 	AUDIT_ARG_VALUE(uap->flags);
1149 	if (jailed(td->td_ucred) || usermount == 0) {
1150 		error = priv_check(td, PRIV_VFS_UNMOUNT);
1151 		if (error)
1152 			return (error);
1153 	}
1154 
1155 	pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1156 	error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1157 	if (error) {
1158 		free(pathbuf, M_TEMP);
1159 		return (error);
1160 	}
1161 	if (uap->flags & MNT_BYFSID) {
1162 		AUDIT_ARG_TEXT(pathbuf);
1163 		/* Decode the filesystem ID. */
1164 		if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
1165 			free(pathbuf, M_TEMP);
1166 			return (EINVAL);
1167 		}
1168 
1169 		mtx_lock(&mountlist_mtx);
1170 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1171 			if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1172 			    mp->mnt_stat.f_fsid.val[1] == id1) {
1173 				vfs_ref(mp);
1174 				break;
1175 			}
1176 		}
1177 		mtx_unlock(&mountlist_mtx);
1178 	} else {
1179 		/*
1180 		 * Try to find global path for path argument.
1181 		 */
1182 		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1183 		    UIO_SYSSPACE, pathbuf, td);
1184 		if (namei(&nd) == 0) {
1185 			NDFREE(&nd, NDF_ONLY_PNBUF);
1186 			error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
1187 			    MNAMELEN);
1188 			if (error == 0 || error == ENODEV)
1189 				vput(nd.ni_vp);
1190 		}
1191 		mtx_lock(&mountlist_mtx);
1192 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1193 			if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) {
1194 				vfs_ref(mp);
1195 				break;
1196 			}
1197 		}
1198 		mtx_unlock(&mountlist_mtx);
1199 	}
1200 	free(pathbuf, M_TEMP);
1201 	if (mp == NULL) {
1202 		/*
1203 		 * Previously we returned ENOENT for a nonexistent path and
1204 		 * EINVAL for a non-mountpoint.  We cannot tell these apart
1205 		 * now, so in the !MNT_BYFSID case return the more likely
1206 		 * EINVAL for compatibility.
1207 		 */
1208 		return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1209 	}
1210 
1211 	/*
1212 	 * Don't allow unmounting the root filesystem.
1213 	 */
1214 	if (mp->mnt_flag & MNT_ROOTFS) {
1215 		vfs_rel(mp);
1216 		return (EINVAL);
1217 	}
1218 	error = dounmount(mp, uap->flags, td);
1219 	return (error);
1220 }
1221 
1222 /*
1223  * Return error if any of the vnodes, ignoring the root vnode
1224  * and the syncer vnode, have non-zero usecount.
1225  *
1226  * This function is purely advisory - it can return false positives
1227  * and negatives.
1228  */
1229 static int
1230 vfs_check_usecounts(struct mount *mp)
1231 {
1232 	struct vnode *vp, *mvp;
1233 
1234 	MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1235 		if ((vp->v_vflag & VV_ROOT) == 0 && vp->v_type != VNON &&
1236 		    vp->v_usecount != 0) {
1237 			VI_UNLOCK(vp);
1238 			MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1239 			return (EBUSY);
1240 		}
1241 		VI_UNLOCK(vp);
1242 	}
1243 
1244 	return (0);
1245 }
1246 
1247 static void
1248 dounmount_cleanup(struct mount *mp, struct vnode *coveredvp, int mntkflags)
1249 {
1250 
1251 	mtx_assert(MNT_MTX(mp), MA_OWNED);
1252 	mp->mnt_kern_flag &= ~mntkflags;
1253 	if ((mp->mnt_kern_flag & MNTK_MWAIT) != 0) {
1254 		mp->mnt_kern_flag &= ~MNTK_MWAIT;
1255 		wakeup(mp);
1256 	}
1257 	MNT_IUNLOCK(mp);
1258 	if (coveredvp != NULL) {
1259 		VOP_UNLOCK(coveredvp, 0);
1260 		vdrop(coveredvp);
1261 	}
1262 	vn_finished_write(mp);
1263 }
1264 
1265 /*
1266  * Do the actual filesystem unmount.
1267  */
1268 int
1269 dounmount(struct mount *mp, int flags, struct thread *td)
1270 {
1271 	struct vnode *coveredvp;
1272 	int error;
1273 	uint64_t async_flag;
1274 	int mnt_gen_r;
1275 
1276 	if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
1277 		mnt_gen_r = mp->mnt_gen;
1278 		VI_LOCK(coveredvp);
1279 		vholdl(coveredvp);
1280 		vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
1281 		/*
1282 		 * Check for mp being unmounted while waiting for the
1283 		 * covered vnode lock.
1284 		 */
1285 		if (coveredvp->v_mountedhere != mp ||
1286 		    coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
1287 			VOP_UNLOCK(coveredvp, 0);
1288 			vdrop(coveredvp);
1289 			vfs_rel(mp);
1290 			return (EBUSY);
1291 		}
1292 	}
1293 
1294 	/*
1295 	 * Only privileged root, or (if MNT_USER is set) the user that did the
1296 	 * original mount is permitted to unmount this filesystem.
1297 	 */
1298 	error = vfs_suser(mp, td);
1299 	if (error != 0) {
1300 		if (coveredvp != NULL) {
1301 			VOP_UNLOCK(coveredvp, 0);
1302 			vdrop(coveredvp);
1303 		}
1304 		vfs_rel(mp);
1305 		return (error);
1306 	}
1307 
1308 	vn_start_write(NULL, &mp, V_WAIT | V_MNTREF);
1309 	MNT_ILOCK(mp);
1310 	if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 ||
1311 	    (mp->mnt_flag & MNT_UPDATE) != 0 ||
1312 	    !TAILQ_EMPTY(&mp->mnt_uppers)) {
1313 		dounmount_cleanup(mp, coveredvp, 0);
1314 		return (EBUSY);
1315 	}
1316 	mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ;
1317 	if (flags & MNT_NONBUSY) {
1318 		MNT_IUNLOCK(mp);
1319 		error = vfs_check_usecounts(mp);
1320 		MNT_ILOCK(mp);
1321 		if (error != 0) {
1322 			dounmount_cleanup(mp, coveredvp, MNTK_UNMOUNT |
1323 			    MNTK_NOINSMNTQ);
1324 			return (error);
1325 		}
1326 	}
1327 	/* Allow filesystems to detect that a forced unmount is in progress. */
1328 	if (flags & MNT_FORCE) {
1329 		mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1330 		MNT_IUNLOCK(mp);
1331 		/*
1332 		 * Must be done after setting MNTK_UNMOUNTF and before
1333 		 * waiting for mnt_lockref to become 0.
1334 		 */
1335 		VFS_PURGE(mp);
1336 		MNT_ILOCK(mp);
1337 	}
1338 	error = 0;
1339 	if (mp->mnt_lockref) {
1340 		mp->mnt_kern_flag |= MNTK_DRAINING;
1341 		error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
1342 		    "mount drain", 0);
1343 	}
1344 	MNT_IUNLOCK(mp);
1345 	KASSERT(mp->mnt_lockref == 0,
1346 	    ("%s: invalid lock refcount in the drain path @ %s:%d",
1347 	    __func__, __FILE__, __LINE__));
1348 	KASSERT(error == 0,
1349 	    ("%s: invalid return value for msleep in the drain path @ %s:%d",
1350 	    __func__, __FILE__, __LINE__));
1351 
1352 	if (mp->mnt_flag & MNT_EXPUBLIC)
1353 		vfs_setpublicfs(NULL, NULL, NULL);
1354 
1355 	/*
1356 	 * From now, we can claim that the use reference on the
1357 	 * coveredvp is ours, and the ref can be released only by
1358 	 * successfull unmount by us, or left for later unmount
1359 	 * attempt.  The previously acquired hold reference is no
1360 	 * longer needed to protect the vnode from reuse.
1361 	 */
1362 	if (coveredvp != NULL)
1363 		vdrop(coveredvp);
1364 
1365 	vfs_msync(mp, MNT_WAIT);
1366 	MNT_ILOCK(mp);
1367 	async_flag = mp->mnt_flag & MNT_ASYNC;
1368 	mp->mnt_flag &= ~MNT_ASYNC;
1369 	mp->mnt_kern_flag &= ~MNTK_ASYNC;
1370 	MNT_IUNLOCK(mp);
1371 	cache_purgevfs(mp, false); /* remove cache entries for this file sys */
1372 	vfs_deallocate_syncvnode(mp);
1373 	if ((mp->mnt_flag & MNT_RDONLY) != 0 || (flags & MNT_FORCE) != 0 ||
1374 	    (error = VFS_SYNC(mp, MNT_WAIT)) == 0)
1375 		error = VFS_UNMOUNT(mp, flags);
1376 	vn_finished_write(mp);
1377 	/*
1378 	 * If we failed to flush the dirty blocks for this mount point,
1379 	 * undo all the cdir/rdir and rootvnode changes we made above.
1380 	 * Unless we failed to do so because the device is reporting that
1381 	 * it doesn't exist anymore.
1382 	 */
1383 	if (error && error != ENXIO) {
1384 		MNT_ILOCK(mp);
1385 		mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ;
1386 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
1387 			MNT_IUNLOCK(mp);
1388 			vfs_allocate_syncvnode(mp);
1389 			MNT_ILOCK(mp);
1390 		}
1391 		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1392 		mp->mnt_flag |= async_flag;
1393 		if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1394 		    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1395 			mp->mnt_kern_flag |= MNTK_ASYNC;
1396 		if (mp->mnt_kern_flag & MNTK_MWAIT) {
1397 			mp->mnt_kern_flag &= ~MNTK_MWAIT;
1398 			wakeup(mp);
1399 		}
1400 		MNT_IUNLOCK(mp);
1401 		if (coveredvp)
1402 			VOP_UNLOCK(coveredvp, 0);
1403 		return (error);
1404 	}
1405 	mtx_lock(&mountlist_mtx);
1406 	TAILQ_REMOVE(&mountlist, mp, mnt_list);
1407 	mtx_unlock(&mountlist_mtx);
1408 	EVENTHANDLER_DIRECT_INVOKE(vfs_unmounted, mp, td);
1409 	if (coveredvp != NULL) {
1410 		coveredvp->v_mountedhere = NULL;
1411 		VOP_UNLOCK(coveredvp, 0);
1412 	}
1413 	vfs_event_signal(NULL, VQ_UNMOUNT, 0);
1414 	if (rootvnode != NULL && mp == rootvnode->v_mount) {
1415 		vrele(rootvnode);
1416 		rootvnode = NULL;
1417 	}
1418 	if (mp == rootdevmp)
1419 		rootdevmp = NULL;
1420 	vfs_mount_destroy(mp);
1421 	return (0);
1422 }
1423 
1424 /*
1425  * Report errors during filesystem mounting.
1426  */
1427 void
1428 vfs_mount_error(struct mount *mp, const char *fmt, ...)
1429 {
1430 	struct vfsoptlist *moptlist = mp->mnt_optnew;
1431 	va_list ap;
1432 	int error, len;
1433 	char *errmsg;
1434 
1435 	error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
1436 	if (error || errmsg == NULL || len <= 0)
1437 		return;
1438 
1439 	va_start(ap, fmt);
1440 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1441 	va_end(ap);
1442 }
1443 
1444 void
1445 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
1446 {
1447 	va_list ap;
1448 	int error, len;
1449 	char *errmsg;
1450 
1451 	error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
1452 	if (error || errmsg == NULL || len <= 0)
1453 		return;
1454 
1455 	va_start(ap, fmt);
1456 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1457 	va_end(ap);
1458 }
1459 
1460 /*
1461  * ---------------------------------------------------------------------
1462  * Functions for querying mount options/arguments from filesystems.
1463  */
1464 
1465 /*
1466  * Check that no unknown options are given
1467  */
1468 int
1469 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1470 {
1471 	struct vfsopt *opt;
1472 	char errmsg[255];
1473 	const char **t, *p, *q;
1474 	int ret = 0;
1475 
1476 	TAILQ_FOREACH(opt, opts, link) {
1477 		p = opt->name;
1478 		q = NULL;
1479 		if (p[0] == 'n' && p[1] == 'o')
1480 			q = p + 2;
1481 		for(t = global_opts; *t != NULL; t++) {
1482 			if (strcmp(*t, p) == 0)
1483 				break;
1484 			if (q != NULL) {
1485 				if (strcmp(*t, q) == 0)
1486 					break;
1487 			}
1488 		}
1489 		if (*t != NULL)
1490 			continue;
1491 		for(t = legal; *t != NULL; t++) {
1492 			if (strcmp(*t, p) == 0)
1493 				break;
1494 			if (q != NULL) {
1495 				if (strcmp(*t, q) == 0)
1496 					break;
1497 			}
1498 		}
1499 		if (*t != NULL)
1500 			continue;
1501 		snprintf(errmsg, sizeof(errmsg),
1502 		    "mount option <%s> is unknown", p);
1503 		ret = EINVAL;
1504 	}
1505 	if (ret != 0) {
1506 		TAILQ_FOREACH(opt, opts, link) {
1507 			if (strcmp(opt->name, "errmsg") == 0) {
1508 				strncpy((char *)opt->value, errmsg, opt->len);
1509 				break;
1510 			}
1511 		}
1512 		if (opt == NULL)
1513 			printf("%s\n", errmsg);
1514 	}
1515 	return (ret);
1516 }
1517 
1518 /*
1519  * Get a mount option by its name.
1520  *
1521  * Return 0 if the option was found, ENOENT otherwise.
1522  * If len is non-NULL it will be filled with the length
1523  * of the option. If buf is non-NULL, it will be filled
1524  * with the address of the option.
1525  */
1526 int
1527 vfs_getopt(struct vfsoptlist *opts, const char *name, void **buf, int *len)
1528 {
1529 	struct vfsopt *opt;
1530 
1531 	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1532 
1533 	TAILQ_FOREACH(opt, opts, link) {
1534 		if (strcmp(name, opt->name) == 0) {
1535 			opt->seen = 1;
1536 			if (len != NULL)
1537 				*len = opt->len;
1538 			if (buf != NULL)
1539 				*buf = opt->value;
1540 			return (0);
1541 		}
1542 	}
1543 	return (ENOENT);
1544 }
1545 
1546 int
1547 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
1548 {
1549 	struct vfsopt *opt;
1550 
1551 	if (opts == NULL)
1552 		return (-1);
1553 
1554 	TAILQ_FOREACH(opt, opts, link) {
1555 		if (strcmp(name, opt->name) == 0) {
1556 			opt->seen = 1;
1557 			return (opt->pos);
1558 		}
1559 	}
1560 	return (-1);
1561 }
1562 
1563 int
1564 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value)
1565 {
1566 	char *opt_value, *vtp;
1567 	quad_t iv;
1568 	int error, opt_len;
1569 
1570 	error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len);
1571 	if (error != 0)
1572 		return (error);
1573 	if (opt_len == 0 || opt_value == NULL)
1574 		return (EINVAL);
1575 	if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0')
1576 		return (EINVAL);
1577 	iv = strtoq(opt_value, &vtp, 0);
1578 	if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0'))
1579 		return (EINVAL);
1580 	if (iv < 0)
1581 		return (EINVAL);
1582 	switch (vtp[0]) {
1583 	case 't':
1584 	case 'T':
1585 		iv *= 1024;
1586 	case 'g':
1587 	case 'G':
1588 		iv *= 1024;
1589 	case 'm':
1590 	case 'M':
1591 		iv *= 1024;
1592 	case 'k':
1593 	case 'K':
1594 		iv *= 1024;
1595 	case '\0':
1596 		break;
1597 	default:
1598 		return (EINVAL);
1599 	}
1600 	*value = iv;
1601 
1602 	return (0);
1603 }
1604 
1605 char *
1606 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
1607 {
1608 	struct vfsopt *opt;
1609 
1610 	*error = 0;
1611 	TAILQ_FOREACH(opt, opts, link) {
1612 		if (strcmp(name, opt->name) != 0)
1613 			continue;
1614 		opt->seen = 1;
1615 		if (opt->len == 0 ||
1616 		    ((char *)opt->value)[opt->len - 1] != '\0') {
1617 			*error = EINVAL;
1618 			return (NULL);
1619 		}
1620 		return (opt->value);
1621 	}
1622 	*error = ENOENT;
1623 	return (NULL);
1624 }
1625 
1626 int
1627 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
1628 	uint64_t val)
1629 {
1630 	struct vfsopt *opt;
1631 
1632 	TAILQ_FOREACH(opt, opts, link) {
1633 		if (strcmp(name, opt->name) == 0) {
1634 			opt->seen = 1;
1635 			if (w != NULL)
1636 				*w |= val;
1637 			return (1);
1638 		}
1639 	}
1640 	if (w != NULL)
1641 		*w &= ~val;
1642 	return (0);
1643 }
1644 
1645 int
1646 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
1647 {
1648 	va_list ap;
1649 	struct vfsopt *opt;
1650 	int ret;
1651 
1652 	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1653 
1654 	TAILQ_FOREACH(opt, opts, link) {
1655 		if (strcmp(name, opt->name) != 0)
1656 			continue;
1657 		opt->seen = 1;
1658 		if (opt->len == 0 || opt->value == NULL)
1659 			return (0);
1660 		if (((char *)opt->value)[opt->len - 1] != '\0')
1661 			return (0);
1662 		va_start(ap, fmt);
1663 		ret = vsscanf(opt->value, fmt, ap);
1664 		va_end(ap);
1665 		return (ret);
1666 	}
1667 	return (0);
1668 }
1669 
1670 int
1671 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len)
1672 {
1673 	struct vfsopt *opt;
1674 
1675 	TAILQ_FOREACH(opt, opts, link) {
1676 		if (strcmp(name, opt->name) != 0)
1677 			continue;
1678 		opt->seen = 1;
1679 		if (opt->value == NULL)
1680 			opt->len = len;
1681 		else {
1682 			if (opt->len != len)
1683 				return (EINVAL);
1684 			bcopy(value, opt->value, len);
1685 		}
1686 		return (0);
1687 	}
1688 	return (ENOENT);
1689 }
1690 
1691 int
1692 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len)
1693 {
1694 	struct vfsopt *opt;
1695 
1696 	TAILQ_FOREACH(opt, opts, link) {
1697 		if (strcmp(name, opt->name) != 0)
1698 			continue;
1699 		opt->seen = 1;
1700 		if (opt->value == NULL)
1701 			opt->len = len;
1702 		else {
1703 			if (opt->len < len)
1704 				return (EINVAL);
1705 			opt->len = len;
1706 			bcopy(value, opt->value, len);
1707 		}
1708 		return (0);
1709 	}
1710 	return (ENOENT);
1711 }
1712 
1713 int
1714 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value)
1715 {
1716 	struct vfsopt *opt;
1717 
1718 	TAILQ_FOREACH(opt, opts, link) {
1719 		if (strcmp(name, opt->name) != 0)
1720 			continue;
1721 		opt->seen = 1;
1722 		if (opt->value == NULL)
1723 			opt->len = strlen(value) + 1;
1724 		else if (strlcpy(opt->value, value, opt->len) >= opt->len)
1725 			return (EINVAL);
1726 		return (0);
1727 	}
1728 	return (ENOENT);
1729 }
1730 
1731 /*
1732  * Find and copy a mount option.
1733  *
1734  * The size of the buffer has to be specified
1735  * in len, if it is not the same length as the
1736  * mount option, EINVAL is returned.
1737  * Returns ENOENT if the option is not found.
1738  */
1739 int
1740 vfs_copyopt(struct vfsoptlist *opts, const char *name, void *dest, int len)
1741 {
1742 	struct vfsopt *opt;
1743 
1744 	KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1745 
1746 	TAILQ_FOREACH(opt, opts, link) {
1747 		if (strcmp(name, opt->name) == 0) {
1748 			opt->seen = 1;
1749 			if (len != opt->len)
1750 				return (EINVAL);
1751 			bcopy(opt->value, dest, opt->len);
1752 			return (0);
1753 		}
1754 	}
1755 	return (ENOENT);
1756 }
1757 
1758 int
1759 __vfs_statfs(struct mount *mp, struct statfs *sbp)
1760 {
1761 	int error;
1762 
1763 	error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat);
1764 	if (sbp != &mp->mnt_stat)
1765 		*sbp = mp->mnt_stat;
1766 	return (error);
1767 }
1768 
1769 void
1770 vfs_mountedfrom(struct mount *mp, const char *from)
1771 {
1772 
1773 	bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
1774 	strlcpy(mp->mnt_stat.f_mntfromname, from,
1775 	    sizeof mp->mnt_stat.f_mntfromname);
1776 }
1777 
1778 /*
1779  * ---------------------------------------------------------------------
1780  * This is the api for building mount args and mounting filesystems from
1781  * inside the kernel.
1782  *
1783  * The API works by accumulation of individual args.  First error is
1784  * latched.
1785  *
1786  * XXX: should be documented in new manpage kernel_mount(9)
1787  */
1788 
1789 /* A memory allocation which must be freed when we are done */
1790 struct mntaarg {
1791 	SLIST_ENTRY(mntaarg)	next;
1792 };
1793 
1794 /* The header for the mount arguments */
1795 struct mntarg {
1796 	struct iovec *v;
1797 	int len;
1798 	int error;
1799 	SLIST_HEAD(, mntaarg)	list;
1800 };
1801 
1802 /*
1803  * Add a boolean argument.
1804  *
1805  * flag is the boolean value.
1806  * name must start with "no".
1807  */
1808 struct mntarg *
1809 mount_argb(struct mntarg *ma, int flag, const char *name)
1810 {
1811 
1812 	KASSERT(name[0] == 'n' && name[1] == 'o',
1813 	    ("mount_argb(...,%s): name must start with 'no'", name));
1814 
1815 	return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
1816 }
1817 
1818 /*
1819  * Add an argument printf style
1820  */
1821 struct mntarg *
1822 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
1823 {
1824 	va_list ap;
1825 	struct mntaarg *maa;
1826 	struct sbuf *sb;
1827 	int len;
1828 
1829 	if (ma == NULL) {
1830 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1831 		SLIST_INIT(&ma->list);
1832 	}
1833 	if (ma->error)
1834 		return (ma);
1835 
1836 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1837 	    M_MOUNT, M_WAITOK);
1838 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1839 	ma->v[ma->len].iov_len = strlen(name) + 1;
1840 	ma->len++;
1841 
1842 	sb = sbuf_new_auto();
1843 	va_start(ap, fmt);
1844 	sbuf_vprintf(sb, fmt, ap);
1845 	va_end(ap);
1846 	sbuf_finish(sb);
1847 	len = sbuf_len(sb) + 1;
1848 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1849 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1850 	bcopy(sbuf_data(sb), maa + 1, len);
1851 	sbuf_delete(sb);
1852 
1853 	ma->v[ma->len].iov_base = maa + 1;
1854 	ma->v[ma->len].iov_len = len;
1855 	ma->len++;
1856 
1857 	return (ma);
1858 }
1859 
1860 /*
1861  * Add an argument which is a userland string.
1862  */
1863 struct mntarg *
1864 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
1865 {
1866 	struct mntaarg *maa;
1867 	char *tbuf;
1868 
1869 	if (val == NULL)
1870 		return (ma);
1871 	if (ma == NULL) {
1872 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1873 		SLIST_INIT(&ma->list);
1874 	}
1875 	if (ma->error)
1876 		return (ma);
1877 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1878 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1879 	tbuf = (void *)(maa + 1);
1880 	ma->error = copyinstr(val, tbuf, len, NULL);
1881 	return (mount_arg(ma, name, tbuf, -1));
1882 }
1883 
1884 /*
1885  * Plain argument.
1886  *
1887  * If length is -1, treat value as a C string.
1888  */
1889 struct mntarg *
1890 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
1891 {
1892 
1893 	if (ma == NULL) {
1894 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1895 		SLIST_INIT(&ma->list);
1896 	}
1897 	if (ma->error)
1898 		return (ma);
1899 
1900 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1901 	    M_MOUNT, M_WAITOK);
1902 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1903 	ma->v[ma->len].iov_len = strlen(name) + 1;
1904 	ma->len++;
1905 
1906 	ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
1907 	if (len < 0)
1908 		ma->v[ma->len].iov_len = strlen(val) + 1;
1909 	else
1910 		ma->v[ma->len].iov_len = len;
1911 	ma->len++;
1912 	return (ma);
1913 }
1914 
1915 /*
1916  * Free a mntarg structure
1917  */
1918 static void
1919 free_mntarg(struct mntarg *ma)
1920 {
1921 	struct mntaarg *maa;
1922 
1923 	while (!SLIST_EMPTY(&ma->list)) {
1924 		maa = SLIST_FIRST(&ma->list);
1925 		SLIST_REMOVE_HEAD(&ma->list, next);
1926 		free(maa, M_MOUNT);
1927 	}
1928 	free(ma->v, M_MOUNT);
1929 	free(ma, M_MOUNT);
1930 }
1931 
1932 /*
1933  * Mount a filesystem
1934  */
1935 int
1936 kernel_mount(struct mntarg *ma, uint64_t flags)
1937 {
1938 	struct uio auio;
1939 	int error;
1940 
1941 	KASSERT(ma != NULL, ("kernel_mount NULL ma"));
1942 	KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
1943 	KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
1944 
1945 	auio.uio_iov = ma->v;
1946 	auio.uio_iovcnt = ma->len;
1947 	auio.uio_segflg = UIO_SYSSPACE;
1948 
1949 	error = ma->error;
1950 	if (!error)
1951 		error = vfs_donmount(curthread, flags, &auio);
1952 	free_mntarg(ma);
1953 	return (error);
1954 }
1955 
1956 /*
1957  * A printflike function to mount a filesystem.
1958  */
1959 int
1960 kernel_vmount(int flags, ...)
1961 {
1962 	struct mntarg *ma = NULL;
1963 	va_list ap;
1964 	const char *cp;
1965 	const void *vp;
1966 	int error;
1967 
1968 	va_start(ap, flags);
1969 	for (;;) {
1970 		cp = va_arg(ap, const char *);
1971 		if (cp == NULL)
1972 			break;
1973 		vp = va_arg(ap, const void *);
1974 		ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0));
1975 	}
1976 	va_end(ap);
1977 
1978 	error = kernel_mount(ma, flags);
1979 	return (error);
1980 }
1981 
1982 void
1983 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp)
1984 {
1985 
1986 	bcopy(oexp, exp, sizeof(*oexp));
1987 	exp->ex_numsecflavors = 0;
1988 }
1989