xref: /freebsd/sys/kern/vfs_mount.c (revision f56f82e0)
1 /*-
2  * Copyright (c) 1999-2004 Poul-Henning Kamp
3  * Copyright (c) 1999 Michael Smith
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include <sys/param.h>
41 #include <sys/conf.h>
42 #include <sys/fcntl.h>
43 #include <sys/jail.h>
44 #include <sys/kernel.h>
45 #include <sys/libkern.h>
46 #include <sys/malloc.h>
47 #include <sys/mount.h>
48 #include <sys/mutex.h>
49 #include <sys/namei.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/filedesc.h>
53 #include <sys/reboot.h>
54 #include <sys/sbuf.h>
55 #include <sys/syscallsubr.h>
56 #include <sys/sysproto.h>
57 #include <sys/sx.h>
58 #include <sys/sysctl.h>
59 #include <sys/sysent.h>
60 #include <sys/systm.h>
61 #include <sys/vnode.h>
62 #include <vm/uma.h>
63 
64 #include <geom/geom.h>
65 
66 #include <machine/stdarg.h>
67 
68 #include <security/audit/audit.h>
69 #include <security/mac/mac_framework.h>
70 
71 #define	VFS_MOUNTARG_SIZE_MAX	(1024 * 64)
72 
73 static int	vfs_domount(struct thread *td, const char *fstype, char *fspath,
74 		    uint64_t fsflags, struct vfsoptlist **optlist);
75 static void	free_mntarg(struct mntarg *ma);
76 
77 static int	usermount = 0;
78 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0,
79     "Unprivileged users may mount and unmount file systems");
80 
81 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
82 MALLOC_DEFINE(M_STATFS, "statfs", "statfs structure");
83 static uma_zone_t mount_zone;
84 
85 /* List of mounted filesystems. */
86 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
87 
88 /* For any iteration/modification of mountlist */
89 struct mtx mountlist_mtx;
90 MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF);
91 
92 /*
93  * Global opts, taken by all filesystems
94  */
95 static const char *global_opts[] = {
96 	"errmsg",
97 	"fstype",
98 	"fspath",
99 	"ro",
100 	"rw",
101 	"nosuid",
102 	"noexec",
103 	NULL
104 };
105 
106 static int
107 mount_init(void *mem, int size, int flags)
108 {
109 	struct mount *mp;
110 
111 	mp = (struct mount *)mem;
112 	mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF);
113 	mtx_init(&mp->mnt_listmtx, "struct mount vlist mtx", NULL, MTX_DEF);
114 	lockinit(&mp->mnt_explock, PVFS, "explock", 0, 0);
115 	return (0);
116 }
117 
118 static void
119 mount_fini(void *mem, int size)
120 {
121 	struct mount *mp;
122 
123 	mp = (struct mount *)mem;
124 	lockdestroy(&mp->mnt_explock);
125 	mtx_destroy(&mp->mnt_listmtx);
126 	mtx_destroy(&mp->mnt_mtx);
127 }
128 
129 static void
130 vfs_mount_init(void *dummy __unused)
131 {
132 
133 	mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount), NULL,
134 	    NULL, mount_init, mount_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
135 }
136 SYSINIT(vfs_mount, SI_SUB_VFS, SI_ORDER_ANY, vfs_mount_init, NULL);
137 
138 /*
139  * ---------------------------------------------------------------------
140  * Functions for building and sanitizing the mount options
141  */
142 
143 /* Remove one mount option. */
144 static void
145 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
146 {
147 
148 	TAILQ_REMOVE(opts, opt, link);
149 	free(opt->name, M_MOUNT);
150 	if (opt->value != NULL)
151 		free(opt->value, M_MOUNT);
152 	free(opt, M_MOUNT);
153 }
154 
155 /* Release all resources related to the mount options. */
156 void
157 vfs_freeopts(struct vfsoptlist *opts)
158 {
159 	struct vfsopt *opt;
160 
161 	while (!TAILQ_EMPTY(opts)) {
162 		opt = TAILQ_FIRST(opts);
163 		vfs_freeopt(opts, opt);
164 	}
165 	free(opts, M_MOUNT);
166 }
167 
168 void
169 vfs_deleteopt(struct vfsoptlist *opts, const char *name)
170 {
171 	struct vfsopt *opt, *temp;
172 
173 	if (opts == NULL)
174 		return;
175 	TAILQ_FOREACH_SAFE(opt, opts, link, temp)  {
176 		if (strcmp(opt->name, name) == 0)
177 			vfs_freeopt(opts, opt);
178 	}
179 }
180 
181 static int
182 vfs_isopt_ro(const char *opt)
183 {
184 
185 	if (strcmp(opt, "ro") == 0 || strcmp(opt, "rdonly") == 0 ||
186 	    strcmp(opt, "norw") == 0)
187 		return (1);
188 	return (0);
189 }
190 
191 static int
192 vfs_isopt_rw(const char *opt)
193 {
194 
195 	if (strcmp(opt, "rw") == 0 || strcmp(opt, "noro") == 0)
196 		return (1);
197 	return (0);
198 }
199 
200 /*
201  * Check if options are equal (with or without the "no" prefix).
202  */
203 static int
204 vfs_equalopts(const char *opt1, const char *opt2)
205 {
206 	char *p;
207 
208 	/* "opt" vs. "opt" or "noopt" vs. "noopt" */
209 	if (strcmp(opt1, opt2) == 0)
210 		return (1);
211 	/* "noopt" vs. "opt" */
212 	if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
213 		return (1);
214 	/* "opt" vs. "noopt" */
215 	if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
216 		return (1);
217 	while ((p = strchr(opt1, '.')) != NULL &&
218 	    !strncmp(opt1, opt2, ++p - opt1)) {
219 		opt2 += p - opt1;
220 		opt1 = p;
221 		/* "foo.noopt" vs. "foo.opt" */
222 		if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
223 			return (1);
224 		/* "foo.opt" vs. "foo.noopt" */
225 		if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
226 			return (1);
227 	}
228 	/* "ro" / "rdonly" / "norw" / "rw" / "noro" */
229 	if ((vfs_isopt_ro(opt1) || vfs_isopt_rw(opt1)) &&
230 	    (vfs_isopt_ro(opt2) || vfs_isopt_rw(opt2)))
231 		return (1);
232 	return (0);
233 }
234 
235 /*
236  * If a mount option is specified several times,
237  * (with or without the "no" prefix) only keep
238  * the last occurrence of it.
239  */
240 static void
241 vfs_sanitizeopts(struct vfsoptlist *opts)
242 {
243 	struct vfsopt *opt, *opt2, *tmp;
244 
245 	TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
246 		opt2 = TAILQ_PREV(opt, vfsoptlist, link);
247 		while (opt2 != NULL) {
248 			if (vfs_equalopts(opt->name, opt2->name)) {
249 				tmp = TAILQ_PREV(opt2, vfsoptlist, link);
250 				vfs_freeopt(opts, opt2);
251 				opt2 = tmp;
252 			} else {
253 				opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
254 			}
255 		}
256 	}
257 }
258 
259 /*
260  * Build a linked list of mount options from a struct uio.
261  */
262 int
263 vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
264 {
265 	struct vfsoptlist *opts;
266 	struct vfsopt *opt;
267 	size_t memused, namelen, optlen;
268 	unsigned int i, iovcnt;
269 	int error;
270 
271 	opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
272 	TAILQ_INIT(opts);
273 	memused = 0;
274 	iovcnt = auio->uio_iovcnt;
275 	for (i = 0; i < iovcnt; i += 2) {
276 		namelen = auio->uio_iov[i].iov_len;
277 		optlen = auio->uio_iov[i + 1].iov_len;
278 		memused += sizeof(struct vfsopt) + optlen + namelen;
279 		/*
280 		 * Avoid consuming too much memory, and attempts to overflow
281 		 * memused.
282 		 */
283 		if (memused > VFS_MOUNTARG_SIZE_MAX ||
284 		    optlen > VFS_MOUNTARG_SIZE_MAX ||
285 		    namelen > VFS_MOUNTARG_SIZE_MAX) {
286 			error = EINVAL;
287 			goto bad;
288 		}
289 
290 		opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
291 		opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
292 		opt->value = NULL;
293 		opt->len = 0;
294 		opt->pos = i / 2;
295 		opt->seen = 0;
296 
297 		/*
298 		 * Do this early, so jumps to "bad" will free the current
299 		 * option.
300 		 */
301 		TAILQ_INSERT_TAIL(opts, opt, link);
302 
303 		if (auio->uio_segflg == UIO_SYSSPACE) {
304 			bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
305 		} else {
306 			error = copyin(auio->uio_iov[i].iov_base, opt->name,
307 			    namelen);
308 			if (error)
309 				goto bad;
310 		}
311 		/* Ensure names are null-terminated strings. */
312 		if (namelen == 0 || opt->name[namelen - 1] != '\0') {
313 			error = EINVAL;
314 			goto bad;
315 		}
316 		if (optlen != 0) {
317 			opt->len = optlen;
318 			opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
319 			if (auio->uio_segflg == UIO_SYSSPACE) {
320 				bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
321 				    optlen);
322 			} else {
323 				error = copyin(auio->uio_iov[i + 1].iov_base,
324 				    opt->value, optlen);
325 				if (error)
326 					goto bad;
327 			}
328 		}
329 	}
330 	vfs_sanitizeopts(opts);
331 	*options = opts;
332 	return (0);
333 bad:
334 	vfs_freeopts(opts);
335 	return (error);
336 }
337 
338 /*
339  * Merge the old mount options with the new ones passed
340  * in the MNT_UPDATE case.
341  *
342  * XXX: This function will keep a "nofoo" option in the new
343  * options.  E.g, if the option's canonical name is "foo",
344  * "nofoo" ends up in the mount point's active options.
345  */
346 static void
347 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *oldopts)
348 {
349 	struct vfsopt *opt, *new;
350 
351 	TAILQ_FOREACH(opt, oldopts, link) {
352 		new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
353 		new->name = strdup(opt->name, M_MOUNT);
354 		if (opt->len != 0) {
355 			new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
356 			bcopy(opt->value, new->value, opt->len);
357 		} else
358 			new->value = NULL;
359 		new->len = opt->len;
360 		new->seen = opt->seen;
361 		TAILQ_INSERT_HEAD(toopts, new, link);
362 	}
363 	vfs_sanitizeopts(toopts);
364 }
365 
366 /*
367  * Mount a filesystem.
368  */
369 #ifndef _SYS_SYSPROTO_H_
370 struct nmount_args {
371 	struct iovec *iovp;
372 	unsigned int iovcnt;
373 	int flags;
374 };
375 #endif
376 int
377 sys_nmount(struct thread *td, struct nmount_args *uap)
378 {
379 	struct uio *auio;
380 	int error;
381 	u_int iovcnt;
382 	uint64_t flags;
383 
384 	/*
385 	 * Mount flags are now 64-bits. On 32-bit archtectures only
386 	 * 32-bits are passed in, but from here on everything handles
387 	 * 64-bit flags correctly.
388 	 */
389 	flags = uap->flags;
390 
391 	AUDIT_ARG_FFLAGS(flags);
392 	CTR4(KTR_VFS, "%s: iovp %p with iovcnt %d and flags %d", __func__,
393 	    uap->iovp, uap->iovcnt, flags);
394 
395 	/*
396 	 * Filter out MNT_ROOTFS.  We do not want clients of nmount() in
397 	 * userspace to set this flag, but we must filter it out if we want
398 	 * MNT_UPDATE on the root file system to work.
399 	 * MNT_ROOTFS should only be set by the kernel when mounting its
400 	 * root file system.
401 	 */
402 	flags &= ~MNT_ROOTFS;
403 
404 	iovcnt = uap->iovcnt;
405 	/*
406 	 * Check that we have an even number of iovec's
407 	 * and that we have at least two options.
408 	 */
409 	if ((iovcnt & 1) || (iovcnt < 4)) {
410 		CTR2(KTR_VFS, "%s: failed for invalid iovcnt %d", __func__,
411 		    uap->iovcnt);
412 		return (EINVAL);
413 	}
414 
415 	error = copyinuio(uap->iovp, iovcnt, &auio);
416 	if (error) {
417 		CTR2(KTR_VFS, "%s: failed for invalid uio op with %d errno",
418 		    __func__, error);
419 		return (error);
420 	}
421 	error = vfs_donmount(td, flags, auio);
422 
423 	free(auio, M_IOV);
424 	return (error);
425 }
426 
427 /*
428  * ---------------------------------------------------------------------
429  * Various utility functions
430  */
431 
432 void
433 vfs_ref(struct mount *mp)
434 {
435 
436 	CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
437 	MNT_ILOCK(mp);
438 	MNT_REF(mp);
439 	MNT_IUNLOCK(mp);
440 }
441 
442 void
443 vfs_rel(struct mount *mp)
444 {
445 
446 	CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
447 	MNT_ILOCK(mp);
448 	MNT_REL(mp);
449 	MNT_IUNLOCK(mp);
450 }
451 
452 /*
453  * Allocate and initialize the mount point struct.
454  */
455 struct mount *
456 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, const char *fspath,
457     struct ucred *cred)
458 {
459 	struct mount *mp;
460 
461 	mp = uma_zalloc(mount_zone, M_WAITOK);
462 	bzero(&mp->mnt_startzero,
463 	    __rangeof(struct mount, mnt_startzero, mnt_endzero));
464 	TAILQ_INIT(&mp->mnt_nvnodelist);
465 	mp->mnt_nvnodelistsize = 0;
466 	TAILQ_INIT(&mp->mnt_activevnodelist);
467 	mp->mnt_activevnodelistsize = 0;
468 	TAILQ_INIT(&mp->mnt_tmpfreevnodelist);
469 	mp->mnt_tmpfreevnodelistsize = 0;
470 	mp->mnt_ref = 0;
471 	(void) vfs_busy(mp, MBF_NOWAIT);
472 	atomic_add_acq_int(&vfsp->vfc_refcount, 1);
473 	mp->mnt_op = vfsp->vfc_vfsops;
474 	mp->mnt_vfc = vfsp;
475 	mp->mnt_stat.f_type = vfsp->vfc_typenum;
476 	mp->mnt_gen++;
477 	strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
478 	mp->mnt_vnodecovered = vp;
479 	mp->mnt_cred = crdup(cred);
480 	mp->mnt_stat.f_owner = cred->cr_uid;
481 	strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
482 	mp->mnt_iosize_max = DFLTPHYS;
483 #ifdef MAC
484 	mac_mount_init(mp);
485 	mac_mount_create(cred, mp);
486 #endif
487 	arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0);
488 	TAILQ_INIT(&mp->mnt_uppers);
489 	return (mp);
490 }
491 
492 /*
493  * Destroy the mount struct previously allocated by vfs_mount_alloc().
494  */
495 void
496 vfs_mount_destroy(struct mount *mp)
497 {
498 
499 	MNT_ILOCK(mp);
500 	mp->mnt_kern_flag |= MNTK_REFEXPIRE;
501 	if (mp->mnt_kern_flag & MNTK_MWAIT) {
502 		mp->mnt_kern_flag &= ~MNTK_MWAIT;
503 		wakeup(mp);
504 	}
505 	while (mp->mnt_ref)
506 		msleep(mp, MNT_MTX(mp), PVFS, "mntref", 0);
507 	KASSERT(mp->mnt_ref == 0,
508 	    ("%s: invalid refcount in the drain path @ %s:%d", __func__,
509 	    __FILE__, __LINE__));
510 	if (mp->mnt_writeopcount != 0)
511 		panic("vfs_mount_destroy: nonzero writeopcount");
512 	if (mp->mnt_secondary_writes != 0)
513 		panic("vfs_mount_destroy: nonzero secondary_writes");
514 	atomic_subtract_rel_int(&mp->mnt_vfc->vfc_refcount, 1);
515 	if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
516 		struct vnode *vp;
517 
518 		TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
519 			vn_printf(vp, "dangling vnode ");
520 		panic("unmount: dangling vnode");
521 	}
522 	KASSERT(TAILQ_EMPTY(&mp->mnt_uppers), ("mnt_uppers"));
523 	if (mp->mnt_nvnodelistsize != 0)
524 		panic("vfs_mount_destroy: nonzero nvnodelistsize");
525 	if (mp->mnt_activevnodelistsize != 0)
526 		panic("vfs_mount_destroy: nonzero activevnodelistsize");
527 	if (mp->mnt_tmpfreevnodelistsize != 0)
528 		panic("vfs_mount_destroy: nonzero tmpfreevnodelistsize");
529 	if (mp->mnt_lockref != 0)
530 		panic("vfs_mount_destroy: nonzero lock refcount");
531 	MNT_IUNLOCK(mp);
532 #ifdef MAC
533 	mac_mount_destroy(mp);
534 #endif
535 	if (mp->mnt_opt != NULL)
536 		vfs_freeopts(mp->mnt_opt);
537 	crfree(mp->mnt_cred);
538 	uma_zfree(mount_zone, mp);
539 }
540 
541 int
542 vfs_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions)
543 {
544 	struct vfsoptlist *optlist;
545 	struct vfsopt *opt, *tmp_opt;
546 	char *fstype, *fspath, *errmsg;
547 	int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
548 
549 	errmsg = fspath = NULL;
550 	errmsg_len = fspathlen = 0;
551 	errmsg_pos = -1;
552 
553 	error = vfs_buildopts(fsoptions, &optlist);
554 	if (error)
555 		return (error);
556 
557 	if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
558 		errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
559 
560 	/*
561 	 * We need these two options before the others,
562 	 * and they are mandatory for any filesystem.
563 	 * Ensure they are NUL terminated as well.
564 	 */
565 	fstypelen = 0;
566 	error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
567 	if (error || fstype[fstypelen - 1] != '\0') {
568 		error = EINVAL;
569 		if (errmsg != NULL)
570 			strncpy(errmsg, "Invalid fstype", errmsg_len);
571 		goto bail;
572 	}
573 	fspathlen = 0;
574 	error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
575 	if (error || fspath[fspathlen - 1] != '\0') {
576 		error = EINVAL;
577 		if (errmsg != NULL)
578 			strncpy(errmsg, "Invalid fspath", errmsg_len);
579 		goto bail;
580 	}
581 
582 	/*
583 	 * We need to see if we have the "update" option
584 	 * before we call vfs_domount(), since vfs_domount() has special
585 	 * logic based on MNT_UPDATE.  This is very important
586 	 * when we want to update the root filesystem.
587 	 */
588 	TAILQ_FOREACH_SAFE(opt, optlist, link, tmp_opt) {
589 		if (strcmp(opt->name, "update") == 0) {
590 			fsflags |= MNT_UPDATE;
591 			vfs_freeopt(optlist, opt);
592 		}
593 		else if (strcmp(opt->name, "async") == 0)
594 			fsflags |= MNT_ASYNC;
595 		else if (strcmp(opt->name, "force") == 0) {
596 			fsflags |= MNT_FORCE;
597 			vfs_freeopt(optlist, opt);
598 		}
599 		else if (strcmp(opt->name, "reload") == 0) {
600 			fsflags |= MNT_RELOAD;
601 			vfs_freeopt(optlist, opt);
602 		}
603 		else if (strcmp(opt->name, "multilabel") == 0)
604 			fsflags |= MNT_MULTILABEL;
605 		else if (strcmp(opt->name, "noasync") == 0)
606 			fsflags &= ~MNT_ASYNC;
607 		else if (strcmp(opt->name, "noatime") == 0)
608 			fsflags |= MNT_NOATIME;
609 		else if (strcmp(opt->name, "atime") == 0) {
610 			free(opt->name, M_MOUNT);
611 			opt->name = strdup("nonoatime", M_MOUNT);
612 		}
613 		else if (strcmp(opt->name, "noclusterr") == 0)
614 			fsflags |= MNT_NOCLUSTERR;
615 		else if (strcmp(opt->name, "clusterr") == 0) {
616 			free(opt->name, M_MOUNT);
617 			opt->name = strdup("nonoclusterr", M_MOUNT);
618 		}
619 		else if (strcmp(opt->name, "noclusterw") == 0)
620 			fsflags |= MNT_NOCLUSTERW;
621 		else if (strcmp(opt->name, "clusterw") == 0) {
622 			free(opt->name, M_MOUNT);
623 			opt->name = strdup("nonoclusterw", M_MOUNT);
624 		}
625 		else if (strcmp(opt->name, "noexec") == 0)
626 			fsflags |= MNT_NOEXEC;
627 		else if (strcmp(opt->name, "exec") == 0) {
628 			free(opt->name, M_MOUNT);
629 			opt->name = strdup("nonoexec", M_MOUNT);
630 		}
631 		else if (strcmp(opt->name, "nosuid") == 0)
632 			fsflags |= MNT_NOSUID;
633 		else if (strcmp(opt->name, "suid") == 0) {
634 			free(opt->name, M_MOUNT);
635 			opt->name = strdup("nonosuid", M_MOUNT);
636 		}
637 		else if (strcmp(opt->name, "nosymfollow") == 0)
638 			fsflags |= MNT_NOSYMFOLLOW;
639 		else if (strcmp(opt->name, "symfollow") == 0) {
640 			free(opt->name, M_MOUNT);
641 			opt->name = strdup("nonosymfollow", M_MOUNT);
642 		}
643 		else if (strcmp(opt->name, "noro") == 0)
644 			fsflags &= ~MNT_RDONLY;
645 		else if (strcmp(opt->name, "rw") == 0)
646 			fsflags &= ~MNT_RDONLY;
647 		else if (strcmp(opt->name, "ro") == 0)
648 			fsflags |= MNT_RDONLY;
649 		else if (strcmp(opt->name, "rdonly") == 0) {
650 			free(opt->name, M_MOUNT);
651 			opt->name = strdup("ro", M_MOUNT);
652 			fsflags |= MNT_RDONLY;
653 		}
654 		else if (strcmp(opt->name, "suiddir") == 0)
655 			fsflags |= MNT_SUIDDIR;
656 		else if (strcmp(opt->name, "sync") == 0)
657 			fsflags |= MNT_SYNCHRONOUS;
658 		else if (strcmp(opt->name, "union") == 0)
659 			fsflags |= MNT_UNION;
660 		else if (strcmp(opt->name, "automounted") == 0) {
661 			fsflags |= MNT_AUTOMOUNTED;
662 			vfs_freeopt(optlist, opt);
663 		}
664 	}
665 
666 	/*
667 	 * Be ultra-paranoid about making sure the type and fspath
668 	 * variables will fit in our mp buffers, including the
669 	 * terminating NUL.
670 	 */
671 	if (fstypelen > MFSNAMELEN || fspathlen > MNAMELEN) {
672 		error = ENAMETOOLONG;
673 		goto bail;
674 	}
675 
676 	error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
677 bail:
678 	/* copyout the errmsg */
679 	if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
680 	    && errmsg_len > 0 && errmsg != NULL) {
681 		if (fsoptions->uio_segflg == UIO_SYSSPACE) {
682 			bcopy(errmsg,
683 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
684 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
685 		} else {
686 			copyout(errmsg,
687 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
688 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
689 		}
690 	}
691 
692 	if (optlist != NULL)
693 		vfs_freeopts(optlist);
694 	return (error);
695 }
696 
697 /*
698  * Old mount API.
699  */
700 #ifndef _SYS_SYSPROTO_H_
701 struct mount_args {
702 	char	*type;
703 	char	*path;
704 	int	flags;
705 	caddr_t	data;
706 };
707 #endif
708 /* ARGSUSED */
709 int
710 sys_mount(struct thread *td, struct mount_args *uap)
711 {
712 	char *fstype;
713 	struct vfsconf *vfsp = NULL;
714 	struct mntarg *ma = NULL;
715 	uint64_t flags;
716 	int error;
717 
718 	/*
719 	 * Mount flags are now 64-bits. On 32-bit architectures only
720 	 * 32-bits are passed in, but from here on everything handles
721 	 * 64-bit flags correctly.
722 	 */
723 	flags = uap->flags;
724 
725 	AUDIT_ARG_FFLAGS(flags);
726 
727 	/*
728 	 * Filter out MNT_ROOTFS.  We do not want clients of mount() in
729 	 * userspace to set this flag, but we must filter it out if we want
730 	 * MNT_UPDATE on the root file system to work.
731 	 * MNT_ROOTFS should only be set by the kernel when mounting its
732 	 * root file system.
733 	 */
734 	flags &= ~MNT_ROOTFS;
735 
736 	fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
737 	error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
738 	if (error) {
739 		free(fstype, M_TEMP);
740 		return (error);
741 	}
742 
743 	AUDIT_ARG_TEXT(fstype);
744 	vfsp = vfs_byname_kld(fstype, td, &error);
745 	free(fstype, M_TEMP);
746 	if (vfsp == NULL)
747 		return (ENOENT);
748 	if (vfsp->vfc_vfsops->vfs_cmount == NULL)
749 		return (EOPNOTSUPP);
750 
751 	ma = mount_argsu(ma, "fstype", uap->type, MFSNAMELEN);
752 	ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
753 	ma = mount_argb(ma, flags & MNT_RDONLY, "noro");
754 	ma = mount_argb(ma, !(flags & MNT_NOSUID), "nosuid");
755 	ma = mount_argb(ma, !(flags & MNT_NOEXEC), "noexec");
756 
757 	error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags);
758 	return (error);
759 }
760 
761 /*
762  * vfs_domount_first(): first file system mount (not update)
763  */
764 static int
765 vfs_domount_first(
766 	struct thread *td,		/* Calling thread. */
767 	struct vfsconf *vfsp,		/* File system type. */
768 	char *fspath,			/* Mount path. */
769 	struct vnode *vp,		/* Vnode to be covered. */
770 	uint64_t fsflags,		/* Flags common to all filesystems. */
771 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
772 	)
773 {
774 	struct vattr va;
775 	struct mount *mp;
776 	struct vnode *newdp;
777 	int error;
778 
779 	ASSERT_VOP_ELOCKED(vp, __func__);
780 	KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
781 
782 	/*
783 	 * If the user is not root, ensure that they own the directory
784 	 * onto which we are attempting to mount.
785 	 */
786 	error = VOP_GETATTR(vp, &va, td->td_ucred);
787 	if (error == 0 && va.va_uid != td->td_ucred->cr_uid)
788 		error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN, 0);
789 	if (error == 0)
790 		error = vinvalbuf(vp, V_SAVE, 0, 0);
791 	if (error == 0 && vp->v_type != VDIR)
792 		error = ENOTDIR;
793 	if (error == 0) {
794 		VI_LOCK(vp);
795 		if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
796 			vp->v_iflag |= VI_MOUNT;
797 		else
798 			error = EBUSY;
799 		VI_UNLOCK(vp);
800 	}
801 	if (error != 0) {
802 		vput(vp);
803 		return (error);
804 	}
805 	VOP_UNLOCK(vp, 0);
806 
807 	/* Allocate and initialize the filesystem. */
808 	mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred);
809 	/* XXXMAC: pass to vfs_mount_alloc? */
810 	mp->mnt_optnew = *optlist;
811 	/* Set the mount level flags. */
812 	mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY));
813 
814 	/*
815 	 * Mount the filesystem.
816 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
817 	 * get.  No freeing of cn_pnbuf.
818 	 */
819 	error = VFS_MOUNT(mp);
820 	if (error != 0) {
821 		vfs_unbusy(mp);
822 		vfs_mount_destroy(mp);
823 		VI_LOCK(vp);
824 		vp->v_iflag &= ~VI_MOUNT;
825 		VI_UNLOCK(vp);
826 		vrele(vp);
827 		return (error);
828 	}
829 
830 	if (mp->mnt_opt != NULL)
831 		vfs_freeopts(mp->mnt_opt);
832 	mp->mnt_opt = mp->mnt_optnew;
833 	*optlist = NULL;
834 	(void)VFS_STATFS(mp, &mp->mnt_stat);
835 
836 	/*
837 	 * Prevent external consumers of mount options from reading mnt_optnew.
838 	 */
839 	mp->mnt_optnew = NULL;
840 
841 	MNT_ILOCK(mp);
842 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
843 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
844 		mp->mnt_kern_flag |= MNTK_ASYNC;
845 	else
846 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
847 	MNT_IUNLOCK(mp);
848 
849 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
850 	cache_purge(vp);
851 	VI_LOCK(vp);
852 	vp->v_iflag &= ~VI_MOUNT;
853 	VI_UNLOCK(vp);
854 	vp->v_mountedhere = mp;
855 	/* Place the new filesystem at the end of the mount list. */
856 	mtx_lock(&mountlist_mtx);
857 	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
858 	mtx_unlock(&mountlist_mtx);
859 	vfs_event_signal(NULL, VQ_MOUNT, 0);
860 	if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp))
861 		panic("mount: lost mount");
862 	VOP_UNLOCK(vp, 0);
863 	EVENTHANDLER_INVOKE(vfs_mounted, mp, newdp, td);
864 	VOP_UNLOCK(newdp, 0);
865 	mountcheckdirs(vp, newdp);
866 	vrele(newdp);
867 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
868 		vfs_allocate_syncvnode(mp);
869 	vfs_unbusy(mp);
870 	return (0);
871 }
872 
873 /*
874  * vfs_domount_update(): update of mounted file system
875  */
876 static int
877 vfs_domount_update(
878 	struct thread *td,		/* Calling thread. */
879 	struct vnode *vp,		/* Mount point vnode. */
880 	uint64_t fsflags,		/* Flags common to all filesystems. */
881 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
882 	)
883 {
884 	struct export_args export;
885 	void *bufp;
886 	struct mount *mp;
887 	int error, export_error, len;
888 	uint64_t flag;
889 
890 	ASSERT_VOP_ELOCKED(vp, __func__);
891 	KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
892 	mp = vp->v_mount;
893 
894 	if ((vp->v_vflag & VV_ROOT) == 0) {
895 		if (vfs_copyopt(*optlist, "export", &export, sizeof(export))
896 		    == 0)
897 			error = EXDEV;
898 		else
899 			error = EINVAL;
900 		vput(vp);
901 		return (error);
902 	}
903 
904 	/*
905 	 * We only allow the filesystem to be reloaded if it
906 	 * is currently mounted read-only.
907 	 */
908 	flag = mp->mnt_flag;
909 	if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
910 		vput(vp);
911 		return (EOPNOTSUPP);	/* Needs translation */
912 	}
913 	/*
914 	 * Only privileged root, or (if MNT_USER is set) the user that
915 	 * did the original mount is permitted to update it.
916 	 */
917 	error = vfs_suser(mp, td);
918 	if (error != 0) {
919 		vput(vp);
920 		return (error);
921 	}
922 	if (vfs_busy(mp, MBF_NOWAIT)) {
923 		vput(vp);
924 		return (EBUSY);
925 	}
926 	VI_LOCK(vp);
927 	if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
928 		VI_UNLOCK(vp);
929 		vfs_unbusy(mp);
930 		vput(vp);
931 		return (EBUSY);
932 	}
933 	vp->v_iflag |= VI_MOUNT;
934 	VI_UNLOCK(vp);
935 	VOP_UNLOCK(vp, 0);
936 
937 	MNT_ILOCK(mp);
938 	if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0) {
939 		MNT_IUNLOCK(mp);
940 		error = EBUSY;
941 		goto end;
942 	}
943 	mp->mnt_flag &= ~MNT_UPDATEMASK;
944 	mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE |
945 	    MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY);
946 	if ((mp->mnt_flag & MNT_ASYNC) == 0)
947 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
948 	MNT_IUNLOCK(mp);
949 	mp->mnt_optnew = *optlist;
950 	vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
951 
952 	/*
953 	 * Mount the filesystem.
954 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
955 	 * get.  No freeing of cn_pnbuf.
956 	 */
957 	error = VFS_MOUNT(mp);
958 
959 	export_error = 0;
960 	/* Process the export option. */
961 	if (error == 0 && vfs_getopt(mp->mnt_optnew, "export", &bufp,
962 	    &len) == 0) {
963 		/* Assume that there is only 1 ABI for each length. */
964 		switch (len) {
965 		case (sizeof(struct oexport_args)):
966 			bzero(&export, sizeof(export));
967 			/* FALLTHROUGH */
968 		case (sizeof(export)):
969 			bcopy(bufp, &export, len);
970 			export_error = vfs_export(mp, &export);
971 			break;
972 		default:
973 			export_error = EINVAL;
974 			break;
975 		}
976 	}
977 
978 	MNT_ILOCK(mp);
979 	if (error == 0) {
980 		mp->mnt_flag &=	~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
981 		    MNT_SNAPSHOT);
982 	} else {
983 		/*
984 		 * If we fail, restore old mount flags. MNT_QUOTA is special,
985 		 * because it is not part of MNT_UPDATEMASK, but it could have
986 		 * changed in the meantime if quotactl(2) was called.
987 		 * All in all we want current value of MNT_QUOTA, not the old
988 		 * one.
989 		 */
990 		mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
991 	}
992 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
993 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
994 		mp->mnt_kern_flag |= MNTK_ASYNC;
995 	else
996 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
997 	MNT_IUNLOCK(mp);
998 
999 	if (error != 0)
1000 		goto end;
1001 
1002 	if (mp->mnt_opt != NULL)
1003 		vfs_freeopts(mp->mnt_opt);
1004 	mp->mnt_opt = mp->mnt_optnew;
1005 	*optlist = NULL;
1006 	(void)VFS_STATFS(mp, &mp->mnt_stat);
1007 	/*
1008 	 * Prevent external consumers of mount options from reading
1009 	 * mnt_optnew.
1010 	 */
1011 	mp->mnt_optnew = NULL;
1012 
1013 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
1014 		vfs_allocate_syncvnode(mp);
1015 	else
1016 		vfs_deallocate_syncvnode(mp);
1017 end:
1018 	vfs_unbusy(mp);
1019 	VI_LOCK(vp);
1020 	vp->v_iflag &= ~VI_MOUNT;
1021 	VI_UNLOCK(vp);
1022 	vrele(vp);
1023 	return (error != 0 ? error : export_error);
1024 }
1025 
1026 /*
1027  * vfs_domount(): actually attempt a filesystem mount.
1028  */
1029 static int
1030 vfs_domount(
1031 	struct thread *td,		/* Calling thread. */
1032 	const char *fstype,		/* Filesystem type. */
1033 	char *fspath,			/* Mount path. */
1034 	uint64_t fsflags,		/* Flags common to all filesystems. */
1035 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
1036 	)
1037 {
1038 	struct vfsconf *vfsp;
1039 	struct nameidata nd;
1040 	struct vnode *vp;
1041 	char *pathbuf;
1042 	int error;
1043 
1044 	/*
1045 	 * Be ultra-paranoid about making sure the type and fspath
1046 	 * variables will fit in our mp buffers, including the
1047 	 * terminating NUL.
1048 	 */
1049 	if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
1050 		return (ENAMETOOLONG);
1051 
1052 	if (jailed(td->td_ucred) || usermount == 0) {
1053 		if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
1054 			return (error);
1055 	}
1056 
1057 	/*
1058 	 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
1059 	 */
1060 	if (fsflags & MNT_EXPORTED) {
1061 		error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
1062 		if (error)
1063 			return (error);
1064 	}
1065 	if (fsflags & MNT_SUIDDIR) {
1066 		error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
1067 		if (error)
1068 			return (error);
1069 	}
1070 	/*
1071 	 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
1072 	 */
1073 	if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
1074 		if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
1075 			fsflags |= MNT_NOSUID | MNT_USER;
1076 	}
1077 
1078 	/* Load KLDs before we lock the covered vnode to avoid reversals. */
1079 	vfsp = NULL;
1080 	if ((fsflags & MNT_UPDATE) == 0) {
1081 		/* Don't try to load KLDs if we're mounting the root. */
1082 		if (fsflags & MNT_ROOTFS)
1083 			vfsp = vfs_byname(fstype);
1084 		else
1085 			vfsp = vfs_byname_kld(fstype, td, &error);
1086 		if (vfsp == NULL)
1087 			return (ENODEV);
1088 		if (jailed(td->td_ucred) && !(vfsp->vfc_flags & VFCF_JAIL))
1089 			return (EPERM);
1090 	}
1091 
1092 	/*
1093 	 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
1094 	 */
1095 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1096 	    UIO_SYSSPACE, fspath, td);
1097 	error = namei(&nd);
1098 	if (error != 0)
1099 		return (error);
1100 	NDFREE(&nd, NDF_ONLY_PNBUF);
1101 	vp = nd.ni_vp;
1102 	if ((fsflags & MNT_UPDATE) == 0) {
1103 		pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1104 		strcpy(pathbuf, fspath);
1105 		error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN);
1106 		/* debug.disablefullpath == 1 results in ENODEV */
1107 		if (error == 0 || error == ENODEV) {
1108 			error = vfs_domount_first(td, vfsp, pathbuf, vp,
1109 			    fsflags, optlist);
1110 		}
1111 		free(pathbuf, M_TEMP);
1112 	} else
1113 		error = vfs_domount_update(td, vp, fsflags, optlist);
1114 
1115 	return (error);
1116 }
1117 
1118 /*
1119  * Unmount a filesystem.
1120  *
1121  * Note: unmount takes a path to the vnode mounted on as argument, not
1122  * special file (as before).
1123  */
1124 #ifndef _SYS_SYSPROTO_H_
1125 struct unmount_args {
1126 	char	*path;
1127 	int	flags;
1128 };
1129 #endif
1130 /* ARGSUSED */
1131 int
1132 sys_unmount(struct thread *td, struct unmount_args *uap)
1133 {
1134 	struct nameidata nd;
1135 	struct mount *mp;
1136 	char *pathbuf;
1137 	int error, id0, id1;
1138 
1139 	AUDIT_ARG_VALUE(uap->flags);
1140 	if (jailed(td->td_ucred) || usermount == 0) {
1141 		error = priv_check(td, PRIV_VFS_UNMOUNT);
1142 		if (error)
1143 			return (error);
1144 	}
1145 
1146 	pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1147 	error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1148 	if (error) {
1149 		free(pathbuf, M_TEMP);
1150 		return (error);
1151 	}
1152 	if (uap->flags & MNT_BYFSID) {
1153 		AUDIT_ARG_TEXT(pathbuf);
1154 		/* Decode the filesystem ID. */
1155 		if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
1156 			free(pathbuf, M_TEMP);
1157 			return (EINVAL);
1158 		}
1159 
1160 		mtx_lock(&mountlist_mtx);
1161 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1162 			if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1163 			    mp->mnt_stat.f_fsid.val[1] == id1) {
1164 				vfs_ref(mp);
1165 				break;
1166 			}
1167 		}
1168 		mtx_unlock(&mountlist_mtx);
1169 	} else {
1170 		/*
1171 		 * Try to find global path for path argument.
1172 		 */
1173 		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1174 		    UIO_SYSSPACE, pathbuf, td);
1175 		if (namei(&nd) == 0) {
1176 			NDFREE(&nd, NDF_ONLY_PNBUF);
1177 			error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
1178 			    MNAMELEN);
1179 			if (error == 0 || error == ENODEV)
1180 				vput(nd.ni_vp);
1181 		}
1182 		mtx_lock(&mountlist_mtx);
1183 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1184 			if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) {
1185 				vfs_ref(mp);
1186 				break;
1187 			}
1188 		}
1189 		mtx_unlock(&mountlist_mtx);
1190 	}
1191 	free(pathbuf, M_TEMP);
1192 	if (mp == NULL) {
1193 		/*
1194 		 * Previously we returned ENOENT for a nonexistent path and
1195 		 * EINVAL for a non-mountpoint.  We cannot tell these apart
1196 		 * now, so in the !MNT_BYFSID case return the more likely
1197 		 * EINVAL for compatibility.
1198 		 */
1199 		return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1200 	}
1201 
1202 	/*
1203 	 * Don't allow unmounting the root filesystem.
1204 	 */
1205 	if (mp->mnt_flag & MNT_ROOTFS) {
1206 		vfs_rel(mp);
1207 		return (EINVAL);
1208 	}
1209 	error = dounmount(mp, uap->flags, td);
1210 	return (error);
1211 }
1212 
1213 /*
1214  * Return error if any of the vnodes, ignoring the root vnode
1215  * and the syncer vnode, have non-zero usecount.
1216  *
1217  * This function is purely advisory - it can return false positives
1218  * and negatives.
1219  */
1220 static int
1221 vfs_check_usecounts(struct mount *mp)
1222 {
1223 	struct vnode *vp, *mvp;
1224 
1225 	MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1226 		if ((vp->v_vflag & VV_ROOT) == 0 && vp->v_type != VNON &&
1227 		    vp->v_usecount != 0) {
1228 			VI_UNLOCK(vp);
1229 			MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1230 			return (EBUSY);
1231 		}
1232 		VI_UNLOCK(vp);
1233 	}
1234 
1235 	return (0);
1236 }
1237 
1238 static void
1239 dounmount_cleanup(struct mount *mp, struct vnode *coveredvp, int mntkflags)
1240 {
1241 
1242 	mtx_assert(MNT_MTX(mp), MA_OWNED);
1243 	mp->mnt_kern_flag &= ~mntkflags;
1244 	if ((mp->mnt_kern_flag & MNTK_MWAIT) != 0) {
1245 		mp->mnt_kern_flag &= ~MNTK_MWAIT;
1246 		wakeup(mp);
1247 	}
1248 	MNT_IUNLOCK(mp);
1249 	if (coveredvp != NULL) {
1250 		VOP_UNLOCK(coveredvp, 0);
1251 		vdrop(coveredvp);
1252 	}
1253 	vn_finished_write(mp);
1254 }
1255 
1256 /*
1257  * Do the actual filesystem unmount.
1258  */
1259 int
1260 dounmount(struct mount *mp, int flags, struct thread *td)
1261 {
1262 	struct vnode *coveredvp, *fsrootvp;
1263 	int error;
1264 	uint64_t async_flag;
1265 	int mnt_gen_r;
1266 
1267 	if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
1268 		mnt_gen_r = mp->mnt_gen;
1269 		VI_LOCK(coveredvp);
1270 		vholdl(coveredvp);
1271 		vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
1272 		/*
1273 		 * Check for mp being unmounted while waiting for the
1274 		 * covered vnode lock.
1275 		 */
1276 		if (coveredvp->v_mountedhere != mp ||
1277 		    coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
1278 			VOP_UNLOCK(coveredvp, 0);
1279 			vdrop(coveredvp);
1280 			vfs_rel(mp);
1281 			return (EBUSY);
1282 		}
1283 	}
1284 
1285 	/*
1286 	 * Only privileged root, or (if MNT_USER is set) the user that did the
1287 	 * original mount is permitted to unmount this filesystem.
1288 	 */
1289 	error = vfs_suser(mp, td);
1290 	if (error != 0) {
1291 		if (coveredvp != NULL) {
1292 			VOP_UNLOCK(coveredvp, 0);
1293 			vdrop(coveredvp);
1294 		}
1295 		vfs_rel(mp);
1296 		return (error);
1297 	}
1298 
1299 	vn_start_write(NULL, &mp, V_WAIT | V_MNTREF);
1300 	MNT_ILOCK(mp);
1301 	if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 ||
1302 	    (mp->mnt_flag & MNT_UPDATE) != 0 ||
1303 	    !TAILQ_EMPTY(&mp->mnt_uppers)) {
1304 		dounmount_cleanup(mp, coveredvp, 0);
1305 		return (EBUSY);
1306 	}
1307 	mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ;
1308 	if (flags & MNT_NONBUSY) {
1309 		MNT_IUNLOCK(mp);
1310 		error = vfs_check_usecounts(mp);
1311 		MNT_ILOCK(mp);
1312 		if (error != 0) {
1313 			dounmount_cleanup(mp, coveredvp, MNTK_UNMOUNT |
1314 			    MNTK_NOINSMNTQ);
1315 			return (error);
1316 		}
1317 	}
1318 	/* Allow filesystems to detect that a forced unmount is in progress. */
1319 	if (flags & MNT_FORCE) {
1320 		mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1321 		MNT_IUNLOCK(mp);
1322 		/*
1323 		 * Must be done after setting MNTK_UNMOUNTF and before
1324 		 * waiting for mnt_lockref to become 0.
1325 		 */
1326 		VFS_PURGE(mp);
1327 		MNT_ILOCK(mp);
1328 	}
1329 	error = 0;
1330 	if (mp->mnt_lockref) {
1331 		mp->mnt_kern_flag |= MNTK_DRAINING;
1332 		error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
1333 		    "mount drain", 0);
1334 	}
1335 	MNT_IUNLOCK(mp);
1336 	KASSERT(mp->mnt_lockref == 0,
1337 	    ("%s: invalid lock refcount in the drain path @ %s:%d",
1338 	    __func__, __FILE__, __LINE__));
1339 	KASSERT(error == 0,
1340 	    ("%s: invalid return value for msleep in the drain path @ %s:%d",
1341 	    __func__, __FILE__, __LINE__));
1342 
1343 	if (mp->mnt_flag & MNT_EXPUBLIC)
1344 		vfs_setpublicfs(NULL, NULL, NULL);
1345 
1346 	/*
1347 	 * From now, we can claim that the use reference on the
1348 	 * coveredvp is ours, and the ref can be released only by
1349 	 * successfull unmount by us, or left for later unmount
1350 	 * attempt.  The previously acquired hold reference is no
1351 	 * longer needed to protect the vnode from reuse.
1352 	 */
1353 	if (coveredvp != NULL)
1354 		vdrop(coveredvp);
1355 
1356 	vfs_msync(mp, MNT_WAIT);
1357 	MNT_ILOCK(mp);
1358 	async_flag = mp->mnt_flag & MNT_ASYNC;
1359 	mp->mnt_flag &= ~MNT_ASYNC;
1360 	mp->mnt_kern_flag &= ~MNTK_ASYNC;
1361 	MNT_IUNLOCK(mp);
1362 	cache_purgevfs(mp, false); /* remove cache entries for this file sys */
1363 	vfs_deallocate_syncvnode(mp);
1364 	/*
1365 	 * For forced unmounts, move process cdir/rdir refs on the fs root
1366 	 * vnode to the covered vnode.  For non-forced unmounts we want
1367 	 * such references to cause an EBUSY error.
1368 	 */
1369 	if ((flags & MNT_FORCE) &&
1370 	    VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1371 		if (mp->mnt_vnodecovered != NULL &&
1372 		    (mp->mnt_flag & MNT_IGNORE) == 0)
1373 			mountcheckdirs(fsrootvp, mp->mnt_vnodecovered);
1374 		if (fsrootvp == rootvnode) {
1375 			vrele(rootvnode);
1376 			rootvnode = NULL;
1377 		}
1378 		vput(fsrootvp);
1379 	}
1380 	if ((mp->mnt_flag & MNT_RDONLY) != 0 || (flags & MNT_FORCE) != 0 ||
1381 	    (error = VFS_SYNC(mp, MNT_WAIT)) == 0)
1382 		error = VFS_UNMOUNT(mp, flags);
1383 	vn_finished_write(mp);
1384 	/*
1385 	 * If we failed to flush the dirty blocks for this mount point,
1386 	 * undo all the cdir/rdir and rootvnode changes we made above.
1387 	 * Unless we failed to do so because the device is reporting that
1388 	 * it doesn't exist anymore.
1389 	 */
1390 	if (error && error != ENXIO) {
1391 		if ((flags & MNT_FORCE) &&
1392 		    VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1393 			if (mp->mnt_vnodecovered != NULL &&
1394 			    (mp->mnt_flag & MNT_IGNORE) == 0)
1395 				mountcheckdirs(mp->mnt_vnodecovered, fsrootvp);
1396 			if (rootvnode == NULL) {
1397 				rootvnode = fsrootvp;
1398 				vref(rootvnode);
1399 			}
1400 			vput(fsrootvp);
1401 		}
1402 		MNT_ILOCK(mp);
1403 		mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ;
1404 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
1405 			MNT_IUNLOCK(mp);
1406 			vfs_allocate_syncvnode(mp);
1407 			MNT_ILOCK(mp);
1408 		}
1409 		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1410 		mp->mnt_flag |= async_flag;
1411 		if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1412 		    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1413 			mp->mnt_kern_flag |= MNTK_ASYNC;
1414 		if (mp->mnt_kern_flag & MNTK_MWAIT) {
1415 			mp->mnt_kern_flag &= ~MNTK_MWAIT;
1416 			wakeup(mp);
1417 		}
1418 		MNT_IUNLOCK(mp);
1419 		if (coveredvp)
1420 			VOP_UNLOCK(coveredvp, 0);
1421 		return (error);
1422 	}
1423 	mtx_lock(&mountlist_mtx);
1424 	TAILQ_REMOVE(&mountlist, mp, mnt_list);
1425 	mtx_unlock(&mountlist_mtx);
1426 	EVENTHANDLER_INVOKE(vfs_unmounted, mp, td);
1427 	if (coveredvp != NULL) {
1428 		coveredvp->v_mountedhere = NULL;
1429 		vput(coveredvp);
1430 	}
1431 	vfs_event_signal(NULL, VQ_UNMOUNT, 0);
1432 	if (mp == rootdevmp)
1433 		rootdevmp = NULL;
1434 	vfs_mount_destroy(mp);
1435 	return (0);
1436 }
1437 
1438 /*
1439  * Report errors during filesystem mounting.
1440  */
1441 void
1442 vfs_mount_error(struct mount *mp, const char *fmt, ...)
1443 {
1444 	struct vfsoptlist *moptlist = mp->mnt_optnew;
1445 	va_list ap;
1446 	int error, len;
1447 	char *errmsg;
1448 
1449 	error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
1450 	if (error || errmsg == NULL || len <= 0)
1451 		return;
1452 
1453 	va_start(ap, fmt);
1454 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1455 	va_end(ap);
1456 }
1457 
1458 void
1459 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
1460 {
1461 	va_list ap;
1462 	int error, len;
1463 	char *errmsg;
1464 
1465 	error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
1466 	if (error || errmsg == NULL || len <= 0)
1467 		return;
1468 
1469 	va_start(ap, fmt);
1470 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1471 	va_end(ap);
1472 }
1473 
1474 /*
1475  * ---------------------------------------------------------------------
1476  * Functions for querying mount options/arguments from filesystems.
1477  */
1478 
1479 /*
1480  * Check that no unknown options are given
1481  */
1482 int
1483 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1484 {
1485 	struct vfsopt *opt;
1486 	char errmsg[255];
1487 	const char **t, *p, *q;
1488 	int ret = 0;
1489 
1490 	TAILQ_FOREACH(opt, opts, link) {
1491 		p = opt->name;
1492 		q = NULL;
1493 		if (p[0] == 'n' && p[1] == 'o')
1494 			q = p + 2;
1495 		for(t = global_opts; *t != NULL; t++) {
1496 			if (strcmp(*t, p) == 0)
1497 				break;
1498 			if (q != NULL) {
1499 				if (strcmp(*t, q) == 0)
1500 					break;
1501 			}
1502 		}
1503 		if (*t != NULL)
1504 			continue;
1505 		for(t = legal; *t != NULL; t++) {
1506 			if (strcmp(*t, p) == 0)
1507 				break;
1508 			if (q != NULL) {
1509 				if (strcmp(*t, q) == 0)
1510 					break;
1511 			}
1512 		}
1513 		if (*t != NULL)
1514 			continue;
1515 		snprintf(errmsg, sizeof(errmsg),
1516 		    "mount option <%s> is unknown", p);
1517 		ret = EINVAL;
1518 	}
1519 	if (ret != 0) {
1520 		TAILQ_FOREACH(opt, opts, link) {
1521 			if (strcmp(opt->name, "errmsg") == 0) {
1522 				strncpy((char *)opt->value, errmsg, opt->len);
1523 				break;
1524 			}
1525 		}
1526 		if (opt == NULL)
1527 			printf("%s\n", errmsg);
1528 	}
1529 	return (ret);
1530 }
1531 
1532 /*
1533  * Get a mount option by its name.
1534  *
1535  * Return 0 if the option was found, ENOENT otherwise.
1536  * If len is non-NULL it will be filled with the length
1537  * of the option. If buf is non-NULL, it will be filled
1538  * with the address of the option.
1539  */
1540 int
1541 vfs_getopt(struct vfsoptlist *opts, const char *name, void **buf, int *len)
1542 {
1543 	struct vfsopt *opt;
1544 
1545 	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1546 
1547 	TAILQ_FOREACH(opt, opts, link) {
1548 		if (strcmp(name, opt->name) == 0) {
1549 			opt->seen = 1;
1550 			if (len != NULL)
1551 				*len = opt->len;
1552 			if (buf != NULL)
1553 				*buf = opt->value;
1554 			return (0);
1555 		}
1556 	}
1557 	return (ENOENT);
1558 }
1559 
1560 int
1561 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
1562 {
1563 	struct vfsopt *opt;
1564 
1565 	if (opts == NULL)
1566 		return (-1);
1567 
1568 	TAILQ_FOREACH(opt, opts, link) {
1569 		if (strcmp(name, opt->name) == 0) {
1570 			opt->seen = 1;
1571 			return (opt->pos);
1572 		}
1573 	}
1574 	return (-1);
1575 }
1576 
1577 int
1578 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value)
1579 {
1580 	char *opt_value, *vtp;
1581 	quad_t iv;
1582 	int error, opt_len;
1583 
1584 	error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len);
1585 	if (error != 0)
1586 		return (error);
1587 	if (opt_len == 0 || opt_value == NULL)
1588 		return (EINVAL);
1589 	if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0')
1590 		return (EINVAL);
1591 	iv = strtoq(opt_value, &vtp, 0);
1592 	if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0'))
1593 		return (EINVAL);
1594 	if (iv < 0)
1595 		return (EINVAL);
1596 	switch (vtp[0]) {
1597 	case 't':
1598 	case 'T':
1599 		iv *= 1024;
1600 	case 'g':
1601 	case 'G':
1602 		iv *= 1024;
1603 	case 'm':
1604 	case 'M':
1605 		iv *= 1024;
1606 	case 'k':
1607 	case 'K':
1608 		iv *= 1024;
1609 	case '\0':
1610 		break;
1611 	default:
1612 		return (EINVAL);
1613 	}
1614 	*value = iv;
1615 
1616 	return (0);
1617 }
1618 
1619 char *
1620 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
1621 {
1622 	struct vfsopt *opt;
1623 
1624 	*error = 0;
1625 	TAILQ_FOREACH(opt, opts, link) {
1626 		if (strcmp(name, opt->name) != 0)
1627 			continue;
1628 		opt->seen = 1;
1629 		if (opt->len == 0 ||
1630 		    ((char *)opt->value)[opt->len - 1] != '\0') {
1631 			*error = EINVAL;
1632 			return (NULL);
1633 		}
1634 		return (opt->value);
1635 	}
1636 	*error = ENOENT;
1637 	return (NULL);
1638 }
1639 
1640 int
1641 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
1642 	uint64_t val)
1643 {
1644 	struct vfsopt *opt;
1645 
1646 	TAILQ_FOREACH(opt, opts, link) {
1647 		if (strcmp(name, opt->name) == 0) {
1648 			opt->seen = 1;
1649 			if (w != NULL)
1650 				*w |= val;
1651 			return (1);
1652 		}
1653 	}
1654 	if (w != NULL)
1655 		*w &= ~val;
1656 	return (0);
1657 }
1658 
1659 int
1660 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
1661 {
1662 	va_list ap;
1663 	struct vfsopt *opt;
1664 	int ret;
1665 
1666 	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1667 
1668 	TAILQ_FOREACH(opt, opts, link) {
1669 		if (strcmp(name, opt->name) != 0)
1670 			continue;
1671 		opt->seen = 1;
1672 		if (opt->len == 0 || opt->value == NULL)
1673 			return (0);
1674 		if (((char *)opt->value)[opt->len - 1] != '\0')
1675 			return (0);
1676 		va_start(ap, fmt);
1677 		ret = vsscanf(opt->value, fmt, ap);
1678 		va_end(ap);
1679 		return (ret);
1680 	}
1681 	return (0);
1682 }
1683 
1684 int
1685 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len)
1686 {
1687 	struct vfsopt *opt;
1688 
1689 	TAILQ_FOREACH(opt, opts, link) {
1690 		if (strcmp(name, opt->name) != 0)
1691 			continue;
1692 		opt->seen = 1;
1693 		if (opt->value == NULL)
1694 			opt->len = len;
1695 		else {
1696 			if (opt->len != len)
1697 				return (EINVAL);
1698 			bcopy(value, opt->value, len);
1699 		}
1700 		return (0);
1701 	}
1702 	return (ENOENT);
1703 }
1704 
1705 int
1706 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len)
1707 {
1708 	struct vfsopt *opt;
1709 
1710 	TAILQ_FOREACH(opt, opts, link) {
1711 		if (strcmp(name, opt->name) != 0)
1712 			continue;
1713 		opt->seen = 1;
1714 		if (opt->value == NULL)
1715 			opt->len = len;
1716 		else {
1717 			if (opt->len < len)
1718 				return (EINVAL);
1719 			opt->len = len;
1720 			bcopy(value, opt->value, len);
1721 		}
1722 		return (0);
1723 	}
1724 	return (ENOENT);
1725 }
1726 
1727 int
1728 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value)
1729 {
1730 	struct vfsopt *opt;
1731 
1732 	TAILQ_FOREACH(opt, opts, link) {
1733 		if (strcmp(name, opt->name) != 0)
1734 			continue;
1735 		opt->seen = 1;
1736 		if (opt->value == NULL)
1737 			opt->len = strlen(value) + 1;
1738 		else if (strlcpy(opt->value, value, opt->len) >= opt->len)
1739 			return (EINVAL);
1740 		return (0);
1741 	}
1742 	return (ENOENT);
1743 }
1744 
1745 /*
1746  * Find and copy a mount option.
1747  *
1748  * The size of the buffer has to be specified
1749  * in len, if it is not the same length as the
1750  * mount option, EINVAL is returned.
1751  * Returns ENOENT if the option is not found.
1752  */
1753 int
1754 vfs_copyopt(struct vfsoptlist *opts, const char *name, void *dest, int len)
1755 {
1756 	struct vfsopt *opt;
1757 
1758 	KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1759 
1760 	TAILQ_FOREACH(opt, opts, link) {
1761 		if (strcmp(name, opt->name) == 0) {
1762 			opt->seen = 1;
1763 			if (len != opt->len)
1764 				return (EINVAL);
1765 			bcopy(opt->value, dest, opt->len);
1766 			return (0);
1767 		}
1768 	}
1769 	return (ENOENT);
1770 }
1771 
1772 int
1773 __vfs_statfs(struct mount *mp, struct statfs *sbp)
1774 {
1775 	int error;
1776 
1777 	error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat);
1778 	if (sbp != &mp->mnt_stat)
1779 		*sbp = mp->mnt_stat;
1780 	return (error);
1781 }
1782 
1783 void
1784 vfs_mountedfrom(struct mount *mp, const char *from)
1785 {
1786 
1787 	bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
1788 	strlcpy(mp->mnt_stat.f_mntfromname, from,
1789 	    sizeof mp->mnt_stat.f_mntfromname);
1790 }
1791 
1792 /*
1793  * ---------------------------------------------------------------------
1794  * This is the api for building mount args and mounting filesystems from
1795  * inside the kernel.
1796  *
1797  * The API works by accumulation of individual args.  First error is
1798  * latched.
1799  *
1800  * XXX: should be documented in new manpage kernel_mount(9)
1801  */
1802 
1803 /* A memory allocation which must be freed when we are done */
1804 struct mntaarg {
1805 	SLIST_ENTRY(mntaarg)	next;
1806 };
1807 
1808 /* The header for the mount arguments */
1809 struct mntarg {
1810 	struct iovec *v;
1811 	int len;
1812 	int error;
1813 	SLIST_HEAD(, mntaarg)	list;
1814 };
1815 
1816 /*
1817  * Add a boolean argument.
1818  *
1819  * flag is the boolean value.
1820  * name must start with "no".
1821  */
1822 struct mntarg *
1823 mount_argb(struct mntarg *ma, int flag, const char *name)
1824 {
1825 
1826 	KASSERT(name[0] == 'n' && name[1] == 'o',
1827 	    ("mount_argb(...,%s): name must start with 'no'", name));
1828 
1829 	return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
1830 }
1831 
1832 /*
1833  * Add an argument printf style
1834  */
1835 struct mntarg *
1836 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
1837 {
1838 	va_list ap;
1839 	struct mntaarg *maa;
1840 	struct sbuf *sb;
1841 	int len;
1842 
1843 	if (ma == NULL) {
1844 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1845 		SLIST_INIT(&ma->list);
1846 	}
1847 	if (ma->error)
1848 		return (ma);
1849 
1850 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1851 	    M_MOUNT, M_WAITOK);
1852 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1853 	ma->v[ma->len].iov_len = strlen(name) + 1;
1854 	ma->len++;
1855 
1856 	sb = sbuf_new_auto();
1857 	va_start(ap, fmt);
1858 	sbuf_vprintf(sb, fmt, ap);
1859 	va_end(ap);
1860 	sbuf_finish(sb);
1861 	len = sbuf_len(sb) + 1;
1862 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1863 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1864 	bcopy(sbuf_data(sb), maa + 1, len);
1865 	sbuf_delete(sb);
1866 
1867 	ma->v[ma->len].iov_base = maa + 1;
1868 	ma->v[ma->len].iov_len = len;
1869 	ma->len++;
1870 
1871 	return (ma);
1872 }
1873 
1874 /*
1875  * Add an argument which is a userland string.
1876  */
1877 struct mntarg *
1878 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
1879 {
1880 	struct mntaarg *maa;
1881 	char *tbuf;
1882 
1883 	if (val == NULL)
1884 		return (ma);
1885 	if (ma == NULL) {
1886 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1887 		SLIST_INIT(&ma->list);
1888 	}
1889 	if (ma->error)
1890 		return (ma);
1891 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1892 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1893 	tbuf = (void *)(maa + 1);
1894 	ma->error = copyinstr(val, tbuf, len, NULL);
1895 	return (mount_arg(ma, name, tbuf, -1));
1896 }
1897 
1898 /*
1899  * Plain argument.
1900  *
1901  * If length is -1, treat value as a C string.
1902  */
1903 struct mntarg *
1904 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
1905 {
1906 
1907 	if (ma == NULL) {
1908 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1909 		SLIST_INIT(&ma->list);
1910 	}
1911 	if (ma->error)
1912 		return (ma);
1913 
1914 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1915 	    M_MOUNT, M_WAITOK);
1916 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1917 	ma->v[ma->len].iov_len = strlen(name) + 1;
1918 	ma->len++;
1919 
1920 	ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
1921 	if (len < 0)
1922 		ma->v[ma->len].iov_len = strlen(val) + 1;
1923 	else
1924 		ma->v[ma->len].iov_len = len;
1925 	ma->len++;
1926 	return (ma);
1927 }
1928 
1929 /*
1930  * Free a mntarg structure
1931  */
1932 static void
1933 free_mntarg(struct mntarg *ma)
1934 {
1935 	struct mntaarg *maa;
1936 
1937 	while (!SLIST_EMPTY(&ma->list)) {
1938 		maa = SLIST_FIRST(&ma->list);
1939 		SLIST_REMOVE_HEAD(&ma->list, next);
1940 		free(maa, M_MOUNT);
1941 	}
1942 	free(ma->v, M_MOUNT);
1943 	free(ma, M_MOUNT);
1944 }
1945 
1946 /*
1947  * Mount a filesystem
1948  */
1949 int
1950 kernel_mount(struct mntarg *ma, uint64_t flags)
1951 {
1952 	struct uio auio;
1953 	int error;
1954 
1955 	KASSERT(ma != NULL, ("kernel_mount NULL ma"));
1956 	KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
1957 	KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
1958 
1959 	auio.uio_iov = ma->v;
1960 	auio.uio_iovcnt = ma->len;
1961 	auio.uio_segflg = UIO_SYSSPACE;
1962 
1963 	error = ma->error;
1964 	if (!error)
1965 		error = vfs_donmount(curthread, flags, &auio);
1966 	free_mntarg(ma);
1967 	return (error);
1968 }
1969 
1970 /*
1971  * A printflike function to mount a filesystem.
1972  */
1973 int
1974 kernel_vmount(int flags, ...)
1975 {
1976 	struct mntarg *ma = NULL;
1977 	va_list ap;
1978 	const char *cp;
1979 	const void *vp;
1980 	int error;
1981 
1982 	va_start(ap, flags);
1983 	for (;;) {
1984 		cp = va_arg(ap, const char *);
1985 		if (cp == NULL)
1986 			break;
1987 		vp = va_arg(ap, const void *);
1988 		ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0));
1989 	}
1990 	va_end(ap);
1991 
1992 	error = kernel_mount(ma, flags);
1993 	return (error);
1994 }
1995 
1996 void
1997 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp)
1998 {
1999 
2000 	bcopy(oexp, exp, sizeof(*oexp));
2001 	exp->ex_numsecflavors = 0;
2002 }
2003