xref: /freebsd/sys/kern/vfs_mount.c (revision 3fc36ee018bb836bd1796067cf4ef8683f166ebc)
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  * 4. 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 static uma_zone_t mount_zone;
83 
84 /* List of mounted filesystems. */
85 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
86 
87 /* For any iteration/modification of mountlist */
88 struct mtx mountlist_mtx;
89 MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF);
90 
91 /*
92  * Global opts, taken by all filesystems
93  */
94 static const char *global_opts[] = {
95 	"errmsg",
96 	"fstype",
97 	"fspath",
98 	"ro",
99 	"rw",
100 	"nosuid",
101 	"noexec",
102 	NULL
103 };
104 
105 static int
106 mount_init(void *mem, int size, int flags)
107 {
108 	struct mount *mp;
109 
110 	mp = (struct mount *)mem;
111 	mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF);
112 	lockinit(&mp->mnt_explock, PVFS, "explock", 0, 0);
113 	return (0);
114 }
115 
116 static void
117 mount_fini(void *mem, int size)
118 {
119 	struct mount *mp;
120 
121 	mp = (struct mount *)mem;
122 	lockdestroy(&mp->mnt_explock);
123 	mtx_destroy(&mp->mnt_mtx);
124 }
125 
126 static void
127 vfs_mount_init(void *dummy __unused)
128 {
129 
130 	mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount), NULL,
131 	    NULL, mount_init, mount_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
132 }
133 SYSINIT(vfs_mount, SI_SUB_VFS, SI_ORDER_ANY, vfs_mount_init, NULL);
134 
135 /*
136  * ---------------------------------------------------------------------
137  * Functions for building and sanitizing the mount options
138  */
139 
140 /* Remove one mount option. */
141 static void
142 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
143 {
144 
145 	TAILQ_REMOVE(opts, opt, link);
146 	free(opt->name, M_MOUNT);
147 	if (opt->value != NULL)
148 		free(opt->value, M_MOUNT);
149 	free(opt, M_MOUNT);
150 }
151 
152 /* Release all resources related to the mount options. */
153 void
154 vfs_freeopts(struct vfsoptlist *opts)
155 {
156 	struct vfsopt *opt;
157 
158 	while (!TAILQ_EMPTY(opts)) {
159 		opt = TAILQ_FIRST(opts);
160 		vfs_freeopt(opts, opt);
161 	}
162 	free(opts, M_MOUNT);
163 }
164 
165 void
166 vfs_deleteopt(struct vfsoptlist *opts, const char *name)
167 {
168 	struct vfsopt *opt, *temp;
169 
170 	if (opts == NULL)
171 		return;
172 	TAILQ_FOREACH_SAFE(opt, opts, link, temp)  {
173 		if (strcmp(opt->name, name) == 0)
174 			vfs_freeopt(opts, opt);
175 	}
176 }
177 
178 static int
179 vfs_isopt_ro(const char *opt)
180 {
181 
182 	if (strcmp(opt, "ro") == 0 || strcmp(opt, "rdonly") == 0 ||
183 	    strcmp(opt, "norw") == 0)
184 		return (1);
185 	return (0);
186 }
187 
188 static int
189 vfs_isopt_rw(const char *opt)
190 {
191 
192 	if (strcmp(opt, "rw") == 0 || strcmp(opt, "noro") == 0)
193 		return (1);
194 	return (0);
195 }
196 
197 /*
198  * Check if options are equal (with or without the "no" prefix).
199  */
200 static int
201 vfs_equalopts(const char *opt1, const char *opt2)
202 {
203 	char *p;
204 
205 	/* "opt" vs. "opt" or "noopt" vs. "noopt" */
206 	if (strcmp(opt1, opt2) == 0)
207 		return (1);
208 	/* "noopt" vs. "opt" */
209 	if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
210 		return (1);
211 	/* "opt" vs. "noopt" */
212 	if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
213 		return (1);
214 	while ((p = strchr(opt1, '.')) != NULL &&
215 	    !strncmp(opt1, opt2, ++p - opt1)) {
216 		opt2 += p - opt1;
217 		opt1 = p;
218 		/* "foo.noopt" vs. "foo.opt" */
219 		if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
220 			return (1);
221 		/* "foo.opt" vs. "foo.noopt" */
222 		if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
223 			return (1);
224 	}
225 	/* "ro" / "rdonly" / "norw" / "rw" / "noro" */
226 	if ((vfs_isopt_ro(opt1) || vfs_isopt_rw(opt1)) &&
227 	    (vfs_isopt_ro(opt2) || vfs_isopt_rw(opt2)))
228 		return (1);
229 	return (0);
230 }
231 
232 /*
233  * If a mount option is specified several times,
234  * (with or without the "no" prefix) only keep
235  * the last occurrence of it.
236  */
237 static void
238 vfs_sanitizeopts(struct vfsoptlist *opts)
239 {
240 	struct vfsopt *opt, *opt2, *tmp;
241 
242 	TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
243 		opt2 = TAILQ_PREV(opt, vfsoptlist, link);
244 		while (opt2 != NULL) {
245 			if (vfs_equalopts(opt->name, opt2->name)) {
246 				tmp = TAILQ_PREV(opt2, vfsoptlist, link);
247 				vfs_freeopt(opts, opt2);
248 				opt2 = tmp;
249 			} else {
250 				opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
251 			}
252 		}
253 	}
254 }
255 
256 /*
257  * Build a linked list of mount options from a struct uio.
258  */
259 int
260 vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
261 {
262 	struct vfsoptlist *opts;
263 	struct vfsopt *opt;
264 	size_t memused, namelen, optlen;
265 	unsigned int i, iovcnt;
266 	int error;
267 
268 	opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
269 	TAILQ_INIT(opts);
270 	memused = 0;
271 	iovcnt = auio->uio_iovcnt;
272 	for (i = 0; i < iovcnt; i += 2) {
273 		namelen = auio->uio_iov[i].iov_len;
274 		optlen = auio->uio_iov[i + 1].iov_len;
275 		memused += sizeof(struct vfsopt) + optlen + namelen;
276 		/*
277 		 * Avoid consuming too much memory, and attempts to overflow
278 		 * memused.
279 		 */
280 		if (memused > VFS_MOUNTARG_SIZE_MAX ||
281 		    optlen > VFS_MOUNTARG_SIZE_MAX ||
282 		    namelen > VFS_MOUNTARG_SIZE_MAX) {
283 			error = EINVAL;
284 			goto bad;
285 		}
286 
287 		opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
288 		opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
289 		opt->value = NULL;
290 		opt->len = 0;
291 		opt->pos = i / 2;
292 		opt->seen = 0;
293 
294 		/*
295 		 * Do this early, so jumps to "bad" will free the current
296 		 * option.
297 		 */
298 		TAILQ_INSERT_TAIL(opts, opt, link);
299 
300 		if (auio->uio_segflg == UIO_SYSSPACE) {
301 			bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
302 		} else {
303 			error = copyin(auio->uio_iov[i].iov_base, opt->name,
304 			    namelen);
305 			if (error)
306 				goto bad;
307 		}
308 		/* Ensure names are null-terminated strings. */
309 		if (namelen == 0 || opt->name[namelen - 1] != '\0') {
310 			error = EINVAL;
311 			goto bad;
312 		}
313 		if (optlen != 0) {
314 			opt->len = optlen;
315 			opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
316 			if (auio->uio_segflg == UIO_SYSSPACE) {
317 				bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
318 				    optlen);
319 			} else {
320 				error = copyin(auio->uio_iov[i + 1].iov_base,
321 				    opt->value, optlen);
322 				if (error)
323 					goto bad;
324 			}
325 		}
326 	}
327 	vfs_sanitizeopts(opts);
328 	*options = opts;
329 	return (0);
330 bad:
331 	vfs_freeopts(opts);
332 	return (error);
333 }
334 
335 /*
336  * Merge the old mount options with the new ones passed
337  * in the MNT_UPDATE case.
338  *
339  * XXX: This function will keep a "nofoo" option in the new
340  * options.  E.g, if the option's canonical name is "foo",
341  * "nofoo" ends up in the mount point's active options.
342  */
343 static void
344 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *oldopts)
345 {
346 	struct vfsopt *opt, *new;
347 
348 	TAILQ_FOREACH(opt, oldopts, link) {
349 		new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
350 		new->name = strdup(opt->name, M_MOUNT);
351 		if (opt->len != 0) {
352 			new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
353 			bcopy(opt->value, new->value, opt->len);
354 		} else
355 			new->value = NULL;
356 		new->len = opt->len;
357 		new->seen = opt->seen;
358 		TAILQ_INSERT_HEAD(toopts, new, link);
359 	}
360 	vfs_sanitizeopts(toopts);
361 }
362 
363 /*
364  * Mount a filesystem.
365  */
366 int
367 sys_nmount(td, uap)
368 	struct thread *td;
369 	struct nmount_args /* {
370 		struct iovec *iovp;
371 		unsigned int iovcnt;
372 		int flags;
373 	} */ *uap;
374 {
375 	struct uio *auio;
376 	int error;
377 	u_int iovcnt;
378 	uint64_t flags;
379 
380 	/*
381 	 * Mount flags are now 64-bits. On 32-bit archtectures only
382 	 * 32-bits are passed in, but from here on everything handles
383 	 * 64-bit flags correctly.
384 	 */
385 	flags = uap->flags;
386 
387 	AUDIT_ARG_FFLAGS(flags);
388 	CTR4(KTR_VFS, "%s: iovp %p with iovcnt %d and flags %d", __func__,
389 	    uap->iovp, uap->iovcnt, flags);
390 
391 	/*
392 	 * Filter out MNT_ROOTFS.  We do not want clients of nmount() in
393 	 * userspace to set this flag, but we must filter it out if we want
394 	 * MNT_UPDATE on the root file system to work.
395 	 * MNT_ROOTFS should only be set by the kernel when mounting its
396 	 * root file system.
397 	 */
398 	flags &= ~MNT_ROOTFS;
399 
400 	iovcnt = uap->iovcnt;
401 	/*
402 	 * Check that we have an even number of iovec's
403 	 * and that we have at least two options.
404 	 */
405 	if ((iovcnt & 1) || (iovcnt < 4)) {
406 		CTR2(KTR_VFS, "%s: failed for invalid iovcnt %d", __func__,
407 		    uap->iovcnt);
408 		return (EINVAL);
409 	}
410 
411 	error = copyinuio(uap->iovp, iovcnt, &auio);
412 	if (error) {
413 		CTR2(KTR_VFS, "%s: failed for invalid uio op with %d errno",
414 		    __func__, error);
415 		return (error);
416 	}
417 	error = vfs_donmount(td, flags, auio);
418 
419 	free(auio, M_IOV);
420 	return (error);
421 }
422 
423 /*
424  * ---------------------------------------------------------------------
425  * Various utility functions
426  */
427 
428 void
429 vfs_ref(struct mount *mp)
430 {
431 
432 	CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
433 	MNT_ILOCK(mp);
434 	MNT_REF(mp);
435 	MNT_IUNLOCK(mp);
436 }
437 
438 void
439 vfs_rel(struct mount *mp)
440 {
441 
442 	CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
443 	MNT_ILOCK(mp);
444 	MNT_REL(mp);
445 	MNT_IUNLOCK(mp);
446 }
447 
448 /*
449  * Allocate and initialize the mount point struct.
450  */
451 struct mount *
452 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, const char *fspath,
453     struct ucred *cred)
454 {
455 	struct mount *mp;
456 
457 	mp = uma_zalloc(mount_zone, M_WAITOK);
458 	bzero(&mp->mnt_startzero,
459 	    __rangeof(struct mount, mnt_startzero, mnt_endzero));
460 	TAILQ_INIT(&mp->mnt_nvnodelist);
461 	mp->mnt_nvnodelistsize = 0;
462 	TAILQ_INIT(&mp->mnt_activevnodelist);
463 	mp->mnt_activevnodelistsize = 0;
464 	mp->mnt_ref = 0;
465 	(void) vfs_busy(mp, MBF_NOWAIT);
466 	atomic_add_acq_int(&vfsp->vfc_refcount, 1);
467 	mp->mnt_op = vfsp->vfc_vfsops;
468 	mp->mnt_vfc = vfsp;
469 	mp->mnt_stat.f_type = vfsp->vfc_typenum;
470 	mp->mnt_gen++;
471 	strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
472 	mp->mnt_vnodecovered = vp;
473 	mp->mnt_cred = crdup(cred);
474 	mp->mnt_stat.f_owner = cred->cr_uid;
475 	strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
476 	mp->mnt_iosize_max = DFLTPHYS;
477 #ifdef MAC
478 	mac_mount_init(mp);
479 	mac_mount_create(cred, mp);
480 #endif
481 	arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0);
482 	TAILQ_INIT(&mp->mnt_uppers);
483 	return (mp);
484 }
485 
486 /*
487  * Destroy the mount struct previously allocated by vfs_mount_alloc().
488  */
489 void
490 vfs_mount_destroy(struct mount *mp)
491 {
492 
493 	MNT_ILOCK(mp);
494 	mp->mnt_kern_flag |= MNTK_REFEXPIRE;
495 	if (mp->mnt_kern_flag & MNTK_MWAIT) {
496 		mp->mnt_kern_flag &= ~MNTK_MWAIT;
497 		wakeup(mp);
498 	}
499 	while (mp->mnt_ref)
500 		msleep(mp, MNT_MTX(mp), PVFS, "mntref", 0);
501 	KASSERT(mp->mnt_ref == 0,
502 	    ("%s: invalid refcount in the drain path @ %s:%d", __func__,
503 	    __FILE__, __LINE__));
504 	if (mp->mnt_writeopcount != 0)
505 		panic("vfs_mount_destroy: nonzero writeopcount");
506 	if (mp->mnt_secondary_writes != 0)
507 		panic("vfs_mount_destroy: nonzero secondary_writes");
508 	atomic_subtract_rel_int(&mp->mnt_vfc->vfc_refcount, 1);
509 	if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
510 		struct vnode *vp;
511 
512 		TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
513 			vn_printf(vp, "dangling vnode ");
514 		panic("unmount: dangling vnode");
515 	}
516 	KASSERT(TAILQ_EMPTY(&mp->mnt_uppers), ("mnt_uppers"));
517 	if (mp->mnt_nvnodelistsize != 0)
518 		panic("vfs_mount_destroy: nonzero nvnodelistsize");
519 	if (mp->mnt_activevnodelistsize != 0)
520 		panic("vfs_mount_destroy: nonzero activevnodelistsize");
521 	if (mp->mnt_lockref != 0)
522 		panic("vfs_mount_destroy: nonzero lock refcount");
523 	MNT_IUNLOCK(mp);
524 #ifdef MAC
525 	mac_mount_destroy(mp);
526 #endif
527 	if (mp->mnt_opt != NULL)
528 		vfs_freeopts(mp->mnt_opt);
529 	crfree(mp->mnt_cred);
530 	uma_zfree(mount_zone, mp);
531 }
532 
533 int
534 vfs_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions)
535 {
536 	struct vfsoptlist *optlist;
537 	struct vfsopt *opt, *tmp_opt;
538 	char *fstype, *fspath, *errmsg;
539 	int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
540 
541 	errmsg = fspath = NULL;
542 	errmsg_len = fspathlen = 0;
543 	errmsg_pos = -1;
544 
545 	error = vfs_buildopts(fsoptions, &optlist);
546 	if (error)
547 		return (error);
548 
549 	if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
550 		errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
551 
552 	/*
553 	 * We need these two options before the others,
554 	 * and they are mandatory for any filesystem.
555 	 * Ensure they are NUL terminated as well.
556 	 */
557 	fstypelen = 0;
558 	error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
559 	if (error || fstype[fstypelen - 1] != '\0') {
560 		error = EINVAL;
561 		if (errmsg != NULL)
562 			strncpy(errmsg, "Invalid fstype", errmsg_len);
563 		goto bail;
564 	}
565 	fspathlen = 0;
566 	error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
567 	if (error || fspath[fspathlen - 1] != '\0') {
568 		error = EINVAL;
569 		if (errmsg != NULL)
570 			strncpy(errmsg, "Invalid fspath", errmsg_len);
571 		goto bail;
572 	}
573 
574 	/*
575 	 * We need to see if we have the "update" option
576 	 * before we call vfs_domount(), since vfs_domount() has special
577 	 * logic based on MNT_UPDATE.  This is very important
578 	 * when we want to update the root filesystem.
579 	 */
580 	TAILQ_FOREACH_SAFE(opt, optlist, link, tmp_opt) {
581 		if (strcmp(opt->name, "update") == 0) {
582 			fsflags |= MNT_UPDATE;
583 			vfs_freeopt(optlist, opt);
584 		}
585 		else if (strcmp(opt->name, "async") == 0)
586 			fsflags |= MNT_ASYNC;
587 		else if (strcmp(opt->name, "force") == 0) {
588 			fsflags |= MNT_FORCE;
589 			vfs_freeopt(optlist, opt);
590 		}
591 		else if (strcmp(opt->name, "reload") == 0) {
592 			fsflags |= MNT_RELOAD;
593 			vfs_freeopt(optlist, opt);
594 		}
595 		else if (strcmp(opt->name, "multilabel") == 0)
596 			fsflags |= MNT_MULTILABEL;
597 		else if (strcmp(opt->name, "noasync") == 0)
598 			fsflags &= ~MNT_ASYNC;
599 		else if (strcmp(opt->name, "noatime") == 0)
600 			fsflags |= MNT_NOATIME;
601 		else if (strcmp(opt->name, "atime") == 0) {
602 			free(opt->name, M_MOUNT);
603 			opt->name = strdup("nonoatime", M_MOUNT);
604 		}
605 		else if (strcmp(opt->name, "noclusterr") == 0)
606 			fsflags |= MNT_NOCLUSTERR;
607 		else if (strcmp(opt->name, "clusterr") == 0) {
608 			free(opt->name, M_MOUNT);
609 			opt->name = strdup("nonoclusterr", M_MOUNT);
610 		}
611 		else if (strcmp(opt->name, "noclusterw") == 0)
612 			fsflags |= MNT_NOCLUSTERW;
613 		else if (strcmp(opt->name, "clusterw") == 0) {
614 			free(opt->name, M_MOUNT);
615 			opt->name = strdup("nonoclusterw", M_MOUNT);
616 		}
617 		else if (strcmp(opt->name, "noexec") == 0)
618 			fsflags |= MNT_NOEXEC;
619 		else if (strcmp(opt->name, "exec") == 0) {
620 			free(opt->name, M_MOUNT);
621 			opt->name = strdup("nonoexec", M_MOUNT);
622 		}
623 		else if (strcmp(opt->name, "nosuid") == 0)
624 			fsflags |= MNT_NOSUID;
625 		else if (strcmp(opt->name, "suid") == 0) {
626 			free(opt->name, M_MOUNT);
627 			opt->name = strdup("nonosuid", M_MOUNT);
628 		}
629 		else if (strcmp(opt->name, "nosymfollow") == 0)
630 			fsflags |= MNT_NOSYMFOLLOW;
631 		else if (strcmp(opt->name, "symfollow") == 0) {
632 			free(opt->name, M_MOUNT);
633 			opt->name = strdup("nonosymfollow", M_MOUNT);
634 		}
635 		else if (strcmp(opt->name, "noro") == 0)
636 			fsflags &= ~MNT_RDONLY;
637 		else if (strcmp(opt->name, "rw") == 0)
638 			fsflags &= ~MNT_RDONLY;
639 		else if (strcmp(opt->name, "ro") == 0)
640 			fsflags |= MNT_RDONLY;
641 		else if (strcmp(opt->name, "rdonly") == 0) {
642 			free(opt->name, M_MOUNT);
643 			opt->name = strdup("ro", M_MOUNT);
644 			fsflags |= MNT_RDONLY;
645 		}
646 		else if (strcmp(opt->name, "suiddir") == 0)
647 			fsflags |= MNT_SUIDDIR;
648 		else if (strcmp(opt->name, "sync") == 0)
649 			fsflags |= MNT_SYNCHRONOUS;
650 		else if (strcmp(opt->name, "union") == 0)
651 			fsflags |= MNT_UNION;
652 		else if (strcmp(opt->name, "automounted") == 0) {
653 			fsflags |= MNT_AUTOMOUNTED;
654 			vfs_freeopt(optlist, opt);
655 		}
656 	}
657 
658 	/*
659 	 * Be ultra-paranoid about making sure the type and fspath
660 	 * variables will fit in our mp buffers, including the
661 	 * terminating NUL.
662 	 */
663 	if (fstypelen > MFSNAMELEN || fspathlen > MNAMELEN) {
664 		error = ENAMETOOLONG;
665 		goto bail;
666 	}
667 
668 	error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
669 bail:
670 	/* copyout the errmsg */
671 	if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
672 	    && errmsg_len > 0 && errmsg != NULL) {
673 		if (fsoptions->uio_segflg == UIO_SYSSPACE) {
674 			bcopy(errmsg,
675 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
676 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
677 		} else {
678 			copyout(errmsg,
679 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
680 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
681 		}
682 	}
683 
684 	if (optlist != NULL)
685 		vfs_freeopts(optlist);
686 	return (error);
687 }
688 
689 /*
690  * Old mount API.
691  */
692 #ifndef _SYS_SYSPROTO_H_
693 struct mount_args {
694 	char	*type;
695 	char	*path;
696 	int	flags;
697 	caddr_t	data;
698 };
699 #endif
700 /* ARGSUSED */
701 int
702 sys_mount(td, uap)
703 	struct thread *td;
704 	struct mount_args /* {
705 		char *type;
706 		char *path;
707 		int flags;
708 		caddr_t data;
709 	} */ *uap;
710 {
711 	char *fstype;
712 	struct vfsconf *vfsp = NULL;
713 	struct mntarg *ma = NULL;
714 	uint64_t flags;
715 	int error;
716 
717 	/*
718 	 * Mount flags are now 64-bits. On 32-bit architectures only
719 	 * 32-bits are passed in, but from here on everything handles
720 	 * 64-bit flags correctly.
721 	 */
722 	flags = uap->flags;
723 
724 	AUDIT_ARG_FFLAGS(flags);
725 
726 	/*
727 	 * Filter out MNT_ROOTFS.  We do not want clients of mount() in
728 	 * userspace to set this flag, but we must filter it out if we want
729 	 * MNT_UPDATE on the root file system to work.
730 	 * MNT_ROOTFS should only be set by the kernel when mounting its
731 	 * root file system.
732 	 */
733 	flags &= ~MNT_ROOTFS;
734 
735 	fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
736 	error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
737 	if (error) {
738 		free(fstype, M_TEMP);
739 		return (error);
740 	}
741 
742 	AUDIT_ARG_TEXT(fstype);
743 	vfsp = vfs_byname_kld(fstype, td, &error);
744 	free(fstype, M_TEMP);
745 	if (vfsp == NULL)
746 		return (ENOENT);
747 	if (vfsp->vfc_vfsops->vfs_cmount == NULL)
748 		return (EOPNOTSUPP);
749 
750 	ma = mount_argsu(ma, "fstype", uap->type, MFSNAMELEN);
751 	ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
752 	ma = mount_argb(ma, flags & MNT_RDONLY, "noro");
753 	ma = mount_argb(ma, !(flags & MNT_NOSUID), "nosuid");
754 	ma = mount_argb(ma, !(flags & MNT_NOEXEC), "noexec");
755 
756 	error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags);
757 	return (error);
758 }
759 
760 /*
761  * vfs_domount_first(): first file system mount (not update)
762  */
763 static int
764 vfs_domount_first(
765 	struct thread *td,		/* Calling thread. */
766 	struct vfsconf *vfsp,		/* File system type. */
767 	char *fspath,			/* Mount path. */
768 	struct vnode *vp,		/* Vnode to be covered. */
769 	uint64_t fsflags,		/* Flags common to all filesystems. */
770 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
771 	)
772 {
773 	struct vattr va;
774 	struct mount *mp;
775 	struct vnode *newdp;
776 	int error;
777 
778 	ASSERT_VOP_ELOCKED(vp, __func__);
779 	KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
780 
781 	/*
782 	 * If the user is not root, ensure that they own the directory
783 	 * onto which we are attempting to mount.
784 	 */
785 	error = VOP_GETATTR(vp, &va, td->td_ucred);
786 	if (error == 0 && va.va_uid != td->td_ucred->cr_uid)
787 		error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN, 0);
788 	if (error == 0)
789 		error = vinvalbuf(vp, V_SAVE, 0, 0);
790 	if (error == 0 && vp->v_type != VDIR)
791 		error = ENOTDIR;
792 	if (error == 0) {
793 		VI_LOCK(vp);
794 		if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
795 			vp->v_iflag |= VI_MOUNT;
796 		else
797 			error = EBUSY;
798 		VI_UNLOCK(vp);
799 	}
800 	if (error != 0) {
801 		vput(vp);
802 		return (error);
803 	}
804 	VOP_UNLOCK(vp, 0);
805 
806 	/* Allocate and initialize the filesystem. */
807 	mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred);
808 	/* XXXMAC: pass to vfs_mount_alloc? */
809 	mp->mnt_optnew = *optlist;
810 	/* Set the mount level flags. */
811 	mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY));
812 
813 	/*
814 	 * Mount the filesystem.
815 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
816 	 * get.  No freeing of cn_pnbuf.
817 	 */
818 	error = VFS_MOUNT(mp);
819 	if (error != 0) {
820 		vfs_unbusy(mp);
821 		vfs_mount_destroy(mp);
822 		VI_LOCK(vp);
823 		vp->v_iflag &= ~VI_MOUNT;
824 		VI_UNLOCK(vp);
825 		vrele(vp);
826 		return (error);
827 	}
828 
829 	if (mp->mnt_opt != NULL)
830 		vfs_freeopts(mp->mnt_opt);
831 	mp->mnt_opt = mp->mnt_optnew;
832 	*optlist = NULL;
833 	(void)VFS_STATFS(mp, &mp->mnt_stat);
834 
835 	/*
836 	 * Prevent external consumers of mount options from reading mnt_optnew.
837 	 */
838 	mp->mnt_optnew = NULL;
839 
840 	MNT_ILOCK(mp);
841 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
842 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
843 		mp->mnt_kern_flag |= MNTK_ASYNC;
844 	else
845 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
846 	MNT_IUNLOCK(mp);
847 
848 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
849 	cache_purge(vp);
850 	VI_LOCK(vp);
851 	vp->v_iflag &= ~VI_MOUNT;
852 	VI_UNLOCK(vp);
853 	vp->v_mountedhere = mp;
854 	/* Place the new filesystem at the end of the mount list. */
855 	mtx_lock(&mountlist_mtx);
856 	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
857 	mtx_unlock(&mountlist_mtx);
858 	vfs_event_signal(NULL, VQ_MOUNT, 0);
859 	if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp))
860 		panic("mount: lost mount");
861 	VOP_UNLOCK(vp, 0);
862 	EVENTHANDLER_INVOKE(vfs_mounted, mp, newdp, td);
863 	VOP_UNLOCK(newdp, 0);
864 	mountcheckdirs(vp, newdp);
865 	vrele(newdp);
866 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
867 		vfs_allocate_syncvnode(mp);
868 	vfs_unbusy(mp);
869 	return (0);
870 }
871 
872 /*
873  * vfs_domount_update(): update of mounted file system
874  */
875 static int
876 vfs_domount_update(
877 	struct thread *td,		/* Calling thread. */
878 	struct vnode *vp,		/* Mount point vnode. */
879 	uint64_t fsflags,		/* Flags common to all filesystems. */
880 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
881 	)
882 {
883 	struct export_args export;
884 	void *bufp;
885 	struct mount *mp;
886 	int error, export_error, len;
887 	uint64_t flag;
888 
889 	ASSERT_VOP_ELOCKED(vp, __func__);
890 	KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
891 	mp = vp->v_mount;
892 
893 	if ((vp->v_vflag & VV_ROOT) == 0) {
894 		if (vfs_copyopt(*optlist, "export", &export, sizeof(export))
895 		    == 0)
896 			error = EXDEV;
897 		else
898 			error = EINVAL;
899 		vput(vp);
900 		return (error);
901 	}
902 
903 	/*
904 	 * We only allow the filesystem to be reloaded if it
905 	 * is currently mounted read-only.
906 	 */
907 	flag = mp->mnt_flag;
908 	if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
909 		vput(vp);
910 		return (EOPNOTSUPP);	/* Needs translation */
911 	}
912 	/*
913 	 * Only privileged root, or (if MNT_USER is set) the user that
914 	 * did the original mount is permitted to update it.
915 	 */
916 	error = vfs_suser(mp, td);
917 	if (error != 0) {
918 		vput(vp);
919 		return (error);
920 	}
921 	if (vfs_busy(mp, MBF_NOWAIT)) {
922 		vput(vp);
923 		return (EBUSY);
924 	}
925 	VI_LOCK(vp);
926 	if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
927 		VI_UNLOCK(vp);
928 		vfs_unbusy(mp);
929 		vput(vp);
930 		return (EBUSY);
931 	}
932 	vp->v_iflag |= VI_MOUNT;
933 	VI_UNLOCK(vp);
934 	VOP_UNLOCK(vp, 0);
935 
936 	MNT_ILOCK(mp);
937 	mp->mnt_flag &= ~MNT_UPDATEMASK;
938 	mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE |
939 	    MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY);
940 	if ((mp->mnt_flag & MNT_ASYNC) == 0)
941 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
942 	MNT_IUNLOCK(mp);
943 	mp->mnt_optnew = *optlist;
944 	vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
945 
946 	/*
947 	 * Mount the filesystem.
948 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
949 	 * get.  No freeing of cn_pnbuf.
950 	 */
951 	error = VFS_MOUNT(mp);
952 
953 	export_error = 0;
954 	/* Process the export option. */
955 	if (error == 0 && vfs_getopt(mp->mnt_optnew, "export", &bufp,
956 	    &len) == 0) {
957 		/* Assume that there is only 1 ABI for each length. */
958 		switch (len) {
959 		case (sizeof(struct oexport_args)):
960 			bzero(&export, sizeof(export));
961 			/* FALLTHROUGH */
962 		case (sizeof(export)):
963 			bcopy(bufp, &export, len);
964 			export_error = vfs_export(mp, &export);
965 			break;
966 		default:
967 			export_error = EINVAL;
968 			break;
969 		}
970 	}
971 
972 	MNT_ILOCK(mp);
973 	if (error == 0) {
974 		mp->mnt_flag &=	~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
975 		    MNT_SNAPSHOT);
976 	} else {
977 		/*
978 		 * If we fail, restore old mount flags. MNT_QUOTA is special,
979 		 * because it is not part of MNT_UPDATEMASK, but it could have
980 		 * changed in the meantime if quotactl(2) was called.
981 		 * All in all we want current value of MNT_QUOTA, not the old
982 		 * one.
983 		 */
984 		mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
985 	}
986 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
987 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
988 		mp->mnt_kern_flag |= MNTK_ASYNC;
989 	else
990 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
991 	MNT_IUNLOCK(mp);
992 
993 	if (error != 0)
994 		goto end;
995 
996 	if (mp->mnt_opt != NULL)
997 		vfs_freeopts(mp->mnt_opt);
998 	mp->mnt_opt = mp->mnt_optnew;
999 	*optlist = NULL;
1000 	(void)VFS_STATFS(mp, &mp->mnt_stat);
1001 	/*
1002 	 * Prevent external consumers of mount options from reading
1003 	 * mnt_optnew.
1004 	 */
1005 	mp->mnt_optnew = NULL;
1006 
1007 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
1008 		vfs_allocate_syncvnode(mp);
1009 	else
1010 		vfs_deallocate_syncvnode(mp);
1011 end:
1012 	vfs_unbusy(mp);
1013 	VI_LOCK(vp);
1014 	vp->v_iflag &= ~VI_MOUNT;
1015 	VI_UNLOCK(vp);
1016 	vrele(vp);
1017 	return (error != 0 ? error : export_error);
1018 }
1019 
1020 /*
1021  * vfs_domount(): actually attempt a filesystem mount.
1022  */
1023 static int
1024 vfs_domount(
1025 	struct thread *td,		/* Calling thread. */
1026 	const char *fstype,		/* Filesystem type. */
1027 	char *fspath,			/* Mount path. */
1028 	uint64_t fsflags,		/* Flags common to all filesystems. */
1029 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
1030 	)
1031 {
1032 	struct vfsconf *vfsp;
1033 	struct nameidata nd;
1034 	struct vnode *vp;
1035 	char *pathbuf;
1036 	int error;
1037 
1038 	/*
1039 	 * Be ultra-paranoid about making sure the type and fspath
1040 	 * variables will fit in our mp buffers, including the
1041 	 * terminating NUL.
1042 	 */
1043 	if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
1044 		return (ENAMETOOLONG);
1045 
1046 	if (jailed(td->td_ucred) || usermount == 0) {
1047 		if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
1048 			return (error);
1049 	}
1050 
1051 	/*
1052 	 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
1053 	 */
1054 	if (fsflags & MNT_EXPORTED) {
1055 		error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
1056 		if (error)
1057 			return (error);
1058 	}
1059 	if (fsflags & MNT_SUIDDIR) {
1060 		error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
1061 		if (error)
1062 			return (error);
1063 	}
1064 	/*
1065 	 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
1066 	 */
1067 	if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
1068 		if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
1069 			fsflags |= MNT_NOSUID | MNT_USER;
1070 	}
1071 
1072 	/* Load KLDs before we lock the covered vnode to avoid reversals. */
1073 	vfsp = NULL;
1074 	if ((fsflags & MNT_UPDATE) == 0) {
1075 		/* Don't try to load KLDs if we're mounting the root. */
1076 		if (fsflags & MNT_ROOTFS)
1077 			vfsp = vfs_byname(fstype);
1078 		else
1079 			vfsp = vfs_byname_kld(fstype, td, &error);
1080 		if (vfsp == NULL)
1081 			return (ENODEV);
1082 		if (jailed(td->td_ucred) && !(vfsp->vfc_flags & VFCF_JAIL))
1083 			return (EPERM);
1084 	}
1085 
1086 	/*
1087 	 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
1088 	 */
1089 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1090 	    UIO_SYSSPACE, fspath, td);
1091 	error = namei(&nd);
1092 	if (error != 0)
1093 		return (error);
1094 	NDFREE(&nd, NDF_ONLY_PNBUF);
1095 	vp = nd.ni_vp;
1096 	if ((fsflags & MNT_UPDATE) == 0) {
1097 		pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1098 		strcpy(pathbuf, fspath);
1099 		error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN);
1100 		/* debug.disablefullpath == 1 results in ENODEV */
1101 		if (error == 0 || error == ENODEV) {
1102 			error = vfs_domount_first(td, vfsp, pathbuf, vp,
1103 			    fsflags, optlist);
1104 		}
1105 		free(pathbuf, M_TEMP);
1106 	} else
1107 		error = vfs_domount_update(td, vp, fsflags, optlist);
1108 
1109 	return (error);
1110 }
1111 
1112 /*
1113  * Unmount a filesystem.
1114  *
1115  * Note: unmount takes a path to the vnode mounted on as argument, not
1116  * special file (as before).
1117  */
1118 #ifndef _SYS_SYSPROTO_H_
1119 struct unmount_args {
1120 	char	*path;
1121 	int	flags;
1122 };
1123 #endif
1124 /* ARGSUSED */
1125 int
1126 sys_unmount(struct thread *td, struct unmount_args *uap)
1127 {
1128 	struct nameidata nd;
1129 	struct mount *mp;
1130 	char *pathbuf;
1131 	int error, id0, id1;
1132 
1133 	AUDIT_ARG_VALUE(uap->flags);
1134 	if (jailed(td->td_ucred) || usermount == 0) {
1135 		error = priv_check(td, PRIV_VFS_UNMOUNT);
1136 		if (error)
1137 			return (error);
1138 	}
1139 
1140 	pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1141 	error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1142 	if (error) {
1143 		free(pathbuf, M_TEMP);
1144 		return (error);
1145 	}
1146 	if (uap->flags & MNT_BYFSID) {
1147 		AUDIT_ARG_TEXT(pathbuf);
1148 		/* Decode the filesystem ID. */
1149 		if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
1150 			free(pathbuf, M_TEMP);
1151 			return (EINVAL);
1152 		}
1153 
1154 		mtx_lock(&mountlist_mtx);
1155 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1156 			if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1157 			    mp->mnt_stat.f_fsid.val[1] == id1) {
1158 				vfs_ref(mp);
1159 				break;
1160 			}
1161 		}
1162 		mtx_unlock(&mountlist_mtx);
1163 	} else {
1164 		/*
1165 		 * Try to find global path for path argument.
1166 		 */
1167 		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1168 		    UIO_SYSSPACE, pathbuf, td);
1169 		if (namei(&nd) == 0) {
1170 			NDFREE(&nd, NDF_ONLY_PNBUF);
1171 			error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
1172 			    MNAMELEN);
1173 			if (error == 0 || error == ENODEV)
1174 				vput(nd.ni_vp);
1175 		}
1176 		mtx_lock(&mountlist_mtx);
1177 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1178 			if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) {
1179 				vfs_ref(mp);
1180 				break;
1181 			}
1182 		}
1183 		mtx_unlock(&mountlist_mtx);
1184 	}
1185 	free(pathbuf, M_TEMP);
1186 	if (mp == NULL) {
1187 		/*
1188 		 * Previously we returned ENOENT for a nonexistent path and
1189 		 * EINVAL for a non-mountpoint.  We cannot tell these apart
1190 		 * now, so in the !MNT_BYFSID case return the more likely
1191 		 * EINVAL for compatibility.
1192 		 */
1193 		return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1194 	}
1195 
1196 	/*
1197 	 * Don't allow unmounting the root filesystem.
1198 	 */
1199 	if (mp->mnt_flag & MNT_ROOTFS) {
1200 		vfs_rel(mp);
1201 		return (EINVAL);
1202 	}
1203 	error = dounmount(mp, uap->flags, td);
1204 	return (error);
1205 }
1206 
1207 /*
1208  * Return error if any of the vnodes, ignoring the root vnode
1209  * and the syncer vnode, have non-zero usecount.
1210  */
1211 static int
1212 vfs_check_usecounts(struct mount *mp)
1213 {
1214 	struct vnode *vp, *mvp;
1215 
1216 	MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1217 		if ((vp->v_vflag & VV_ROOT) == 0 && vp->v_type != VNON &&
1218 		    vp->v_usecount != 0) {
1219 			VI_UNLOCK(vp);
1220 			MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1221 			return (EBUSY);
1222 		}
1223 		VI_UNLOCK(vp);
1224 	}
1225 
1226 	return (0);
1227 }
1228 
1229 /*
1230  * Do the actual filesystem unmount.
1231  */
1232 int
1233 dounmount(struct mount *mp, int flags, struct thread *td)
1234 {
1235 	struct vnode *coveredvp, *fsrootvp;
1236 	int error;
1237 	uint64_t async_flag;
1238 	int mnt_gen_r;
1239 
1240 	if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
1241 		mnt_gen_r = mp->mnt_gen;
1242 		VI_LOCK(coveredvp);
1243 		vholdl(coveredvp);
1244 		vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
1245 		/*
1246 		 * Check for mp being unmounted while waiting for the
1247 		 * covered vnode lock.
1248 		 */
1249 		if (coveredvp->v_mountedhere != mp ||
1250 		    coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
1251 			VOP_UNLOCK(coveredvp, 0);
1252 			vdrop(coveredvp);
1253 			vfs_rel(mp);
1254 			return (EBUSY);
1255 		}
1256 	}
1257 
1258 	/*
1259 	 * Only privileged root, or (if MNT_USER is set) the user that did the
1260 	 * original mount is permitted to unmount this filesystem.
1261 	 */
1262 	error = vfs_suser(mp, td);
1263 	if (error != 0) {
1264 		if (coveredvp != NULL) {
1265 			VOP_UNLOCK(coveredvp, 0);
1266 			vdrop(coveredvp);
1267 		}
1268 		vfs_rel(mp);
1269 		return (error);
1270 	}
1271 
1272 	vn_start_write(NULL, &mp, V_WAIT | V_MNTREF);
1273 	MNT_ILOCK(mp);
1274 	if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 ||
1275 	    !TAILQ_EMPTY(&mp->mnt_uppers)) {
1276 		MNT_IUNLOCK(mp);
1277 		if (coveredvp != NULL) {
1278 			VOP_UNLOCK(coveredvp, 0);
1279 			vdrop(coveredvp);
1280 		}
1281 		vn_finished_write(mp);
1282 		return (EBUSY);
1283 	}
1284 	mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ;
1285 	if (flags & MNT_NONBUSY) {
1286 		MNT_IUNLOCK(mp);
1287 		error = vfs_check_usecounts(mp);
1288 		MNT_ILOCK(mp);
1289 		if (error != 0) {
1290 			mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_NOINSMNTQ);
1291 			MNT_IUNLOCK(mp);
1292 			if (coveredvp != NULL) {
1293 				VOP_UNLOCK(coveredvp, 0);
1294 				vdrop(coveredvp);
1295 			}
1296 			vn_finished_write(mp);
1297 			return (error);
1298 		}
1299 	}
1300 	/* Allow filesystems to detect that a forced unmount is in progress. */
1301 	if (flags & MNT_FORCE) {
1302 		mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1303 		MNT_IUNLOCK(mp);
1304 		/*
1305 		 * Must be done after setting MNTK_UNMOUNTF and before
1306 		 * waiting for mnt_lockref to become 0.
1307 		 */
1308 		VFS_PURGE(mp);
1309 		MNT_ILOCK(mp);
1310 	}
1311 	error = 0;
1312 	if (mp->mnt_lockref) {
1313 		mp->mnt_kern_flag |= MNTK_DRAINING;
1314 		error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
1315 		    "mount drain", 0);
1316 	}
1317 	MNT_IUNLOCK(mp);
1318 	KASSERT(mp->mnt_lockref == 0,
1319 	    ("%s: invalid lock refcount in the drain path @ %s:%d",
1320 	    __func__, __FILE__, __LINE__));
1321 	KASSERT(error == 0,
1322 	    ("%s: invalid return value for msleep in the drain path @ %s:%d",
1323 	    __func__, __FILE__, __LINE__));
1324 
1325 	if (mp->mnt_flag & MNT_EXPUBLIC)
1326 		vfs_setpublicfs(NULL, NULL, NULL);
1327 
1328 	/*
1329 	 * From now, we can claim that the use reference on the
1330 	 * coveredvp is ours, and the ref can be released only by
1331 	 * successfull unmount by us, or left for later unmount
1332 	 * attempt.  The previously acquired hold reference is no
1333 	 * longer needed to protect the vnode from reuse.
1334 	 */
1335 	if (coveredvp != NULL)
1336 		vdrop(coveredvp);
1337 
1338 	vfs_msync(mp, MNT_WAIT);
1339 	MNT_ILOCK(mp);
1340 	async_flag = mp->mnt_flag & MNT_ASYNC;
1341 	mp->mnt_flag &= ~MNT_ASYNC;
1342 	mp->mnt_kern_flag &= ~MNTK_ASYNC;
1343 	MNT_IUNLOCK(mp);
1344 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
1345 	vfs_deallocate_syncvnode(mp);
1346 	/*
1347 	 * For forced unmounts, move process cdir/rdir refs on the fs root
1348 	 * vnode to the covered vnode.  For non-forced unmounts we want
1349 	 * such references to cause an EBUSY error.
1350 	 */
1351 	if ((flags & MNT_FORCE) &&
1352 	    VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1353 		if (mp->mnt_vnodecovered != NULL &&
1354 		    (mp->mnt_flag & MNT_IGNORE) == 0)
1355 			mountcheckdirs(fsrootvp, mp->mnt_vnodecovered);
1356 		if (fsrootvp == rootvnode) {
1357 			vrele(rootvnode);
1358 			rootvnode = NULL;
1359 		}
1360 		vput(fsrootvp);
1361 	}
1362 	if ((mp->mnt_flag & MNT_RDONLY) != 0 || (flags & MNT_FORCE) != 0 ||
1363 	    (error = VFS_SYNC(mp, MNT_WAIT)) == 0)
1364 		error = VFS_UNMOUNT(mp, flags);
1365 	vn_finished_write(mp);
1366 	/*
1367 	 * If we failed to flush the dirty blocks for this mount point,
1368 	 * undo all the cdir/rdir and rootvnode changes we made above.
1369 	 * Unless we failed to do so because the device is reporting that
1370 	 * it doesn't exist anymore.
1371 	 */
1372 	if (error && error != ENXIO) {
1373 		if ((flags & MNT_FORCE) &&
1374 		    VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1375 			if (mp->mnt_vnodecovered != NULL &&
1376 			    (mp->mnt_flag & MNT_IGNORE) == 0)
1377 				mountcheckdirs(mp->mnt_vnodecovered, fsrootvp);
1378 			if (rootvnode == NULL) {
1379 				rootvnode = fsrootvp;
1380 				vref(rootvnode);
1381 			}
1382 			vput(fsrootvp);
1383 		}
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_INVOKE(vfs_unmounted, mp, td);
1409 	if (coveredvp != NULL) {
1410 		coveredvp->v_mountedhere = NULL;
1411 		vput(coveredvp);
1412 	}
1413 	vfs_event_signal(NULL, VQ_UNMOUNT, 0);
1414 	if (mp == rootdevmp)
1415 		rootdevmp = NULL;
1416 	vfs_mount_destroy(mp);
1417 	return (0);
1418 }
1419 
1420 /*
1421  * Report errors during filesystem mounting.
1422  */
1423 void
1424 vfs_mount_error(struct mount *mp, const char *fmt, ...)
1425 {
1426 	struct vfsoptlist *moptlist = mp->mnt_optnew;
1427 	va_list ap;
1428 	int error, len;
1429 	char *errmsg;
1430 
1431 	error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
1432 	if (error || errmsg == NULL || len <= 0)
1433 		return;
1434 
1435 	va_start(ap, fmt);
1436 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1437 	va_end(ap);
1438 }
1439 
1440 void
1441 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
1442 {
1443 	va_list ap;
1444 	int error, len;
1445 	char *errmsg;
1446 
1447 	error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
1448 	if (error || errmsg == NULL || len <= 0)
1449 		return;
1450 
1451 	va_start(ap, fmt);
1452 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1453 	va_end(ap);
1454 }
1455 
1456 /*
1457  * ---------------------------------------------------------------------
1458  * Functions for querying mount options/arguments from filesystems.
1459  */
1460 
1461 /*
1462  * Check that no unknown options are given
1463  */
1464 int
1465 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1466 {
1467 	struct vfsopt *opt;
1468 	char errmsg[255];
1469 	const char **t, *p, *q;
1470 	int ret = 0;
1471 
1472 	TAILQ_FOREACH(opt, opts, link) {
1473 		p = opt->name;
1474 		q = NULL;
1475 		if (p[0] == 'n' && p[1] == 'o')
1476 			q = p + 2;
1477 		for(t = global_opts; *t != NULL; t++) {
1478 			if (strcmp(*t, p) == 0)
1479 				break;
1480 			if (q != NULL) {
1481 				if (strcmp(*t, q) == 0)
1482 					break;
1483 			}
1484 		}
1485 		if (*t != NULL)
1486 			continue;
1487 		for(t = legal; *t != NULL; t++) {
1488 			if (strcmp(*t, p) == 0)
1489 				break;
1490 			if (q != NULL) {
1491 				if (strcmp(*t, q) == 0)
1492 					break;
1493 			}
1494 		}
1495 		if (*t != NULL)
1496 			continue;
1497 		snprintf(errmsg, sizeof(errmsg),
1498 		    "mount option <%s> is unknown", p);
1499 		ret = EINVAL;
1500 	}
1501 	if (ret != 0) {
1502 		TAILQ_FOREACH(opt, opts, link) {
1503 			if (strcmp(opt->name, "errmsg") == 0) {
1504 				strncpy((char *)opt->value, errmsg, opt->len);
1505 				break;
1506 			}
1507 		}
1508 		if (opt == NULL)
1509 			printf("%s\n", errmsg);
1510 	}
1511 	return (ret);
1512 }
1513 
1514 /*
1515  * Get a mount option by its name.
1516  *
1517  * Return 0 if the option was found, ENOENT otherwise.
1518  * If len is non-NULL it will be filled with the length
1519  * of the option. If buf is non-NULL, it will be filled
1520  * with the address of the option.
1521  */
1522 int
1523 vfs_getopt(opts, name, buf, len)
1524 	struct vfsoptlist *opts;
1525 	const char *name;
1526 	void **buf;
1527 	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(opts, name, dest, len)
1741 	struct vfsoptlist *opts;
1742 	const char *name;
1743 	void *dest;
1744 	int len;
1745 {
1746 	struct vfsopt *opt;
1747 
1748 	KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1749 
1750 	TAILQ_FOREACH(opt, opts, link) {
1751 		if (strcmp(name, opt->name) == 0) {
1752 			opt->seen = 1;
1753 			if (len != opt->len)
1754 				return (EINVAL);
1755 			bcopy(opt->value, dest, opt->len);
1756 			return (0);
1757 		}
1758 	}
1759 	return (ENOENT);
1760 }
1761 
1762 int
1763 __vfs_statfs(struct mount *mp, struct statfs *sbp)
1764 {
1765 	int error;
1766 
1767 	error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat);
1768 	if (sbp != &mp->mnt_stat)
1769 		*sbp = mp->mnt_stat;
1770 	return (error);
1771 }
1772 
1773 void
1774 vfs_mountedfrom(struct mount *mp, const char *from)
1775 {
1776 
1777 	bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
1778 	strlcpy(mp->mnt_stat.f_mntfromname, from,
1779 	    sizeof mp->mnt_stat.f_mntfromname);
1780 }
1781 
1782 /*
1783  * ---------------------------------------------------------------------
1784  * This is the api for building mount args and mounting filesystems from
1785  * inside the kernel.
1786  *
1787  * The API works by accumulation of individual args.  First error is
1788  * latched.
1789  *
1790  * XXX: should be documented in new manpage kernel_mount(9)
1791  */
1792 
1793 /* A memory allocation which must be freed when we are done */
1794 struct mntaarg {
1795 	SLIST_ENTRY(mntaarg)	next;
1796 };
1797 
1798 /* The header for the mount arguments */
1799 struct mntarg {
1800 	struct iovec *v;
1801 	int len;
1802 	int error;
1803 	SLIST_HEAD(, mntaarg)	list;
1804 };
1805 
1806 /*
1807  * Add a boolean argument.
1808  *
1809  * flag is the boolean value.
1810  * name must start with "no".
1811  */
1812 struct mntarg *
1813 mount_argb(struct mntarg *ma, int flag, const char *name)
1814 {
1815 
1816 	KASSERT(name[0] == 'n' && name[1] == 'o',
1817 	    ("mount_argb(...,%s): name must start with 'no'", name));
1818 
1819 	return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
1820 }
1821 
1822 /*
1823  * Add an argument printf style
1824  */
1825 struct mntarg *
1826 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
1827 {
1828 	va_list ap;
1829 	struct mntaarg *maa;
1830 	struct sbuf *sb;
1831 	int len;
1832 
1833 	if (ma == NULL) {
1834 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1835 		SLIST_INIT(&ma->list);
1836 	}
1837 	if (ma->error)
1838 		return (ma);
1839 
1840 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1841 	    M_MOUNT, M_WAITOK);
1842 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1843 	ma->v[ma->len].iov_len = strlen(name) + 1;
1844 	ma->len++;
1845 
1846 	sb = sbuf_new_auto();
1847 	va_start(ap, fmt);
1848 	sbuf_vprintf(sb, fmt, ap);
1849 	va_end(ap);
1850 	sbuf_finish(sb);
1851 	len = sbuf_len(sb) + 1;
1852 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1853 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1854 	bcopy(sbuf_data(sb), maa + 1, len);
1855 	sbuf_delete(sb);
1856 
1857 	ma->v[ma->len].iov_base = maa + 1;
1858 	ma->v[ma->len].iov_len = len;
1859 	ma->len++;
1860 
1861 	return (ma);
1862 }
1863 
1864 /*
1865  * Add an argument which is a userland string.
1866  */
1867 struct mntarg *
1868 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
1869 {
1870 	struct mntaarg *maa;
1871 	char *tbuf;
1872 
1873 	if (val == NULL)
1874 		return (ma);
1875 	if (ma == NULL) {
1876 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1877 		SLIST_INIT(&ma->list);
1878 	}
1879 	if (ma->error)
1880 		return (ma);
1881 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1882 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1883 	tbuf = (void *)(maa + 1);
1884 	ma->error = copyinstr(val, tbuf, len, NULL);
1885 	return (mount_arg(ma, name, tbuf, -1));
1886 }
1887 
1888 /*
1889  * Plain argument.
1890  *
1891  * If length is -1, treat value as a C string.
1892  */
1893 struct mntarg *
1894 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
1895 {
1896 
1897 	if (ma == NULL) {
1898 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1899 		SLIST_INIT(&ma->list);
1900 	}
1901 	if (ma->error)
1902 		return (ma);
1903 
1904 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1905 	    M_MOUNT, M_WAITOK);
1906 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1907 	ma->v[ma->len].iov_len = strlen(name) + 1;
1908 	ma->len++;
1909 
1910 	ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
1911 	if (len < 0)
1912 		ma->v[ma->len].iov_len = strlen(val) + 1;
1913 	else
1914 		ma->v[ma->len].iov_len = len;
1915 	ma->len++;
1916 	return (ma);
1917 }
1918 
1919 /*
1920  * Free a mntarg structure
1921  */
1922 static void
1923 free_mntarg(struct mntarg *ma)
1924 {
1925 	struct mntaarg *maa;
1926 
1927 	while (!SLIST_EMPTY(&ma->list)) {
1928 		maa = SLIST_FIRST(&ma->list);
1929 		SLIST_REMOVE_HEAD(&ma->list, next);
1930 		free(maa, M_MOUNT);
1931 	}
1932 	free(ma->v, M_MOUNT);
1933 	free(ma, M_MOUNT);
1934 }
1935 
1936 /*
1937  * Mount a filesystem
1938  */
1939 int
1940 kernel_mount(struct mntarg *ma, uint64_t flags)
1941 {
1942 	struct uio auio;
1943 	int error;
1944 
1945 	KASSERT(ma != NULL, ("kernel_mount NULL ma"));
1946 	KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
1947 	KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
1948 
1949 	auio.uio_iov = ma->v;
1950 	auio.uio_iovcnt = ma->len;
1951 	auio.uio_segflg = UIO_SYSSPACE;
1952 
1953 	error = ma->error;
1954 	if (!error)
1955 		error = vfs_donmount(curthread, flags, &auio);
1956 	free_mntarg(ma);
1957 	return (error);
1958 }
1959 
1960 /*
1961  * A printflike function to mount a filesystem.
1962  */
1963 int
1964 kernel_vmount(int flags, ...)
1965 {
1966 	struct mntarg *ma = NULL;
1967 	va_list ap;
1968 	const char *cp;
1969 	const void *vp;
1970 	int error;
1971 
1972 	va_start(ap, flags);
1973 	for (;;) {
1974 		cp = va_arg(ap, const char *);
1975 		if (cp == NULL)
1976 			break;
1977 		vp = va_arg(ap, const void *);
1978 		ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0));
1979 	}
1980 	va_end(ap);
1981 
1982 	error = kernel_mount(ma, flags);
1983 	return (error);
1984 }
1985 
1986 void
1987 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp)
1988 {
1989 
1990 	bcopy(oexp, exp, sizeof(*oexp));
1991 	exp->ex_numsecflavors = 0;
1992 }
1993