xref: /freebsd/sys/kern/vfs_mount.c (revision 595e514d0df2bac5b813d35f83e32875dbf16a83)
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 	mp->mnt_op = vfsp->vfc_vfsops;
467 	mp->mnt_vfc = vfsp;
468 	vfsp->vfc_refcount++;	/* XXX Unlocked */
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 	mp->mnt_vfc->vfc_refcount--;
509 	if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
510 		struct vnode *vp;
511 
512 		TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
513 			vprint("", vp);
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 	}
653 
654 	/*
655 	 * Be ultra-paranoid about making sure the type and fspath
656 	 * variables will fit in our mp buffers, including the
657 	 * terminating NUL.
658 	 */
659 	if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) {
660 		error = ENAMETOOLONG;
661 		goto bail;
662 	}
663 
664 	error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
665 bail:
666 	/* copyout the errmsg */
667 	if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
668 	    && errmsg_len > 0 && errmsg != NULL) {
669 		if (fsoptions->uio_segflg == UIO_SYSSPACE) {
670 			bcopy(errmsg,
671 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
672 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
673 		} else {
674 			copyout(errmsg,
675 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
676 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
677 		}
678 	}
679 
680 	if (optlist != NULL)
681 		vfs_freeopts(optlist);
682 	return (error);
683 }
684 
685 /*
686  * Old mount API.
687  */
688 #ifndef _SYS_SYSPROTO_H_
689 struct mount_args {
690 	char	*type;
691 	char	*path;
692 	int	flags;
693 	caddr_t	data;
694 };
695 #endif
696 /* ARGSUSED */
697 int
698 sys_mount(td, uap)
699 	struct thread *td;
700 	struct mount_args /* {
701 		char *type;
702 		char *path;
703 		int flags;
704 		caddr_t data;
705 	} */ *uap;
706 {
707 	char *fstype;
708 	struct vfsconf *vfsp = NULL;
709 	struct mntarg *ma = NULL;
710 	uint64_t flags;
711 	int error;
712 
713 	/*
714 	 * Mount flags are now 64-bits. On 32-bit architectures only
715 	 * 32-bits are passed in, but from here on everything handles
716 	 * 64-bit flags correctly.
717 	 */
718 	flags = uap->flags;
719 
720 	AUDIT_ARG_FFLAGS(flags);
721 
722 	/*
723 	 * Filter out MNT_ROOTFS.  We do not want clients of mount() in
724 	 * userspace to set this flag, but we must filter it out if we want
725 	 * MNT_UPDATE on the root file system to work.
726 	 * MNT_ROOTFS should only be set by the kernel when mounting its
727 	 * root file system.
728 	 */
729 	flags &= ~MNT_ROOTFS;
730 
731 	fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
732 	error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
733 	if (error) {
734 		free(fstype, M_TEMP);
735 		return (error);
736 	}
737 
738 	AUDIT_ARG_TEXT(fstype);
739 	mtx_lock(&Giant);
740 	vfsp = vfs_byname_kld(fstype, td, &error);
741 	free(fstype, M_TEMP);
742 	if (vfsp == NULL) {
743 		mtx_unlock(&Giant);
744 		return (ENOENT);
745 	}
746 	if (vfsp->vfc_vfsops->vfs_cmount == NULL) {
747 		mtx_unlock(&Giant);
748 		return (EOPNOTSUPP);
749 	}
750 
751 	ma = mount_argsu(ma, "fstype", uap->type, MNAMELEN);
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 	mtx_unlock(&Giant);
759 	return (error);
760 }
761 
762 /*
763  * vfs_domount_first(): first file system mount (not update)
764  */
765 static int
766 vfs_domount_first(
767 	struct thread *td,		/* Calling thread. */
768 	struct vfsconf *vfsp,		/* File system type. */
769 	char *fspath,			/* Mount path. */
770 	struct vnode *vp,		/* Vnode to be covered. */
771 	uint64_t fsflags,		/* Flags common to all filesystems. */
772 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
773 	)
774 {
775 	struct vattr va;
776 	struct mount *mp;
777 	struct vnode *newdp;
778 	int error;
779 
780 	mtx_assert(&Giant, MA_OWNED);
781 	ASSERT_VOP_ELOCKED(vp, __func__);
782 	KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
783 
784 	/*
785 	 * If the user is not root, ensure that they own the directory
786 	 * onto which we are attempting to mount.
787 	 */
788 	error = VOP_GETATTR(vp, &va, td->td_ucred);
789 	if (error == 0 && va.va_uid != td->td_ucred->cr_uid)
790 		error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN, 0);
791 	if (error == 0)
792 		error = vinvalbuf(vp, V_SAVE, 0, 0);
793 	if (error == 0 && vp->v_type != VDIR)
794 		error = ENOTDIR;
795 	if (error == 0) {
796 		VI_LOCK(vp);
797 		if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
798 			vp->v_iflag |= VI_MOUNT;
799 		else
800 			error = EBUSY;
801 		VI_UNLOCK(vp);
802 	}
803 	if (error != 0) {
804 		vput(vp);
805 		return (error);
806 	}
807 	VOP_UNLOCK(vp, 0);
808 
809 	/* Allocate and initialize the filesystem. */
810 	mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred);
811 	/* XXXMAC: pass to vfs_mount_alloc? */
812 	mp->mnt_optnew = *optlist;
813 	/* Set the mount level flags. */
814 	mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY));
815 
816 	/*
817 	 * Mount the filesystem.
818 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
819 	 * get.  No freeing of cn_pnbuf.
820 	 */
821 	error = VFS_MOUNT(mp);
822 	if (error != 0) {
823 		vfs_unbusy(mp);
824 		vfs_mount_destroy(mp);
825 		VI_LOCK(vp);
826 		vp->v_iflag &= ~VI_MOUNT;
827 		VI_UNLOCK(vp);
828 		vrele(vp);
829 		return (error);
830 	}
831 
832 	if (mp->mnt_opt != NULL)
833 		vfs_freeopts(mp->mnt_opt);
834 	mp->mnt_opt = mp->mnt_optnew;
835 	*optlist = NULL;
836 	(void)VFS_STATFS(mp, &mp->mnt_stat);
837 
838 	/*
839 	 * Prevent external consumers of mount options from reading mnt_optnew.
840 	 */
841 	mp->mnt_optnew = NULL;
842 
843 	MNT_ILOCK(mp);
844 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
845 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
846 		mp->mnt_kern_flag |= MNTK_ASYNC;
847 	else
848 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
849 	MNT_IUNLOCK(mp);
850 
851 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
852 	cache_purge(vp);
853 	VI_LOCK(vp);
854 	vp->v_iflag &= ~VI_MOUNT;
855 	VI_UNLOCK(vp);
856 	vp->v_mountedhere = mp;
857 	/* Place the new filesystem at the end of the mount list. */
858 	mtx_lock(&mountlist_mtx);
859 	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
860 	mtx_unlock(&mountlist_mtx);
861 	vfs_event_signal(NULL, VQ_MOUNT, 0);
862 	if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp))
863 		panic("mount: lost mount");
864 	VOP_UNLOCK(newdp, 0);
865 	VOP_UNLOCK(vp, 0);
866 	mountcheckdirs(vp, newdp);
867 	vrele(newdp);
868 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
869 		vfs_allocate_syncvnode(mp);
870 	vfs_unbusy(mp);
871 	return (0);
872 }
873 
874 /*
875  * vfs_domount_update(): update of mounted file system
876  */
877 static int
878 vfs_domount_update(
879 	struct thread *td,		/* Calling thread. */
880 	struct vnode *vp,		/* Mount point vnode. */
881 	uint64_t fsflags,		/* Flags common to all filesystems. */
882 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
883 	)
884 {
885 	struct oexport_args oexport;
886 	struct export_args export;
887 	struct mount *mp;
888 	int error, export_error;
889 	uint64_t flag;
890 
891 	mtx_assert(&Giant, MA_OWNED);
892 	ASSERT_VOP_ELOCKED(vp, __func__);
893 	KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
894 
895 	if ((vp->v_vflag & VV_ROOT) == 0) {
896 		vput(vp);
897 		return (EINVAL);
898 	}
899 	mp = vp->v_mount;
900 	/*
901 	 * We only allow the filesystem to be reloaded if it
902 	 * is currently mounted read-only.
903 	 */
904 	flag = mp->mnt_flag;
905 	if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
906 		vput(vp);
907 		return (EOPNOTSUPP);	/* Needs translation */
908 	}
909 	/*
910 	 * Only privileged root, or (if MNT_USER is set) the user that
911 	 * did the original mount is permitted to update it.
912 	 */
913 	error = vfs_suser(mp, td);
914 	if (error != 0) {
915 		vput(vp);
916 		return (error);
917 	}
918 	if (vfs_busy(mp, MBF_NOWAIT)) {
919 		vput(vp);
920 		return (EBUSY);
921 	}
922 	VI_LOCK(vp);
923 	if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
924 		VI_UNLOCK(vp);
925 		vfs_unbusy(mp);
926 		vput(vp);
927 		return (EBUSY);
928 	}
929 	vp->v_iflag |= VI_MOUNT;
930 	VI_UNLOCK(vp);
931 	VOP_UNLOCK(vp, 0);
932 
933 	MNT_ILOCK(mp);
934 	mp->mnt_flag &= ~MNT_UPDATEMASK;
935 	mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE |
936 	    MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY);
937 	if ((mp->mnt_flag & MNT_ASYNC) == 0)
938 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
939 	MNT_IUNLOCK(mp);
940 	mp->mnt_optnew = *optlist;
941 	vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
942 
943 	/*
944 	 * Mount the filesystem.
945 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
946 	 * get.  No freeing of cn_pnbuf.
947 	 */
948 	error = VFS_MOUNT(mp);
949 
950 	export_error = 0;
951 	if (error == 0) {
952 		/* Process the export option. */
953 		if (vfs_copyopt(mp->mnt_optnew, "export", &export,
954 		    sizeof(export)) == 0) {
955 			export_error = vfs_export(mp, &export);
956 		} else if (vfs_copyopt(mp->mnt_optnew, "export", &oexport,
957 		    sizeof(oexport)) == 0) {
958 			export.ex_flags = oexport.ex_flags;
959 			export.ex_root = oexport.ex_root;
960 			export.ex_anon = oexport.ex_anon;
961 			export.ex_addr = oexport.ex_addr;
962 			export.ex_addrlen = oexport.ex_addrlen;
963 			export.ex_mask = oexport.ex_mask;
964 			export.ex_masklen = oexport.ex_masklen;
965 			export.ex_indexfile = oexport.ex_indexfile;
966 			export.ex_numsecflavors = 0;
967 			export_error = vfs_export(mp, &export);
968 		}
969 	}
970 
971 	MNT_ILOCK(mp);
972 	if (error == 0) {
973 		mp->mnt_flag &=	~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
974 		    MNT_SNAPSHOT);
975 	} else {
976 		/*
977 		 * If we fail, restore old mount flags. MNT_QUOTA is special,
978 		 * because it is not part of MNT_UPDATEMASK, but it could have
979 		 * changed in the meantime if quotactl(2) was called.
980 		 * All in all we want current value of MNT_QUOTA, not the old
981 		 * one.
982 		 */
983 		mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
984 	}
985 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
986 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
987 		mp->mnt_kern_flag |= MNTK_ASYNC;
988 	else
989 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
990 	MNT_IUNLOCK(mp);
991 
992 	if (error != 0)
993 		goto end;
994 
995 	if (mp->mnt_opt != NULL)
996 		vfs_freeopts(mp->mnt_opt);
997 	mp->mnt_opt = mp->mnt_optnew;
998 	*optlist = NULL;
999 	(void)VFS_STATFS(mp, &mp->mnt_stat);
1000 	/*
1001 	 * Prevent external consumers of mount options from reading
1002 	 * mnt_optnew.
1003 	 */
1004 	mp->mnt_optnew = NULL;
1005 
1006 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
1007 		vfs_allocate_syncvnode(mp);
1008 	else
1009 		vfs_deallocate_syncvnode(mp);
1010 end:
1011 	vfs_unbusy(mp);
1012 	VI_LOCK(vp);
1013 	vp->v_iflag &= ~VI_MOUNT;
1014 	VI_UNLOCK(vp);
1015 	vrele(vp);
1016 	return (error != 0 ? error : export_error);
1017 }
1018 
1019 /*
1020  * vfs_domount(): actually attempt a filesystem mount.
1021  */
1022 static int
1023 vfs_domount(
1024 	struct thread *td,		/* Calling thread. */
1025 	const char *fstype,		/* Filesystem type. */
1026 	char *fspath,			/* Mount path. */
1027 	uint64_t fsflags,		/* Flags common to all filesystems. */
1028 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
1029 	)
1030 {
1031 	struct vfsconf *vfsp;
1032 	struct nameidata nd;
1033 	struct vnode *vp;
1034 	char *pathbuf;
1035 	int error;
1036 
1037 	/*
1038 	 * Be ultra-paranoid about making sure the type and fspath
1039 	 * variables will fit in our mp buffers, including the
1040 	 * terminating NUL.
1041 	 */
1042 	if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
1043 		return (ENAMETOOLONG);
1044 
1045 	if (jailed(td->td_ucred) || usermount == 0) {
1046 		if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
1047 			return (error);
1048 	}
1049 
1050 	/*
1051 	 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
1052 	 */
1053 	if (fsflags & MNT_EXPORTED) {
1054 		error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
1055 		if (error)
1056 			return (error);
1057 	}
1058 	if (fsflags & MNT_SUIDDIR) {
1059 		error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
1060 		if (error)
1061 			return (error);
1062 	}
1063 	/*
1064 	 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
1065 	 */
1066 	if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
1067 		if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
1068 			fsflags |= MNT_NOSUID | MNT_USER;
1069 	}
1070 
1071 	/* Load KLDs before we lock the covered vnode to avoid reversals. */
1072 	vfsp = NULL;
1073 	if ((fsflags & MNT_UPDATE) == 0) {
1074 		/* Don't try to load KLDs if we're mounting the root. */
1075 		if (fsflags & MNT_ROOTFS)
1076 			vfsp = vfs_byname(fstype);
1077 		else
1078 			vfsp = vfs_byname_kld(fstype, td, &error);
1079 		if (vfsp == NULL)
1080 			return (ENODEV);
1081 		if (jailed(td->td_ucred) && !(vfsp->vfc_flags & VFCF_JAIL))
1082 			return (EPERM);
1083 	}
1084 
1085 	/*
1086 	 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
1087 	 */
1088 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1089 	    UIO_SYSSPACE, fspath, td);
1090 	error = namei(&nd);
1091 	if (error != 0)
1092 		return (error);
1093 	mtx_lock(&Giant);
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 	mtx_unlock(&Giant);
1109 
1110 	ASSERT_VI_UNLOCKED(vp, __func__);
1111 	ASSERT_VOP_UNLOCKED(vp, __func__);
1112 
1113 	return (error);
1114 }
1115 
1116 /*
1117  * Unmount a filesystem.
1118  *
1119  * Note: unmount takes a path to the vnode mounted on as argument, not
1120  * special file (as before).
1121  */
1122 #ifndef _SYS_SYSPROTO_H_
1123 struct unmount_args {
1124 	char	*path;
1125 	int	flags;
1126 };
1127 #endif
1128 /* ARGSUSED */
1129 int
1130 sys_unmount(td, uap)
1131 	struct thread *td;
1132 	register struct unmount_args /* {
1133 		char *path;
1134 		int flags;
1135 	} */ *uap;
1136 {
1137 	struct nameidata nd;
1138 	struct mount *mp;
1139 	char *pathbuf;
1140 	int error, id0, id1;
1141 
1142 	AUDIT_ARG_VALUE(uap->flags);
1143 	if (jailed(td->td_ucred) || usermount == 0) {
1144 		error = priv_check(td, PRIV_VFS_UNMOUNT);
1145 		if (error)
1146 			return (error);
1147 	}
1148 
1149 	pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1150 	error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1151 	if (error) {
1152 		free(pathbuf, M_TEMP);
1153 		return (error);
1154 	}
1155 	mtx_lock(&Giant);
1156 	if (uap->flags & MNT_BYFSID) {
1157 		AUDIT_ARG_TEXT(pathbuf);
1158 		/* Decode the filesystem ID. */
1159 		if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
1160 			mtx_unlock(&Giant);
1161 			free(pathbuf, M_TEMP);
1162 			return (EINVAL);
1163 		}
1164 
1165 		mtx_lock(&mountlist_mtx);
1166 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1167 			if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1168 			    mp->mnt_stat.f_fsid.val[1] == id1)
1169 				break;
1170 		}
1171 		mtx_unlock(&mountlist_mtx);
1172 	} else {
1173 		/*
1174 		 * Try to find global path for path argument.
1175 		 */
1176 		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1177 		    UIO_SYSSPACE, pathbuf, td);
1178 		if (namei(&nd) == 0) {
1179 			NDFREE(&nd, NDF_ONLY_PNBUF);
1180 			error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
1181 			    MNAMELEN);
1182 			if (error == 0 || error == ENODEV)
1183 				vput(nd.ni_vp);
1184 		}
1185 		mtx_lock(&mountlist_mtx);
1186 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1187 			if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0)
1188 				break;
1189 		}
1190 		mtx_unlock(&mountlist_mtx);
1191 	}
1192 	free(pathbuf, M_TEMP);
1193 	if (mp == NULL) {
1194 		/*
1195 		 * Previously we returned ENOENT for a nonexistent path and
1196 		 * EINVAL for a non-mountpoint.  We cannot tell these apart
1197 		 * now, so in the !MNT_BYFSID case return the more likely
1198 		 * EINVAL for compatibility.
1199 		 */
1200 		mtx_unlock(&Giant);
1201 		return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1202 	}
1203 
1204 	/*
1205 	 * Don't allow unmounting the root filesystem.
1206 	 */
1207 	if (mp->mnt_flag & MNT_ROOTFS) {
1208 		mtx_unlock(&Giant);
1209 		return (EINVAL);
1210 	}
1211 	error = dounmount(mp, uap->flags, td);
1212 	mtx_unlock(&Giant);
1213 	return (error);
1214 }
1215 
1216 /*
1217  * Do the actual filesystem unmount.
1218  */
1219 int
1220 dounmount(mp, flags, td)
1221 	struct mount *mp;
1222 	int flags;
1223 	struct thread *td;
1224 {
1225 	struct vnode *coveredvp, *fsrootvp;
1226 	int error;
1227 	uint64_t async_flag;
1228 	int mnt_gen_r;
1229 
1230 	mtx_assert(&Giant, MA_OWNED);
1231 
1232 	if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
1233 		mnt_gen_r = mp->mnt_gen;
1234 		VI_LOCK(coveredvp);
1235 		vholdl(coveredvp);
1236 		vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
1237 		vdrop(coveredvp);
1238 		/*
1239 		 * Check for mp being unmounted while waiting for the
1240 		 * covered vnode lock.
1241 		 */
1242 		if (coveredvp->v_mountedhere != mp ||
1243 		    coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
1244 			VOP_UNLOCK(coveredvp, 0);
1245 			return (EBUSY);
1246 		}
1247 	}
1248 	/*
1249 	 * Only privileged root, or (if MNT_USER is set) the user that did the
1250 	 * original mount is permitted to unmount this filesystem.
1251 	 */
1252 	error = vfs_suser(mp, td);
1253 	if (error) {
1254 		if (coveredvp)
1255 			VOP_UNLOCK(coveredvp, 0);
1256 		return (error);
1257 	}
1258 
1259 	vn_start_write(NULL, &mp, V_WAIT);
1260 	MNT_ILOCK(mp);
1261 	if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 ||
1262 	    !TAILQ_EMPTY(&mp->mnt_uppers)) {
1263 		MNT_IUNLOCK(mp);
1264 		if (coveredvp)
1265 			VOP_UNLOCK(coveredvp, 0);
1266 		vn_finished_write(mp);
1267 		return (EBUSY);
1268 	}
1269 	mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ;
1270 	/* Allow filesystems to detect that a forced unmount is in progress. */
1271 	if (flags & MNT_FORCE)
1272 		mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1273 	error = 0;
1274 	if (mp->mnt_lockref) {
1275 		mp->mnt_kern_flag |= MNTK_DRAINING;
1276 		error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
1277 		    "mount drain", 0);
1278 	}
1279 	MNT_IUNLOCK(mp);
1280 	KASSERT(mp->mnt_lockref == 0,
1281 	    ("%s: invalid lock refcount in the drain path @ %s:%d",
1282 	    __func__, __FILE__, __LINE__));
1283 	KASSERT(error == 0,
1284 	    ("%s: invalid return value for msleep in the drain path @ %s:%d",
1285 	    __func__, __FILE__, __LINE__));
1286 
1287 	if (mp->mnt_flag & MNT_EXPUBLIC)
1288 		vfs_setpublicfs(NULL, NULL, NULL);
1289 
1290 	vfs_msync(mp, MNT_WAIT);
1291 	MNT_ILOCK(mp);
1292 	async_flag = mp->mnt_flag & MNT_ASYNC;
1293 	mp->mnt_flag &= ~MNT_ASYNC;
1294 	mp->mnt_kern_flag &= ~MNTK_ASYNC;
1295 	MNT_IUNLOCK(mp);
1296 	cache_purgevfs(mp);	/* remove cache entries for this file sys */
1297 	vfs_deallocate_syncvnode(mp);
1298 	/*
1299 	 * For forced unmounts, move process cdir/rdir refs on the fs root
1300 	 * vnode to the covered vnode.  For non-forced unmounts we want
1301 	 * such references to cause an EBUSY error.
1302 	 */
1303 	if ((flags & MNT_FORCE) &&
1304 	    VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1305 		if (mp->mnt_vnodecovered != NULL)
1306 			mountcheckdirs(fsrootvp, mp->mnt_vnodecovered);
1307 		if (fsrootvp == rootvnode) {
1308 			vrele(rootvnode);
1309 			rootvnode = NULL;
1310 		}
1311 		vput(fsrootvp);
1312 	}
1313 	if (((mp->mnt_flag & MNT_RDONLY) ||
1314 	     (error = VFS_SYNC(mp, MNT_WAIT)) == 0) || (flags & MNT_FORCE) != 0)
1315 		error = VFS_UNMOUNT(mp, flags);
1316 	vn_finished_write(mp);
1317 	/*
1318 	 * If we failed to flush the dirty blocks for this mount point,
1319 	 * undo all the cdir/rdir and rootvnode changes we made above.
1320 	 * Unless we failed to do so because the device is reporting that
1321 	 * it doesn't exist anymore.
1322 	 */
1323 	if (error && error != ENXIO) {
1324 		if ((flags & MNT_FORCE) &&
1325 		    VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1326 			if (mp->mnt_vnodecovered != NULL)
1327 				mountcheckdirs(mp->mnt_vnodecovered, fsrootvp);
1328 			if (rootvnode == NULL) {
1329 				rootvnode = fsrootvp;
1330 				vref(rootvnode);
1331 			}
1332 			vput(fsrootvp);
1333 		}
1334 		MNT_ILOCK(mp);
1335 		mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ;
1336 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
1337 			MNT_IUNLOCK(mp);
1338 			vfs_allocate_syncvnode(mp);
1339 			MNT_ILOCK(mp);
1340 		}
1341 		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1342 		mp->mnt_flag |= async_flag;
1343 		if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1344 		    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1345 			mp->mnt_kern_flag |= MNTK_ASYNC;
1346 		if (mp->mnt_kern_flag & MNTK_MWAIT) {
1347 			mp->mnt_kern_flag &= ~MNTK_MWAIT;
1348 			wakeup(mp);
1349 		}
1350 		MNT_IUNLOCK(mp);
1351 		if (coveredvp)
1352 			VOP_UNLOCK(coveredvp, 0);
1353 		return (error);
1354 	}
1355 	mtx_lock(&mountlist_mtx);
1356 	TAILQ_REMOVE(&mountlist, mp, mnt_list);
1357 	mtx_unlock(&mountlist_mtx);
1358 	if (coveredvp != NULL) {
1359 		coveredvp->v_mountedhere = NULL;
1360 		vput(coveredvp);
1361 	}
1362 	vfs_event_signal(NULL, VQ_UNMOUNT, 0);
1363 	vfs_mount_destroy(mp);
1364 	return (0);
1365 }
1366 
1367 /*
1368  * Report errors during filesystem mounting.
1369  */
1370 void
1371 vfs_mount_error(struct mount *mp, const char *fmt, ...)
1372 {
1373 	struct vfsoptlist *moptlist = mp->mnt_optnew;
1374 	va_list ap;
1375 	int error, len;
1376 	char *errmsg;
1377 
1378 	error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
1379 	if (error || errmsg == NULL || len <= 0)
1380 		return;
1381 
1382 	va_start(ap, fmt);
1383 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1384 	va_end(ap);
1385 }
1386 
1387 void
1388 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
1389 {
1390 	va_list ap;
1391 	int error, len;
1392 	char *errmsg;
1393 
1394 	error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
1395 	if (error || errmsg == NULL || len <= 0)
1396 		return;
1397 
1398 	va_start(ap, fmt);
1399 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1400 	va_end(ap);
1401 }
1402 
1403 /*
1404  * ---------------------------------------------------------------------
1405  * Functions for querying mount options/arguments from filesystems.
1406  */
1407 
1408 /*
1409  * Check that no unknown options are given
1410  */
1411 int
1412 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1413 {
1414 	struct vfsopt *opt;
1415 	char errmsg[255];
1416 	const char **t, *p, *q;
1417 	int ret = 0;
1418 
1419 	TAILQ_FOREACH(opt, opts, link) {
1420 		p = opt->name;
1421 		q = NULL;
1422 		if (p[0] == 'n' && p[1] == 'o')
1423 			q = p + 2;
1424 		for(t = global_opts; *t != NULL; t++) {
1425 			if (strcmp(*t, p) == 0)
1426 				break;
1427 			if (q != NULL) {
1428 				if (strcmp(*t, q) == 0)
1429 					break;
1430 			}
1431 		}
1432 		if (*t != NULL)
1433 			continue;
1434 		for(t = legal; *t != NULL; t++) {
1435 			if (strcmp(*t, p) == 0)
1436 				break;
1437 			if (q != NULL) {
1438 				if (strcmp(*t, q) == 0)
1439 					break;
1440 			}
1441 		}
1442 		if (*t != NULL)
1443 			continue;
1444 		snprintf(errmsg, sizeof(errmsg),
1445 		    "mount option <%s> is unknown", p);
1446 		ret = EINVAL;
1447 	}
1448 	if (ret != 0) {
1449 		TAILQ_FOREACH(opt, opts, link) {
1450 			if (strcmp(opt->name, "errmsg") == 0) {
1451 				strncpy((char *)opt->value, errmsg, opt->len);
1452 				break;
1453 			}
1454 		}
1455 		if (opt == NULL)
1456 			printf("%s\n", errmsg);
1457 	}
1458 	return (ret);
1459 }
1460 
1461 /*
1462  * Get a mount option by its name.
1463  *
1464  * Return 0 if the option was found, ENOENT otherwise.
1465  * If len is non-NULL it will be filled with the length
1466  * of the option. If buf is non-NULL, it will be filled
1467  * with the address of the option.
1468  */
1469 int
1470 vfs_getopt(opts, name, buf, len)
1471 	struct vfsoptlist *opts;
1472 	const char *name;
1473 	void **buf;
1474 	int *len;
1475 {
1476 	struct vfsopt *opt;
1477 
1478 	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1479 
1480 	TAILQ_FOREACH(opt, opts, link) {
1481 		if (strcmp(name, opt->name) == 0) {
1482 			opt->seen = 1;
1483 			if (len != NULL)
1484 				*len = opt->len;
1485 			if (buf != NULL)
1486 				*buf = opt->value;
1487 			return (0);
1488 		}
1489 	}
1490 	return (ENOENT);
1491 }
1492 
1493 int
1494 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
1495 {
1496 	struct vfsopt *opt;
1497 
1498 	if (opts == NULL)
1499 		return (-1);
1500 
1501 	TAILQ_FOREACH(opt, opts, link) {
1502 		if (strcmp(name, opt->name) == 0) {
1503 			opt->seen = 1;
1504 			return (opt->pos);
1505 		}
1506 	}
1507 	return (-1);
1508 }
1509 
1510 int
1511 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value)
1512 {
1513 	char *opt_value, *vtp;
1514 	quad_t iv;
1515 	int error, opt_len;
1516 
1517 	error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len);
1518 	if (error != 0)
1519 		return (error);
1520 	if (opt_len == 0 || opt_value == NULL)
1521 		return (EINVAL);
1522 	if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0')
1523 		return (EINVAL);
1524 	iv = strtoq(opt_value, &vtp, 0);
1525 	if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0'))
1526 		return (EINVAL);
1527 	if (iv < 0)
1528 		return (EINVAL);
1529 	switch (vtp[0]) {
1530 	case 't':
1531 	case 'T':
1532 		iv *= 1024;
1533 	case 'g':
1534 	case 'G':
1535 		iv *= 1024;
1536 	case 'm':
1537 	case 'M':
1538 		iv *= 1024;
1539 	case 'k':
1540 	case 'K':
1541 		iv *= 1024;
1542 	case '\0':
1543 		break;
1544 	default:
1545 		return (EINVAL);
1546 	}
1547 	*value = iv;
1548 
1549 	return (0);
1550 }
1551 
1552 char *
1553 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
1554 {
1555 	struct vfsopt *opt;
1556 
1557 	*error = 0;
1558 	TAILQ_FOREACH(opt, opts, link) {
1559 		if (strcmp(name, opt->name) != 0)
1560 			continue;
1561 		opt->seen = 1;
1562 		if (opt->len == 0 ||
1563 		    ((char *)opt->value)[opt->len - 1] != '\0') {
1564 			*error = EINVAL;
1565 			return (NULL);
1566 		}
1567 		return (opt->value);
1568 	}
1569 	*error = ENOENT;
1570 	return (NULL);
1571 }
1572 
1573 int
1574 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
1575 	uint64_t val)
1576 {
1577 	struct vfsopt *opt;
1578 
1579 	TAILQ_FOREACH(opt, opts, link) {
1580 		if (strcmp(name, opt->name) == 0) {
1581 			opt->seen = 1;
1582 			if (w != NULL)
1583 				*w |= val;
1584 			return (1);
1585 		}
1586 	}
1587 	if (w != NULL)
1588 		*w &= ~val;
1589 	return (0);
1590 }
1591 
1592 int
1593 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
1594 {
1595 	va_list ap;
1596 	struct vfsopt *opt;
1597 	int ret;
1598 
1599 	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1600 
1601 	TAILQ_FOREACH(opt, opts, link) {
1602 		if (strcmp(name, opt->name) != 0)
1603 			continue;
1604 		opt->seen = 1;
1605 		if (opt->len == 0 || opt->value == NULL)
1606 			return (0);
1607 		if (((char *)opt->value)[opt->len - 1] != '\0')
1608 			return (0);
1609 		va_start(ap, fmt);
1610 		ret = vsscanf(opt->value, fmt, ap);
1611 		va_end(ap);
1612 		return (ret);
1613 	}
1614 	return (0);
1615 }
1616 
1617 int
1618 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len)
1619 {
1620 	struct vfsopt *opt;
1621 
1622 	TAILQ_FOREACH(opt, opts, link) {
1623 		if (strcmp(name, opt->name) != 0)
1624 			continue;
1625 		opt->seen = 1;
1626 		if (opt->value == NULL)
1627 			opt->len = len;
1628 		else {
1629 			if (opt->len != len)
1630 				return (EINVAL);
1631 			bcopy(value, opt->value, len);
1632 		}
1633 		return (0);
1634 	}
1635 	return (ENOENT);
1636 }
1637 
1638 int
1639 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len)
1640 {
1641 	struct vfsopt *opt;
1642 
1643 	TAILQ_FOREACH(opt, opts, link) {
1644 		if (strcmp(name, opt->name) != 0)
1645 			continue;
1646 		opt->seen = 1;
1647 		if (opt->value == NULL)
1648 			opt->len = len;
1649 		else {
1650 			if (opt->len < len)
1651 				return (EINVAL);
1652 			opt->len = len;
1653 			bcopy(value, opt->value, len);
1654 		}
1655 		return (0);
1656 	}
1657 	return (ENOENT);
1658 }
1659 
1660 int
1661 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value)
1662 {
1663 	struct vfsopt *opt;
1664 
1665 	TAILQ_FOREACH(opt, opts, link) {
1666 		if (strcmp(name, opt->name) != 0)
1667 			continue;
1668 		opt->seen = 1;
1669 		if (opt->value == NULL)
1670 			opt->len = strlen(value) + 1;
1671 		else if (strlcpy(opt->value, value, opt->len) >= opt->len)
1672 			return (EINVAL);
1673 		return (0);
1674 	}
1675 	return (ENOENT);
1676 }
1677 
1678 /*
1679  * Find and copy a mount option.
1680  *
1681  * The size of the buffer has to be specified
1682  * in len, if it is not the same length as the
1683  * mount option, EINVAL is returned.
1684  * Returns ENOENT if the option is not found.
1685  */
1686 int
1687 vfs_copyopt(opts, name, dest, len)
1688 	struct vfsoptlist *opts;
1689 	const char *name;
1690 	void *dest;
1691 	int len;
1692 {
1693 	struct vfsopt *opt;
1694 
1695 	KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1696 
1697 	TAILQ_FOREACH(opt, opts, link) {
1698 		if (strcmp(name, opt->name) == 0) {
1699 			opt->seen = 1;
1700 			if (len != opt->len)
1701 				return (EINVAL);
1702 			bcopy(opt->value, dest, opt->len);
1703 			return (0);
1704 		}
1705 	}
1706 	return (ENOENT);
1707 }
1708 
1709 int
1710 __vfs_statfs(struct mount *mp, struct statfs *sbp)
1711 {
1712 	int error;
1713 
1714 	error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat);
1715 	if (sbp != &mp->mnt_stat)
1716 		*sbp = mp->mnt_stat;
1717 	return (error);
1718 }
1719 
1720 void
1721 vfs_mountedfrom(struct mount *mp, const char *from)
1722 {
1723 
1724 	bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
1725 	strlcpy(mp->mnt_stat.f_mntfromname, from,
1726 	    sizeof mp->mnt_stat.f_mntfromname);
1727 }
1728 
1729 /*
1730  * ---------------------------------------------------------------------
1731  * This is the api for building mount args and mounting filesystems from
1732  * inside the kernel.
1733  *
1734  * The API works by accumulation of individual args.  First error is
1735  * latched.
1736  *
1737  * XXX: should be documented in new manpage kernel_mount(9)
1738  */
1739 
1740 /* A memory allocation which must be freed when we are done */
1741 struct mntaarg {
1742 	SLIST_ENTRY(mntaarg)	next;
1743 };
1744 
1745 /* The header for the mount arguments */
1746 struct mntarg {
1747 	struct iovec *v;
1748 	int len;
1749 	int error;
1750 	SLIST_HEAD(, mntaarg)	list;
1751 };
1752 
1753 /*
1754  * Add a boolean argument.
1755  *
1756  * flag is the boolean value.
1757  * name must start with "no".
1758  */
1759 struct mntarg *
1760 mount_argb(struct mntarg *ma, int flag, const char *name)
1761 {
1762 
1763 	KASSERT(name[0] == 'n' && name[1] == 'o',
1764 	    ("mount_argb(...,%s): name must start with 'no'", name));
1765 
1766 	return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
1767 }
1768 
1769 /*
1770  * Add an argument printf style
1771  */
1772 struct mntarg *
1773 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
1774 {
1775 	va_list ap;
1776 	struct mntaarg *maa;
1777 	struct sbuf *sb;
1778 	int len;
1779 
1780 	if (ma == NULL) {
1781 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1782 		SLIST_INIT(&ma->list);
1783 	}
1784 	if (ma->error)
1785 		return (ma);
1786 
1787 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1788 	    M_MOUNT, M_WAITOK);
1789 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1790 	ma->v[ma->len].iov_len = strlen(name) + 1;
1791 	ma->len++;
1792 
1793 	sb = sbuf_new_auto();
1794 	va_start(ap, fmt);
1795 	sbuf_vprintf(sb, fmt, ap);
1796 	va_end(ap);
1797 	sbuf_finish(sb);
1798 	len = sbuf_len(sb) + 1;
1799 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1800 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1801 	bcopy(sbuf_data(sb), maa + 1, len);
1802 	sbuf_delete(sb);
1803 
1804 	ma->v[ma->len].iov_base = maa + 1;
1805 	ma->v[ma->len].iov_len = len;
1806 	ma->len++;
1807 
1808 	return (ma);
1809 }
1810 
1811 /*
1812  * Add an argument which is a userland string.
1813  */
1814 struct mntarg *
1815 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
1816 {
1817 	struct mntaarg *maa;
1818 	char *tbuf;
1819 
1820 	if (val == NULL)
1821 		return (ma);
1822 	if (ma == NULL) {
1823 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1824 		SLIST_INIT(&ma->list);
1825 	}
1826 	if (ma->error)
1827 		return (ma);
1828 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1829 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1830 	tbuf = (void *)(maa + 1);
1831 	ma->error = copyinstr(val, tbuf, len, NULL);
1832 	return (mount_arg(ma, name, tbuf, -1));
1833 }
1834 
1835 /*
1836  * Plain argument.
1837  *
1838  * If length is -1, treat value as a C string.
1839  */
1840 struct mntarg *
1841 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
1842 {
1843 
1844 	if (ma == NULL) {
1845 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1846 		SLIST_INIT(&ma->list);
1847 	}
1848 	if (ma->error)
1849 		return (ma);
1850 
1851 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1852 	    M_MOUNT, M_WAITOK);
1853 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1854 	ma->v[ma->len].iov_len = strlen(name) + 1;
1855 	ma->len++;
1856 
1857 	ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
1858 	if (len < 0)
1859 		ma->v[ma->len].iov_len = strlen(val) + 1;
1860 	else
1861 		ma->v[ma->len].iov_len = len;
1862 	ma->len++;
1863 	return (ma);
1864 }
1865 
1866 /*
1867  * Free a mntarg structure
1868  */
1869 static void
1870 free_mntarg(struct mntarg *ma)
1871 {
1872 	struct mntaarg *maa;
1873 
1874 	while (!SLIST_EMPTY(&ma->list)) {
1875 		maa = SLIST_FIRST(&ma->list);
1876 		SLIST_REMOVE_HEAD(&ma->list, next);
1877 		free(maa, M_MOUNT);
1878 	}
1879 	free(ma->v, M_MOUNT);
1880 	free(ma, M_MOUNT);
1881 }
1882 
1883 /*
1884  * Mount a filesystem
1885  */
1886 int
1887 kernel_mount(struct mntarg *ma, uint64_t flags)
1888 {
1889 	struct uio auio;
1890 	int error;
1891 
1892 	KASSERT(ma != NULL, ("kernel_mount NULL ma"));
1893 	KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
1894 	KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
1895 
1896 	auio.uio_iov = ma->v;
1897 	auio.uio_iovcnt = ma->len;
1898 	auio.uio_segflg = UIO_SYSSPACE;
1899 
1900 	error = ma->error;
1901 	if (!error)
1902 		error = vfs_donmount(curthread, flags, &auio);
1903 	free_mntarg(ma);
1904 	return (error);
1905 }
1906 
1907 /*
1908  * A printflike function to mount a filesystem.
1909  */
1910 int
1911 kernel_vmount(int flags, ...)
1912 {
1913 	struct mntarg *ma = NULL;
1914 	va_list ap;
1915 	const char *cp;
1916 	const void *vp;
1917 	int error;
1918 
1919 	va_start(ap, flags);
1920 	for (;;) {
1921 		cp = va_arg(ap, const char *);
1922 		if (cp == NULL)
1923 			break;
1924 		vp = va_arg(ap, const void *);
1925 		ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0));
1926 	}
1927 	va_end(ap);
1928 
1929 	error = kernel_mount(ma, flags);
1930 	return (error);
1931 }
1932 
1933 void
1934 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp)
1935 {
1936 
1937 	bcopy(oexp, exp, sizeof(*oexp));
1938 	exp->ex_numsecflavors = 0;
1939 }
1940