xref: /freebsd/sys/kern/vfs_mount.c (revision 5dae51da3da0cc94d17bd67b308fad304ebec7e0)
1 /*-
2  * Copyright (c) 1999-2004 Poul-Henning Kamp
3  * Copyright (c) 1999 Michael Smith
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  * (c) UNIX System Laboratories, Inc.
7  * All or some portions of this file are derived from material licensed
8  * to the University of California by American Telephone and Telegraph
9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10  * the permission of UNIX System Laboratories, Inc.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  */
36 
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 #include <sys/param.h>
41 #include <sys/conf.h>
42 #include <sys/fcntl.h>
43 #include <sys/jail.h>
44 #include <sys/kernel.h>
45 #include <sys/libkern.h>
46 #include <sys/malloc.h>
47 #include <sys/mount.h>
48 #include <sys/mutex.h>
49 #include <sys/namei.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/filedesc.h>
53 #include <sys/reboot.h>
54 #include <sys/sbuf.h>
55 #include <sys/syscallsubr.h>
56 #include <sys/sysproto.h>
57 #include <sys/sx.h>
58 #include <sys/sysctl.h>
59 #include <sys/sysent.h>
60 #include <sys/systm.h>
61 #include <sys/vnode.h>
62 #include <vm/uma.h>
63 
64 #include <geom/geom.h>
65 
66 #include <machine/stdarg.h>
67 
68 #include <security/audit/audit.h>
69 #include <security/mac/mac_framework.h>
70 
71 #define	VFS_MOUNTARG_SIZE_MAX	(1024 * 64)
72 
73 static int	vfs_domount(struct thread *td, const char *fstype, char *fspath,
74 		    uint64_t fsflags, struct vfsoptlist **optlist);
75 static void	free_mntarg(struct mntarg *ma);
76 
77 static int	usermount = 0;
78 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0,
79     "Unprivileged users may mount and unmount file systems");
80 
81 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
82 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 	mtx_init(&mp->mnt_listmtx, "struct mount vlist mtx", NULL, MTX_DEF);
113 	lockinit(&mp->mnt_explock, PVFS, "explock", 0, 0);
114 	return (0);
115 }
116 
117 static void
118 mount_fini(void *mem, int size)
119 {
120 	struct mount *mp;
121 
122 	mp = (struct mount *)mem;
123 	lockdestroy(&mp->mnt_explock);
124 	mtx_destroy(&mp->mnt_listmtx);
125 	mtx_destroy(&mp->mnt_mtx);
126 }
127 
128 static void
129 vfs_mount_init(void *dummy __unused)
130 {
131 
132 	mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount), NULL,
133 	    NULL, mount_init, mount_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
134 }
135 SYSINIT(vfs_mount, SI_SUB_VFS, SI_ORDER_ANY, vfs_mount_init, NULL);
136 
137 /*
138  * ---------------------------------------------------------------------
139  * Functions for building and sanitizing the mount options
140  */
141 
142 /* Remove one mount option. */
143 static void
144 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
145 {
146 
147 	TAILQ_REMOVE(opts, opt, link);
148 	free(opt->name, M_MOUNT);
149 	if (opt->value != NULL)
150 		free(opt->value, M_MOUNT);
151 	free(opt, M_MOUNT);
152 }
153 
154 /* Release all resources related to the mount options. */
155 void
156 vfs_freeopts(struct vfsoptlist *opts)
157 {
158 	struct vfsopt *opt;
159 
160 	while (!TAILQ_EMPTY(opts)) {
161 		opt = TAILQ_FIRST(opts);
162 		vfs_freeopt(opts, opt);
163 	}
164 	free(opts, M_MOUNT);
165 }
166 
167 void
168 vfs_deleteopt(struct vfsoptlist *opts, const char *name)
169 {
170 	struct vfsopt *opt, *temp;
171 
172 	if (opts == NULL)
173 		return;
174 	TAILQ_FOREACH_SAFE(opt, opts, link, temp)  {
175 		if (strcmp(opt->name, name) == 0)
176 			vfs_freeopt(opts, opt);
177 	}
178 }
179 
180 static int
181 vfs_isopt_ro(const char *opt)
182 {
183 
184 	if (strcmp(opt, "ro") == 0 || strcmp(opt, "rdonly") == 0 ||
185 	    strcmp(opt, "norw") == 0)
186 		return (1);
187 	return (0);
188 }
189 
190 static int
191 vfs_isopt_rw(const char *opt)
192 {
193 
194 	if (strcmp(opt, "rw") == 0 || strcmp(opt, "noro") == 0)
195 		return (1);
196 	return (0);
197 }
198 
199 /*
200  * Check if options are equal (with or without the "no" prefix).
201  */
202 static int
203 vfs_equalopts(const char *opt1, const char *opt2)
204 {
205 	char *p;
206 
207 	/* "opt" vs. "opt" or "noopt" vs. "noopt" */
208 	if (strcmp(opt1, opt2) == 0)
209 		return (1);
210 	/* "noopt" vs. "opt" */
211 	if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
212 		return (1);
213 	/* "opt" vs. "noopt" */
214 	if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
215 		return (1);
216 	while ((p = strchr(opt1, '.')) != NULL &&
217 	    !strncmp(opt1, opt2, ++p - opt1)) {
218 		opt2 += p - opt1;
219 		opt1 = p;
220 		/* "foo.noopt" vs. "foo.opt" */
221 		if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
222 			return (1);
223 		/* "foo.opt" vs. "foo.noopt" */
224 		if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
225 			return (1);
226 	}
227 	/* "ro" / "rdonly" / "norw" / "rw" / "noro" */
228 	if ((vfs_isopt_ro(opt1) || vfs_isopt_rw(opt1)) &&
229 	    (vfs_isopt_ro(opt2) || vfs_isopt_rw(opt2)))
230 		return (1);
231 	return (0);
232 }
233 
234 /*
235  * If a mount option is specified several times,
236  * (with or without the "no" prefix) only keep
237  * the last occurrence of it.
238  */
239 static void
240 vfs_sanitizeopts(struct vfsoptlist *opts)
241 {
242 	struct vfsopt *opt, *opt2, *tmp;
243 
244 	TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
245 		opt2 = TAILQ_PREV(opt, vfsoptlist, link);
246 		while (opt2 != NULL) {
247 			if (vfs_equalopts(opt->name, opt2->name)) {
248 				tmp = TAILQ_PREV(opt2, vfsoptlist, link);
249 				vfs_freeopt(opts, opt2);
250 				opt2 = tmp;
251 			} else {
252 				opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
253 			}
254 		}
255 	}
256 }
257 
258 /*
259  * Build a linked list of mount options from a struct uio.
260  */
261 int
262 vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
263 {
264 	struct vfsoptlist *opts;
265 	struct vfsopt *opt;
266 	size_t memused, namelen, optlen;
267 	unsigned int i, iovcnt;
268 	int error;
269 
270 	opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
271 	TAILQ_INIT(opts);
272 	memused = 0;
273 	iovcnt = auio->uio_iovcnt;
274 	for (i = 0; i < iovcnt; i += 2) {
275 		namelen = auio->uio_iov[i].iov_len;
276 		optlen = auio->uio_iov[i + 1].iov_len;
277 		memused += sizeof(struct vfsopt) + optlen + namelen;
278 		/*
279 		 * Avoid consuming too much memory, and attempts to overflow
280 		 * memused.
281 		 */
282 		if (memused > VFS_MOUNTARG_SIZE_MAX ||
283 		    optlen > VFS_MOUNTARG_SIZE_MAX ||
284 		    namelen > VFS_MOUNTARG_SIZE_MAX) {
285 			error = EINVAL;
286 			goto bad;
287 		}
288 
289 		opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
290 		opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
291 		opt->value = NULL;
292 		opt->len = 0;
293 		opt->pos = i / 2;
294 		opt->seen = 0;
295 
296 		/*
297 		 * Do this early, so jumps to "bad" will free the current
298 		 * option.
299 		 */
300 		TAILQ_INSERT_TAIL(opts, opt, link);
301 
302 		if (auio->uio_segflg == UIO_SYSSPACE) {
303 			bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
304 		} else {
305 			error = copyin(auio->uio_iov[i].iov_base, opt->name,
306 			    namelen);
307 			if (error)
308 				goto bad;
309 		}
310 		/* Ensure names are null-terminated strings. */
311 		if (namelen == 0 || opt->name[namelen - 1] != '\0') {
312 			error = EINVAL;
313 			goto bad;
314 		}
315 		if (optlen != 0) {
316 			opt->len = optlen;
317 			opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
318 			if (auio->uio_segflg == UIO_SYSSPACE) {
319 				bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
320 				    optlen);
321 			} else {
322 				error = copyin(auio->uio_iov[i + 1].iov_base,
323 				    opt->value, optlen);
324 				if (error)
325 					goto bad;
326 			}
327 		}
328 	}
329 	vfs_sanitizeopts(opts);
330 	*options = opts;
331 	return (0);
332 bad:
333 	vfs_freeopts(opts);
334 	return (error);
335 }
336 
337 /*
338  * Merge the old mount options with the new ones passed
339  * in the MNT_UPDATE case.
340  *
341  * XXX: This function will keep a "nofoo" option in the new
342  * options.  E.g, if the option's canonical name is "foo",
343  * "nofoo" ends up in the mount point's active options.
344  */
345 static void
346 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *oldopts)
347 {
348 	struct vfsopt *opt, *new;
349 
350 	TAILQ_FOREACH(opt, oldopts, link) {
351 		new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
352 		new->name = strdup(opt->name, M_MOUNT);
353 		if (opt->len != 0) {
354 			new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
355 			bcopy(opt->value, new->value, opt->len);
356 		} else
357 			new->value = NULL;
358 		new->len = opt->len;
359 		new->seen = opt->seen;
360 		TAILQ_INSERT_HEAD(toopts, new, link);
361 	}
362 	vfs_sanitizeopts(toopts);
363 }
364 
365 /*
366  * Mount a filesystem.
367  */
368 int
369 sys_nmount(td, uap)
370 	struct thread *td;
371 	struct nmount_args /* {
372 		struct iovec *iovp;
373 		unsigned int iovcnt;
374 		int flags;
375 	} */ *uap;
376 {
377 	struct uio *auio;
378 	int error;
379 	u_int iovcnt;
380 	uint64_t flags;
381 
382 	/*
383 	 * Mount flags are now 64-bits. On 32-bit archtectures only
384 	 * 32-bits are passed in, but from here on everything handles
385 	 * 64-bit flags correctly.
386 	 */
387 	flags = uap->flags;
388 
389 	AUDIT_ARG_FFLAGS(flags);
390 	CTR4(KTR_VFS, "%s: iovp %p with iovcnt %d and flags %d", __func__,
391 	    uap->iovp, uap->iovcnt, flags);
392 
393 	/*
394 	 * Filter out MNT_ROOTFS.  We do not want clients of nmount() in
395 	 * userspace to set this flag, but we must filter it out if we want
396 	 * MNT_UPDATE on the root file system to work.
397 	 * MNT_ROOTFS should only be set by the kernel when mounting its
398 	 * root file system.
399 	 */
400 	flags &= ~MNT_ROOTFS;
401 
402 	iovcnt = uap->iovcnt;
403 	/*
404 	 * Check that we have an even number of iovec's
405 	 * and that we have at least two options.
406 	 */
407 	if ((iovcnt & 1) || (iovcnt < 4)) {
408 		CTR2(KTR_VFS, "%s: failed for invalid iovcnt %d", __func__,
409 		    uap->iovcnt);
410 		return (EINVAL);
411 	}
412 
413 	error = copyinuio(uap->iovp, iovcnt, &auio);
414 	if (error) {
415 		CTR2(KTR_VFS, "%s: failed for invalid uio op with %d errno",
416 		    __func__, error);
417 		return (error);
418 	}
419 	error = vfs_donmount(td, flags, auio);
420 
421 	free(auio, M_IOV);
422 	return (error);
423 }
424 
425 /*
426  * ---------------------------------------------------------------------
427  * Various utility functions
428  */
429 
430 void
431 vfs_ref(struct mount *mp)
432 {
433 
434 	CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
435 	MNT_ILOCK(mp);
436 	MNT_REF(mp);
437 	MNT_IUNLOCK(mp);
438 }
439 
440 void
441 vfs_rel(struct mount *mp)
442 {
443 
444 	CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
445 	MNT_ILOCK(mp);
446 	MNT_REL(mp);
447 	MNT_IUNLOCK(mp);
448 }
449 
450 /*
451  * Allocate and initialize the mount point struct.
452  */
453 struct mount *
454 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, const char *fspath,
455     struct ucred *cred)
456 {
457 	struct mount *mp;
458 
459 	mp = uma_zalloc(mount_zone, M_WAITOK);
460 	bzero(&mp->mnt_startzero,
461 	    __rangeof(struct mount, mnt_startzero, mnt_endzero));
462 	TAILQ_INIT(&mp->mnt_nvnodelist);
463 	mp->mnt_nvnodelistsize = 0;
464 	TAILQ_INIT(&mp->mnt_activevnodelist);
465 	mp->mnt_activevnodelistsize = 0;
466 	TAILQ_INIT(&mp->mnt_tmpfreevnodelist);
467 	mp->mnt_tmpfreevnodelistsize = 0;
468 	mp->mnt_ref = 0;
469 	(void) vfs_busy(mp, MBF_NOWAIT);
470 	atomic_add_acq_int(&vfsp->vfc_refcount, 1);
471 	mp->mnt_op = vfsp->vfc_vfsops;
472 	mp->mnt_vfc = vfsp;
473 	mp->mnt_stat.f_type = vfsp->vfc_typenum;
474 	mp->mnt_gen++;
475 	strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
476 	mp->mnt_vnodecovered = vp;
477 	mp->mnt_cred = crdup(cred);
478 	mp->mnt_stat.f_owner = cred->cr_uid;
479 	strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
480 	mp->mnt_iosize_max = DFLTPHYS;
481 #ifdef MAC
482 	mac_mount_init(mp);
483 	mac_mount_create(cred, mp);
484 #endif
485 	arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0);
486 	TAILQ_INIT(&mp->mnt_uppers);
487 	return (mp);
488 }
489 
490 /*
491  * Destroy the mount struct previously allocated by vfs_mount_alloc().
492  */
493 void
494 vfs_mount_destroy(struct mount *mp)
495 {
496 
497 	MNT_ILOCK(mp);
498 	mp->mnt_kern_flag |= MNTK_REFEXPIRE;
499 	if (mp->mnt_kern_flag & MNTK_MWAIT) {
500 		mp->mnt_kern_flag &= ~MNTK_MWAIT;
501 		wakeup(mp);
502 	}
503 	while (mp->mnt_ref)
504 		msleep(mp, MNT_MTX(mp), PVFS, "mntref", 0);
505 	KASSERT(mp->mnt_ref == 0,
506 	    ("%s: invalid refcount in the drain path @ %s:%d", __func__,
507 	    __FILE__, __LINE__));
508 	if (mp->mnt_writeopcount != 0)
509 		panic("vfs_mount_destroy: nonzero writeopcount");
510 	if (mp->mnt_secondary_writes != 0)
511 		panic("vfs_mount_destroy: nonzero secondary_writes");
512 	atomic_subtract_rel_int(&mp->mnt_vfc->vfc_refcount, 1);
513 	if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
514 		struct vnode *vp;
515 
516 		TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
517 			vn_printf(vp, "dangling vnode ");
518 		panic("unmount: dangling vnode");
519 	}
520 	KASSERT(TAILQ_EMPTY(&mp->mnt_uppers), ("mnt_uppers"));
521 	if (mp->mnt_nvnodelistsize != 0)
522 		panic("vfs_mount_destroy: nonzero nvnodelistsize");
523 	if (mp->mnt_activevnodelistsize != 0)
524 		panic("vfs_mount_destroy: nonzero activevnodelistsize");
525 	if (mp->mnt_tmpfreevnodelistsize != 0)
526 		panic("vfs_mount_destroy: nonzero tmpfreevnodelistsize");
527 	if (mp->mnt_lockref != 0)
528 		panic("vfs_mount_destroy: nonzero lock refcount");
529 	MNT_IUNLOCK(mp);
530 #ifdef MAC
531 	mac_mount_destroy(mp);
532 #endif
533 	if (mp->mnt_opt != NULL)
534 		vfs_freeopts(mp->mnt_opt);
535 	crfree(mp->mnt_cred);
536 	uma_zfree(mount_zone, mp);
537 }
538 
539 int
540 vfs_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions)
541 {
542 	struct vfsoptlist *optlist;
543 	struct vfsopt *opt, *tmp_opt;
544 	char *fstype, *fspath, *errmsg;
545 	int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
546 
547 	errmsg = fspath = NULL;
548 	errmsg_len = fspathlen = 0;
549 	errmsg_pos = -1;
550 
551 	error = vfs_buildopts(fsoptions, &optlist);
552 	if (error)
553 		return (error);
554 
555 	if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
556 		errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
557 
558 	/*
559 	 * We need these two options before the others,
560 	 * and they are mandatory for any filesystem.
561 	 * Ensure they are NUL terminated as well.
562 	 */
563 	fstypelen = 0;
564 	error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
565 	if (error || fstype[fstypelen - 1] != '\0') {
566 		error = EINVAL;
567 		if (errmsg != NULL)
568 			strncpy(errmsg, "Invalid fstype", errmsg_len);
569 		goto bail;
570 	}
571 	fspathlen = 0;
572 	error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
573 	if (error || fspath[fspathlen - 1] != '\0') {
574 		error = EINVAL;
575 		if (errmsg != NULL)
576 			strncpy(errmsg, "Invalid fspath", errmsg_len);
577 		goto bail;
578 	}
579 
580 	/*
581 	 * We need to see if we have the "update" option
582 	 * before we call vfs_domount(), since vfs_domount() has special
583 	 * logic based on MNT_UPDATE.  This is very important
584 	 * when we want to update the root filesystem.
585 	 */
586 	TAILQ_FOREACH_SAFE(opt, optlist, link, tmp_opt) {
587 		if (strcmp(opt->name, "update") == 0) {
588 			fsflags |= MNT_UPDATE;
589 			vfs_freeopt(optlist, opt);
590 		}
591 		else if (strcmp(opt->name, "async") == 0)
592 			fsflags |= MNT_ASYNC;
593 		else if (strcmp(opt->name, "force") == 0) {
594 			fsflags |= MNT_FORCE;
595 			vfs_freeopt(optlist, opt);
596 		}
597 		else if (strcmp(opt->name, "reload") == 0) {
598 			fsflags |= MNT_RELOAD;
599 			vfs_freeopt(optlist, opt);
600 		}
601 		else if (strcmp(opt->name, "multilabel") == 0)
602 			fsflags |= MNT_MULTILABEL;
603 		else if (strcmp(opt->name, "noasync") == 0)
604 			fsflags &= ~MNT_ASYNC;
605 		else if (strcmp(opt->name, "noatime") == 0)
606 			fsflags |= MNT_NOATIME;
607 		else if (strcmp(opt->name, "atime") == 0) {
608 			free(opt->name, M_MOUNT);
609 			opt->name = strdup("nonoatime", M_MOUNT);
610 		}
611 		else if (strcmp(opt->name, "noclusterr") == 0)
612 			fsflags |= MNT_NOCLUSTERR;
613 		else if (strcmp(opt->name, "clusterr") == 0) {
614 			free(opt->name, M_MOUNT);
615 			opt->name = strdup("nonoclusterr", M_MOUNT);
616 		}
617 		else if (strcmp(opt->name, "noclusterw") == 0)
618 			fsflags |= MNT_NOCLUSTERW;
619 		else if (strcmp(opt->name, "clusterw") == 0) {
620 			free(opt->name, M_MOUNT);
621 			opt->name = strdup("nonoclusterw", M_MOUNT);
622 		}
623 		else if (strcmp(opt->name, "noexec") == 0)
624 			fsflags |= MNT_NOEXEC;
625 		else if (strcmp(opt->name, "exec") == 0) {
626 			free(opt->name, M_MOUNT);
627 			opt->name = strdup("nonoexec", M_MOUNT);
628 		}
629 		else if (strcmp(opt->name, "nosuid") == 0)
630 			fsflags |= MNT_NOSUID;
631 		else if (strcmp(opt->name, "suid") == 0) {
632 			free(opt->name, M_MOUNT);
633 			opt->name = strdup("nonosuid", M_MOUNT);
634 		}
635 		else if (strcmp(opt->name, "nosymfollow") == 0)
636 			fsflags |= MNT_NOSYMFOLLOW;
637 		else if (strcmp(opt->name, "symfollow") == 0) {
638 			free(opt->name, M_MOUNT);
639 			opt->name = strdup("nonosymfollow", M_MOUNT);
640 		}
641 		else if (strcmp(opt->name, "noro") == 0)
642 			fsflags &= ~MNT_RDONLY;
643 		else if (strcmp(opt->name, "rw") == 0)
644 			fsflags &= ~MNT_RDONLY;
645 		else if (strcmp(opt->name, "ro") == 0)
646 			fsflags |= MNT_RDONLY;
647 		else if (strcmp(opt->name, "rdonly") == 0) {
648 			free(opt->name, M_MOUNT);
649 			opt->name = strdup("ro", M_MOUNT);
650 			fsflags |= MNT_RDONLY;
651 		}
652 		else if (strcmp(opt->name, "suiddir") == 0)
653 			fsflags |= MNT_SUIDDIR;
654 		else if (strcmp(opt->name, "sync") == 0)
655 			fsflags |= MNT_SYNCHRONOUS;
656 		else if (strcmp(opt->name, "union") == 0)
657 			fsflags |= MNT_UNION;
658 		else if (strcmp(opt->name, "automounted") == 0) {
659 			fsflags |= MNT_AUTOMOUNTED;
660 			vfs_freeopt(optlist, opt);
661 		}
662 	}
663 
664 	/*
665 	 * Be ultra-paranoid about making sure the type and fspath
666 	 * variables will fit in our mp buffers, including the
667 	 * terminating NUL.
668 	 */
669 	if (fstypelen > MFSNAMELEN || fspathlen > MNAMELEN) {
670 		error = ENAMETOOLONG;
671 		goto bail;
672 	}
673 
674 	error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
675 bail:
676 	/* copyout the errmsg */
677 	if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
678 	    && errmsg_len > 0 && errmsg != NULL) {
679 		if (fsoptions->uio_segflg == UIO_SYSSPACE) {
680 			bcopy(errmsg,
681 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
682 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
683 		} else {
684 			copyout(errmsg,
685 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
686 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
687 		}
688 	}
689 
690 	if (optlist != NULL)
691 		vfs_freeopts(optlist);
692 	return (error);
693 }
694 
695 /*
696  * Old mount API.
697  */
698 #ifndef _SYS_SYSPROTO_H_
699 struct mount_args {
700 	char	*type;
701 	char	*path;
702 	int	flags;
703 	caddr_t	data;
704 };
705 #endif
706 /* ARGSUSED */
707 int
708 sys_mount(td, uap)
709 	struct thread *td;
710 	struct mount_args /* {
711 		char *type;
712 		char *path;
713 		int flags;
714 		caddr_t data;
715 	} */ *uap;
716 {
717 	char *fstype;
718 	struct vfsconf *vfsp = NULL;
719 	struct mntarg *ma = NULL;
720 	uint64_t flags;
721 	int error;
722 
723 	/*
724 	 * Mount flags are now 64-bits. On 32-bit architectures only
725 	 * 32-bits are passed in, but from here on everything handles
726 	 * 64-bit flags correctly.
727 	 */
728 	flags = uap->flags;
729 
730 	AUDIT_ARG_FFLAGS(flags);
731 
732 	/*
733 	 * Filter out MNT_ROOTFS.  We do not want clients of mount() in
734 	 * userspace to set this flag, but we must filter it out if we want
735 	 * MNT_UPDATE on the root file system to work.
736 	 * MNT_ROOTFS should only be set by the kernel when mounting its
737 	 * root file system.
738 	 */
739 	flags &= ~MNT_ROOTFS;
740 
741 	fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
742 	error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
743 	if (error) {
744 		free(fstype, M_TEMP);
745 		return (error);
746 	}
747 
748 	AUDIT_ARG_TEXT(fstype);
749 	vfsp = vfs_byname_kld(fstype, td, &error);
750 	free(fstype, M_TEMP);
751 	if (vfsp == NULL)
752 		return (ENOENT);
753 	if (vfsp->vfc_vfsops->vfs_cmount == NULL)
754 		return (EOPNOTSUPP);
755 
756 	ma = mount_argsu(ma, "fstype", uap->type, MFSNAMELEN);
757 	ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
758 	ma = mount_argb(ma, flags & MNT_RDONLY, "noro");
759 	ma = mount_argb(ma, !(flags & MNT_NOSUID), "nosuid");
760 	ma = mount_argb(ma, !(flags & MNT_NOEXEC), "noexec");
761 
762 	error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags);
763 	return (error);
764 }
765 
766 /*
767  * vfs_domount_first(): first file system mount (not update)
768  */
769 static int
770 vfs_domount_first(
771 	struct thread *td,		/* Calling thread. */
772 	struct vfsconf *vfsp,		/* File system type. */
773 	char *fspath,			/* Mount path. */
774 	struct vnode *vp,		/* Vnode to be covered. */
775 	uint64_t fsflags,		/* Flags common to all filesystems. */
776 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
777 	)
778 {
779 	struct vattr va;
780 	struct mount *mp;
781 	struct vnode *newdp;
782 	int error;
783 
784 	ASSERT_VOP_ELOCKED(vp, __func__);
785 	KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
786 
787 	/*
788 	 * If the user is not root, ensure that they own the directory
789 	 * onto which we are attempting to mount.
790 	 */
791 	error = VOP_GETATTR(vp, &va, td->td_ucred);
792 	if (error == 0 && va.va_uid != td->td_ucred->cr_uid)
793 		error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN, 0);
794 	if (error == 0)
795 		error = vinvalbuf(vp, V_SAVE, 0, 0);
796 	if (error == 0 && vp->v_type != VDIR)
797 		error = ENOTDIR;
798 	if (error == 0) {
799 		VI_LOCK(vp);
800 		if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
801 			vp->v_iflag |= VI_MOUNT;
802 		else
803 			error = EBUSY;
804 		VI_UNLOCK(vp);
805 	}
806 	if (error != 0) {
807 		vput(vp);
808 		return (error);
809 	}
810 	VOP_UNLOCK(vp, 0);
811 
812 	/* Allocate and initialize the filesystem. */
813 	mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred);
814 	/* XXXMAC: pass to vfs_mount_alloc? */
815 	mp->mnt_optnew = *optlist;
816 	/* Set the mount level flags. */
817 	mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY));
818 
819 	/*
820 	 * Mount the filesystem.
821 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
822 	 * get.  No freeing of cn_pnbuf.
823 	 */
824 	error = VFS_MOUNT(mp);
825 	if (error != 0) {
826 		vfs_unbusy(mp);
827 		vfs_mount_destroy(mp);
828 		VI_LOCK(vp);
829 		vp->v_iflag &= ~VI_MOUNT;
830 		VI_UNLOCK(vp);
831 		vrele(vp);
832 		return (error);
833 	}
834 
835 	if (mp->mnt_opt != NULL)
836 		vfs_freeopts(mp->mnt_opt);
837 	mp->mnt_opt = mp->mnt_optnew;
838 	*optlist = NULL;
839 	(void)VFS_STATFS(mp, &mp->mnt_stat);
840 
841 	/*
842 	 * Prevent external consumers of mount options from reading mnt_optnew.
843 	 */
844 	mp->mnt_optnew = NULL;
845 
846 	MNT_ILOCK(mp);
847 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
848 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
849 		mp->mnt_kern_flag |= MNTK_ASYNC;
850 	else
851 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
852 	MNT_IUNLOCK(mp);
853 
854 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
855 	cache_purge(vp);
856 	VI_LOCK(vp);
857 	vp->v_iflag &= ~VI_MOUNT;
858 	VI_UNLOCK(vp);
859 	vp->v_mountedhere = mp;
860 	/* Place the new filesystem at the end of the mount list. */
861 	mtx_lock(&mountlist_mtx);
862 	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
863 	mtx_unlock(&mountlist_mtx);
864 	vfs_event_signal(NULL, VQ_MOUNT, 0);
865 	if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp))
866 		panic("mount: lost mount");
867 	VOP_UNLOCK(vp, 0);
868 	EVENTHANDLER_INVOKE(vfs_mounted, mp, newdp, td);
869 	VOP_UNLOCK(newdp, 0);
870 	mountcheckdirs(vp, newdp);
871 	vrele(newdp);
872 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
873 		vfs_allocate_syncvnode(mp);
874 	vfs_unbusy(mp);
875 	return (0);
876 }
877 
878 /*
879  * vfs_domount_update(): update of mounted file system
880  */
881 static int
882 vfs_domount_update(
883 	struct thread *td,		/* Calling thread. */
884 	struct vnode *vp,		/* Mount point vnode. */
885 	uint64_t fsflags,		/* Flags common to all filesystems. */
886 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
887 	)
888 {
889 	struct export_args export;
890 	void *bufp;
891 	struct mount *mp;
892 	int error, export_error, len;
893 	uint64_t flag;
894 
895 	ASSERT_VOP_ELOCKED(vp, __func__);
896 	KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
897 	mp = vp->v_mount;
898 
899 	if ((vp->v_vflag & VV_ROOT) == 0) {
900 		if (vfs_copyopt(*optlist, "export", &export, sizeof(export))
901 		    == 0)
902 			error = EXDEV;
903 		else
904 			error = EINVAL;
905 		vput(vp);
906 		return (error);
907 	}
908 
909 	/*
910 	 * We only allow the filesystem to be reloaded if it
911 	 * is currently mounted read-only.
912 	 */
913 	flag = mp->mnt_flag;
914 	if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
915 		vput(vp);
916 		return (EOPNOTSUPP);	/* Needs translation */
917 	}
918 	/*
919 	 * Only privileged root, or (if MNT_USER is set) the user that
920 	 * did the original mount is permitted to update it.
921 	 */
922 	error = vfs_suser(mp, td);
923 	if (error != 0) {
924 		vput(vp);
925 		return (error);
926 	}
927 	if (vfs_busy(mp, MBF_NOWAIT)) {
928 		vput(vp);
929 		return (EBUSY);
930 	}
931 	VI_LOCK(vp);
932 	if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
933 		VI_UNLOCK(vp);
934 		vfs_unbusy(mp);
935 		vput(vp);
936 		return (EBUSY);
937 	}
938 	vp->v_iflag |= VI_MOUNT;
939 	VI_UNLOCK(vp);
940 	VOP_UNLOCK(vp, 0);
941 
942 	MNT_ILOCK(mp);
943 	mp->mnt_flag &= ~MNT_UPDATEMASK;
944 	mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE |
945 	    MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY);
946 	if ((mp->mnt_flag & MNT_ASYNC) == 0)
947 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
948 	MNT_IUNLOCK(mp);
949 	mp->mnt_optnew = *optlist;
950 	vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
951 
952 	/*
953 	 * Mount the filesystem.
954 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
955 	 * get.  No freeing of cn_pnbuf.
956 	 */
957 	error = VFS_MOUNT(mp);
958 
959 	export_error = 0;
960 	/* Process the export option. */
961 	if (error == 0 && vfs_getopt(mp->mnt_optnew, "export", &bufp,
962 	    &len) == 0) {
963 		/* Assume that there is only 1 ABI for each length. */
964 		switch (len) {
965 		case (sizeof(struct oexport_args)):
966 			bzero(&export, sizeof(export));
967 			/* FALLTHROUGH */
968 		case (sizeof(export)):
969 			bcopy(bufp, &export, len);
970 			export_error = vfs_export(mp, &export);
971 			break;
972 		default:
973 			export_error = EINVAL;
974 			break;
975 		}
976 	}
977 
978 	MNT_ILOCK(mp);
979 	if (error == 0) {
980 		mp->mnt_flag &=	~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
981 		    MNT_SNAPSHOT);
982 	} else {
983 		/*
984 		 * If we fail, restore old mount flags. MNT_QUOTA is special,
985 		 * because it is not part of MNT_UPDATEMASK, but it could have
986 		 * changed in the meantime if quotactl(2) was called.
987 		 * All in all we want current value of MNT_QUOTA, not the old
988 		 * one.
989 		 */
990 		mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
991 	}
992 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
993 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
994 		mp->mnt_kern_flag |= MNTK_ASYNC;
995 	else
996 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
997 	MNT_IUNLOCK(mp);
998 
999 	if (error != 0)
1000 		goto end;
1001 
1002 	if (mp->mnt_opt != NULL)
1003 		vfs_freeopts(mp->mnt_opt);
1004 	mp->mnt_opt = mp->mnt_optnew;
1005 	*optlist = NULL;
1006 	(void)VFS_STATFS(mp, &mp->mnt_stat);
1007 	/*
1008 	 * Prevent external consumers of mount options from reading
1009 	 * mnt_optnew.
1010 	 */
1011 	mp->mnt_optnew = NULL;
1012 
1013 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
1014 		vfs_allocate_syncvnode(mp);
1015 	else
1016 		vfs_deallocate_syncvnode(mp);
1017 end:
1018 	vfs_unbusy(mp);
1019 	VI_LOCK(vp);
1020 	vp->v_iflag &= ~VI_MOUNT;
1021 	VI_UNLOCK(vp);
1022 	vrele(vp);
1023 	return (error != 0 ? error : export_error);
1024 }
1025 
1026 /*
1027  * vfs_domount(): actually attempt a filesystem mount.
1028  */
1029 static int
1030 vfs_domount(
1031 	struct thread *td,		/* Calling thread. */
1032 	const char *fstype,		/* Filesystem type. */
1033 	char *fspath,			/* Mount path. */
1034 	uint64_t fsflags,		/* Flags common to all filesystems. */
1035 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
1036 	)
1037 {
1038 	struct vfsconf *vfsp;
1039 	struct nameidata nd;
1040 	struct vnode *vp;
1041 	char *pathbuf;
1042 	int error;
1043 
1044 	/*
1045 	 * Be ultra-paranoid about making sure the type and fspath
1046 	 * variables will fit in our mp buffers, including the
1047 	 * terminating NUL.
1048 	 */
1049 	if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
1050 		return (ENAMETOOLONG);
1051 
1052 	if (jailed(td->td_ucred) || usermount == 0) {
1053 		if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
1054 			return (error);
1055 	}
1056 
1057 	/*
1058 	 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
1059 	 */
1060 	if (fsflags & MNT_EXPORTED) {
1061 		error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
1062 		if (error)
1063 			return (error);
1064 	}
1065 	if (fsflags & MNT_SUIDDIR) {
1066 		error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
1067 		if (error)
1068 			return (error);
1069 	}
1070 	/*
1071 	 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
1072 	 */
1073 	if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
1074 		if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
1075 			fsflags |= MNT_NOSUID | MNT_USER;
1076 	}
1077 
1078 	/* Load KLDs before we lock the covered vnode to avoid reversals. */
1079 	vfsp = NULL;
1080 	if ((fsflags & MNT_UPDATE) == 0) {
1081 		/* Don't try to load KLDs if we're mounting the root. */
1082 		if (fsflags & MNT_ROOTFS)
1083 			vfsp = vfs_byname(fstype);
1084 		else
1085 			vfsp = vfs_byname_kld(fstype, td, &error);
1086 		if (vfsp == NULL)
1087 			return (ENODEV);
1088 		if (jailed(td->td_ucred) && !(vfsp->vfc_flags & VFCF_JAIL))
1089 			return (EPERM);
1090 	}
1091 
1092 	/*
1093 	 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
1094 	 */
1095 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1096 	    UIO_SYSSPACE, fspath, td);
1097 	error = namei(&nd);
1098 	if (error != 0)
1099 		return (error);
1100 	NDFREE(&nd, NDF_ONLY_PNBUF);
1101 	vp = nd.ni_vp;
1102 	if ((fsflags & MNT_UPDATE) == 0) {
1103 		pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1104 		strcpy(pathbuf, fspath);
1105 		error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN);
1106 		/* debug.disablefullpath == 1 results in ENODEV */
1107 		if (error == 0 || error == ENODEV) {
1108 			error = vfs_domount_first(td, vfsp, pathbuf, vp,
1109 			    fsflags, optlist);
1110 		}
1111 		free(pathbuf, M_TEMP);
1112 	} else
1113 		error = vfs_domount_update(td, vp, fsflags, optlist);
1114 
1115 	return (error);
1116 }
1117 
1118 /*
1119  * Unmount a filesystem.
1120  *
1121  * Note: unmount takes a path to the vnode mounted on as argument, not
1122  * special file (as before).
1123  */
1124 #ifndef _SYS_SYSPROTO_H_
1125 struct unmount_args {
1126 	char	*path;
1127 	int	flags;
1128 };
1129 #endif
1130 /* ARGSUSED */
1131 int
1132 sys_unmount(struct thread *td, struct unmount_args *uap)
1133 {
1134 	struct nameidata nd;
1135 	struct mount *mp;
1136 	char *pathbuf;
1137 	int error, id0, id1;
1138 
1139 	AUDIT_ARG_VALUE(uap->flags);
1140 	if (jailed(td->td_ucred) || usermount == 0) {
1141 		error = priv_check(td, PRIV_VFS_UNMOUNT);
1142 		if (error)
1143 			return (error);
1144 	}
1145 
1146 	pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1147 	error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1148 	if (error) {
1149 		free(pathbuf, M_TEMP);
1150 		return (error);
1151 	}
1152 	if (uap->flags & MNT_BYFSID) {
1153 		AUDIT_ARG_TEXT(pathbuf);
1154 		/* Decode the filesystem ID. */
1155 		if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
1156 			free(pathbuf, M_TEMP);
1157 			return (EINVAL);
1158 		}
1159 
1160 		mtx_lock(&mountlist_mtx);
1161 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1162 			if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1163 			    mp->mnt_stat.f_fsid.val[1] == id1) {
1164 				vfs_ref(mp);
1165 				break;
1166 			}
1167 		}
1168 		mtx_unlock(&mountlist_mtx);
1169 	} else {
1170 		/*
1171 		 * Try to find global path for path argument.
1172 		 */
1173 		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1174 		    UIO_SYSSPACE, pathbuf, td);
1175 		if (namei(&nd) == 0) {
1176 			NDFREE(&nd, NDF_ONLY_PNBUF);
1177 			error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
1178 			    MNAMELEN);
1179 			if (error == 0 || error == ENODEV)
1180 				vput(nd.ni_vp);
1181 		}
1182 		mtx_lock(&mountlist_mtx);
1183 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1184 			if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) {
1185 				vfs_ref(mp);
1186 				break;
1187 			}
1188 		}
1189 		mtx_unlock(&mountlist_mtx);
1190 	}
1191 	free(pathbuf, M_TEMP);
1192 	if (mp == NULL) {
1193 		/*
1194 		 * Previously we returned ENOENT for a nonexistent path and
1195 		 * EINVAL for a non-mountpoint.  We cannot tell these apart
1196 		 * now, so in the !MNT_BYFSID case return the more likely
1197 		 * EINVAL for compatibility.
1198 		 */
1199 		return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1200 	}
1201 
1202 	/*
1203 	 * Don't allow unmounting the root filesystem.
1204 	 */
1205 	if (mp->mnt_flag & MNT_ROOTFS) {
1206 		vfs_rel(mp);
1207 		return (EINVAL);
1208 	}
1209 	error = dounmount(mp, uap->flags, td);
1210 	return (error);
1211 }
1212 
1213 /*
1214  * Return error if any of the vnodes, ignoring the root vnode
1215  * and the syncer vnode, have non-zero usecount.
1216  *
1217  * This function is purely advisory - it can return false positives
1218  * and negatives.
1219  */
1220 static int
1221 vfs_check_usecounts(struct mount *mp)
1222 {
1223 	struct vnode *vp, *mvp;
1224 
1225 	MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1226 		if ((vp->v_vflag & VV_ROOT) == 0 && vp->v_type != VNON &&
1227 		    vp->v_usecount != 0) {
1228 			VI_UNLOCK(vp);
1229 			MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1230 			return (EBUSY);
1231 		}
1232 		VI_UNLOCK(vp);
1233 	}
1234 
1235 	return (0);
1236 }
1237 
1238 /*
1239  * Do the actual filesystem unmount.
1240  */
1241 int
1242 dounmount(struct mount *mp, int flags, struct thread *td)
1243 {
1244 	struct vnode *coveredvp, *fsrootvp;
1245 	int error;
1246 	uint64_t async_flag;
1247 	int mnt_gen_r;
1248 
1249 	if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
1250 		mnt_gen_r = mp->mnt_gen;
1251 		VI_LOCK(coveredvp);
1252 		vholdl(coveredvp);
1253 		vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
1254 		/*
1255 		 * Check for mp being unmounted while waiting for the
1256 		 * covered vnode lock.
1257 		 */
1258 		if (coveredvp->v_mountedhere != mp ||
1259 		    coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
1260 			VOP_UNLOCK(coveredvp, 0);
1261 			vdrop(coveredvp);
1262 			vfs_rel(mp);
1263 			return (EBUSY);
1264 		}
1265 	}
1266 
1267 	/*
1268 	 * Only privileged root, or (if MNT_USER is set) the user that did the
1269 	 * original mount is permitted to unmount this filesystem.
1270 	 */
1271 	error = vfs_suser(mp, td);
1272 	if (error != 0) {
1273 		if (coveredvp != NULL) {
1274 			VOP_UNLOCK(coveredvp, 0);
1275 			vdrop(coveredvp);
1276 		}
1277 		vfs_rel(mp);
1278 		return (error);
1279 	}
1280 
1281 	vn_start_write(NULL, &mp, V_WAIT | V_MNTREF);
1282 	MNT_ILOCK(mp);
1283 	if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 ||
1284 	    !TAILQ_EMPTY(&mp->mnt_uppers)) {
1285 		MNT_IUNLOCK(mp);
1286 		if (coveredvp != NULL) {
1287 			VOP_UNLOCK(coveredvp, 0);
1288 			vdrop(coveredvp);
1289 		}
1290 		vn_finished_write(mp);
1291 		return (EBUSY);
1292 	}
1293 	mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ;
1294 	if (flags & MNT_NONBUSY) {
1295 		MNT_IUNLOCK(mp);
1296 		error = vfs_check_usecounts(mp);
1297 		MNT_ILOCK(mp);
1298 		if (error != 0) {
1299 			mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_NOINSMNTQ);
1300 			if (mp->mnt_kern_flag & MNTK_MWAIT) {
1301 				mp->mnt_kern_flag &= ~MNTK_MWAIT;
1302 				wakeup(mp);
1303 			}
1304 			MNT_IUNLOCK(mp);
1305 			if (coveredvp != NULL) {
1306 				VOP_UNLOCK(coveredvp, 0);
1307 				vdrop(coveredvp);
1308 			}
1309 			vn_finished_write(mp);
1310 			return (error);
1311 		}
1312 	}
1313 	/* Allow filesystems to detect that a forced unmount is in progress. */
1314 	if (flags & MNT_FORCE) {
1315 		mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1316 		MNT_IUNLOCK(mp);
1317 		/*
1318 		 * Must be done after setting MNTK_UNMOUNTF and before
1319 		 * waiting for mnt_lockref to become 0.
1320 		 */
1321 		VFS_PURGE(mp);
1322 		MNT_ILOCK(mp);
1323 	}
1324 	error = 0;
1325 	if (mp->mnt_lockref) {
1326 		mp->mnt_kern_flag |= MNTK_DRAINING;
1327 		error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
1328 		    "mount drain", 0);
1329 	}
1330 	MNT_IUNLOCK(mp);
1331 	KASSERT(mp->mnt_lockref == 0,
1332 	    ("%s: invalid lock refcount in the drain path @ %s:%d",
1333 	    __func__, __FILE__, __LINE__));
1334 	KASSERT(error == 0,
1335 	    ("%s: invalid return value for msleep in the drain path @ %s:%d",
1336 	    __func__, __FILE__, __LINE__));
1337 
1338 	if (mp->mnt_flag & MNT_EXPUBLIC)
1339 		vfs_setpublicfs(NULL, NULL, NULL);
1340 
1341 	/*
1342 	 * From now, we can claim that the use reference on the
1343 	 * coveredvp is ours, and the ref can be released only by
1344 	 * successfull unmount by us, or left for later unmount
1345 	 * attempt.  The previously acquired hold reference is no
1346 	 * longer needed to protect the vnode from reuse.
1347 	 */
1348 	if (coveredvp != NULL)
1349 		vdrop(coveredvp);
1350 
1351 	vfs_msync(mp, MNT_WAIT);
1352 	MNT_ILOCK(mp);
1353 	async_flag = mp->mnt_flag & MNT_ASYNC;
1354 	mp->mnt_flag &= ~MNT_ASYNC;
1355 	mp->mnt_kern_flag &= ~MNTK_ASYNC;
1356 	MNT_IUNLOCK(mp);
1357 	cache_purgevfs(mp, false); /* remove cache entries for this file sys */
1358 	vfs_deallocate_syncvnode(mp);
1359 	/*
1360 	 * For forced unmounts, move process cdir/rdir refs on the fs root
1361 	 * vnode to the covered vnode.  For non-forced unmounts we want
1362 	 * such references to cause an EBUSY error.
1363 	 */
1364 	if ((flags & MNT_FORCE) &&
1365 	    VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1366 		if (mp->mnt_vnodecovered != NULL &&
1367 		    (mp->mnt_flag & MNT_IGNORE) == 0)
1368 			mountcheckdirs(fsrootvp, mp->mnt_vnodecovered);
1369 		if (fsrootvp == rootvnode) {
1370 			vrele(rootvnode);
1371 			rootvnode = NULL;
1372 		}
1373 		vput(fsrootvp);
1374 	}
1375 	if ((mp->mnt_flag & MNT_RDONLY) != 0 || (flags & MNT_FORCE) != 0 ||
1376 	    (error = VFS_SYNC(mp, MNT_WAIT)) == 0)
1377 		error = VFS_UNMOUNT(mp, flags);
1378 	vn_finished_write(mp);
1379 	/*
1380 	 * If we failed to flush the dirty blocks for this mount point,
1381 	 * undo all the cdir/rdir and rootvnode changes we made above.
1382 	 * Unless we failed to do so because the device is reporting that
1383 	 * it doesn't exist anymore.
1384 	 */
1385 	if (error && error != ENXIO) {
1386 		if ((flags & MNT_FORCE) &&
1387 		    VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp) == 0) {
1388 			if (mp->mnt_vnodecovered != NULL &&
1389 			    (mp->mnt_flag & MNT_IGNORE) == 0)
1390 				mountcheckdirs(mp->mnt_vnodecovered, fsrootvp);
1391 			if (rootvnode == NULL) {
1392 				rootvnode = fsrootvp;
1393 				vref(rootvnode);
1394 			}
1395 			vput(fsrootvp);
1396 		}
1397 		MNT_ILOCK(mp);
1398 		mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ;
1399 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
1400 			MNT_IUNLOCK(mp);
1401 			vfs_allocate_syncvnode(mp);
1402 			MNT_ILOCK(mp);
1403 		}
1404 		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1405 		mp->mnt_flag |= async_flag;
1406 		if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1407 		    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1408 			mp->mnt_kern_flag |= MNTK_ASYNC;
1409 		if (mp->mnt_kern_flag & MNTK_MWAIT) {
1410 			mp->mnt_kern_flag &= ~MNTK_MWAIT;
1411 			wakeup(mp);
1412 		}
1413 		MNT_IUNLOCK(mp);
1414 		if (coveredvp)
1415 			VOP_UNLOCK(coveredvp, 0);
1416 		return (error);
1417 	}
1418 	mtx_lock(&mountlist_mtx);
1419 	TAILQ_REMOVE(&mountlist, mp, mnt_list);
1420 	mtx_unlock(&mountlist_mtx);
1421 	EVENTHANDLER_INVOKE(vfs_unmounted, mp, td);
1422 	if (coveredvp != NULL) {
1423 		coveredvp->v_mountedhere = NULL;
1424 		vput(coveredvp);
1425 	}
1426 	vfs_event_signal(NULL, VQ_UNMOUNT, 0);
1427 	if (mp == rootdevmp)
1428 		rootdevmp = NULL;
1429 	vfs_mount_destroy(mp);
1430 	return (0);
1431 }
1432 
1433 /*
1434  * Report errors during filesystem mounting.
1435  */
1436 void
1437 vfs_mount_error(struct mount *mp, const char *fmt, ...)
1438 {
1439 	struct vfsoptlist *moptlist = mp->mnt_optnew;
1440 	va_list ap;
1441 	int error, len;
1442 	char *errmsg;
1443 
1444 	error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
1445 	if (error || errmsg == NULL || len <= 0)
1446 		return;
1447 
1448 	va_start(ap, fmt);
1449 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1450 	va_end(ap);
1451 }
1452 
1453 void
1454 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
1455 {
1456 	va_list ap;
1457 	int error, len;
1458 	char *errmsg;
1459 
1460 	error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
1461 	if (error || errmsg == NULL || len <= 0)
1462 		return;
1463 
1464 	va_start(ap, fmt);
1465 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1466 	va_end(ap);
1467 }
1468 
1469 /*
1470  * ---------------------------------------------------------------------
1471  * Functions for querying mount options/arguments from filesystems.
1472  */
1473 
1474 /*
1475  * Check that no unknown options are given
1476  */
1477 int
1478 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1479 {
1480 	struct vfsopt *opt;
1481 	char errmsg[255];
1482 	const char **t, *p, *q;
1483 	int ret = 0;
1484 
1485 	TAILQ_FOREACH(opt, opts, link) {
1486 		p = opt->name;
1487 		q = NULL;
1488 		if (p[0] == 'n' && p[1] == 'o')
1489 			q = p + 2;
1490 		for(t = global_opts; *t != NULL; t++) {
1491 			if (strcmp(*t, p) == 0)
1492 				break;
1493 			if (q != NULL) {
1494 				if (strcmp(*t, q) == 0)
1495 					break;
1496 			}
1497 		}
1498 		if (*t != NULL)
1499 			continue;
1500 		for(t = legal; *t != NULL; t++) {
1501 			if (strcmp(*t, p) == 0)
1502 				break;
1503 			if (q != NULL) {
1504 				if (strcmp(*t, q) == 0)
1505 					break;
1506 			}
1507 		}
1508 		if (*t != NULL)
1509 			continue;
1510 		snprintf(errmsg, sizeof(errmsg),
1511 		    "mount option <%s> is unknown", p);
1512 		ret = EINVAL;
1513 	}
1514 	if (ret != 0) {
1515 		TAILQ_FOREACH(opt, opts, link) {
1516 			if (strcmp(opt->name, "errmsg") == 0) {
1517 				strncpy((char *)opt->value, errmsg, opt->len);
1518 				break;
1519 			}
1520 		}
1521 		if (opt == NULL)
1522 			printf("%s\n", errmsg);
1523 	}
1524 	return (ret);
1525 }
1526 
1527 /*
1528  * Get a mount option by its name.
1529  *
1530  * Return 0 if the option was found, ENOENT otherwise.
1531  * If len is non-NULL it will be filled with the length
1532  * of the option. If buf is non-NULL, it will be filled
1533  * with the address of the option.
1534  */
1535 int
1536 vfs_getopt(opts, name, buf, len)
1537 	struct vfsoptlist *opts;
1538 	const char *name;
1539 	void **buf;
1540 	int *len;
1541 {
1542 	struct vfsopt *opt;
1543 
1544 	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1545 
1546 	TAILQ_FOREACH(opt, opts, link) {
1547 		if (strcmp(name, opt->name) == 0) {
1548 			opt->seen = 1;
1549 			if (len != NULL)
1550 				*len = opt->len;
1551 			if (buf != NULL)
1552 				*buf = opt->value;
1553 			return (0);
1554 		}
1555 	}
1556 	return (ENOENT);
1557 }
1558 
1559 int
1560 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
1561 {
1562 	struct vfsopt *opt;
1563 
1564 	if (opts == NULL)
1565 		return (-1);
1566 
1567 	TAILQ_FOREACH(opt, opts, link) {
1568 		if (strcmp(name, opt->name) == 0) {
1569 			opt->seen = 1;
1570 			return (opt->pos);
1571 		}
1572 	}
1573 	return (-1);
1574 }
1575 
1576 int
1577 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value)
1578 {
1579 	char *opt_value, *vtp;
1580 	quad_t iv;
1581 	int error, opt_len;
1582 
1583 	error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len);
1584 	if (error != 0)
1585 		return (error);
1586 	if (opt_len == 0 || opt_value == NULL)
1587 		return (EINVAL);
1588 	if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0')
1589 		return (EINVAL);
1590 	iv = strtoq(opt_value, &vtp, 0);
1591 	if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0'))
1592 		return (EINVAL);
1593 	if (iv < 0)
1594 		return (EINVAL);
1595 	switch (vtp[0]) {
1596 	case 't':
1597 	case 'T':
1598 		iv *= 1024;
1599 	case 'g':
1600 	case 'G':
1601 		iv *= 1024;
1602 	case 'm':
1603 	case 'M':
1604 		iv *= 1024;
1605 	case 'k':
1606 	case 'K':
1607 		iv *= 1024;
1608 	case '\0':
1609 		break;
1610 	default:
1611 		return (EINVAL);
1612 	}
1613 	*value = iv;
1614 
1615 	return (0);
1616 }
1617 
1618 char *
1619 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
1620 {
1621 	struct vfsopt *opt;
1622 
1623 	*error = 0;
1624 	TAILQ_FOREACH(opt, opts, link) {
1625 		if (strcmp(name, opt->name) != 0)
1626 			continue;
1627 		opt->seen = 1;
1628 		if (opt->len == 0 ||
1629 		    ((char *)opt->value)[opt->len - 1] != '\0') {
1630 			*error = EINVAL;
1631 			return (NULL);
1632 		}
1633 		return (opt->value);
1634 	}
1635 	*error = ENOENT;
1636 	return (NULL);
1637 }
1638 
1639 int
1640 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
1641 	uint64_t val)
1642 {
1643 	struct vfsopt *opt;
1644 
1645 	TAILQ_FOREACH(opt, opts, link) {
1646 		if (strcmp(name, opt->name) == 0) {
1647 			opt->seen = 1;
1648 			if (w != NULL)
1649 				*w |= val;
1650 			return (1);
1651 		}
1652 	}
1653 	if (w != NULL)
1654 		*w &= ~val;
1655 	return (0);
1656 }
1657 
1658 int
1659 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
1660 {
1661 	va_list ap;
1662 	struct vfsopt *opt;
1663 	int ret;
1664 
1665 	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1666 
1667 	TAILQ_FOREACH(opt, opts, link) {
1668 		if (strcmp(name, opt->name) != 0)
1669 			continue;
1670 		opt->seen = 1;
1671 		if (opt->len == 0 || opt->value == NULL)
1672 			return (0);
1673 		if (((char *)opt->value)[opt->len - 1] != '\0')
1674 			return (0);
1675 		va_start(ap, fmt);
1676 		ret = vsscanf(opt->value, fmt, ap);
1677 		va_end(ap);
1678 		return (ret);
1679 	}
1680 	return (0);
1681 }
1682 
1683 int
1684 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len)
1685 {
1686 	struct vfsopt *opt;
1687 
1688 	TAILQ_FOREACH(opt, opts, link) {
1689 		if (strcmp(name, opt->name) != 0)
1690 			continue;
1691 		opt->seen = 1;
1692 		if (opt->value == NULL)
1693 			opt->len = len;
1694 		else {
1695 			if (opt->len != len)
1696 				return (EINVAL);
1697 			bcopy(value, opt->value, len);
1698 		}
1699 		return (0);
1700 	}
1701 	return (ENOENT);
1702 }
1703 
1704 int
1705 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len)
1706 {
1707 	struct vfsopt *opt;
1708 
1709 	TAILQ_FOREACH(opt, opts, link) {
1710 		if (strcmp(name, opt->name) != 0)
1711 			continue;
1712 		opt->seen = 1;
1713 		if (opt->value == NULL)
1714 			opt->len = len;
1715 		else {
1716 			if (opt->len < len)
1717 				return (EINVAL);
1718 			opt->len = len;
1719 			bcopy(value, opt->value, len);
1720 		}
1721 		return (0);
1722 	}
1723 	return (ENOENT);
1724 }
1725 
1726 int
1727 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value)
1728 {
1729 	struct vfsopt *opt;
1730 
1731 	TAILQ_FOREACH(opt, opts, link) {
1732 		if (strcmp(name, opt->name) != 0)
1733 			continue;
1734 		opt->seen = 1;
1735 		if (opt->value == NULL)
1736 			opt->len = strlen(value) + 1;
1737 		else if (strlcpy(opt->value, value, opt->len) >= opt->len)
1738 			return (EINVAL);
1739 		return (0);
1740 	}
1741 	return (ENOENT);
1742 }
1743 
1744 /*
1745  * Find and copy a mount option.
1746  *
1747  * The size of the buffer has to be specified
1748  * in len, if it is not the same length as the
1749  * mount option, EINVAL is returned.
1750  * Returns ENOENT if the option is not found.
1751  */
1752 int
1753 vfs_copyopt(opts, name, dest, len)
1754 	struct vfsoptlist *opts;
1755 	const char *name;
1756 	void *dest;
1757 	int len;
1758 {
1759 	struct vfsopt *opt;
1760 
1761 	KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1762 
1763 	TAILQ_FOREACH(opt, opts, link) {
1764 		if (strcmp(name, opt->name) == 0) {
1765 			opt->seen = 1;
1766 			if (len != opt->len)
1767 				return (EINVAL);
1768 			bcopy(opt->value, dest, opt->len);
1769 			return (0);
1770 		}
1771 	}
1772 	return (ENOENT);
1773 }
1774 
1775 int
1776 __vfs_statfs(struct mount *mp, struct statfs *sbp)
1777 {
1778 	int error;
1779 
1780 	error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat);
1781 	if (sbp != &mp->mnt_stat)
1782 		*sbp = mp->mnt_stat;
1783 	return (error);
1784 }
1785 
1786 void
1787 vfs_mountedfrom(struct mount *mp, const char *from)
1788 {
1789 
1790 	bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
1791 	strlcpy(mp->mnt_stat.f_mntfromname, from,
1792 	    sizeof mp->mnt_stat.f_mntfromname);
1793 }
1794 
1795 /*
1796  * ---------------------------------------------------------------------
1797  * This is the api for building mount args and mounting filesystems from
1798  * inside the kernel.
1799  *
1800  * The API works by accumulation of individual args.  First error is
1801  * latched.
1802  *
1803  * XXX: should be documented in new manpage kernel_mount(9)
1804  */
1805 
1806 /* A memory allocation which must be freed when we are done */
1807 struct mntaarg {
1808 	SLIST_ENTRY(mntaarg)	next;
1809 };
1810 
1811 /* The header for the mount arguments */
1812 struct mntarg {
1813 	struct iovec *v;
1814 	int len;
1815 	int error;
1816 	SLIST_HEAD(, mntaarg)	list;
1817 };
1818 
1819 /*
1820  * Add a boolean argument.
1821  *
1822  * flag is the boolean value.
1823  * name must start with "no".
1824  */
1825 struct mntarg *
1826 mount_argb(struct mntarg *ma, int flag, const char *name)
1827 {
1828 
1829 	KASSERT(name[0] == 'n' && name[1] == 'o',
1830 	    ("mount_argb(...,%s): name must start with 'no'", name));
1831 
1832 	return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
1833 }
1834 
1835 /*
1836  * Add an argument printf style
1837  */
1838 struct mntarg *
1839 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
1840 {
1841 	va_list ap;
1842 	struct mntaarg *maa;
1843 	struct sbuf *sb;
1844 	int len;
1845 
1846 	if (ma == NULL) {
1847 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1848 		SLIST_INIT(&ma->list);
1849 	}
1850 	if (ma->error)
1851 		return (ma);
1852 
1853 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1854 	    M_MOUNT, M_WAITOK);
1855 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1856 	ma->v[ma->len].iov_len = strlen(name) + 1;
1857 	ma->len++;
1858 
1859 	sb = sbuf_new_auto();
1860 	va_start(ap, fmt);
1861 	sbuf_vprintf(sb, fmt, ap);
1862 	va_end(ap);
1863 	sbuf_finish(sb);
1864 	len = sbuf_len(sb) + 1;
1865 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1866 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1867 	bcopy(sbuf_data(sb), maa + 1, len);
1868 	sbuf_delete(sb);
1869 
1870 	ma->v[ma->len].iov_base = maa + 1;
1871 	ma->v[ma->len].iov_len = len;
1872 	ma->len++;
1873 
1874 	return (ma);
1875 }
1876 
1877 /*
1878  * Add an argument which is a userland string.
1879  */
1880 struct mntarg *
1881 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
1882 {
1883 	struct mntaarg *maa;
1884 	char *tbuf;
1885 
1886 	if (val == NULL)
1887 		return (ma);
1888 	if (ma == NULL) {
1889 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1890 		SLIST_INIT(&ma->list);
1891 	}
1892 	if (ma->error)
1893 		return (ma);
1894 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1895 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1896 	tbuf = (void *)(maa + 1);
1897 	ma->error = copyinstr(val, tbuf, len, NULL);
1898 	return (mount_arg(ma, name, tbuf, -1));
1899 }
1900 
1901 /*
1902  * Plain argument.
1903  *
1904  * If length is -1, treat value as a C string.
1905  */
1906 struct mntarg *
1907 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
1908 {
1909 
1910 	if (ma == NULL) {
1911 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1912 		SLIST_INIT(&ma->list);
1913 	}
1914 	if (ma->error)
1915 		return (ma);
1916 
1917 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1918 	    M_MOUNT, M_WAITOK);
1919 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1920 	ma->v[ma->len].iov_len = strlen(name) + 1;
1921 	ma->len++;
1922 
1923 	ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
1924 	if (len < 0)
1925 		ma->v[ma->len].iov_len = strlen(val) + 1;
1926 	else
1927 		ma->v[ma->len].iov_len = len;
1928 	ma->len++;
1929 	return (ma);
1930 }
1931 
1932 /*
1933  * Free a mntarg structure
1934  */
1935 static void
1936 free_mntarg(struct mntarg *ma)
1937 {
1938 	struct mntaarg *maa;
1939 
1940 	while (!SLIST_EMPTY(&ma->list)) {
1941 		maa = SLIST_FIRST(&ma->list);
1942 		SLIST_REMOVE_HEAD(&ma->list, next);
1943 		free(maa, M_MOUNT);
1944 	}
1945 	free(ma->v, M_MOUNT);
1946 	free(ma, M_MOUNT);
1947 }
1948 
1949 /*
1950  * Mount a filesystem
1951  */
1952 int
1953 kernel_mount(struct mntarg *ma, uint64_t flags)
1954 {
1955 	struct uio auio;
1956 	int error;
1957 
1958 	KASSERT(ma != NULL, ("kernel_mount NULL ma"));
1959 	KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
1960 	KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
1961 
1962 	auio.uio_iov = ma->v;
1963 	auio.uio_iovcnt = ma->len;
1964 	auio.uio_segflg = UIO_SYSSPACE;
1965 
1966 	error = ma->error;
1967 	if (!error)
1968 		error = vfs_donmount(curthread, flags, &auio);
1969 	free_mntarg(ma);
1970 	return (error);
1971 }
1972 
1973 /*
1974  * A printflike function to mount a filesystem.
1975  */
1976 int
1977 kernel_vmount(int flags, ...)
1978 {
1979 	struct mntarg *ma = NULL;
1980 	va_list ap;
1981 	const char *cp;
1982 	const void *vp;
1983 	int error;
1984 
1985 	va_start(ap, flags);
1986 	for (;;) {
1987 		cp = va_arg(ap, const char *);
1988 		if (cp == NULL)
1989 			break;
1990 		vp = va_arg(ap, const void *);
1991 		ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0));
1992 	}
1993 	va_end(ap);
1994 
1995 	error = kernel_mount(ma, flags);
1996 	return (error);
1997 }
1998 
1999 void
2000 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp)
2001 {
2002 
2003 	bcopy(oexp, exp, sizeof(*oexp));
2004 	exp->ex_numsecflavors = 0;
2005 }
2006