xref: /freebsd/sys/kern/vfs_mount.c (revision b9f654b163bce26de79705e77b872427c9f2afa1)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1999-2004 Poul-Henning Kamp
5  * Copyright (c) 1999 Michael Smith
6  * Copyright (c) 1989, 1993
7  *	The Regents of the University of California.  All rights reserved.
8  * (c) UNIX System Laboratories, Inc.
9  * All or some portions of this file are derived from material licensed
10  * to the University of California by American Telephone and Telegraph
11  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
12  * the permission of UNIX System Laboratories, Inc.
13  *
14  * Redistribution and use in source and binary forms, with or without
15  * modification, are permitted provided that the following conditions
16  * are met:
17  * 1. Redistributions of source code must retain the above copyright
18  *    notice, this list of conditions and the following disclaimer.
19  * 2. Redistributions in binary form must reproduce the above copyright
20  *    notice, this list of conditions and the following disclaimer in the
21  *    documentation and/or other materials provided with the distribution.
22  * 3. Neither the name of the University nor the names of its contributors
23  *    may be used to endorse or promote products derived from this software
24  *    without specific prior written permission.
25  *
26  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
27  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
30  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36  * SUCH DAMAGE.
37  */
38 
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41 
42 #include <sys/param.h>
43 #include <sys/conf.h>
44 #include <sys/eventhandler.h>
45 #include <sys/fcntl.h>
46 #include <sys/jail.h>
47 #include <sys/kernel.h>
48 #include <sys/ktr.h>
49 #include <sys/libkern.h>
50 #include <sys/malloc.h>
51 #include <sys/mount.h>
52 #include <sys/mutex.h>
53 #include <sys/namei.h>
54 #include <sys/priv.h>
55 #include <sys/proc.h>
56 #include <sys/filedesc.h>
57 #include <sys/reboot.h>
58 #include <sys/sbuf.h>
59 #include <sys/syscallsubr.h>
60 #include <sys/sysproto.h>
61 #include <sys/sx.h>
62 #include <sys/sysctl.h>
63 #include <sys/sysent.h>
64 #include <sys/systm.h>
65 #include <sys/vnode.h>
66 #include <vm/uma.h>
67 
68 #include <geom/geom.h>
69 
70 #include <machine/stdarg.h>
71 
72 #include <security/audit/audit.h>
73 #include <security/mac/mac_framework.h>
74 
75 #define	VFS_MOUNTARG_SIZE_MAX	(1024 * 64)
76 
77 static int	vfs_domount(struct thread *td, const char *fstype, char *fspath,
78 		    uint64_t fsflags, struct vfsoptlist **optlist);
79 static void	free_mntarg(struct mntarg *ma);
80 
81 static int	usermount = 0;
82 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0,
83     "Unprivileged users may mount and unmount file systems");
84 
85 static bool	default_autoro = false;
86 SYSCTL_BOOL(_vfs, OID_AUTO, default_autoro, CTLFLAG_RW, &default_autoro, 0,
87     "Retry failed r/w mount as r/o if no explicit ro/rw option is specified");
88 
89 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
90 MALLOC_DEFINE(M_STATFS, "statfs", "statfs structure");
91 static uma_zone_t mount_zone;
92 
93 /* List of mounted filesystems. */
94 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
95 
96 /* For any iteration/modification of mountlist */
97 struct mtx mountlist_mtx;
98 MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF);
99 
100 EVENTHANDLER_LIST_DEFINE(vfs_mounted);
101 EVENTHANDLER_LIST_DEFINE(vfs_unmounted);
102 
103 /*
104  * Global opts, taken by all filesystems
105  */
106 static const char *global_opts[] = {
107 	"errmsg",
108 	"fstype",
109 	"fspath",
110 	"ro",
111 	"rw",
112 	"nosuid",
113 	"noexec",
114 	NULL
115 };
116 
117 static int
118 mount_init(void *mem, int size, int flags)
119 {
120 	struct mount *mp;
121 
122 	mp = (struct mount *)mem;
123 	mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF);
124 	mtx_init(&mp->mnt_listmtx, "struct mount vlist mtx", NULL, MTX_DEF);
125 	lockinit(&mp->mnt_explock, PVFS, "explock", 0, 0);
126 	return (0);
127 }
128 
129 static void
130 mount_fini(void *mem, int size)
131 {
132 	struct mount *mp;
133 
134 	mp = (struct mount *)mem;
135 	lockdestroy(&mp->mnt_explock);
136 	mtx_destroy(&mp->mnt_listmtx);
137 	mtx_destroy(&mp->mnt_mtx);
138 }
139 
140 static void
141 vfs_mount_init(void *dummy __unused)
142 {
143 
144 	mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount), NULL,
145 	    NULL, mount_init, mount_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
146 }
147 SYSINIT(vfs_mount, SI_SUB_VFS, SI_ORDER_ANY, vfs_mount_init, NULL);
148 
149 /*
150  * ---------------------------------------------------------------------
151  * Functions for building and sanitizing the mount options
152  */
153 
154 /* Remove one mount option. */
155 static void
156 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
157 {
158 
159 	TAILQ_REMOVE(opts, opt, link);
160 	free(opt->name, M_MOUNT);
161 	if (opt->value != NULL)
162 		free(opt->value, M_MOUNT);
163 	free(opt, M_MOUNT);
164 }
165 
166 /* Release all resources related to the mount options. */
167 void
168 vfs_freeopts(struct vfsoptlist *opts)
169 {
170 	struct vfsopt *opt;
171 
172 	while (!TAILQ_EMPTY(opts)) {
173 		opt = TAILQ_FIRST(opts);
174 		vfs_freeopt(opts, opt);
175 	}
176 	free(opts, M_MOUNT);
177 }
178 
179 void
180 vfs_deleteopt(struct vfsoptlist *opts, const char *name)
181 {
182 	struct vfsopt *opt, *temp;
183 
184 	if (opts == NULL)
185 		return;
186 	TAILQ_FOREACH_SAFE(opt, opts, link, temp)  {
187 		if (strcmp(opt->name, name) == 0)
188 			vfs_freeopt(opts, opt);
189 	}
190 }
191 
192 static int
193 vfs_isopt_ro(const char *opt)
194 {
195 
196 	if (strcmp(opt, "ro") == 0 || strcmp(opt, "rdonly") == 0 ||
197 	    strcmp(opt, "norw") == 0)
198 		return (1);
199 	return (0);
200 }
201 
202 static int
203 vfs_isopt_rw(const char *opt)
204 {
205 
206 	if (strcmp(opt, "rw") == 0 || strcmp(opt, "noro") == 0)
207 		return (1);
208 	return (0);
209 }
210 
211 /*
212  * Check if options are equal (with or without the "no" prefix).
213  */
214 static int
215 vfs_equalopts(const char *opt1, const char *opt2)
216 {
217 	char *p;
218 
219 	/* "opt" vs. "opt" or "noopt" vs. "noopt" */
220 	if (strcmp(opt1, opt2) == 0)
221 		return (1);
222 	/* "noopt" vs. "opt" */
223 	if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
224 		return (1);
225 	/* "opt" vs. "noopt" */
226 	if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
227 		return (1);
228 	while ((p = strchr(opt1, '.')) != NULL &&
229 	    !strncmp(opt1, opt2, ++p - opt1)) {
230 		opt2 += p - opt1;
231 		opt1 = p;
232 		/* "foo.noopt" vs. "foo.opt" */
233 		if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
234 			return (1);
235 		/* "foo.opt" vs. "foo.noopt" */
236 		if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
237 			return (1);
238 	}
239 	/* "ro" / "rdonly" / "norw" / "rw" / "noro" */
240 	if ((vfs_isopt_ro(opt1) || vfs_isopt_rw(opt1)) &&
241 	    (vfs_isopt_ro(opt2) || vfs_isopt_rw(opt2)))
242 		return (1);
243 	return (0);
244 }
245 
246 /*
247  * If a mount option is specified several times,
248  * (with or without the "no" prefix) only keep
249  * the last occurrence of it.
250  */
251 static void
252 vfs_sanitizeopts(struct vfsoptlist *opts)
253 {
254 	struct vfsopt *opt, *opt2, *tmp;
255 
256 	TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
257 		opt2 = TAILQ_PREV(opt, vfsoptlist, link);
258 		while (opt2 != NULL) {
259 			if (vfs_equalopts(opt->name, opt2->name)) {
260 				tmp = TAILQ_PREV(opt2, vfsoptlist, link);
261 				vfs_freeopt(opts, opt2);
262 				opt2 = tmp;
263 			} else {
264 				opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
265 			}
266 		}
267 	}
268 }
269 
270 /*
271  * Build a linked list of mount options from a struct uio.
272  */
273 int
274 vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
275 {
276 	struct vfsoptlist *opts;
277 	struct vfsopt *opt;
278 	size_t memused, namelen, optlen;
279 	unsigned int i, iovcnt;
280 	int error;
281 
282 	opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
283 	TAILQ_INIT(opts);
284 	memused = 0;
285 	iovcnt = auio->uio_iovcnt;
286 	for (i = 0; i < iovcnt; i += 2) {
287 		namelen = auio->uio_iov[i].iov_len;
288 		optlen = auio->uio_iov[i + 1].iov_len;
289 		memused += sizeof(struct vfsopt) + optlen + namelen;
290 		/*
291 		 * Avoid consuming too much memory, and attempts to overflow
292 		 * memused.
293 		 */
294 		if (memused > VFS_MOUNTARG_SIZE_MAX ||
295 		    optlen > VFS_MOUNTARG_SIZE_MAX ||
296 		    namelen > VFS_MOUNTARG_SIZE_MAX) {
297 			error = EINVAL;
298 			goto bad;
299 		}
300 
301 		opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
302 		opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
303 		opt->value = NULL;
304 		opt->len = 0;
305 		opt->pos = i / 2;
306 		opt->seen = 0;
307 
308 		/*
309 		 * Do this early, so jumps to "bad" will free the current
310 		 * option.
311 		 */
312 		TAILQ_INSERT_TAIL(opts, opt, link);
313 
314 		if (auio->uio_segflg == UIO_SYSSPACE) {
315 			bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
316 		} else {
317 			error = copyin(auio->uio_iov[i].iov_base, opt->name,
318 			    namelen);
319 			if (error)
320 				goto bad;
321 		}
322 		/* Ensure names are null-terminated strings. */
323 		if (namelen == 0 || opt->name[namelen - 1] != '\0') {
324 			error = EINVAL;
325 			goto bad;
326 		}
327 		if (optlen != 0) {
328 			opt->len = optlen;
329 			opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
330 			if (auio->uio_segflg == UIO_SYSSPACE) {
331 				bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
332 				    optlen);
333 			} else {
334 				error = copyin(auio->uio_iov[i + 1].iov_base,
335 				    opt->value, optlen);
336 				if (error)
337 					goto bad;
338 			}
339 		}
340 	}
341 	vfs_sanitizeopts(opts);
342 	*options = opts;
343 	return (0);
344 bad:
345 	vfs_freeopts(opts);
346 	return (error);
347 }
348 
349 /*
350  * Merge the old mount options with the new ones passed
351  * in the MNT_UPDATE case.
352  *
353  * XXX: This function will keep a "nofoo" option in the new
354  * options.  E.g, if the option's canonical name is "foo",
355  * "nofoo" ends up in the mount point's active options.
356  */
357 static void
358 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *oldopts)
359 {
360 	struct vfsopt *opt, *new;
361 
362 	TAILQ_FOREACH(opt, oldopts, link) {
363 		new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
364 		new->name = strdup(opt->name, M_MOUNT);
365 		if (opt->len != 0) {
366 			new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
367 			bcopy(opt->value, new->value, opt->len);
368 		} else
369 			new->value = NULL;
370 		new->len = opt->len;
371 		new->seen = opt->seen;
372 		TAILQ_INSERT_HEAD(toopts, new, link);
373 	}
374 	vfs_sanitizeopts(toopts);
375 }
376 
377 /*
378  * Mount a filesystem.
379  */
380 #ifndef _SYS_SYSPROTO_H_
381 struct nmount_args {
382 	struct iovec *iovp;
383 	unsigned int iovcnt;
384 	int flags;
385 };
386 #endif
387 int
388 sys_nmount(struct thread *td, struct nmount_args *uap)
389 {
390 	struct uio *auio;
391 	int error;
392 	u_int iovcnt;
393 	uint64_t flags;
394 
395 	/*
396 	 * Mount flags are now 64-bits. On 32-bit archtectures only
397 	 * 32-bits are passed in, but from here on everything handles
398 	 * 64-bit flags correctly.
399 	 */
400 	flags = uap->flags;
401 
402 	AUDIT_ARG_FFLAGS(flags);
403 	CTR4(KTR_VFS, "%s: iovp %p with iovcnt %d and flags %d", __func__,
404 	    uap->iovp, uap->iovcnt, flags);
405 
406 	/*
407 	 * Filter out MNT_ROOTFS.  We do not want clients of nmount() in
408 	 * userspace to set this flag, but we must filter it out if we want
409 	 * MNT_UPDATE on the root file system to work.
410 	 * MNT_ROOTFS should only be set by the kernel when mounting its
411 	 * root file system.
412 	 */
413 	flags &= ~MNT_ROOTFS;
414 
415 	iovcnt = uap->iovcnt;
416 	/*
417 	 * Check that we have an even number of iovec's
418 	 * and that we have at least two options.
419 	 */
420 	if ((iovcnt & 1) || (iovcnt < 4)) {
421 		CTR2(KTR_VFS, "%s: failed for invalid iovcnt %d", __func__,
422 		    uap->iovcnt);
423 		return (EINVAL);
424 	}
425 
426 	error = copyinuio(uap->iovp, iovcnt, &auio);
427 	if (error) {
428 		CTR2(KTR_VFS, "%s: failed for invalid uio op with %d errno",
429 		    __func__, error);
430 		return (error);
431 	}
432 	error = vfs_donmount(td, flags, auio);
433 
434 	free(auio, M_IOV);
435 	return (error);
436 }
437 
438 /*
439  * ---------------------------------------------------------------------
440  * Various utility functions
441  */
442 
443 void
444 vfs_ref(struct mount *mp)
445 {
446 
447 	CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
448 	MNT_ILOCK(mp);
449 	MNT_REF(mp);
450 	MNT_IUNLOCK(mp);
451 }
452 
453 void
454 vfs_rel(struct mount *mp)
455 {
456 
457 	CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
458 	MNT_ILOCK(mp);
459 	MNT_REL(mp);
460 	MNT_IUNLOCK(mp);
461 }
462 
463 /*
464  * Allocate and initialize the mount point struct.
465  */
466 struct mount *
467 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, const char *fspath,
468     struct ucred *cred)
469 {
470 	struct mount *mp;
471 
472 	mp = uma_zalloc(mount_zone, M_WAITOK);
473 	bzero(&mp->mnt_startzero,
474 	    __rangeof(struct mount, mnt_startzero, mnt_endzero));
475 	TAILQ_INIT(&mp->mnt_nvnodelist);
476 	mp->mnt_nvnodelistsize = 0;
477 	TAILQ_INIT(&mp->mnt_activevnodelist);
478 	mp->mnt_activevnodelistsize = 0;
479 	TAILQ_INIT(&mp->mnt_tmpfreevnodelist);
480 	mp->mnt_tmpfreevnodelistsize = 0;
481 	mp->mnt_ref = 0;
482 	(void) vfs_busy(mp, MBF_NOWAIT);
483 	atomic_add_acq_int(&vfsp->vfc_refcount, 1);
484 	mp->mnt_op = vfsp->vfc_vfsops;
485 	mp->mnt_vfc = vfsp;
486 	mp->mnt_stat.f_type = vfsp->vfc_typenum;
487 	mp->mnt_gen++;
488 	strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
489 	mp->mnt_vnodecovered = vp;
490 	mp->mnt_cred = crdup(cred);
491 	mp->mnt_stat.f_owner = cred->cr_uid;
492 	strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
493 	mp->mnt_iosize_max = DFLTPHYS;
494 #ifdef MAC
495 	mac_mount_init(mp);
496 	mac_mount_create(cred, mp);
497 #endif
498 	arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0);
499 	TAILQ_INIT(&mp->mnt_uppers);
500 	return (mp);
501 }
502 
503 /*
504  * Destroy the mount struct previously allocated by vfs_mount_alloc().
505  */
506 void
507 vfs_mount_destroy(struct mount *mp)
508 {
509 
510 	MNT_ILOCK(mp);
511 	mp->mnt_kern_flag |= MNTK_REFEXPIRE;
512 	if (mp->mnt_kern_flag & MNTK_MWAIT) {
513 		mp->mnt_kern_flag &= ~MNTK_MWAIT;
514 		wakeup(mp);
515 	}
516 	while (mp->mnt_ref)
517 		msleep(mp, MNT_MTX(mp), PVFS, "mntref", 0);
518 	KASSERT(mp->mnt_ref == 0,
519 	    ("%s: invalid refcount in the drain path @ %s:%d", __func__,
520 	    __FILE__, __LINE__));
521 	if (mp->mnt_writeopcount != 0)
522 		panic("vfs_mount_destroy: nonzero writeopcount");
523 	if (mp->mnt_secondary_writes != 0)
524 		panic("vfs_mount_destroy: nonzero secondary_writes");
525 	atomic_subtract_rel_int(&mp->mnt_vfc->vfc_refcount, 1);
526 	if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
527 		struct vnode *vp;
528 
529 		TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
530 			vn_printf(vp, "dangling vnode ");
531 		panic("unmount: dangling vnode");
532 	}
533 	KASSERT(TAILQ_EMPTY(&mp->mnt_uppers), ("mnt_uppers"));
534 	if (mp->mnt_nvnodelistsize != 0)
535 		panic("vfs_mount_destroy: nonzero nvnodelistsize");
536 	if (mp->mnt_activevnodelistsize != 0)
537 		panic("vfs_mount_destroy: nonzero activevnodelistsize");
538 	if (mp->mnt_tmpfreevnodelistsize != 0)
539 		panic("vfs_mount_destroy: nonzero tmpfreevnodelistsize");
540 	if (mp->mnt_lockref != 0)
541 		panic("vfs_mount_destroy: nonzero lock refcount");
542 	MNT_IUNLOCK(mp);
543 	if (mp->mnt_vnodecovered != NULL)
544 		vrele(mp->mnt_vnodecovered);
545 #ifdef MAC
546 	mac_mount_destroy(mp);
547 #endif
548 	if (mp->mnt_opt != NULL)
549 		vfs_freeopts(mp->mnt_opt);
550 	crfree(mp->mnt_cred);
551 	uma_zfree(mount_zone, mp);
552 }
553 
554 static bool
555 vfs_should_downgrade_to_ro_mount(uint64_t fsflags, int error)
556 {
557 	/* This is an upgrade of an exisiting mount. */
558 	if ((fsflags & MNT_UPDATE) != 0)
559 		return (false);
560 	/* This is already an R/O mount. */
561 	if ((fsflags & MNT_RDONLY) != 0)
562 		return (false);
563 
564 	switch (error) {
565 	case ENODEV:	/* generic, geom, ... */
566 	case EACCES:	/* cam/scsi, ... */
567 	case EROFS:	/* md, mmcsd, ... */
568 		/*
569 		 * These errors can be returned by the storage layer to signal
570 		 * that the media is read-only.  No harm in the R/O mount
571 		 * attempt if the error was returned for some other reason.
572 		 */
573 		return (true);
574 	default:
575 		return (false);
576 	}
577 }
578 
579 int
580 vfs_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions)
581 {
582 	struct vfsoptlist *optlist;
583 	struct vfsopt *opt, *tmp_opt;
584 	char *fstype, *fspath, *errmsg;
585 	int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
586 	bool autoro;
587 
588 	errmsg = fspath = NULL;
589 	errmsg_len = fspathlen = 0;
590 	errmsg_pos = -1;
591 	autoro = default_autoro;
592 
593 	error = vfs_buildopts(fsoptions, &optlist);
594 	if (error)
595 		return (error);
596 
597 	if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
598 		errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
599 
600 	/*
601 	 * We need these two options before the others,
602 	 * and they are mandatory for any filesystem.
603 	 * Ensure they are NUL terminated as well.
604 	 */
605 	fstypelen = 0;
606 	error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
607 	if (error || fstype[fstypelen - 1] != '\0') {
608 		error = EINVAL;
609 		if (errmsg != NULL)
610 			strncpy(errmsg, "Invalid fstype", errmsg_len);
611 		goto bail;
612 	}
613 	fspathlen = 0;
614 	error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
615 	if (error || fspath[fspathlen - 1] != '\0') {
616 		error = EINVAL;
617 		if (errmsg != NULL)
618 			strncpy(errmsg, "Invalid fspath", errmsg_len);
619 		goto bail;
620 	}
621 
622 	/*
623 	 * We need to see if we have the "update" option
624 	 * before we call vfs_domount(), since vfs_domount() has special
625 	 * logic based on MNT_UPDATE.  This is very important
626 	 * when we want to update the root filesystem.
627 	 */
628 	TAILQ_FOREACH_SAFE(opt, optlist, link, tmp_opt) {
629 		if (strcmp(opt->name, "update") == 0) {
630 			fsflags |= MNT_UPDATE;
631 			vfs_freeopt(optlist, opt);
632 		}
633 		else if (strcmp(opt->name, "async") == 0)
634 			fsflags |= MNT_ASYNC;
635 		else if (strcmp(opt->name, "force") == 0) {
636 			fsflags |= MNT_FORCE;
637 			vfs_freeopt(optlist, opt);
638 		}
639 		else if (strcmp(opt->name, "reload") == 0) {
640 			fsflags |= MNT_RELOAD;
641 			vfs_freeopt(optlist, opt);
642 		}
643 		else if (strcmp(opt->name, "multilabel") == 0)
644 			fsflags |= MNT_MULTILABEL;
645 		else if (strcmp(opt->name, "noasync") == 0)
646 			fsflags &= ~MNT_ASYNC;
647 		else if (strcmp(opt->name, "noatime") == 0)
648 			fsflags |= MNT_NOATIME;
649 		else if (strcmp(opt->name, "atime") == 0) {
650 			free(opt->name, M_MOUNT);
651 			opt->name = strdup("nonoatime", M_MOUNT);
652 		}
653 		else if (strcmp(opt->name, "noclusterr") == 0)
654 			fsflags |= MNT_NOCLUSTERR;
655 		else if (strcmp(opt->name, "clusterr") == 0) {
656 			free(opt->name, M_MOUNT);
657 			opt->name = strdup("nonoclusterr", M_MOUNT);
658 		}
659 		else if (strcmp(opt->name, "noclusterw") == 0)
660 			fsflags |= MNT_NOCLUSTERW;
661 		else if (strcmp(opt->name, "clusterw") == 0) {
662 			free(opt->name, M_MOUNT);
663 			opt->name = strdup("nonoclusterw", M_MOUNT);
664 		}
665 		else if (strcmp(opt->name, "noexec") == 0)
666 			fsflags |= MNT_NOEXEC;
667 		else if (strcmp(opt->name, "exec") == 0) {
668 			free(opt->name, M_MOUNT);
669 			opt->name = strdup("nonoexec", M_MOUNT);
670 		}
671 		else if (strcmp(opt->name, "nosuid") == 0)
672 			fsflags |= MNT_NOSUID;
673 		else if (strcmp(opt->name, "suid") == 0) {
674 			free(opt->name, M_MOUNT);
675 			opt->name = strdup("nonosuid", M_MOUNT);
676 		}
677 		else if (strcmp(opt->name, "nosymfollow") == 0)
678 			fsflags |= MNT_NOSYMFOLLOW;
679 		else if (strcmp(opt->name, "symfollow") == 0) {
680 			free(opt->name, M_MOUNT);
681 			opt->name = strdup("nonosymfollow", M_MOUNT);
682 		}
683 		else if (strcmp(opt->name, "noro") == 0) {
684 			fsflags &= ~MNT_RDONLY;
685 			autoro = false;
686 		}
687 		else if (strcmp(opt->name, "rw") == 0) {
688 			fsflags &= ~MNT_RDONLY;
689 			autoro = false;
690 		}
691 		else if (strcmp(opt->name, "ro") == 0) {
692 			fsflags |= MNT_RDONLY;
693 			autoro = false;
694 		}
695 		else if (strcmp(opt->name, "rdonly") == 0) {
696 			free(opt->name, M_MOUNT);
697 			opt->name = strdup("ro", M_MOUNT);
698 			fsflags |= MNT_RDONLY;
699 			autoro = false;
700 		}
701 		else if (strcmp(opt->name, "autoro") == 0) {
702 			vfs_freeopt(optlist, opt);
703 			autoro = true;
704 		}
705 		else if (strcmp(opt->name, "suiddir") == 0)
706 			fsflags |= MNT_SUIDDIR;
707 		else if (strcmp(opt->name, "sync") == 0)
708 			fsflags |= MNT_SYNCHRONOUS;
709 		else if (strcmp(opt->name, "union") == 0)
710 			fsflags |= MNT_UNION;
711 		else if (strcmp(opt->name, "automounted") == 0) {
712 			fsflags |= MNT_AUTOMOUNTED;
713 			vfs_freeopt(optlist, opt);
714 		}
715 	}
716 
717 	/*
718 	 * Be ultra-paranoid about making sure the type and fspath
719 	 * variables will fit in our mp buffers, including the
720 	 * terminating NUL.
721 	 */
722 	if (fstypelen > MFSNAMELEN || fspathlen > MNAMELEN) {
723 		error = ENAMETOOLONG;
724 		goto bail;
725 	}
726 
727 	error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
728 
729 	/*
730 	 * See if we can mount in the read-only mode if the error code suggests
731 	 * that it could be possible and the mount options allow for that.
732 	 * Never try it if "[no]{ro|rw}" has been explicitly requested and not
733 	 * overridden by "autoro".
734 	 */
735 	if (autoro && vfs_should_downgrade_to_ro_mount(fsflags, error)) {
736 		printf("%s: R/W mount failed, possibly R/O media,"
737 		    " trying R/O mount\n", __func__);
738 		fsflags |= MNT_RDONLY;
739 		error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
740 	}
741 bail:
742 	/* copyout the errmsg */
743 	if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
744 	    && errmsg_len > 0 && errmsg != NULL) {
745 		if (fsoptions->uio_segflg == UIO_SYSSPACE) {
746 			bcopy(errmsg,
747 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
748 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
749 		} else {
750 			copyout(errmsg,
751 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
752 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
753 		}
754 	}
755 
756 	if (optlist != NULL)
757 		vfs_freeopts(optlist);
758 	return (error);
759 }
760 
761 /*
762  * Old mount API.
763  */
764 #ifndef _SYS_SYSPROTO_H_
765 struct mount_args {
766 	char	*type;
767 	char	*path;
768 	int	flags;
769 	caddr_t	data;
770 };
771 #endif
772 /* ARGSUSED */
773 int
774 sys_mount(struct thread *td, struct mount_args *uap)
775 {
776 	char *fstype;
777 	struct vfsconf *vfsp = NULL;
778 	struct mntarg *ma = NULL;
779 	uint64_t flags;
780 	int error;
781 
782 	/*
783 	 * Mount flags are now 64-bits. On 32-bit architectures only
784 	 * 32-bits are passed in, but from here on everything handles
785 	 * 64-bit flags correctly.
786 	 */
787 	flags = uap->flags;
788 
789 	AUDIT_ARG_FFLAGS(flags);
790 
791 	/*
792 	 * Filter out MNT_ROOTFS.  We do not want clients of mount() in
793 	 * userspace to set this flag, but we must filter it out if we want
794 	 * MNT_UPDATE on the root file system to work.
795 	 * MNT_ROOTFS should only be set by the kernel when mounting its
796 	 * root file system.
797 	 */
798 	flags &= ~MNT_ROOTFS;
799 
800 	fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
801 	error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
802 	if (error) {
803 		free(fstype, M_TEMP);
804 		return (error);
805 	}
806 
807 	AUDIT_ARG_TEXT(fstype);
808 	vfsp = vfs_byname_kld(fstype, td, &error);
809 	free(fstype, M_TEMP);
810 	if (vfsp == NULL)
811 		return (ENOENT);
812 	if (((vfsp->vfc_flags & VFCF_SBDRY) != 0 &&
813 	    vfsp->vfc_vfsops_sd->vfs_cmount == NULL) ||
814 	    ((vfsp->vfc_flags & VFCF_SBDRY) == 0 &&
815 	    vfsp->vfc_vfsops->vfs_cmount == NULL))
816 		return (EOPNOTSUPP);
817 
818 	ma = mount_argsu(ma, "fstype", uap->type, MFSNAMELEN);
819 	ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
820 	ma = mount_argb(ma, flags & MNT_RDONLY, "noro");
821 	ma = mount_argb(ma, !(flags & MNT_NOSUID), "nosuid");
822 	ma = mount_argb(ma, !(flags & MNT_NOEXEC), "noexec");
823 
824 	if ((vfsp->vfc_flags & VFCF_SBDRY) != 0)
825 		return (vfsp->vfc_vfsops_sd->vfs_cmount(ma, uap->data, flags));
826 	return (vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags));
827 }
828 
829 /*
830  * vfs_domount_first(): first file system mount (not update)
831  */
832 static int
833 vfs_domount_first(
834 	struct thread *td,		/* Calling thread. */
835 	struct vfsconf *vfsp,		/* File system type. */
836 	char *fspath,			/* Mount path. */
837 	struct vnode *vp,		/* Vnode to be covered. */
838 	uint64_t fsflags,		/* Flags common to all filesystems. */
839 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
840 	)
841 {
842 	struct vattr va;
843 	struct mount *mp;
844 	struct vnode *newdp;
845 	int error, error1;
846 
847 	ASSERT_VOP_ELOCKED(vp, __func__);
848 	KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
849 
850 	/*
851 	 * If the jail of the calling thread lacks permission for this type of
852 	 * file system, deny immediately.
853 	 */
854 	if (jailed(td->td_ucred) && !prison_allow(td->td_ucred,
855 	    vfsp->vfc_prison_flag)) {
856 		vput(vp);
857 		return (EPERM);
858 	}
859 
860 	/*
861 	 * If the user is not root, ensure that they own the directory
862 	 * onto which we are attempting to mount.
863 	 */
864 	error = VOP_GETATTR(vp, &va, td->td_ucred);
865 	if (error == 0 && va.va_uid != td->td_ucred->cr_uid)
866 		error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN);
867 	if (error == 0)
868 		error = vinvalbuf(vp, V_SAVE, 0, 0);
869 	if (error == 0 && vp->v_type != VDIR)
870 		error = ENOTDIR;
871 	if (error == 0) {
872 		VI_LOCK(vp);
873 		if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
874 			vp->v_iflag |= VI_MOUNT;
875 		else
876 			error = EBUSY;
877 		VI_UNLOCK(vp);
878 	}
879 	if (error != 0) {
880 		vput(vp);
881 		return (error);
882 	}
883 	VOP_UNLOCK(vp, 0);
884 
885 	/* Allocate and initialize the filesystem. */
886 	mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred);
887 	/* XXXMAC: pass to vfs_mount_alloc? */
888 	mp->mnt_optnew = *optlist;
889 	/* Set the mount level flags. */
890 	mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY));
891 
892 	/*
893 	 * Mount the filesystem.
894 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
895 	 * get.  No freeing of cn_pnbuf.
896 	 */
897 	error1 = 0;
898 	if ((error = VFS_MOUNT(mp)) != 0 ||
899 	    (error1 = VFS_STATFS(mp, &mp->mnt_stat)) != 0 ||
900 	    (error1 = VFS_ROOT(mp, LK_EXCLUSIVE, &newdp)) != 0) {
901 		if (error1 != 0) {
902 			error = error1;
903 			if ((error1 = VFS_UNMOUNT(mp, 0)) != 0)
904 				printf("VFS_UNMOUNT returned %d\n", error1);
905 		}
906 		vfs_unbusy(mp);
907 		mp->mnt_vnodecovered = NULL;
908 		vfs_mount_destroy(mp);
909 		VI_LOCK(vp);
910 		vp->v_iflag &= ~VI_MOUNT;
911 		VI_UNLOCK(vp);
912 		vrele(vp);
913 		return (error);
914 	}
915 	VOP_UNLOCK(newdp, 0);
916 
917 	if (mp->mnt_opt != NULL)
918 		vfs_freeopts(mp->mnt_opt);
919 	mp->mnt_opt = mp->mnt_optnew;
920 	*optlist = NULL;
921 
922 	/*
923 	 * Prevent external consumers of mount options from reading mnt_optnew.
924 	 */
925 	mp->mnt_optnew = NULL;
926 
927 	MNT_ILOCK(mp);
928 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
929 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
930 		mp->mnt_kern_flag |= MNTK_ASYNC;
931 	else
932 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
933 	MNT_IUNLOCK(mp);
934 
935 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
936 	cache_purge(vp);
937 	VI_LOCK(vp);
938 	vp->v_iflag &= ~VI_MOUNT;
939 	VI_UNLOCK(vp);
940 	vp->v_mountedhere = mp;
941 	/* Place the new filesystem at the end of the mount list. */
942 	mtx_lock(&mountlist_mtx);
943 	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
944 	mtx_unlock(&mountlist_mtx);
945 	vfs_event_signal(NULL, VQ_MOUNT, 0);
946 	vn_lock(newdp, LK_EXCLUSIVE | LK_RETRY);
947 	VOP_UNLOCK(vp, 0);
948 	EVENTHANDLER_DIRECT_INVOKE(vfs_mounted, mp, newdp, td);
949 	VOP_UNLOCK(newdp, 0);
950 	mountcheckdirs(vp, newdp);
951 	vrele(newdp);
952 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
953 		vfs_allocate_syncvnode(mp);
954 	vfs_unbusy(mp);
955 	return (0);
956 }
957 
958 /*
959  * vfs_domount_update(): update of mounted file system
960  */
961 static int
962 vfs_domount_update(
963 	struct thread *td,		/* Calling thread. */
964 	struct vnode *vp,		/* Mount point vnode. */
965 	uint64_t fsflags,		/* Flags common to all filesystems. */
966 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
967 	)
968 {
969 	struct export_args export;
970 	void *bufp;
971 	struct mount *mp;
972 	int error, export_error, len;
973 	uint64_t flag;
974 
975 	ASSERT_VOP_ELOCKED(vp, __func__);
976 	KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
977 	mp = vp->v_mount;
978 
979 	if ((vp->v_vflag & VV_ROOT) == 0) {
980 		if (vfs_copyopt(*optlist, "export", &export, sizeof(export))
981 		    == 0)
982 			error = EXDEV;
983 		else
984 			error = EINVAL;
985 		vput(vp);
986 		return (error);
987 	}
988 
989 	/*
990 	 * We only allow the filesystem to be reloaded if it
991 	 * is currently mounted read-only.
992 	 */
993 	flag = mp->mnt_flag;
994 	if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
995 		vput(vp);
996 		return (EOPNOTSUPP);	/* Needs translation */
997 	}
998 	/*
999 	 * Only privileged root, or (if MNT_USER is set) the user that
1000 	 * did the original mount is permitted to update it.
1001 	 */
1002 	error = vfs_suser(mp, td);
1003 	if (error != 0) {
1004 		vput(vp);
1005 		return (error);
1006 	}
1007 	if (vfs_busy(mp, MBF_NOWAIT)) {
1008 		vput(vp);
1009 		return (EBUSY);
1010 	}
1011 	VI_LOCK(vp);
1012 	if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
1013 		VI_UNLOCK(vp);
1014 		vfs_unbusy(mp);
1015 		vput(vp);
1016 		return (EBUSY);
1017 	}
1018 	vp->v_iflag |= VI_MOUNT;
1019 	VI_UNLOCK(vp);
1020 	VOP_UNLOCK(vp, 0);
1021 
1022 	MNT_ILOCK(mp);
1023 	if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0) {
1024 		MNT_IUNLOCK(mp);
1025 		error = EBUSY;
1026 		goto end;
1027 	}
1028 	mp->mnt_flag &= ~MNT_UPDATEMASK;
1029 	mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE |
1030 	    MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY);
1031 	if ((mp->mnt_flag & MNT_ASYNC) == 0)
1032 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
1033 	MNT_IUNLOCK(mp);
1034 	mp->mnt_optnew = *optlist;
1035 	vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
1036 
1037 	/*
1038 	 * Mount the filesystem.
1039 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
1040 	 * get.  No freeing of cn_pnbuf.
1041 	 */
1042 	error = VFS_MOUNT(mp);
1043 
1044 	export_error = 0;
1045 	/* Process the export option. */
1046 	if (error == 0 && vfs_getopt(mp->mnt_optnew, "export", &bufp,
1047 	    &len) == 0) {
1048 		/* Assume that there is only 1 ABI for each length. */
1049 		switch (len) {
1050 		case (sizeof(struct oexport_args)):
1051 			bzero(&export, sizeof(export));
1052 			/* FALLTHROUGH */
1053 		case (sizeof(export)):
1054 			bcopy(bufp, &export, len);
1055 			export_error = vfs_export(mp, &export);
1056 			break;
1057 		default:
1058 			export_error = EINVAL;
1059 			break;
1060 		}
1061 	}
1062 
1063 	MNT_ILOCK(mp);
1064 	if (error == 0) {
1065 		mp->mnt_flag &=	~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
1066 		    MNT_SNAPSHOT);
1067 	} else {
1068 		/*
1069 		 * If we fail, restore old mount flags. MNT_QUOTA is special,
1070 		 * because it is not part of MNT_UPDATEMASK, but it could have
1071 		 * changed in the meantime if quotactl(2) was called.
1072 		 * All in all we want current value of MNT_QUOTA, not the old
1073 		 * one.
1074 		 */
1075 		mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
1076 	}
1077 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1078 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1079 		mp->mnt_kern_flag |= MNTK_ASYNC;
1080 	else
1081 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
1082 	MNT_IUNLOCK(mp);
1083 
1084 	if (error != 0)
1085 		goto end;
1086 
1087 	if (mp->mnt_opt != NULL)
1088 		vfs_freeopts(mp->mnt_opt);
1089 	mp->mnt_opt = mp->mnt_optnew;
1090 	*optlist = NULL;
1091 	(void)VFS_STATFS(mp, &mp->mnt_stat);
1092 	/*
1093 	 * Prevent external consumers of mount options from reading
1094 	 * mnt_optnew.
1095 	 */
1096 	mp->mnt_optnew = NULL;
1097 
1098 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
1099 		vfs_allocate_syncvnode(mp);
1100 	else
1101 		vfs_deallocate_syncvnode(mp);
1102 end:
1103 	vfs_unbusy(mp);
1104 	VI_LOCK(vp);
1105 	vp->v_iflag &= ~VI_MOUNT;
1106 	VI_UNLOCK(vp);
1107 	vrele(vp);
1108 	return (error != 0 ? error : export_error);
1109 }
1110 
1111 /*
1112  * vfs_domount(): actually attempt a filesystem mount.
1113  */
1114 static int
1115 vfs_domount(
1116 	struct thread *td,		/* Calling thread. */
1117 	const char *fstype,		/* Filesystem type. */
1118 	char *fspath,			/* Mount path. */
1119 	uint64_t fsflags,		/* Flags common to all filesystems. */
1120 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
1121 	)
1122 {
1123 	struct vfsconf *vfsp;
1124 	struct nameidata nd;
1125 	struct vnode *vp;
1126 	char *pathbuf;
1127 	int error;
1128 
1129 	/*
1130 	 * Be ultra-paranoid about making sure the type and fspath
1131 	 * variables will fit in our mp buffers, including the
1132 	 * terminating NUL.
1133 	 */
1134 	if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
1135 		return (ENAMETOOLONG);
1136 
1137 	if (jailed(td->td_ucred) || usermount == 0) {
1138 		if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
1139 			return (error);
1140 	}
1141 
1142 	/*
1143 	 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
1144 	 */
1145 	if (fsflags & MNT_EXPORTED) {
1146 		error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
1147 		if (error)
1148 			return (error);
1149 	}
1150 	if (fsflags & MNT_SUIDDIR) {
1151 		error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
1152 		if (error)
1153 			return (error);
1154 	}
1155 	/*
1156 	 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
1157 	 */
1158 	if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
1159 		if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
1160 			fsflags |= MNT_NOSUID | MNT_USER;
1161 	}
1162 
1163 	/* Load KLDs before we lock the covered vnode to avoid reversals. */
1164 	vfsp = NULL;
1165 	if ((fsflags & MNT_UPDATE) == 0) {
1166 		/* Don't try to load KLDs if we're mounting the root. */
1167 		if (fsflags & MNT_ROOTFS)
1168 			vfsp = vfs_byname(fstype);
1169 		else
1170 			vfsp = vfs_byname_kld(fstype, td, &error);
1171 		if (vfsp == NULL)
1172 			return (ENODEV);
1173 	}
1174 
1175 	/*
1176 	 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
1177 	 */
1178 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1179 	    UIO_SYSSPACE, fspath, td);
1180 	error = namei(&nd);
1181 	if (error != 0)
1182 		return (error);
1183 	NDFREE(&nd, NDF_ONLY_PNBUF);
1184 	vp = nd.ni_vp;
1185 	if ((fsflags & MNT_UPDATE) == 0) {
1186 		pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1187 		strcpy(pathbuf, fspath);
1188 		error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN);
1189 		/* debug.disablefullpath == 1 results in ENODEV */
1190 		if (error == 0 || error == ENODEV) {
1191 			error = vfs_domount_first(td, vfsp, pathbuf, vp,
1192 			    fsflags, optlist);
1193 		}
1194 		free(pathbuf, M_TEMP);
1195 	} else
1196 		error = vfs_domount_update(td, vp, fsflags, optlist);
1197 
1198 	return (error);
1199 }
1200 
1201 /*
1202  * Unmount a filesystem.
1203  *
1204  * Note: unmount takes a path to the vnode mounted on as argument, not
1205  * special file (as before).
1206  */
1207 #ifndef _SYS_SYSPROTO_H_
1208 struct unmount_args {
1209 	char	*path;
1210 	int	flags;
1211 };
1212 #endif
1213 /* ARGSUSED */
1214 int
1215 sys_unmount(struct thread *td, struct unmount_args *uap)
1216 {
1217 	struct nameidata nd;
1218 	struct mount *mp;
1219 	char *pathbuf;
1220 	int error, id0, id1;
1221 
1222 	AUDIT_ARG_VALUE(uap->flags);
1223 	if (jailed(td->td_ucred) || usermount == 0) {
1224 		error = priv_check(td, PRIV_VFS_UNMOUNT);
1225 		if (error)
1226 			return (error);
1227 	}
1228 
1229 	pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1230 	error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1231 	if (error) {
1232 		free(pathbuf, M_TEMP);
1233 		return (error);
1234 	}
1235 	if (uap->flags & MNT_BYFSID) {
1236 		AUDIT_ARG_TEXT(pathbuf);
1237 		/* Decode the filesystem ID. */
1238 		if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
1239 			free(pathbuf, M_TEMP);
1240 			return (EINVAL);
1241 		}
1242 
1243 		mtx_lock(&mountlist_mtx);
1244 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1245 			if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1246 			    mp->mnt_stat.f_fsid.val[1] == id1) {
1247 				vfs_ref(mp);
1248 				break;
1249 			}
1250 		}
1251 		mtx_unlock(&mountlist_mtx);
1252 	} else {
1253 		/*
1254 		 * Try to find global path for path argument.
1255 		 */
1256 		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1257 		    UIO_SYSSPACE, pathbuf, td);
1258 		if (namei(&nd) == 0) {
1259 			NDFREE(&nd, NDF_ONLY_PNBUF);
1260 			error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
1261 			    MNAMELEN);
1262 			if (error == 0 || error == ENODEV)
1263 				vput(nd.ni_vp);
1264 		}
1265 		mtx_lock(&mountlist_mtx);
1266 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1267 			if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) {
1268 				vfs_ref(mp);
1269 				break;
1270 			}
1271 		}
1272 		mtx_unlock(&mountlist_mtx);
1273 	}
1274 	free(pathbuf, M_TEMP);
1275 	if (mp == NULL) {
1276 		/*
1277 		 * Previously we returned ENOENT for a nonexistent path and
1278 		 * EINVAL for a non-mountpoint.  We cannot tell these apart
1279 		 * now, so in the !MNT_BYFSID case return the more likely
1280 		 * EINVAL for compatibility.
1281 		 */
1282 		return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1283 	}
1284 
1285 	/*
1286 	 * Don't allow unmounting the root filesystem.
1287 	 */
1288 	if (mp->mnt_flag & MNT_ROOTFS) {
1289 		vfs_rel(mp);
1290 		return (EINVAL);
1291 	}
1292 	error = dounmount(mp, uap->flags, td);
1293 	return (error);
1294 }
1295 
1296 /*
1297  * Return error if any of the vnodes, ignoring the root vnode
1298  * and the syncer vnode, have non-zero usecount.
1299  *
1300  * This function is purely advisory - it can return false positives
1301  * and negatives.
1302  */
1303 static int
1304 vfs_check_usecounts(struct mount *mp)
1305 {
1306 	struct vnode *vp, *mvp;
1307 
1308 	MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1309 		if ((vp->v_vflag & VV_ROOT) == 0 && vp->v_type != VNON &&
1310 		    vp->v_usecount != 0) {
1311 			VI_UNLOCK(vp);
1312 			MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1313 			return (EBUSY);
1314 		}
1315 		VI_UNLOCK(vp);
1316 	}
1317 
1318 	return (0);
1319 }
1320 
1321 static void
1322 dounmount_cleanup(struct mount *mp, struct vnode *coveredvp, int mntkflags)
1323 {
1324 
1325 	mtx_assert(MNT_MTX(mp), MA_OWNED);
1326 	mp->mnt_kern_flag &= ~mntkflags;
1327 	if ((mp->mnt_kern_flag & MNTK_MWAIT) != 0) {
1328 		mp->mnt_kern_flag &= ~MNTK_MWAIT;
1329 		wakeup(mp);
1330 	}
1331 	MNT_IUNLOCK(mp);
1332 	if (coveredvp != NULL) {
1333 		VOP_UNLOCK(coveredvp, 0);
1334 		vdrop(coveredvp);
1335 	}
1336 	vn_finished_write(mp);
1337 }
1338 
1339 /*
1340  * Do the actual filesystem unmount.
1341  */
1342 int
1343 dounmount(struct mount *mp, int flags, struct thread *td)
1344 {
1345 	struct vnode *coveredvp;
1346 	int error;
1347 	uint64_t async_flag;
1348 	int mnt_gen_r;
1349 
1350 	if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
1351 		mnt_gen_r = mp->mnt_gen;
1352 		VI_LOCK(coveredvp);
1353 		vholdl(coveredvp);
1354 		vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
1355 		/*
1356 		 * Check for mp being unmounted while waiting for the
1357 		 * covered vnode lock.
1358 		 */
1359 		if (coveredvp->v_mountedhere != mp ||
1360 		    coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
1361 			VOP_UNLOCK(coveredvp, 0);
1362 			vdrop(coveredvp);
1363 			vfs_rel(mp);
1364 			return (EBUSY);
1365 		}
1366 	}
1367 
1368 	/*
1369 	 * Only privileged root, or (if MNT_USER is set) the user that did the
1370 	 * original mount is permitted to unmount this filesystem.
1371 	 */
1372 	error = vfs_suser(mp, td);
1373 	if (error != 0) {
1374 		if (coveredvp != NULL) {
1375 			VOP_UNLOCK(coveredvp, 0);
1376 			vdrop(coveredvp);
1377 		}
1378 		vfs_rel(mp);
1379 		return (error);
1380 	}
1381 
1382 	vn_start_write(NULL, &mp, V_WAIT | V_MNTREF);
1383 	MNT_ILOCK(mp);
1384 	if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 ||
1385 	    (mp->mnt_flag & MNT_UPDATE) != 0 ||
1386 	    !TAILQ_EMPTY(&mp->mnt_uppers)) {
1387 		dounmount_cleanup(mp, coveredvp, 0);
1388 		return (EBUSY);
1389 	}
1390 	mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ;
1391 	if (flags & MNT_NONBUSY) {
1392 		MNT_IUNLOCK(mp);
1393 		error = vfs_check_usecounts(mp);
1394 		MNT_ILOCK(mp);
1395 		if (error != 0) {
1396 			dounmount_cleanup(mp, coveredvp, MNTK_UNMOUNT |
1397 			    MNTK_NOINSMNTQ);
1398 			return (error);
1399 		}
1400 	}
1401 	/* Allow filesystems to detect that a forced unmount is in progress. */
1402 	if (flags & MNT_FORCE) {
1403 		mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1404 		MNT_IUNLOCK(mp);
1405 		/*
1406 		 * Must be done after setting MNTK_UNMOUNTF and before
1407 		 * waiting for mnt_lockref to become 0.
1408 		 */
1409 		VFS_PURGE(mp);
1410 		MNT_ILOCK(mp);
1411 	}
1412 	error = 0;
1413 	if (mp->mnt_lockref) {
1414 		mp->mnt_kern_flag |= MNTK_DRAINING;
1415 		error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
1416 		    "mount drain", 0);
1417 	}
1418 	MNT_IUNLOCK(mp);
1419 	KASSERT(mp->mnt_lockref == 0,
1420 	    ("%s: invalid lock refcount in the drain path @ %s:%d",
1421 	    __func__, __FILE__, __LINE__));
1422 	KASSERT(error == 0,
1423 	    ("%s: invalid return value for msleep in the drain path @ %s:%d",
1424 	    __func__, __FILE__, __LINE__));
1425 
1426 	if (mp->mnt_flag & MNT_EXPUBLIC)
1427 		vfs_setpublicfs(NULL, NULL, NULL);
1428 
1429 	/*
1430 	 * From now, we can claim that the use reference on the
1431 	 * coveredvp is ours, and the ref can be released only by
1432 	 * successfull unmount by us, or left for later unmount
1433 	 * attempt.  The previously acquired hold reference is no
1434 	 * longer needed to protect the vnode from reuse.
1435 	 */
1436 	if (coveredvp != NULL)
1437 		vdrop(coveredvp);
1438 
1439 	vfs_msync(mp, MNT_WAIT);
1440 	MNT_ILOCK(mp);
1441 	async_flag = mp->mnt_flag & MNT_ASYNC;
1442 	mp->mnt_flag &= ~MNT_ASYNC;
1443 	mp->mnt_kern_flag &= ~MNTK_ASYNC;
1444 	MNT_IUNLOCK(mp);
1445 	cache_purgevfs(mp, false); /* remove cache entries for this file sys */
1446 	vfs_deallocate_syncvnode(mp);
1447 	if ((mp->mnt_flag & MNT_RDONLY) != 0 || (flags & MNT_FORCE) != 0 ||
1448 	    (error = VFS_SYNC(mp, MNT_WAIT)) == 0)
1449 		error = VFS_UNMOUNT(mp, flags);
1450 	vn_finished_write(mp);
1451 	/*
1452 	 * If we failed to flush the dirty blocks for this mount point,
1453 	 * undo all the cdir/rdir and rootvnode changes we made above.
1454 	 * Unless we failed to do so because the device is reporting that
1455 	 * it doesn't exist anymore.
1456 	 */
1457 	if (error && error != ENXIO) {
1458 		MNT_ILOCK(mp);
1459 		mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ;
1460 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
1461 			MNT_IUNLOCK(mp);
1462 			vfs_allocate_syncvnode(mp);
1463 			MNT_ILOCK(mp);
1464 		}
1465 		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1466 		mp->mnt_flag |= async_flag;
1467 		if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1468 		    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1469 			mp->mnt_kern_flag |= MNTK_ASYNC;
1470 		if (mp->mnt_kern_flag & MNTK_MWAIT) {
1471 			mp->mnt_kern_flag &= ~MNTK_MWAIT;
1472 			wakeup(mp);
1473 		}
1474 		MNT_IUNLOCK(mp);
1475 		if (coveredvp)
1476 			VOP_UNLOCK(coveredvp, 0);
1477 		return (error);
1478 	}
1479 	mtx_lock(&mountlist_mtx);
1480 	TAILQ_REMOVE(&mountlist, mp, mnt_list);
1481 	mtx_unlock(&mountlist_mtx);
1482 	EVENTHANDLER_DIRECT_INVOKE(vfs_unmounted, mp, td);
1483 	if (coveredvp != NULL) {
1484 		coveredvp->v_mountedhere = NULL;
1485 		VOP_UNLOCK(coveredvp, 0);
1486 	}
1487 	vfs_event_signal(NULL, VQ_UNMOUNT, 0);
1488 	if (rootvnode != NULL && mp == rootvnode->v_mount) {
1489 		vrele(rootvnode);
1490 		rootvnode = NULL;
1491 	}
1492 	if (mp == rootdevmp)
1493 		rootdevmp = NULL;
1494 	vfs_mount_destroy(mp);
1495 	return (0);
1496 }
1497 
1498 /*
1499  * Report errors during filesystem mounting.
1500  */
1501 void
1502 vfs_mount_error(struct mount *mp, const char *fmt, ...)
1503 {
1504 	struct vfsoptlist *moptlist = mp->mnt_optnew;
1505 	va_list ap;
1506 	int error, len;
1507 	char *errmsg;
1508 
1509 	error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
1510 	if (error || errmsg == NULL || len <= 0)
1511 		return;
1512 
1513 	va_start(ap, fmt);
1514 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1515 	va_end(ap);
1516 }
1517 
1518 void
1519 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
1520 {
1521 	va_list ap;
1522 	int error, len;
1523 	char *errmsg;
1524 
1525 	error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
1526 	if (error || errmsg == NULL || len <= 0)
1527 		return;
1528 
1529 	va_start(ap, fmt);
1530 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1531 	va_end(ap);
1532 }
1533 
1534 /*
1535  * ---------------------------------------------------------------------
1536  * Functions for querying mount options/arguments from filesystems.
1537  */
1538 
1539 /*
1540  * Check that no unknown options are given
1541  */
1542 int
1543 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1544 {
1545 	struct vfsopt *opt;
1546 	char errmsg[255];
1547 	const char **t, *p, *q;
1548 	int ret = 0;
1549 
1550 	TAILQ_FOREACH(opt, opts, link) {
1551 		p = opt->name;
1552 		q = NULL;
1553 		if (p[0] == 'n' && p[1] == 'o')
1554 			q = p + 2;
1555 		for(t = global_opts; *t != NULL; t++) {
1556 			if (strcmp(*t, p) == 0)
1557 				break;
1558 			if (q != NULL) {
1559 				if (strcmp(*t, q) == 0)
1560 					break;
1561 			}
1562 		}
1563 		if (*t != NULL)
1564 			continue;
1565 		for(t = legal; *t != NULL; t++) {
1566 			if (strcmp(*t, p) == 0)
1567 				break;
1568 			if (q != NULL) {
1569 				if (strcmp(*t, q) == 0)
1570 					break;
1571 			}
1572 		}
1573 		if (*t != NULL)
1574 			continue;
1575 		snprintf(errmsg, sizeof(errmsg),
1576 		    "mount option <%s> is unknown", p);
1577 		ret = EINVAL;
1578 	}
1579 	if (ret != 0) {
1580 		TAILQ_FOREACH(opt, opts, link) {
1581 			if (strcmp(opt->name, "errmsg") == 0) {
1582 				strncpy((char *)opt->value, errmsg, opt->len);
1583 				break;
1584 			}
1585 		}
1586 		if (opt == NULL)
1587 			printf("%s\n", errmsg);
1588 	}
1589 	return (ret);
1590 }
1591 
1592 /*
1593  * Get a mount option by its name.
1594  *
1595  * Return 0 if the option was found, ENOENT otherwise.
1596  * If len is non-NULL it will be filled with the length
1597  * of the option. If buf is non-NULL, it will be filled
1598  * with the address of the option.
1599  */
1600 int
1601 vfs_getopt(struct vfsoptlist *opts, const char *name, void **buf, int *len)
1602 {
1603 	struct vfsopt *opt;
1604 
1605 	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1606 
1607 	TAILQ_FOREACH(opt, opts, link) {
1608 		if (strcmp(name, opt->name) == 0) {
1609 			opt->seen = 1;
1610 			if (len != NULL)
1611 				*len = opt->len;
1612 			if (buf != NULL)
1613 				*buf = opt->value;
1614 			return (0);
1615 		}
1616 	}
1617 	return (ENOENT);
1618 }
1619 
1620 int
1621 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
1622 {
1623 	struct vfsopt *opt;
1624 
1625 	if (opts == NULL)
1626 		return (-1);
1627 
1628 	TAILQ_FOREACH(opt, opts, link) {
1629 		if (strcmp(name, opt->name) == 0) {
1630 			opt->seen = 1;
1631 			return (opt->pos);
1632 		}
1633 	}
1634 	return (-1);
1635 }
1636 
1637 int
1638 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value)
1639 {
1640 	char *opt_value, *vtp;
1641 	quad_t iv;
1642 	int error, opt_len;
1643 
1644 	error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len);
1645 	if (error != 0)
1646 		return (error);
1647 	if (opt_len == 0 || opt_value == NULL)
1648 		return (EINVAL);
1649 	if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0')
1650 		return (EINVAL);
1651 	iv = strtoq(opt_value, &vtp, 0);
1652 	if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0'))
1653 		return (EINVAL);
1654 	if (iv < 0)
1655 		return (EINVAL);
1656 	switch (vtp[0]) {
1657 	case 't': case 'T':
1658 		iv *= 1024;
1659 		/* FALLTHROUGH */
1660 	case 'g': case 'G':
1661 		iv *= 1024;
1662 		/* FALLTHROUGH */
1663 	case 'm': case 'M':
1664 		iv *= 1024;
1665 		/* FALLTHROUGH */
1666 	case 'k': case 'K':
1667 		iv *= 1024;
1668 	case '\0':
1669 		break;
1670 	default:
1671 		return (EINVAL);
1672 	}
1673 	*value = iv;
1674 
1675 	return (0);
1676 }
1677 
1678 char *
1679 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
1680 {
1681 	struct vfsopt *opt;
1682 
1683 	*error = 0;
1684 	TAILQ_FOREACH(opt, opts, link) {
1685 		if (strcmp(name, opt->name) != 0)
1686 			continue;
1687 		opt->seen = 1;
1688 		if (opt->len == 0 ||
1689 		    ((char *)opt->value)[opt->len - 1] != '\0') {
1690 			*error = EINVAL;
1691 			return (NULL);
1692 		}
1693 		return (opt->value);
1694 	}
1695 	*error = ENOENT;
1696 	return (NULL);
1697 }
1698 
1699 int
1700 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
1701 	uint64_t val)
1702 {
1703 	struct vfsopt *opt;
1704 
1705 	TAILQ_FOREACH(opt, opts, link) {
1706 		if (strcmp(name, opt->name) == 0) {
1707 			opt->seen = 1;
1708 			if (w != NULL)
1709 				*w |= val;
1710 			return (1);
1711 		}
1712 	}
1713 	if (w != NULL)
1714 		*w &= ~val;
1715 	return (0);
1716 }
1717 
1718 int
1719 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
1720 {
1721 	va_list ap;
1722 	struct vfsopt *opt;
1723 	int ret;
1724 
1725 	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1726 
1727 	TAILQ_FOREACH(opt, opts, link) {
1728 		if (strcmp(name, opt->name) != 0)
1729 			continue;
1730 		opt->seen = 1;
1731 		if (opt->len == 0 || opt->value == NULL)
1732 			return (0);
1733 		if (((char *)opt->value)[opt->len - 1] != '\0')
1734 			return (0);
1735 		va_start(ap, fmt);
1736 		ret = vsscanf(opt->value, fmt, ap);
1737 		va_end(ap);
1738 		return (ret);
1739 	}
1740 	return (0);
1741 }
1742 
1743 int
1744 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len)
1745 {
1746 	struct vfsopt *opt;
1747 
1748 	TAILQ_FOREACH(opt, opts, link) {
1749 		if (strcmp(name, opt->name) != 0)
1750 			continue;
1751 		opt->seen = 1;
1752 		if (opt->value == NULL)
1753 			opt->len = len;
1754 		else {
1755 			if (opt->len != len)
1756 				return (EINVAL);
1757 			bcopy(value, opt->value, len);
1758 		}
1759 		return (0);
1760 	}
1761 	return (ENOENT);
1762 }
1763 
1764 int
1765 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len)
1766 {
1767 	struct vfsopt *opt;
1768 
1769 	TAILQ_FOREACH(opt, opts, link) {
1770 		if (strcmp(name, opt->name) != 0)
1771 			continue;
1772 		opt->seen = 1;
1773 		if (opt->value == NULL)
1774 			opt->len = len;
1775 		else {
1776 			if (opt->len < len)
1777 				return (EINVAL);
1778 			opt->len = len;
1779 			bcopy(value, opt->value, len);
1780 		}
1781 		return (0);
1782 	}
1783 	return (ENOENT);
1784 }
1785 
1786 int
1787 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value)
1788 {
1789 	struct vfsopt *opt;
1790 
1791 	TAILQ_FOREACH(opt, opts, link) {
1792 		if (strcmp(name, opt->name) != 0)
1793 			continue;
1794 		opt->seen = 1;
1795 		if (opt->value == NULL)
1796 			opt->len = strlen(value) + 1;
1797 		else if (strlcpy(opt->value, value, opt->len) >= opt->len)
1798 			return (EINVAL);
1799 		return (0);
1800 	}
1801 	return (ENOENT);
1802 }
1803 
1804 /*
1805  * Find and copy a mount option.
1806  *
1807  * The size of the buffer has to be specified
1808  * in len, if it is not the same length as the
1809  * mount option, EINVAL is returned.
1810  * Returns ENOENT if the option is not found.
1811  */
1812 int
1813 vfs_copyopt(struct vfsoptlist *opts, const char *name, void *dest, int len)
1814 {
1815 	struct vfsopt *opt;
1816 
1817 	KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1818 
1819 	TAILQ_FOREACH(opt, opts, link) {
1820 		if (strcmp(name, opt->name) == 0) {
1821 			opt->seen = 1;
1822 			if (len != opt->len)
1823 				return (EINVAL);
1824 			bcopy(opt->value, dest, opt->len);
1825 			return (0);
1826 		}
1827 	}
1828 	return (ENOENT);
1829 }
1830 
1831 int
1832 __vfs_statfs(struct mount *mp, struct statfs *sbp)
1833 {
1834 	int error;
1835 
1836 	error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat);
1837 	if (sbp != &mp->mnt_stat)
1838 		*sbp = mp->mnt_stat;
1839 	return (error);
1840 }
1841 
1842 void
1843 vfs_mountedfrom(struct mount *mp, const char *from)
1844 {
1845 
1846 	bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
1847 	strlcpy(mp->mnt_stat.f_mntfromname, from,
1848 	    sizeof mp->mnt_stat.f_mntfromname);
1849 }
1850 
1851 /*
1852  * ---------------------------------------------------------------------
1853  * This is the api for building mount args and mounting filesystems from
1854  * inside the kernel.
1855  *
1856  * The API works by accumulation of individual args.  First error is
1857  * latched.
1858  *
1859  * XXX: should be documented in new manpage kernel_mount(9)
1860  */
1861 
1862 /* A memory allocation which must be freed when we are done */
1863 struct mntaarg {
1864 	SLIST_ENTRY(mntaarg)	next;
1865 };
1866 
1867 /* The header for the mount arguments */
1868 struct mntarg {
1869 	struct iovec *v;
1870 	int len;
1871 	int error;
1872 	SLIST_HEAD(, mntaarg)	list;
1873 };
1874 
1875 /*
1876  * Add a boolean argument.
1877  *
1878  * flag is the boolean value.
1879  * name must start with "no".
1880  */
1881 struct mntarg *
1882 mount_argb(struct mntarg *ma, int flag, const char *name)
1883 {
1884 
1885 	KASSERT(name[0] == 'n' && name[1] == 'o',
1886 	    ("mount_argb(...,%s): name must start with 'no'", name));
1887 
1888 	return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
1889 }
1890 
1891 /*
1892  * Add an argument printf style
1893  */
1894 struct mntarg *
1895 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
1896 {
1897 	va_list ap;
1898 	struct mntaarg *maa;
1899 	struct sbuf *sb;
1900 	int len;
1901 
1902 	if (ma == NULL) {
1903 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1904 		SLIST_INIT(&ma->list);
1905 	}
1906 	if (ma->error)
1907 		return (ma);
1908 
1909 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1910 	    M_MOUNT, M_WAITOK);
1911 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1912 	ma->v[ma->len].iov_len = strlen(name) + 1;
1913 	ma->len++;
1914 
1915 	sb = sbuf_new_auto();
1916 	va_start(ap, fmt);
1917 	sbuf_vprintf(sb, fmt, ap);
1918 	va_end(ap);
1919 	sbuf_finish(sb);
1920 	len = sbuf_len(sb) + 1;
1921 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1922 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1923 	bcopy(sbuf_data(sb), maa + 1, len);
1924 	sbuf_delete(sb);
1925 
1926 	ma->v[ma->len].iov_base = maa + 1;
1927 	ma->v[ma->len].iov_len = len;
1928 	ma->len++;
1929 
1930 	return (ma);
1931 }
1932 
1933 /*
1934  * Add an argument which is a userland string.
1935  */
1936 struct mntarg *
1937 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
1938 {
1939 	struct mntaarg *maa;
1940 	char *tbuf;
1941 
1942 	if (val == NULL)
1943 		return (ma);
1944 	if (ma == NULL) {
1945 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1946 		SLIST_INIT(&ma->list);
1947 	}
1948 	if (ma->error)
1949 		return (ma);
1950 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1951 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1952 	tbuf = (void *)(maa + 1);
1953 	ma->error = copyinstr(val, tbuf, len, NULL);
1954 	return (mount_arg(ma, name, tbuf, -1));
1955 }
1956 
1957 /*
1958  * Plain argument.
1959  *
1960  * If length is -1, treat value as a C string.
1961  */
1962 struct mntarg *
1963 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
1964 {
1965 
1966 	if (ma == NULL) {
1967 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1968 		SLIST_INIT(&ma->list);
1969 	}
1970 	if (ma->error)
1971 		return (ma);
1972 
1973 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1974 	    M_MOUNT, M_WAITOK);
1975 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1976 	ma->v[ma->len].iov_len = strlen(name) + 1;
1977 	ma->len++;
1978 
1979 	ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
1980 	if (len < 0)
1981 		ma->v[ma->len].iov_len = strlen(val) + 1;
1982 	else
1983 		ma->v[ma->len].iov_len = len;
1984 	ma->len++;
1985 	return (ma);
1986 }
1987 
1988 /*
1989  * Free a mntarg structure
1990  */
1991 static void
1992 free_mntarg(struct mntarg *ma)
1993 {
1994 	struct mntaarg *maa;
1995 
1996 	while (!SLIST_EMPTY(&ma->list)) {
1997 		maa = SLIST_FIRST(&ma->list);
1998 		SLIST_REMOVE_HEAD(&ma->list, next);
1999 		free(maa, M_MOUNT);
2000 	}
2001 	free(ma->v, M_MOUNT);
2002 	free(ma, M_MOUNT);
2003 }
2004 
2005 /*
2006  * Mount a filesystem
2007  */
2008 int
2009 kernel_mount(struct mntarg *ma, uint64_t flags)
2010 {
2011 	struct uio auio;
2012 	int error;
2013 
2014 	KASSERT(ma != NULL, ("kernel_mount NULL ma"));
2015 	KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
2016 	KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
2017 
2018 	auio.uio_iov = ma->v;
2019 	auio.uio_iovcnt = ma->len;
2020 	auio.uio_segflg = UIO_SYSSPACE;
2021 
2022 	error = ma->error;
2023 	if (!error)
2024 		error = vfs_donmount(curthread, flags, &auio);
2025 	free_mntarg(ma);
2026 	return (error);
2027 }
2028 
2029 /*
2030  * A printflike function to mount a filesystem.
2031  */
2032 int
2033 kernel_vmount(int flags, ...)
2034 {
2035 	struct mntarg *ma = NULL;
2036 	va_list ap;
2037 	const char *cp;
2038 	const void *vp;
2039 	int error;
2040 
2041 	va_start(ap, flags);
2042 	for (;;) {
2043 		cp = va_arg(ap, const char *);
2044 		if (cp == NULL)
2045 			break;
2046 		vp = va_arg(ap, const void *);
2047 		ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0));
2048 	}
2049 	va_end(ap);
2050 
2051 	error = kernel_mount(ma, flags);
2052 	return (error);
2053 }
2054 
2055 void
2056 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp)
2057 {
2058 
2059 	bcopy(oexp, exp, sizeof(*oexp));
2060 	exp->ex_numsecflavors = 0;
2061 }
2062