xref: /freebsd/sys/kern/vfs_mount.c (revision 6829dae12bb055451fa467da4589c43bd03b1e64)
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/libkern.h>
49 #include <sys/malloc.h>
50 #include <sys/mount.h>
51 #include <sys/mutex.h>
52 #include <sys/namei.h>
53 #include <sys/priv.h>
54 #include <sys/proc.h>
55 #include <sys/filedesc.h>
56 #include <sys/reboot.h>
57 #include <sys/sbuf.h>
58 #include <sys/syscallsubr.h>
59 #include <sys/sysproto.h>
60 #include <sys/sx.h>
61 #include <sys/sysctl.h>
62 #include <sys/sysent.h>
63 #include <sys/systm.h>
64 #include <sys/vnode.h>
65 #include <vm/uma.h>
66 
67 #include <geom/geom.h>
68 
69 #include <machine/stdarg.h>
70 
71 #include <security/audit/audit.h>
72 #include <security/mac/mac_framework.h>
73 
74 #define	VFS_MOUNTARG_SIZE_MAX	(1024 * 64)
75 
76 static int	vfs_domount(struct thread *td, const char *fstype, char *fspath,
77 		    uint64_t fsflags, struct vfsoptlist **optlist);
78 static void	free_mntarg(struct mntarg *ma);
79 
80 static int	usermount = 0;
81 SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0,
82     "Unprivileged users may mount and unmount file systems");
83 
84 static bool	default_autoro = false;
85 SYSCTL_BOOL(_vfs, OID_AUTO, default_autoro, CTLFLAG_RW, &default_autoro, 0,
86     "Retry failed r/w mount as r/o if no explicit ro/rw option is specified");
87 
88 MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
89 MALLOC_DEFINE(M_STATFS, "statfs", "statfs structure");
90 static uma_zone_t mount_zone;
91 
92 /* List of mounted filesystems. */
93 struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
94 
95 /* For any iteration/modification of mountlist */
96 struct mtx mountlist_mtx;
97 MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF);
98 
99 EVENTHANDLER_LIST_DEFINE(vfs_mounted);
100 EVENTHANDLER_LIST_DEFINE(vfs_unmounted);
101 
102 /*
103  * Global opts, taken by all filesystems
104  */
105 static const char *global_opts[] = {
106 	"errmsg",
107 	"fstype",
108 	"fspath",
109 	"ro",
110 	"rw",
111 	"nosuid",
112 	"noexec",
113 	NULL
114 };
115 
116 static int
117 mount_init(void *mem, int size, int flags)
118 {
119 	struct mount *mp;
120 
121 	mp = (struct mount *)mem;
122 	mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF);
123 	mtx_init(&mp->mnt_listmtx, "struct mount vlist mtx", NULL, MTX_DEF);
124 	lockinit(&mp->mnt_explock, PVFS, "explock", 0, 0);
125 	return (0);
126 }
127 
128 static void
129 mount_fini(void *mem, int size)
130 {
131 	struct mount *mp;
132 
133 	mp = (struct mount *)mem;
134 	lockdestroy(&mp->mnt_explock);
135 	mtx_destroy(&mp->mnt_listmtx);
136 	mtx_destroy(&mp->mnt_mtx);
137 }
138 
139 static void
140 vfs_mount_init(void *dummy __unused)
141 {
142 
143 	mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount), NULL,
144 	    NULL, mount_init, mount_fini, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
145 }
146 SYSINIT(vfs_mount, SI_SUB_VFS, SI_ORDER_ANY, vfs_mount_init, NULL);
147 
148 /*
149  * ---------------------------------------------------------------------
150  * Functions for building and sanitizing the mount options
151  */
152 
153 /* Remove one mount option. */
154 static void
155 vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
156 {
157 
158 	TAILQ_REMOVE(opts, opt, link);
159 	free(opt->name, M_MOUNT);
160 	if (opt->value != NULL)
161 		free(opt->value, M_MOUNT);
162 	free(opt, M_MOUNT);
163 }
164 
165 /* Release all resources related to the mount options. */
166 void
167 vfs_freeopts(struct vfsoptlist *opts)
168 {
169 	struct vfsopt *opt;
170 
171 	while (!TAILQ_EMPTY(opts)) {
172 		opt = TAILQ_FIRST(opts);
173 		vfs_freeopt(opts, opt);
174 	}
175 	free(opts, M_MOUNT);
176 }
177 
178 void
179 vfs_deleteopt(struct vfsoptlist *opts, const char *name)
180 {
181 	struct vfsopt *opt, *temp;
182 
183 	if (opts == NULL)
184 		return;
185 	TAILQ_FOREACH_SAFE(opt, opts, link, temp)  {
186 		if (strcmp(opt->name, name) == 0)
187 			vfs_freeopt(opts, opt);
188 	}
189 }
190 
191 static int
192 vfs_isopt_ro(const char *opt)
193 {
194 
195 	if (strcmp(opt, "ro") == 0 || strcmp(opt, "rdonly") == 0 ||
196 	    strcmp(opt, "norw") == 0)
197 		return (1);
198 	return (0);
199 }
200 
201 static int
202 vfs_isopt_rw(const char *opt)
203 {
204 
205 	if (strcmp(opt, "rw") == 0 || strcmp(opt, "noro") == 0)
206 		return (1);
207 	return (0);
208 }
209 
210 /*
211  * Check if options are equal (with or without the "no" prefix).
212  */
213 static int
214 vfs_equalopts(const char *opt1, const char *opt2)
215 {
216 	char *p;
217 
218 	/* "opt" vs. "opt" or "noopt" vs. "noopt" */
219 	if (strcmp(opt1, opt2) == 0)
220 		return (1);
221 	/* "noopt" vs. "opt" */
222 	if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
223 		return (1);
224 	/* "opt" vs. "noopt" */
225 	if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
226 		return (1);
227 	while ((p = strchr(opt1, '.')) != NULL &&
228 	    !strncmp(opt1, opt2, ++p - opt1)) {
229 		opt2 += p - opt1;
230 		opt1 = p;
231 		/* "foo.noopt" vs. "foo.opt" */
232 		if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
233 			return (1);
234 		/* "foo.opt" vs. "foo.noopt" */
235 		if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
236 			return (1);
237 	}
238 	/* "ro" / "rdonly" / "norw" / "rw" / "noro" */
239 	if ((vfs_isopt_ro(opt1) || vfs_isopt_rw(opt1)) &&
240 	    (vfs_isopt_ro(opt2) || vfs_isopt_rw(opt2)))
241 		return (1);
242 	return (0);
243 }
244 
245 /*
246  * If a mount option is specified several times,
247  * (with or without the "no" prefix) only keep
248  * the last occurrence of it.
249  */
250 static void
251 vfs_sanitizeopts(struct vfsoptlist *opts)
252 {
253 	struct vfsopt *opt, *opt2, *tmp;
254 
255 	TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
256 		opt2 = TAILQ_PREV(opt, vfsoptlist, link);
257 		while (opt2 != NULL) {
258 			if (vfs_equalopts(opt->name, opt2->name)) {
259 				tmp = TAILQ_PREV(opt2, vfsoptlist, link);
260 				vfs_freeopt(opts, opt2);
261 				opt2 = tmp;
262 			} else {
263 				opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
264 			}
265 		}
266 	}
267 }
268 
269 /*
270  * Build a linked list of mount options from a struct uio.
271  */
272 int
273 vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
274 {
275 	struct vfsoptlist *opts;
276 	struct vfsopt *opt;
277 	size_t memused, namelen, optlen;
278 	unsigned int i, iovcnt;
279 	int error;
280 
281 	opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
282 	TAILQ_INIT(opts);
283 	memused = 0;
284 	iovcnt = auio->uio_iovcnt;
285 	for (i = 0; i < iovcnt; i += 2) {
286 		namelen = auio->uio_iov[i].iov_len;
287 		optlen = auio->uio_iov[i + 1].iov_len;
288 		memused += sizeof(struct vfsopt) + optlen + namelen;
289 		/*
290 		 * Avoid consuming too much memory, and attempts to overflow
291 		 * memused.
292 		 */
293 		if (memused > VFS_MOUNTARG_SIZE_MAX ||
294 		    optlen > VFS_MOUNTARG_SIZE_MAX ||
295 		    namelen > VFS_MOUNTARG_SIZE_MAX) {
296 			error = EINVAL;
297 			goto bad;
298 		}
299 
300 		opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
301 		opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
302 		opt->value = NULL;
303 		opt->len = 0;
304 		opt->pos = i / 2;
305 		opt->seen = 0;
306 
307 		/*
308 		 * Do this early, so jumps to "bad" will free the current
309 		 * option.
310 		 */
311 		TAILQ_INSERT_TAIL(opts, opt, link);
312 
313 		if (auio->uio_segflg == UIO_SYSSPACE) {
314 			bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
315 		} else {
316 			error = copyin(auio->uio_iov[i].iov_base, opt->name,
317 			    namelen);
318 			if (error)
319 				goto bad;
320 		}
321 		/* Ensure names are null-terminated strings. */
322 		if (namelen == 0 || opt->name[namelen - 1] != '\0') {
323 			error = EINVAL;
324 			goto bad;
325 		}
326 		if (optlen != 0) {
327 			opt->len = optlen;
328 			opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
329 			if (auio->uio_segflg == UIO_SYSSPACE) {
330 				bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
331 				    optlen);
332 			} else {
333 				error = copyin(auio->uio_iov[i + 1].iov_base,
334 				    opt->value, optlen);
335 				if (error)
336 					goto bad;
337 			}
338 		}
339 	}
340 	vfs_sanitizeopts(opts);
341 	*options = opts;
342 	return (0);
343 bad:
344 	vfs_freeopts(opts);
345 	return (error);
346 }
347 
348 /*
349  * Merge the old mount options with the new ones passed
350  * in the MNT_UPDATE case.
351  *
352  * XXX: This function will keep a "nofoo" option in the new
353  * options.  E.g, if the option's canonical name is "foo",
354  * "nofoo" ends up in the mount point's active options.
355  */
356 static void
357 vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *oldopts)
358 {
359 	struct vfsopt *opt, *new;
360 
361 	TAILQ_FOREACH(opt, oldopts, link) {
362 		new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
363 		new->name = strdup(opt->name, M_MOUNT);
364 		if (opt->len != 0) {
365 			new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
366 			bcopy(opt->value, new->value, opt->len);
367 		} else
368 			new->value = NULL;
369 		new->len = opt->len;
370 		new->seen = opt->seen;
371 		TAILQ_INSERT_HEAD(toopts, new, link);
372 	}
373 	vfs_sanitizeopts(toopts);
374 }
375 
376 /*
377  * Mount a filesystem.
378  */
379 #ifndef _SYS_SYSPROTO_H_
380 struct nmount_args {
381 	struct iovec *iovp;
382 	unsigned int iovcnt;
383 	int flags;
384 };
385 #endif
386 int
387 sys_nmount(struct thread *td, struct nmount_args *uap)
388 {
389 	struct uio *auio;
390 	int error;
391 	u_int iovcnt;
392 	uint64_t flags;
393 
394 	/*
395 	 * Mount flags are now 64-bits. On 32-bit archtectures only
396 	 * 32-bits are passed in, but from here on everything handles
397 	 * 64-bit flags correctly.
398 	 */
399 	flags = uap->flags;
400 
401 	AUDIT_ARG_FFLAGS(flags);
402 	CTR4(KTR_VFS, "%s: iovp %p with iovcnt %d and flags %d", __func__,
403 	    uap->iovp, uap->iovcnt, flags);
404 
405 	/*
406 	 * Filter out MNT_ROOTFS.  We do not want clients of nmount() in
407 	 * userspace to set this flag, but we must filter it out if we want
408 	 * MNT_UPDATE on the root file system to work.
409 	 * MNT_ROOTFS should only be set by the kernel when mounting its
410 	 * root file system.
411 	 */
412 	flags &= ~MNT_ROOTFS;
413 
414 	iovcnt = uap->iovcnt;
415 	/*
416 	 * Check that we have an even number of iovec's
417 	 * and that we have at least two options.
418 	 */
419 	if ((iovcnt & 1) || (iovcnt < 4)) {
420 		CTR2(KTR_VFS, "%s: failed for invalid iovcnt %d", __func__,
421 		    uap->iovcnt);
422 		return (EINVAL);
423 	}
424 
425 	error = copyinuio(uap->iovp, iovcnt, &auio);
426 	if (error) {
427 		CTR2(KTR_VFS, "%s: failed for invalid uio op with %d errno",
428 		    __func__, error);
429 		return (error);
430 	}
431 	error = vfs_donmount(td, flags, auio);
432 
433 	free(auio, M_IOV);
434 	return (error);
435 }
436 
437 /*
438  * ---------------------------------------------------------------------
439  * Various utility functions
440  */
441 
442 void
443 vfs_ref(struct mount *mp)
444 {
445 
446 	CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
447 	MNT_ILOCK(mp);
448 	MNT_REF(mp);
449 	MNT_IUNLOCK(mp);
450 }
451 
452 void
453 vfs_rel(struct mount *mp)
454 {
455 
456 	CTR2(KTR_VFS, "%s: mp %p", __func__, mp);
457 	MNT_ILOCK(mp);
458 	MNT_REL(mp);
459 	MNT_IUNLOCK(mp);
460 }
461 
462 /*
463  * Allocate and initialize the mount point struct.
464  */
465 struct mount *
466 vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, const char *fspath,
467     struct ucred *cred)
468 {
469 	struct mount *mp;
470 
471 	mp = uma_zalloc(mount_zone, M_WAITOK);
472 	bzero(&mp->mnt_startzero,
473 	    __rangeof(struct mount, mnt_startzero, mnt_endzero));
474 	TAILQ_INIT(&mp->mnt_nvnodelist);
475 	mp->mnt_nvnodelistsize = 0;
476 	TAILQ_INIT(&mp->mnt_activevnodelist);
477 	mp->mnt_activevnodelistsize = 0;
478 	TAILQ_INIT(&mp->mnt_tmpfreevnodelist);
479 	mp->mnt_tmpfreevnodelistsize = 0;
480 	mp->mnt_ref = 0;
481 	(void) vfs_busy(mp, MBF_NOWAIT);
482 	atomic_add_acq_int(&vfsp->vfc_refcount, 1);
483 	mp->mnt_op = vfsp->vfc_vfsops;
484 	mp->mnt_vfc = vfsp;
485 	mp->mnt_stat.f_type = vfsp->vfc_typenum;
486 	mp->mnt_gen++;
487 	strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
488 	mp->mnt_vnodecovered = vp;
489 	mp->mnt_cred = crdup(cred);
490 	mp->mnt_stat.f_owner = cred->cr_uid;
491 	strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
492 	mp->mnt_iosize_max = DFLTPHYS;
493 #ifdef MAC
494 	mac_mount_init(mp);
495 	mac_mount_create(cred, mp);
496 #endif
497 	arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0);
498 	TAILQ_INIT(&mp->mnt_uppers);
499 	return (mp);
500 }
501 
502 /*
503  * Destroy the mount struct previously allocated by vfs_mount_alloc().
504  */
505 void
506 vfs_mount_destroy(struct mount *mp)
507 {
508 
509 	MNT_ILOCK(mp);
510 	mp->mnt_kern_flag |= MNTK_REFEXPIRE;
511 	if (mp->mnt_kern_flag & MNTK_MWAIT) {
512 		mp->mnt_kern_flag &= ~MNTK_MWAIT;
513 		wakeup(mp);
514 	}
515 	while (mp->mnt_ref)
516 		msleep(mp, MNT_MTX(mp), PVFS, "mntref", 0);
517 	KASSERT(mp->mnt_ref == 0,
518 	    ("%s: invalid refcount in the drain path @ %s:%d", __func__,
519 	    __FILE__, __LINE__));
520 	if (mp->mnt_writeopcount != 0)
521 		panic("vfs_mount_destroy: nonzero writeopcount");
522 	if (mp->mnt_secondary_writes != 0)
523 		panic("vfs_mount_destroy: nonzero secondary_writes");
524 	atomic_subtract_rel_int(&mp->mnt_vfc->vfc_refcount, 1);
525 	if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
526 		struct vnode *vp;
527 
528 		TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
529 			vn_printf(vp, "dangling vnode ");
530 		panic("unmount: dangling vnode");
531 	}
532 	KASSERT(TAILQ_EMPTY(&mp->mnt_uppers), ("mnt_uppers"));
533 	if (mp->mnt_nvnodelistsize != 0)
534 		panic("vfs_mount_destroy: nonzero nvnodelistsize");
535 	if (mp->mnt_activevnodelistsize != 0)
536 		panic("vfs_mount_destroy: nonzero activevnodelistsize");
537 	if (mp->mnt_tmpfreevnodelistsize != 0)
538 		panic("vfs_mount_destroy: nonzero tmpfreevnodelistsize");
539 	if (mp->mnt_lockref != 0)
540 		panic("vfs_mount_destroy: nonzero lock refcount");
541 	MNT_IUNLOCK(mp);
542 	if (mp->mnt_vnodecovered != NULL)
543 		vrele(mp->mnt_vnodecovered);
544 #ifdef MAC
545 	mac_mount_destroy(mp);
546 #endif
547 	if (mp->mnt_opt != NULL)
548 		vfs_freeopts(mp->mnt_opt);
549 	crfree(mp->mnt_cred);
550 	uma_zfree(mount_zone, mp);
551 }
552 
553 static bool
554 vfs_should_downgrade_to_ro_mount(uint64_t fsflags, int error)
555 {
556 	/* This is an upgrade of an exisiting mount. */
557 	if ((fsflags & MNT_UPDATE) != 0)
558 		return (false);
559 	/* This is already an R/O mount. */
560 	if ((fsflags & MNT_RDONLY) != 0)
561 		return (false);
562 
563 	switch (error) {
564 	case ENODEV:	/* generic, geom, ... */
565 	case EACCES:	/* cam/scsi, ... */
566 	case EROFS:	/* md, mmcsd, ... */
567 		/*
568 		 * These errors can be returned by the storage layer to signal
569 		 * that the media is read-only.  No harm in the R/O mount
570 		 * attempt if the error was returned for some other reason.
571 		 */
572 		return (true);
573 	default:
574 		return (false);
575 	}
576 }
577 
578 int
579 vfs_donmount(struct thread *td, uint64_t fsflags, struct uio *fsoptions)
580 {
581 	struct vfsoptlist *optlist;
582 	struct vfsopt *opt, *tmp_opt;
583 	char *fstype, *fspath, *errmsg;
584 	int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
585 	bool autoro;
586 
587 	errmsg = fspath = NULL;
588 	errmsg_len = fspathlen = 0;
589 	errmsg_pos = -1;
590 	autoro = default_autoro;
591 
592 	error = vfs_buildopts(fsoptions, &optlist);
593 	if (error)
594 		return (error);
595 
596 	if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
597 		errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
598 
599 	/*
600 	 * We need these two options before the others,
601 	 * and they are mandatory for any filesystem.
602 	 * Ensure they are NUL terminated as well.
603 	 */
604 	fstypelen = 0;
605 	error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
606 	if (error || fstype[fstypelen - 1] != '\0') {
607 		error = EINVAL;
608 		if (errmsg != NULL)
609 			strncpy(errmsg, "Invalid fstype", errmsg_len);
610 		goto bail;
611 	}
612 	fspathlen = 0;
613 	error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
614 	if (error || fspath[fspathlen - 1] != '\0') {
615 		error = EINVAL;
616 		if (errmsg != NULL)
617 			strncpy(errmsg, "Invalid fspath", errmsg_len);
618 		goto bail;
619 	}
620 
621 	/*
622 	 * We need to see if we have the "update" option
623 	 * before we call vfs_domount(), since vfs_domount() has special
624 	 * logic based on MNT_UPDATE.  This is very important
625 	 * when we want to update the root filesystem.
626 	 */
627 	TAILQ_FOREACH_SAFE(opt, optlist, link, tmp_opt) {
628 		if (strcmp(opt->name, "update") == 0) {
629 			fsflags |= MNT_UPDATE;
630 			vfs_freeopt(optlist, opt);
631 		}
632 		else if (strcmp(opt->name, "async") == 0)
633 			fsflags |= MNT_ASYNC;
634 		else if (strcmp(opt->name, "force") == 0) {
635 			fsflags |= MNT_FORCE;
636 			vfs_freeopt(optlist, opt);
637 		}
638 		else if (strcmp(opt->name, "reload") == 0) {
639 			fsflags |= MNT_RELOAD;
640 			vfs_freeopt(optlist, opt);
641 		}
642 		else if (strcmp(opt->name, "multilabel") == 0)
643 			fsflags |= MNT_MULTILABEL;
644 		else if (strcmp(opt->name, "noasync") == 0)
645 			fsflags &= ~MNT_ASYNC;
646 		else if (strcmp(opt->name, "noatime") == 0)
647 			fsflags |= MNT_NOATIME;
648 		else if (strcmp(opt->name, "atime") == 0) {
649 			free(opt->name, M_MOUNT);
650 			opt->name = strdup("nonoatime", M_MOUNT);
651 		}
652 		else if (strcmp(opt->name, "noclusterr") == 0)
653 			fsflags |= MNT_NOCLUSTERR;
654 		else if (strcmp(opt->name, "clusterr") == 0) {
655 			free(opt->name, M_MOUNT);
656 			opt->name = strdup("nonoclusterr", M_MOUNT);
657 		}
658 		else if (strcmp(opt->name, "noclusterw") == 0)
659 			fsflags |= MNT_NOCLUSTERW;
660 		else if (strcmp(opt->name, "clusterw") == 0) {
661 			free(opt->name, M_MOUNT);
662 			opt->name = strdup("nonoclusterw", M_MOUNT);
663 		}
664 		else if (strcmp(opt->name, "noexec") == 0)
665 			fsflags |= MNT_NOEXEC;
666 		else if (strcmp(opt->name, "exec") == 0) {
667 			free(opt->name, M_MOUNT);
668 			opt->name = strdup("nonoexec", M_MOUNT);
669 		}
670 		else if (strcmp(opt->name, "nosuid") == 0)
671 			fsflags |= MNT_NOSUID;
672 		else if (strcmp(opt->name, "suid") == 0) {
673 			free(opt->name, M_MOUNT);
674 			opt->name = strdup("nonosuid", M_MOUNT);
675 		}
676 		else if (strcmp(opt->name, "nosymfollow") == 0)
677 			fsflags |= MNT_NOSYMFOLLOW;
678 		else if (strcmp(opt->name, "symfollow") == 0) {
679 			free(opt->name, M_MOUNT);
680 			opt->name = strdup("nonosymfollow", M_MOUNT);
681 		}
682 		else if (strcmp(opt->name, "noro") == 0) {
683 			fsflags &= ~MNT_RDONLY;
684 			autoro = false;
685 		}
686 		else if (strcmp(opt->name, "rw") == 0) {
687 			fsflags &= ~MNT_RDONLY;
688 			autoro = false;
689 		}
690 		else if (strcmp(opt->name, "ro") == 0) {
691 			fsflags |= MNT_RDONLY;
692 			autoro = false;
693 		}
694 		else if (strcmp(opt->name, "rdonly") == 0) {
695 			free(opt->name, M_MOUNT);
696 			opt->name = strdup("ro", M_MOUNT);
697 			fsflags |= MNT_RDONLY;
698 			autoro = false;
699 		}
700 		else if (strcmp(opt->name, "autoro") == 0) {
701 			vfs_freeopt(optlist, opt);
702 			autoro = true;
703 		}
704 		else if (strcmp(opt->name, "suiddir") == 0)
705 			fsflags |= MNT_SUIDDIR;
706 		else if (strcmp(opt->name, "sync") == 0)
707 			fsflags |= MNT_SYNCHRONOUS;
708 		else if (strcmp(opt->name, "union") == 0)
709 			fsflags |= MNT_UNION;
710 		else if (strcmp(opt->name, "automounted") == 0) {
711 			fsflags |= MNT_AUTOMOUNTED;
712 			vfs_freeopt(optlist, opt);
713 		}
714 	}
715 
716 	/*
717 	 * Be ultra-paranoid about making sure the type and fspath
718 	 * variables will fit in our mp buffers, including the
719 	 * terminating NUL.
720 	 */
721 	if (fstypelen > MFSNAMELEN || fspathlen > MNAMELEN) {
722 		error = ENAMETOOLONG;
723 		goto bail;
724 	}
725 
726 	error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
727 
728 	/*
729 	 * See if we can mount in the read-only mode if the error code suggests
730 	 * that it could be possible and the mount options allow for that.
731 	 * Never try it if "[no]{ro|rw}" has been explicitly requested and not
732 	 * overridden by "autoro".
733 	 */
734 	if (autoro && vfs_should_downgrade_to_ro_mount(fsflags, error)) {
735 		printf("%s: R/W mount failed, possibly R/O media,"
736 		    " trying R/O mount\n", __func__);
737 		fsflags |= MNT_RDONLY;
738 		error = vfs_domount(td, fstype, fspath, fsflags, &optlist);
739 	}
740 bail:
741 	/* copyout the errmsg */
742 	if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
743 	    && errmsg_len > 0 && errmsg != NULL) {
744 		if (fsoptions->uio_segflg == UIO_SYSSPACE) {
745 			bcopy(errmsg,
746 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
747 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
748 		} else {
749 			copyout(errmsg,
750 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
751 			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
752 		}
753 	}
754 
755 	if (optlist != NULL)
756 		vfs_freeopts(optlist);
757 	return (error);
758 }
759 
760 /*
761  * Old mount API.
762  */
763 #ifndef _SYS_SYSPROTO_H_
764 struct mount_args {
765 	char	*type;
766 	char	*path;
767 	int	flags;
768 	caddr_t	data;
769 };
770 #endif
771 /* ARGSUSED */
772 int
773 sys_mount(struct thread *td, struct mount_args *uap)
774 {
775 	char *fstype;
776 	struct vfsconf *vfsp = NULL;
777 	struct mntarg *ma = NULL;
778 	uint64_t flags;
779 	int error;
780 
781 	/*
782 	 * Mount flags are now 64-bits. On 32-bit architectures only
783 	 * 32-bits are passed in, but from here on everything handles
784 	 * 64-bit flags correctly.
785 	 */
786 	flags = uap->flags;
787 
788 	AUDIT_ARG_FFLAGS(flags);
789 
790 	/*
791 	 * Filter out MNT_ROOTFS.  We do not want clients of mount() in
792 	 * userspace to set this flag, but we must filter it out if we want
793 	 * MNT_UPDATE on the root file system to work.
794 	 * MNT_ROOTFS should only be set by the kernel when mounting its
795 	 * root file system.
796 	 */
797 	flags &= ~MNT_ROOTFS;
798 
799 	fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
800 	error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
801 	if (error) {
802 		free(fstype, M_TEMP);
803 		return (error);
804 	}
805 
806 	AUDIT_ARG_TEXT(fstype);
807 	vfsp = vfs_byname_kld(fstype, td, &error);
808 	free(fstype, M_TEMP);
809 	if (vfsp == NULL)
810 		return (ENOENT);
811 	if (((vfsp->vfc_flags & VFCF_SBDRY) != 0 &&
812 	    vfsp->vfc_vfsops_sd->vfs_cmount == NULL) ||
813 	    ((vfsp->vfc_flags & VFCF_SBDRY) == 0 &&
814 	    vfsp->vfc_vfsops->vfs_cmount == NULL))
815 		return (EOPNOTSUPP);
816 
817 	ma = mount_argsu(ma, "fstype", uap->type, MFSNAMELEN);
818 	ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
819 	ma = mount_argb(ma, flags & MNT_RDONLY, "noro");
820 	ma = mount_argb(ma, !(flags & MNT_NOSUID), "nosuid");
821 	ma = mount_argb(ma, !(flags & MNT_NOEXEC), "noexec");
822 
823 	if ((vfsp->vfc_flags & VFCF_SBDRY) != 0)
824 		return (vfsp->vfc_vfsops_sd->vfs_cmount(ma, uap->data, flags));
825 	return (vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, flags));
826 }
827 
828 /*
829  * vfs_domount_first(): first file system mount (not update)
830  */
831 static int
832 vfs_domount_first(
833 	struct thread *td,		/* Calling thread. */
834 	struct vfsconf *vfsp,		/* File system type. */
835 	char *fspath,			/* Mount path. */
836 	struct vnode *vp,		/* Vnode to be covered. */
837 	uint64_t fsflags,		/* Flags common to all filesystems. */
838 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
839 	)
840 {
841 	struct vattr va;
842 	struct mount *mp;
843 	struct vnode *newdp;
844 	int error, error1;
845 
846 	ASSERT_VOP_ELOCKED(vp, __func__);
847 	KASSERT((fsflags & MNT_UPDATE) == 0, ("MNT_UPDATE shouldn't be here"));
848 
849 	/*
850 	 * If the jail of the calling thread lacks permission for this type of
851 	 * file system, deny immediately.
852 	 */
853 	if (jailed(td->td_ucred) && !prison_allow(td->td_ucred,
854 	    vfsp->vfc_prison_flag)) {
855 		vput(vp);
856 		return (EPERM);
857 	}
858 
859 	/*
860 	 * If the user is not root, ensure that they own the directory
861 	 * onto which we are attempting to mount.
862 	 */
863 	error = VOP_GETATTR(vp, &va, td->td_ucred);
864 	if (error == 0 && va.va_uid != td->td_ucred->cr_uid)
865 		error = priv_check_cred(td->td_ucred, PRIV_VFS_ADMIN);
866 	if (error == 0)
867 		error = vinvalbuf(vp, V_SAVE, 0, 0);
868 	if (error == 0 && vp->v_type != VDIR)
869 		error = ENOTDIR;
870 	if (error == 0) {
871 		VI_LOCK(vp);
872 		if ((vp->v_iflag & VI_MOUNT) == 0 && vp->v_mountedhere == NULL)
873 			vp->v_iflag |= VI_MOUNT;
874 		else
875 			error = EBUSY;
876 		VI_UNLOCK(vp);
877 	}
878 	if (error != 0) {
879 		vput(vp);
880 		return (error);
881 	}
882 	VOP_UNLOCK(vp, 0);
883 
884 	/* Allocate and initialize the filesystem. */
885 	mp = vfs_mount_alloc(vp, vfsp, fspath, td->td_ucred);
886 	/* XXXMAC: pass to vfs_mount_alloc? */
887 	mp->mnt_optnew = *optlist;
888 	/* Set the mount level flags. */
889 	mp->mnt_flag = (fsflags & (MNT_UPDATEMASK | MNT_ROOTFS | MNT_RDONLY));
890 
891 	/*
892 	 * Mount the filesystem.
893 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
894 	 * get.  No freeing of cn_pnbuf.
895 	 */
896 	error1 = 0;
897 	if ((error = VFS_MOUNT(mp)) != 0 ||
898 	    (error1 = VFS_STATFS(mp, &mp->mnt_stat)) != 0 ||
899 	    (error1 = VFS_ROOT(mp, LK_EXCLUSIVE, &newdp)) != 0) {
900 		if (error1 != 0) {
901 			error = error1;
902 			if ((error1 = VFS_UNMOUNT(mp, 0)) != 0)
903 				printf("VFS_UNMOUNT returned %d\n", error1);
904 		}
905 		vfs_unbusy(mp);
906 		mp->mnt_vnodecovered = NULL;
907 		vfs_mount_destroy(mp);
908 		VI_LOCK(vp);
909 		vp->v_iflag &= ~VI_MOUNT;
910 		VI_UNLOCK(vp);
911 		vrele(vp);
912 		return (error);
913 	}
914 	VOP_UNLOCK(newdp, 0);
915 
916 	if (mp->mnt_opt != NULL)
917 		vfs_freeopts(mp->mnt_opt);
918 	mp->mnt_opt = mp->mnt_optnew;
919 	*optlist = NULL;
920 
921 	/*
922 	 * Prevent external consumers of mount options from reading mnt_optnew.
923 	 */
924 	mp->mnt_optnew = NULL;
925 
926 	MNT_ILOCK(mp);
927 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
928 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
929 		mp->mnt_kern_flag |= MNTK_ASYNC;
930 	else
931 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
932 	MNT_IUNLOCK(mp);
933 
934 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
935 	cache_purge(vp);
936 	VI_LOCK(vp);
937 	vp->v_iflag &= ~VI_MOUNT;
938 	VI_UNLOCK(vp);
939 	vp->v_mountedhere = mp;
940 	/* Place the new filesystem at the end of the mount list. */
941 	mtx_lock(&mountlist_mtx);
942 	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
943 	mtx_unlock(&mountlist_mtx);
944 	vfs_event_signal(NULL, VQ_MOUNT, 0);
945 	vn_lock(newdp, LK_EXCLUSIVE | LK_RETRY);
946 	VOP_UNLOCK(vp, 0);
947 	EVENTHANDLER_DIRECT_INVOKE(vfs_mounted, mp, newdp, td);
948 	VOP_UNLOCK(newdp, 0);
949 	mountcheckdirs(vp, newdp);
950 	vrele(newdp);
951 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
952 		vfs_allocate_syncvnode(mp);
953 	vfs_unbusy(mp);
954 	return (0);
955 }
956 
957 /*
958  * vfs_domount_update(): update of mounted file system
959  */
960 static int
961 vfs_domount_update(
962 	struct thread *td,		/* Calling thread. */
963 	struct vnode *vp,		/* Mount point vnode. */
964 	uint64_t fsflags,		/* Flags common to all filesystems. */
965 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
966 	)
967 {
968 	struct export_args export;
969 	void *bufp;
970 	struct mount *mp;
971 	int error, export_error, len;
972 	uint64_t flag;
973 
974 	ASSERT_VOP_ELOCKED(vp, __func__);
975 	KASSERT((fsflags & MNT_UPDATE) != 0, ("MNT_UPDATE should be here"));
976 	mp = vp->v_mount;
977 
978 	if ((vp->v_vflag & VV_ROOT) == 0) {
979 		if (vfs_copyopt(*optlist, "export", &export, sizeof(export))
980 		    == 0)
981 			error = EXDEV;
982 		else
983 			error = EINVAL;
984 		vput(vp);
985 		return (error);
986 	}
987 
988 	/*
989 	 * We only allow the filesystem to be reloaded if it
990 	 * is currently mounted read-only.
991 	 */
992 	flag = mp->mnt_flag;
993 	if ((fsflags & MNT_RELOAD) != 0 && (flag & MNT_RDONLY) == 0) {
994 		vput(vp);
995 		return (EOPNOTSUPP);	/* Needs translation */
996 	}
997 	/*
998 	 * Only privileged root, or (if MNT_USER is set) the user that
999 	 * did the original mount is permitted to update it.
1000 	 */
1001 	error = vfs_suser(mp, td);
1002 	if (error != 0) {
1003 		vput(vp);
1004 		return (error);
1005 	}
1006 	if (vfs_busy(mp, MBF_NOWAIT)) {
1007 		vput(vp);
1008 		return (EBUSY);
1009 	}
1010 	VI_LOCK(vp);
1011 	if ((vp->v_iflag & VI_MOUNT) != 0 || vp->v_mountedhere != NULL) {
1012 		VI_UNLOCK(vp);
1013 		vfs_unbusy(mp);
1014 		vput(vp);
1015 		return (EBUSY);
1016 	}
1017 	vp->v_iflag |= VI_MOUNT;
1018 	VI_UNLOCK(vp);
1019 	VOP_UNLOCK(vp, 0);
1020 
1021 	MNT_ILOCK(mp);
1022 	if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0) {
1023 		MNT_IUNLOCK(mp);
1024 		error = EBUSY;
1025 		goto end;
1026 	}
1027 	mp->mnt_flag &= ~MNT_UPDATEMASK;
1028 	mp->mnt_flag |= fsflags & (MNT_RELOAD | MNT_FORCE | MNT_UPDATE |
1029 	    MNT_SNAPSHOT | MNT_ROOTFS | MNT_UPDATEMASK | MNT_RDONLY);
1030 	if ((mp->mnt_flag & MNT_ASYNC) == 0)
1031 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
1032 	MNT_IUNLOCK(mp);
1033 	mp->mnt_optnew = *optlist;
1034 	vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
1035 
1036 	/*
1037 	 * Mount the filesystem.
1038 	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
1039 	 * get.  No freeing of cn_pnbuf.
1040 	 */
1041 	error = VFS_MOUNT(mp);
1042 
1043 	export_error = 0;
1044 	/* Process the export option. */
1045 	if (error == 0 && vfs_getopt(mp->mnt_optnew, "export", &bufp,
1046 	    &len) == 0) {
1047 		/* Assume that there is only 1 ABI for each length. */
1048 		switch (len) {
1049 		case (sizeof(struct oexport_args)):
1050 			bzero(&export, sizeof(export));
1051 			/* FALLTHROUGH */
1052 		case (sizeof(export)):
1053 			bcopy(bufp, &export, len);
1054 			export_error = vfs_export(mp, &export);
1055 			break;
1056 		default:
1057 			export_error = EINVAL;
1058 			break;
1059 		}
1060 	}
1061 
1062 	MNT_ILOCK(mp);
1063 	if (error == 0) {
1064 		mp->mnt_flag &=	~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE |
1065 		    MNT_SNAPSHOT);
1066 	} else {
1067 		/*
1068 		 * If we fail, restore old mount flags. MNT_QUOTA is special,
1069 		 * because it is not part of MNT_UPDATEMASK, but it could have
1070 		 * changed in the meantime if quotactl(2) was called.
1071 		 * All in all we want current value of MNT_QUOTA, not the old
1072 		 * one.
1073 		 */
1074 		mp->mnt_flag = (mp->mnt_flag & MNT_QUOTA) | (flag & ~MNT_QUOTA);
1075 	}
1076 	if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1077 	    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1078 		mp->mnt_kern_flag |= MNTK_ASYNC;
1079 	else
1080 		mp->mnt_kern_flag &= ~MNTK_ASYNC;
1081 	MNT_IUNLOCK(mp);
1082 
1083 	if (error != 0)
1084 		goto end;
1085 
1086 	if (mp->mnt_opt != NULL)
1087 		vfs_freeopts(mp->mnt_opt);
1088 	mp->mnt_opt = mp->mnt_optnew;
1089 	*optlist = NULL;
1090 	(void)VFS_STATFS(mp, &mp->mnt_stat);
1091 	/*
1092 	 * Prevent external consumers of mount options from reading
1093 	 * mnt_optnew.
1094 	 */
1095 	mp->mnt_optnew = NULL;
1096 
1097 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
1098 		vfs_allocate_syncvnode(mp);
1099 	else
1100 		vfs_deallocate_syncvnode(mp);
1101 end:
1102 	vfs_unbusy(mp);
1103 	VI_LOCK(vp);
1104 	vp->v_iflag &= ~VI_MOUNT;
1105 	VI_UNLOCK(vp);
1106 	vrele(vp);
1107 	return (error != 0 ? error : export_error);
1108 }
1109 
1110 /*
1111  * vfs_domount(): actually attempt a filesystem mount.
1112  */
1113 static int
1114 vfs_domount(
1115 	struct thread *td,		/* Calling thread. */
1116 	const char *fstype,		/* Filesystem type. */
1117 	char *fspath,			/* Mount path. */
1118 	uint64_t fsflags,		/* Flags common to all filesystems. */
1119 	struct vfsoptlist **optlist	/* Options local to the filesystem. */
1120 	)
1121 {
1122 	struct vfsconf *vfsp;
1123 	struct nameidata nd;
1124 	struct vnode *vp;
1125 	char *pathbuf;
1126 	int error;
1127 
1128 	/*
1129 	 * Be ultra-paranoid about making sure the type and fspath
1130 	 * variables will fit in our mp buffers, including the
1131 	 * terminating NUL.
1132 	 */
1133 	if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
1134 		return (ENAMETOOLONG);
1135 
1136 	if (jailed(td->td_ucred) || usermount == 0) {
1137 		if ((error = priv_check(td, PRIV_VFS_MOUNT)) != 0)
1138 			return (error);
1139 	}
1140 
1141 	/*
1142 	 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
1143 	 */
1144 	if (fsflags & MNT_EXPORTED) {
1145 		error = priv_check(td, PRIV_VFS_MOUNT_EXPORTED);
1146 		if (error)
1147 			return (error);
1148 	}
1149 	if (fsflags & MNT_SUIDDIR) {
1150 		error = priv_check(td, PRIV_VFS_MOUNT_SUIDDIR);
1151 		if (error)
1152 			return (error);
1153 	}
1154 	/*
1155 	 * Silently enforce MNT_NOSUID and MNT_USER for unprivileged users.
1156 	 */
1157 	if ((fsflags & (MNT_NOSUID | MNT_USER)) != (MNT_NOSUID | MNT_USER)) {
1158 		if (priv_check(td, PRIV_VFS_MOUNT_NONUSER) != 0)
1159 			fsflags |= MNT_NOSUID | MNT_USER;
1160 	}
1161 
1162 	/* Load KLDs before we lock the covered vnode to avoid reversals. */
1163 	vfsp = NULL;
1164 	if ((fsflags & MNT_UPDATE) == 0) {
1165 		/* Don't try to load KLDs if we're mounting the root. */
1166 		if (fsflags & MNT_ROOTFS)
1167 			vfsp = vfs_byname(fstype);
1168 		else
1169 			vfsp = vfs_byname_kld(fstype, td, &error);
1170 		if (vfsp == NULL)
1171 			return (ENODEV);
1172 	}
1173 
1174 	/*
1175 	 * Get vnode to be covered or mount point's vnode in case of MNT_UPDATE.
1176 	 */
1177 	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1178 	    UIO_SYSSPACE, fspath, td);
1179 	error = namei(&nd);
1180 	if (error != 0)
1181 		return (error);
1182 	NDFREE(&nd, NDF_ONLY_PNBUF);
1183 	vp = nd.ni_vp;
1184 	if ((fsflags & MNT_UPDATE) == 0) {
1185 		pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1186 		strcpy(pathbuf, fspath);
1187 		error = vn_path_to_global_path(td, vp, pathbuf, MNAMELEN);
1188 		/* debug.disablefullpath == 1 results in ENODEV */
1189 		if (error == 0 || error == ENODEV) {
1190 			error = vfs_domount_first(td, vfsp, pathbuf, vp,
1191 			    fsflags, optlist);
1192 		}
1193 		free(pathbuf, M_TEMP);
1194 	} else
1195 		error = vfs_domount_update(td, vp, fsflags, optlist);
1196 
1197 	return (error);
1198 }
1199 
1200 /*
1201  * Unmount a filesystem.
1202  *
1203  * Note: unmount takes a path to the vnode mounted on as argument, not
1204  * special file (as before).
1205  */
1206 #ifndef _SYS_SYSPROTO_H_
1207 struct unmount_args {
1208 	char	*path;
1209 	int	flags;
1210 };
1211 #endif
1212 /* ARGSUSED */
1213 int
1214 sys_unmount(struct thread *td, struct unmount_args *uap)
1215 {
1216 	struct nameidata nd;
1217 	struct mount *mp;
1218 	char *pathbuf;
1219 	int error, id0, id1;
1220 
1221 	AUDIT_ARG_VALUE(uap->flags);
1222 	if (jailed(td->td_ucred) || usermount == 0) {
1223 		error = priv_check(td, PRIV_VFS_UNMOUNT);
1224 		if (error)
1225 			return (error);
1226 	}
1227 
1228 	pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1229 	error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1230 	if (error) {
1231 		free(pathbuf, M_TEMP);
1232 		return (error);
1233 	}
1234 	if (uap->flags & MNT_BYFSID) {
1235 		AUDIT_ARG_TEXT(pathbuf);
1236 		/* Decode the filesystem ID. */
1237 		if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
1238 			free(pathbuf, M_TEMP);
1239 			return (EINVAL);
1240 		}
1241 
1242 		mtx_lock(&mountlist_mtx);
1243 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1244 			if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1245 			    mp->mnt_stat.f_fsid.val[1] == id1) {
1246 				vfs_ref(mp);
1247 				break;
1248 			}
1249 		}
1250 		mtx_unlock(&mountlist_mtx);
1251 	} else {
1252 		/*
1253 		 * Try to find global path for path argument.
1254 		 */
1255 		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1,
1256 		    UIO_SYSSPACE, pathbuf, td);
1257 		if (namei(&nd) == 0) {
1258 			NDFREE(&nd, NDF_ONLY_PNBUF);
1259 			error = vn_path_to_global_path(td, nd.ni_vp, pathbuf,
1260 			    MNAMELEN);
1261 			if (error == 0 || error == ENODEV)
1262 				vput(nd.ni_vp);
1263 		}
1264 		mtx_lock(&mountlist_mtx);
1265 		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1266 			if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0) {
1267 				vfs_ref(mp);
1268 				break;
1269 			}
1270 		}
1271 		mtx_unlock(&mountlist_mtx);
1272 	}
1273 	free(pathbuf, M_TEMP);
1274 	if (mp == NULL) {
1275 		/*
1276 		 * Previously we returned ENOENT for a nonexistent path and
1277 		 * EINVAL for a non-mountpoint.  We cannot tell these apart
1278 		 * now, so in the !MNT_BYFSID case return the more likely
1279 		 * EINVAL for compatibility.
1280 		 */
1281 		return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1282 	}
1283 
1284 	/*
1285 	 * Don't allow unmounting the root filesystem.
1286 	 */
1287 	if (mp->mnt_flag & MNT_ROOTFS) {
1288 		vfs_rel(mp);
1289 		return (EINVAL);
1290 	}
1291 	error = dounmount(mp, uap->flags, td);
1292 	return (error);
1293 }
1294 
1295 /*
1296  * Return error if any of the vnodes, ignoring the root vnode
1297  * and the syncer vnode, have non-zero usecount.
1298  *
1299  * This function is purely advisory - it can return false positives
1300  * and negatives.
1301  */
1302 static int
1303 vfs_check_usecounts(struct mount *mp)
1304 {
1305 	struct vnode *vp, *mvp;
1306 
1307 	MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
1308 		if ((vp->v_vflag & VV_ROOT) == 0 && vp->v_type != VNON &&
1309 		    vp->v_usecount != 0) {
1310 			VI_UNLOCK(vp);
1311 			MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
1312 			return (EBUSY);
1313 		}
1314 		VI_UNLOCK(vp);
1315 	}
1316 
1317 	return (0);
1318 }
1319 
1320 static void
1321 dounmount_cleanup(struct mount *mp, struct vnode *coveredvp, int mntkflags)
1322 {
1323 
1324 	mtx_assert(MNT_MTX(mp), MA_OWNED);
1325 	mp->mnt_kern_flag &= ~mntkflags;
1326 	if ((mp->mnt_kern_flag & MNTK_MWAIT) != 0) {
1327 		mp->mnt_kern_flag &= ~MNTK_MWAIT;
1328 		wakeup(mp);
1329 	}
1330 	MNT_IUNLOCK(mp);
1331 	if (coveredvp != NULL) {
1332 		VOP_UNLOCK(coveredvp, 0);
1333 		vdrop(coveredvp);
1334 	}
1335 	vn_finished_write(mp);
1336 }
1337 
1338 /*
1339  * Do the actual filesystem unmount.
1340  */
1341 int
1342 dounmount(struct mount *mp, int flags, struct thread *td)
1343 {
1344 	struct vnode *coveredvp;
1345 	int error;
1346 	uint64_t async_flag;
1347 	int mnt_gen_r;
1348 
1349 	if ((coveredvp = mp->mnt_vnodecovered) != NULL) {
1350 		mnt_gen_r = mp->mnt_gen;
1351 		VI_LOCK(coveredvp);
1352 		vholdl(coveredvp);
1353 		vn_lock(coveredvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_RETRY);
1354 		/*
1355 		 * Check for mp being unmounted while waiting for the
1356 		 * covered vnode lock.
1357 		 */
1358 		if (coveredvp->v_mountedhere != mp ||
1359 		    coveredvp->v_mountedhere->mnt_gen != mnt_gen_r) {
1360 			VOP_UNLOCK(coveredvp, 0);
1361 			vdrop(coveredvp);
1362 			vfs_rel(mp);
1363 			return (EBUSY);
1364 		}
1365 	}
1366 
1367 	/*
1368 	 * Only privileged root, or (if MNT_USER is set) the user that did the
1369 	 * original mount is permitted to unmount this filesystem.
1370 	 */
1371 	error = vfs_suser(mp, td);
1372 	if (error != 0) {
1373 		if (coveredvp != NULL) {
1374 			VOP_UNLOCK(coveredvp, 0);
1375 			vdrop(coveredvp);
1376 		}
1377 		vfs_rel(mp);
1378 		return (error);
1379 	}
1380 
1381 	vn_start_write(NULL, &mp, V_WAIT | V_MNTREF);
1382 	MNT_ILOCK(mp);
1383 	if ((mp->mnt_kern_flag & MNTK_UNMOUNT) != 0 ||
1384 	    (mp->mnt_flag & MNT_UPDATE) != 0 ||
1385 	    !TAILQ_EMPTY(&mp->mnt_uppers)) {
1386 		dounmount_cleanup(mp, coveredvp, 0);
1387 		return (EBUSY);
1388 	}
1389 	mp->mnt_kern_flag |= MNTK_UNMOUNT | MNTK_NOINSMNTQ;
1390 	if (flags & MNT_NONBUSY) {
1391 		MNT_IUNLOCK(mp);
1392 		error = vfs_check_usecounts(mp);
1393 		MNT_ILOCK(mp);
1394 		if (error != 0) {
1395 			dounmount_cleanup(mp, coveredvp, MNTK_UNMOUNT |
1396 			    MNTK_NOINSMNTQ);
1397 			return (error);
1398 		}
1399 	}
1400 	/* Allow filesystems to detect that a forced unmount is in progress. */
1401 	if (flags & MNT_FORCE) {
1402 		mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1403 		MNT_IUNLOCK(mp);
1404 		/*
1405 		 * Must be done after setting MNTK_UNMOUNTF and before
1406 		 * waiting for mnt_lockref to become 0.
1407 		 */
1408 		VFS_PURGE(mp);
1409 		MNT_ILOCK(mp);
1410 	}
1411 	error = 0;
1412 	if (mp->mnt_lockref) {
1413 		mp->mnt_kern_flag |= MNTK_DRAINING;
1414 		error = msleep(&mp->mnt_lockref, MNT_MTX(mp), PVFS,
1415 		    "mount drain", 0);
1416 	}
1417 	MNT_IUNLOCK(mp);
1418 	KASSERT(mp->mnt_lockref == 0,
1419 	    ("%s: invalid lock refcount in the drain path @ %s:%d",
1420 	    __func__, __FILE__, __LINE__));
1421 	KASSERT(error == 0,
1422 	    ("%s: invalid return value for msleep in the drain path @ %s:%d",
1423 	    __func__, __FILE__, __LINE__));
1424 
1425 	if (mp->mnt_flag & MNT_EXPUBLIC)
1426 		vfs_setpublicfs(NULL, NULL, NULL);
1427 
1428 	/*
1429 	 * From now, we can claim that the use reference on the
1430 	 * coveredvp is ours, and the ref can be released only by
1431 	 * successfull unmount by us, or left for later unmount
1432 	 * attempt.  The previously acquired hold reference is no
1433 	 * longer needed to protect the vnode from reuse.
1434 	 */
1435 	if (coveredvp != NULL)
1436 		vdrop(coveredvp);
1437 
1438 	vfs_msync(mp, MNT_WAIT);
1439 	MNT_ILOCK(mp);
1440 	async_flag = mp->mnt_flag & MNT_ASYNC;
1441 	mp->mnt_flag &= ~MNT_ASYNC;
1442 	mp->mnt_kern_flag &= ~MNTK_ASYNC;
1443 	MNT_IUNLOCK(mp);
1444 	cache_purgevfs(mp, false); /* remove cache entries for this file sys */
1445 	vfs_deallocate_syncvnode(mp);
1446 	if ((mp->mnt_flag & MNT_RDONLY) != 0 || (flags & MNT_FORCE) != 0 ||
1447 	    (error = VFS_SYNC(mp, MNT_WAIT)) == 0)
1448 		error = VFS_UNMOUNT(mp, flags);
1449 	vn_finished_write(mp);
1450 	/*
1451 	 * If we failed to flush the dirty blocks for this mount point,
1452 	 * undo all the cdir/rdir and rootvnode changes we made above.
1453 	 * Unless we failed to do so because the device is reporting that
1454 	 * it doesn't exist anymore.
1455 	 */
1456 	if (error && error != ENXIO) {
1457 		MNT_ILOCK(mp);
1458 		mp->mnt_kern_flag &= ~MNTK_NOINSMNTQ;
1459 		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
1460 			MNT_IUNLOCK(mp);
1461 			vfs_allocate_syncvnode(mp);
1462 			MNT_ILOCK(mp);
1463 		}
1464 		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1465 		mp->mnt_flag |= async_flag;
1466 		if ((mp->mnt_flag & MNT_ASYNC) != 0 &&
1467 		    (mp->mnt_kern_flag & MNTK_NOASYNC) == 0)
1468 			mp->mnt_kern_flag |= MNTK_ASYNC;
1469 		if (mp->mnt_kern_flag & MNTK_MWAIT) {
1470 			mp->mnt_kern_flag &= ~MNTK_MWAIT;
1471 			wakeup(mp);
1472 		}
1473 		MNT_IUNLOCK(mp);
1474 		if (coveredvp)
1475 			VOP_UNLOCK(coveredvp, 0);
1476 		return (error);
1477 	}
1478 	mtx_lock(&mountlist_mtx);
1479 	TAILQ_REMOVE(&mountlist, mp, mnt_list);
1480 	mtx_unlock(&mountlist_mtx);
1481 	EVENTHANDLER_DIRECT_INVOKE(vfs_unmounted, mp, td);
1482 	if (coveredvp != NULL) {
1483 		coveredvp->v_mountedhere = NULL;
1484 		VOP_UNLOCK(coveredvp, 0);
1485 	}
1486 	vfs_event_signal(NULL, VQ_UNMOUNT, 0);
1487 	if (rootvnode != NULL && mp == rootvnode->v_mount) {
1488 		vrele(rootvnode);
1489 		rootvnode = NULL;
1490 	}
1491 	if (mp == rootdevmp)
1492 		rootdevmp = NULL;
1493 	vfs_mount_destroy(mp);
1494 	return (0);
1495 }
1496 
1497 /*
1498  * Report errors during filesystem mounting.
1499  */
1500 void
1501 vfs_mount_error(struct mount *mp, const char *fmt, ...)
1502 {
1503 	struct vfsoptlist *moptlist = mp->mnt_optnew;
1504 	va_list ap;
1505 	int error, len;
1506 	char *errmsg;
1507 
1508 	error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
1509 	if (error || errmsg == NULL || len <= 0)
1510 		return;
1511 
1512 	va_start(ap, fmt);
1513 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1514 	va_end(ap);
1515 }
1516 
1517 void
1518 vfs_opterror(struct vfsoptlist *opts, const char *fmt, ...)
1519 {
1520 	va_list ap;
1521 	int error, len;
1522 	char *errmsg;
1523 
1524 	error = vfs_getopt(opts, "errmsg", (void **)&errmsg, &len);
1525 	if (error || errmsg == NULL || len <= 0)
1526 		return;
1527 
1528 	va_start(ap, fmt);
1529 	vsnprintf(errmsg, (size_t)len, fmt, ap);
1530 	va_end(ap);
1531 }
1532 
1533 /*
1534  * ---------------------------------------------------------------------
1535  * Functions for querying mount options/arguments from filesystems.
1536  */
1537 
1538 /*
1539  * Check that no unknown options are given
1540  */
1541 int
1542 vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1543 {
1544 	struct vfsopt *opt;
1545 	char errmsg[255];
1546 	const char **t, *p, *q;
1547 	int ret = 0;
1548 
1549 	TAILQ_FOREACH(opt, opts, link) {
1550 		p = opt->name;
1551 		q = NULL;
1552 		if (p[0] == 'n' && p[1] == 'o')
1553 			q = p + 2;
1554 		for(t = global_opts; *t != NULL; t++) {
1555 			if (strcmp(*t, p) == 0)
1556 				break;
1557 			if (q != NULL) {
1558 				if (strcmp(*t, q) == 0)
1559 					break;
1560 			}
1561 		}
1562 		if (*t != NULL)
1563 			continue;
1564 		for(t = legal; *t != NULL; t++) {
1565 			if (strcmp(*t, p) == 0)
1566 				break;
1567 			if (q != NULL) {
1568 				if (strcmp(*t, q) == 0)
1569 					break;
1570 			}
1571 		}
1572 		if (*t != NULL)
1573 			continue;
1574 		snprintf(errmsg, sizeof(errmsg),
1575 		    "mount option <%s> is unknown", p);
1576 		ret = EINVAL;
1577 	}
1578 	if (ret != 0) {
1579 		TAILQ_FOREACH(opt, opts, link) {
1580 			if (strcmp(opt->name, "errmsg") == 0) {
1581 				strncpy((char *)opt->value, errmsg, opt->len);
1582 				break;
1583 			}
1584 		}
1585 		if (opt == NULL)
1586 			printf("%s\n", errmsg);
1587 	}
1588 	return (ret);
1589 }
1590 
1591 /*
1592  * Get a mount option by its name.
1593  *
1594  * Return 0 if the option was found, ENOENT otherwise.
1595  * If len is non-NULL it will be filled with the length
1596  * of the option. If buf is non-NULL, it will be filled
1597  * with the address of the option.
1598  */
1599 int
1600 vfs_getopt(struct vfsoptlist *opts, const char *name, void **buf, int *len)
1601 {
1602 	struct vfsopt *opt;
1603 
1604 	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1605 
1606 	TAILQ_FOREACH(opt, opts, link) {
1607 		if (strcmp(name, opt->name) == 0) {
1608 			opt->seen = 1;
1609 			if (len != NULL)
1610 				*len = opt->len;
1611 			if (buf != NULL)
1612 				*buf = opt->value;
1613 			return (0);
1614 		}
1615 	}
1616 	return (ENOENT);
1617 }
1618 
1619 int
1620 vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
1621 {
1622 	struct vfsopt *opt;
1623 
1624 	if (opts == NULL)
1625 		return (-1);
1626 
1627 	TAILQ_FOREACH(opt, opts, link) {
1628 		if (strcmp(name, opt->name) == 0) {
1629 			opt->seen = 1;
1630 			return (opt->pos);
1631 		}
1632 	}
1633 	return (-1);
1634 }
1635 
1636 int
1637 vfs_getopt_size(struct vfsoptlist *opts, const char *name, off_t *value)
1638 {
1639 	char *opt_value, *vtp;
1640 	quad_t iv;
1641 	int error, opt_len;
1642 
1643 	error = vfs_getopt(opts, name, (void **)&opt_value, &opt_len);
1644 	if (error != 0)
1645 		return (error);
1646 	if (opt_len == 0 || opt_value == NULL)
1647 		return (EINVAL);
1648 	if (opt_value[0] == '\0' || opt_value[opt_len - 1] != '\0')
1649 		return (EINVAL);
1650 	iv = strtoq(opt_value, &vtp, 0);
1651 	if (vtp == opt_value || (vtp[0] != '\0' && vtp[1] != '\0'))
1652 		return (EINVAL);
1653 	if (iv < 0)
1654 		return (EINVAL);
1655 	switch (vtp[0]) {
1656 	case 't': case 'T':
1657 		iv *= 1024;
1658 		/* FALLTHROUGH */
1659 	case 'g': case 'G':
1660 		iv *= 1024;
1661 		/* FALLTHROUGH */
1662 	case 'm': case 'M':
1663 		iv *= 1024;
1664 		/* FALLTHROUGH */
1665 	case 'k': case 'K':
1666 		iv *= 1024;
1667 	case '\0':
1668 		break;
1669 	default:
1670 		return (EINVAL);
1671 	}
1672 	*value = iv;
1673 
1674 	return (0);
1675 }
1676 
1677 char *
1678 vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
1679 {
1680 	struct vfsopt *opt;
1681 
1682 	*error = 0;
1683 	TAILQ_FOREACH(opt, opts, link) {
1684 		if (strcmp(name, opt->name) != 0)
1685 			continue;
1686 		opt->seen = 1;
1687 		if (opt->len == 0 ||
1688 		    ((char *)opt->value)[opt->len - 1] != '\0') {
1689 			*error = EINVAL;
1690 			return (NULL);
1691 		}
1692 		return (opt->value);
1693 	}
1694 	*error = ENOENT;
1695 	return (NULL);
1696 }
1697 
1698 int
1699 vfs_flagopt(struct vfsoptlist *opts, const char *name, uint64_t *w,
1700 	uint64_t val)
1701 {
1702 	struct vfsopt *opt;
1703 
1704 	TAILQ_FOREACH(opt, opts, link) {
1705 		if (strcmp(name, opt->name) == 0) {
1706 			opt->seen = 1;
1707 			if (w != NULL)
1708 				*w |= val;
1709 			return (1);
1710 		}
1711 	}
1712 	if (w != NULL)
1713 		*w &= ~val;
1714 	return (0);
1715 }
1716 
1717 int
1718 vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
1719 {
1720 	va_list ap;
1721 	struct vfsopt *opt;
1722 	int ret;
1723 
1724 	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1725 
1726 	TAILQ_FOREACH(opt, opts, link) {
1727 		if (strcmp(name, opt->name) != 0)
1728 			continue;
1729 		opt->seen = 1;
1730 		if (opt->len == 0 || opt->value == NULL)
1731 			return (0);
1732 		if (((char *)opt->value)[opt->len - 1] != '\0')
1733 			return (0);
1734 		va_start(ap, fmt);
1735 		ret = vsscanf(opt->value, fmt, ap);
1736 		va_end(ap);
1737 		return (ret);
1738 	}
1739 	return (0);
1740 }
1741 
1742 int
1743 vfs_setopt(struct vfsoptlist *opts, const char *name, void *value, int len)
1744 {
1745 	struct vfsopt *opt;
1746 
1747 	TAILQ_FOREACH(opt, opts, link) {
1748 		if (strcmp(name, opt->name) != 0)
1749 			continue;
1750 		opt->seen = 1;
1751 		if (opt->value == NULL)
1752 			opt->len = len;
1753 		else {
1754 			if (opt->len != len)
1755 				return (EINVAL);
1756 			bcopy(value, opt->value, len);
1757 		}
1758 		return (0);
1759 	}
1760 	return (ENOENT);
1761 }
1762 
1763 int
1764 vfs_setopt_part(struct vfsoptlist *opts, const char *name, void *value, int len)
1765 {
1766 	struct vfsopt *opt;
1767 
1768 	TAILQ_FOREACH(opt, opts, link) {
1769 		if (strcmp(name, opt->name) != 0)
1770 			continue;
1771 		opt->seen = 1;
1772 		if (opt->value == NULL)
1773 			opt->len = len;
1774 		else {
1775 			if (opt->len < len)
1776 				return (EINVAL);
1777 			opt->len = len;
1778 			bcopy(value, opt->value, len);
1779 		}
1780 		return (0);
1781 	}
1782 	return (ENOENT);
1783 }
1784 
1785 int
1786 vfs_setopts(struct vfsoptlist *opts, const char *name, const char *value)
1787 {
1788 	struct vfsopt *opt;
1789 
1790 	TAILQ_FOREACH(opt, opts, link) {
1791 		if (strcmp(name, opt->name) != 0)
1792 			continue;
1793 		opt->seen = 1;
1794 		if (opt->value == NULL)
1795 			opt->len = strlen(value) + 1;
1796 		else if (strlcpy(opt->value, value, opt->len) >= opt->len)
1797 			return (EINVAL);
1798 		return (0);
1799 	}
1800 	return (ENOENT);
1801 }
1802 
1803 /*
1804  * Find and copy a mount option.
1805  *
1806  * The size of the buffer has to be specified
1807  * in len, if it is not the same length as the
1808  * mount option, EINVAL is returned.
1809  * Returns ENOENT if the option is not found.
1810  */
1811 int
1812 vfs_copyopt(struct vfsoptlist *opts, const char *name, void *dest, int len)
1813 {
1814 	struct vfsopt *opt;
1815 
1816 	KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1817 
1818 	TAILQ_FOREACH(opt, opts, link) {
1819 		if (strcmp(name, opt->name) == 0) {
1820 			opt->seen = 1;
1821 			if (len != opt->len)
1822 				return (EINVAL);
1823 			bcopy(opt->value, dest, opt->len);
1824 			return (0);
1825 		}
1826 	}
1827 	return (ENOENT);
1828 }
1829 
1830 int
1831 __vfs_statfs(struct mount *mp, struct statfs *sbp)
1832 {
1833 	int error;
1834 
1835 	error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat);
1836 	if (sbp != &mp->mnt_stat)
1837 		*sbp = mp->mnt_stat;
1838 	return (error);
1839 }
1840 
1841 void
1842 vfs_mountedfrom(struct mount *mp, const char *from)
1843 {
1844 
1845 	bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
1846 	strlcpy(mp->mnt_stat.f_mntfromname, from,
1847 	    sizeof mp->mnt_stat.f_mntfromname);
1848 }
1849 
1850 /*
1851  * ---------------------------------------------------------------------
1852  * This is the api for building mount args and mounting filesystems from
1853  * inside the kernel.
1854  *
1855  * The API works by accumulation of individual args.  First error is
1856  * latched.
1857  *
1858  * XXX: should be documented in new manpage kernel_mount(9)
1859  */
1860 
1861 /* A memory allocation which must be freed when we are done */
1862 struct mntaarg {
1863 	SLIST_ENTRY(mntaarg)	next;
1864 };
1865 
1866 /* The header for the mount arguments */
1867 struct mntarg {
1868 	struct iovec *v;
1869 	int len;
1870 	int error;
1871 	SLIST_HEAD(, mntaarg)	list;
1872 };
1873 
1874 /*
1875  * Add a boolean argument.
1876  *
1877  * flag is the boolean value.
1878  * name must start with "no".
1879  */
1880 struct mntarg *
1881 mount_argb(struct mntarg *ma, int flag, const char *name)
1882 {
1883 
1884 	KASSERT(name[0] == 'n' && name[1] == 'o',
1885 	    ("mount_argb(...,%s): name must start with 'no'", name));
1886 
1887 	return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
1888 }
1889 
1890 /*
1891  * Add an argument printf style
1892  */
1893 struct mntarg *
1894 mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
1895 {
1896 	va_list ap;
1897 	struct mntaarg *maa;
1898 	struct sbuf *sb;
1899 	int len;
1900 
1901 	if (ma == NULL) {
1902 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1903 		SLIST_INIT(&ma->list);
1904 	}
1905 	if (ma->error)
1906 		return (ma);
1907 
1908 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1909 	    M_MOUNT, M_WAITOK);
1910 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1911 	ma->v[ma->len].iov_len = strlen(name) + 1;
1912 	ma->len++;
1913 
1914 	sb = sbuf_new_auto();
1915 	va_start(ap, fmt);
1916 	sbuf_vprintf(sb, fmt, ap);
1917 	va_end(ap);
1918 	sbuf_finish(sb);
1919 	len = sbuf_len(sb) + 1;
1920 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1921 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1922 	bcopy(sbuf_data(sb), maa + 1, len);
1923 	sbuf_delete(sb);
1924 
1925 	ma->v[ma->len].iov_base = maa + 1;
1926 	ma->v[ma->len].iov_len = len;
1927 	ma->len++;
1928 
1929 	return (ma);
1930 }
1931 
1932 /*
1933  * Add an argument which is a userland string.
1934  */
1935 struct mntarg *
1936 mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
1937 {
1938 	struct mntaarg *maa;
1939 	char *tbuf;
1940 
1941 	if (val == NULL)
1942 		return (ma);
1943 	if (ma == NULL) {
1944 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1945 		SLIST_INIT(&ma->list);
1946 	}
1947 	if (ma->error)
1948 		return (ma);
1949 	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1950 	SLIST_INSERT_HEAD(&ma->list, maa, next);
1951 	tbuf = (void *)(maa + 1);
1952 	ma->error = copyinstr(val, tbuf, len, NULL);
1953 	return (mount_arg(ma, name, tbuf, -1));
1954 }
1955 
1956 /*
1957  * Plain argument.
1958  *
1959  * If length is -1, treat value as a C string.
1960  */
1961 struct mntarg *
1962 mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
1963 {
1964 
1965 	if (ma == NULL) {
1966 		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1967 		SLIST_INIT(&ma->list);
1968 	}
1969 	if (ma->error)
1970 		return (ma);
1971 
1972 	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1973 	    M_MOUNT, M_WAITOK);
1974 	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1975 	ma->v[ma->len].iov_len = strlen(name) + 1;
1976 	ma->len++;
1977 
1978 	ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
1979 	if (len < 0)
1980 		ma->v[ma->len].iov_len = strlen(val) + 1;
1981 	else
1982 		ma->v[ma->len].iov_len = len;
1983 	ma->len++;
1984 	return (ma);
1985 }
1986 
1987 /*
1988  * Free a mntarg structure
1989  */
1990 static void
1991 free_mntarg(struct mntarg *ma)
1992 {
1993 	struct mntaarg *maa;
1994 
1995 	while (!SLIST_EMPTY(&ma->list)) {
1996 		maa = SLIST_FIRST(&ma->list);
1997 		SLIST_REMOVE_HEAD(&ma->list, next);
1998 		free(maa, M_MOUNT);
1999 	}
2000 	free(ma->v, M_MOUNT);
2001 	free(ma, M_MOUNT);
2002 }
2003 
2004 /*
2005  * Mount a filesystem
2006  */
2007 int
2008 kernel_mount(struct mntarg *ma, uint64_t flags)
2009 {
2010 	struct uio auio;
2011 	int error;
2012 
2013 	KASSERT(ma != NULL, ("kernel_mount NULL ma"));
2014 	KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
2015 	KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
2016 
2017 	auio.uio_iov = ma->v;
2018 	auio.uio_iovcnt = ma->len;
2019 	auio.uio_segflg = UIO_SYSSPACE;
2020 
2021 	error = ma->error;
2022 	if (!error)
2023 		error = vfs_donmount(curthread, flags, &auio);
2024 	free_mntarg(ma);
2025 	return (error);
2026 }
2027 
2028 /*
2029  * A printflike function to mount a filesystem.
2030  */
2031 int
2032 kernel_vmount(int flags, ...)
2033 {
2034 	struct mntarg *ma = NULL;
2035 	va_list ap;
2036 	const char *cp;
2037 	const void *vp;
2038 	int error;
2039 
2040 	va_start(ap, flags);
2041 	for (;;) {
2042 		cp = va_arg(ap, const char *);
2043 		if (cp == NULL)
2044 			break;
2045 		vp = va_arg(ap, const void *);
2046 		ma = mount_arg(ma, cp, vp, (vp != NULL ? -1 : 0));
2047 	}
2048 	va_end(ap);
2049 
2050 	error = kernel_mount(ma, flags);
2051 	return (error);
2052 }
2053 
2054 void
2055 vfs_oexport_conv(const struct oexport_args *oexp, struct export_args *exp)
2056 {
2057 
2058 	bcopy(oexp, exp, sizeof(*oexp));
2059 	exp->ex_numsecflavors = 0;
2060 }
2061