vfs_mount.c revision 157343
1/*-
2 * Copyright (c) 1999-2004 Poul-Henning Kamp
3 * Copyright (c) 1999 Michael Smith
4 * Copyright (c) 1989, 1993
5 *	The Regents of the University of California.  All rights reserved.
6 * (c) UNIX System Laboratories, Inc.
7 * All or some portions of this file are derived from material licensed
8 * to the University of California by American Telephone and Telegraph
9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10 * the permission of UNIX System Laboratories, Inc.
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 * 1. Redistributions of source code must retain the above copyright
16 *    notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 *    notice, this list of conditions and the following disclaimer in the
19 *    documentation and/or other materials provided with the distribution.
20 * 4. Neither the name of the University nor the names of its contributors
21 *    may be used to endorse or promote products derived from this software
22 *    without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 */
36
37#include <sys/cdefs.h>
38__FBSDID("$FreeBSD: head/sys/kern/vfs_mount.c 157343 2006-03-31 23:38:15Z jeff $");
39
40#include <sys/param.h>
41#include <sys/conf.h>
42#include <sys/jail.h>
43#include <sys/kernel.h>
44#include <sys/libkern.h>
45#include <sys/mac.h>
46#include <sys/malloc.h>
47#include <sys/mount.h>
48#include <sys/mutex.h>
49#include <sys/namei.h>
50#include <sys/proc.h>
51#include <sys/filedesc.h>
52#include <sys/reboot.h>
53#include <sys/syscallsubr.h>
54#include <sys/sysproto.h>
55#include <sys/sx.h>
56#include <sys/sysctl.h>
57#include <sys/sysent.h>
58#include <sys/systm.h>
59#include <sys/vnode.h>
60#include <vm/uma.h>
61
62#include <geom/geom.h>
63
64#include <machine/stdarg.h>
65
66#include "opt_rootdevname.h"
67#include "opt_ddb.h"
68#include "opt_mac.h"
69
70#ifdef DDB
71#include <ddb/ddb.h>
72#endif
73
74#define	ROOTNAME		"root_device"
75#define	VFS_MOUNTARG_SIZE_MAX	(1024 * 64)
76
77static int	vfs_domount(struct thread *td, const char *fstype,
78		    char *fspath, int fsflags, void *fsdata);
79static struct mount *vfs_mount_alloc(struct vnode *dvp, struct vfsconf *vfsp,
80		    const char *fspath, struct thread *td);
81static int	vfs_mountroot_ask(void);
82static int	vfs_mountroot_try(const char *mountfrom);
83static int	vfs_donmount(struct thread *td, int fsflags,
84		    struct uio *fsoptions);
85static void	free_mntarg(struct mntarg *ma);
86static void	vfs_mount_destroy(struct mount *, struct thread *);
87static int	vfs_getopt_pos(struct vfsoptlist *opts, const char *name);
88
89static int	usermount = 0;
90SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0,
91    "Unprivileged users may mount and unmount file systems");
92
93MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
94MALLOC_DEFINE(M_VNODE_MARKER, "vnodemarker", "vnode marker");
95static uma_zone_t mount_zone;
96
97/* List of mounted filesystems. */
98struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
99
100/* For any iteration/modification of mountlist */
101struct mtx mountlist_mtx;
102MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF);
103
104TAILQ_HEAD(vfsoptlist, vfsopt);
105struct vfsopt {
106	TAILQ_ENTRY(vfsopt) link;
107	char	*name;
108	void	*value;
109	int	len;
110};
111
112/*
113 * The vnode of the system's root (/ in the filesystem, without chroot
114 * active.)
115 */
116struct vnode	*rootvnode;
117
118/*
119 * The root filesystem is detailed in the kernel environment variable
120 * vfs.root.mountfrom, which is expected to be in the general format
121 *
122 * <vfsname>:[<path>]
123 * vfsname   := the name of a VFS known to the kernel and capable
124 *              of being mounted as root
125 * path      := disk device name or other data used by the filesystem
126 *              to locate its physical store
127 */
128
129/*
130 * Global opts, taken by all filesystems
131 */
132static const char *global_opts[] = {
133	"errmsg",
134	"fstype",
135	"fspath",
136	"rdonly",
137	"ro",
138	"rw",
139	"suid",
140	"exec",
141	NULL
142};
143
144/*
145 * The root specifiers we will try if RB_CDROM is specified.
146 */
147static char *cdrom_rootdevnames[] = {
148	"cd9660:cd0",
149	"cd9660:acd0",
150	NULL
151};
152
153/* legacy find-root code */
154char		*rootdevnames[2] = {NULL, NULL};
155#ifndef ROOTDEVNAME
156#  define ROOTDEVNAME NULL
157#endif
158static const char	*ctrootdevname = ROOTDEVNAME;
159
160/*
161 * ---------------------------------------------------------------------
162 * Functions for building and sanitizing the mount options
163 */
164
165/* Remove one mount option. */
166static void
167vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
168{
169
170	TAILQ_REMOVE(opts, opt, link);
171	free(opt->name, M_MOUNT);
172	if (opt->value != NULL)
173		free(opt->value, M_MOUNT);
174#ifdef INVARIANTS
175	else if (opt->len != 0)
176		panic("%s: mount option with NULL value but length != 0",
177		    __func__);
178#endif
179	free(opt, M_MOUNT);
180}
181
182/* Release all resources related to the mount options. */
183static void
184vfs_freeopts(struct vfsoptlist *opts)
185{
186	struct vfsopt *opt;
187
188	while (!TAILQ_EMPTY(opts)) {
189		opt = TAILQ_FIRST(opts);
190		vfs_freeopt(opts, opt);
191	}
192	free(opts, M_MOUNT);
193}
194
195/*
196 * Check if options are equal (with or without the "no" prefix).
197 */
198static int
199vfs_equalopts(const char *opt1, const char *opt2)
200{
201
202	/* "opt" vs. "opt" or "noopt" vs. "noopt" */
203	if (strcmp(opt1, opt2) == 0)
204		return (1);
205	/* "noopt" vs. "opt" */
206	if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
207		return (1);
208	/* "opt" vs. "noopt" */
209	if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
210		return (1);
211	return (0);
212}
213
214/*
215 * If a mount option is specified several times,
216 * (with or without the "no" prefix) only keep
217 * the last occurence of it.
218 */
219static void
220vfs_sanitizeopts(struct vfsoptlist *opts)
221{
222	struct vfsopt *opt, *opt2, *tmp;
223
224	TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
225		opt2 = TAILQ_PREV(opt, vfsoptlist, link);
226		while (opt2 != NULL) {
227			if (vfs_equalopts(opt->name, opt2->name)) {
228				tmp = TAILQ_PREV(opt2, vfsoptlist, link);
229				vfs_freeopt(opts, opt2);
230				opt2 = tmp;
231			} else {
232				opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
233			}
234		}
235	}
236}
237
238/*
239 * Build a linked list of mount options from a struct uio.
240 */
241static int
242vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
243{
244	struct vfsoptlist *opts;
245	struct vfsopt *opt;
246	size_t memused;
247	unsigned int i, iovcnt;
248	int error, namelen, optlen;
249
250	opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
251	TAILQ_INIT(opts);
252	memused = 0;
253	iovcnt = auio->uio_iovcnt;
254	for (i = 0; i < iovcnt; i += 2) {
255		opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
256		namelen = auio->uio_iov[i].iov_len;
257		optlen = auio->uio_iov[i + 1].iov_len;
258		opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
259		opt->value = NULL;
260		opt->len = 0;
261
262		/*
263		 * Do this early, so jumps to "bad" will free the current
264		 * option.
265		 */
266		TAILQ_INSERT_TAIL(opts, opt, link);
267		memused += sizeof(struct vfsopt) + optlen + namelen;
268
269		/*
270		 * Avoid consuming too much memory, and attempts to overflow
271		 * memused.
272		 */
273		if (memused > VFS_MOUNTARG_SIZE_MAX ||
274		    optlen > VFS_MOUNTARG_SIZE_MAX ||
275		    namelen > VFS_MOUNTARG_SIZE_MAX) {
276			error = EINVAL;
277			goto bad;
278		}
279
280		if (auio->uio_segflg == UIO_SYSSPACE) {
281			bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
282		} else {
283			error = copyin(auio->uio_iov[i].iov_base, opt->name,
284			    namelen);
285			if (error)
286				goto bad;
287		}
288		/* Ensure names are null-terminated strings. */
289		if (opt->name[namelen - 1] != '\0') {
290			error = EINVAL;
291			goto bad;
292		}
293		if (optlen != 0) {
294			opt->len = optlen;
295			opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
296			if (auio->uio_segflg == UIO_SYSSPACE) {
297				bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
298				    optlen);
299			} else {
300				error = copyin(auio->uio_iov[i + 1].iov_base,
301				    opt->value, optlen);
302				if (error)
303					goto bad;
304			}
305		}
306	}
307	vfs_sanitizeopts(opts);
308	*options = opts;
309	return (0);
310bad:
311	vfs_freeopts(opts);
312	return (error);
313}
314
315/*
316 * Merge the old mount options with the new ones passed
317 * in the MNT_UPDATE case.
318 */
319static void
320vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *opts)
321{
322	struct vfsopt *opt, *opt2, *new;
323
324	TAILQ_FOREACH(opt, opts, link) {
325		/*
326		 * Check that this option hasn't been redefined
327		 * nor cancelled with a "no" mount option.
328		 */
329		opt2 = TAILQ_FIRST(toopts);
330		while (opt2 != NULL) {
331			if (strcmp(opt2->name, opt->name) == 0)
332				goto next;
333			if (strncmp(opt2->name, "no", 2) == 0 &&
334			    strcmp(opt2->name + 2, opt->name) == 0) {
335				vfs_freeopt(toopts, opt2);
336				goto next;
337			}
338			opt2 = TAILQ_NEXT(opt2, link);
339		}
340		/* We want this option, duplicate it. */
341		new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
342		new->name = malloc(strlen(opt->name) + 1, M_MOUNT, M_WAITOK);
343		strcpy(new->name, opt->name);
344		if (opt->len != 0) {
345			new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
346			bcopy(opt->value, new->value, opt->len);
347		} else {
348			new->value = NULL;
349		}
350		new->len = opt->len;
351		TAILQ_INSERT_TAIL(toopts, new, link);
352next:
353		continue;
354	}
355}
356
357/*
358 * ---------------------------------------------------------------------
359 * Mount a filesystem
360 */
361int
362nmount(td, uap)
363	struct thread *td;
364	struct nmount_args /* {
365		struct iovec *iovp;
366		unsigned int iovcnt;
367		int flags;
368	} */ *uap;
369{
370	struct uio *auio;
371	struct iovec *iov;
372	unsigned int i;
373	int error;
374	u_int iovcnt;
375
376	/* Kick out MNT_ROOTFS early as it is legal internally */
377	if (uap->flags & MNT_ROOTFS)
378		return (EINVAL);
379
380	iovcnt = uap->iovcnt;
381	/*
382	 * Check that we have an even number of iovec's
383	 * and that we have at least two options.
384	 */
385	if ((iovcnt & 1) || (iovcnt < 4))
386		return (EINVAL);
387
388	error = copyinuio(uap->iovp, iovcnt, &auio);
389	if (error)
390		return (error);
391	iov = auio->uio_iov;
392	for (i = 0; i < iovcnt; i++) {
393		if (iov->iov_len > MMAXOPTIONLEN) {
394			free(auio, M_IOV);
395			return (EINVAL);
396		}
397		iov++;
398	}
399	error = vfs_donmount(td, uap->flags, auio);
400
401	free(auio, M_IOV);
402	return (error);
403}
404
405/*
406 * ---------------------------------------------------------------------
407 * Various utility functions
408 */
409
410void
411vfs_ref(struct mount *mp)
412{
413
414	MNT_ILOCK(mp);
415	MNT_REF(mp);
416	MNT_IUNLOCK(mp);
417}
418
419void
420vfs_rel(struct mount *mp)
421{
422
423	MNT_ILOCK(mp);
424	MNT_REL(mp);
425	MNT_IUNLOCK(mp);
426}
427
428static int
429mount_init(void *mem, int size, int flags)
430{
431	struct mount *mp;
432
433	mp = (struct mount *)mem;
434	mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF);
435	lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
436	return (0);
437}
438
439static void
440mount_fini(void *mem, int size)
441{
442	struct mount *mp;
443
444	mp = (struct mount *)mem;
445	lockdestroy(&mp->mnt_lock);
446	mtx_destroy(&mp->mnt_mtx);
447}
448
449/*
450 * Allocate and initialize the mount point struct.
451 */
452static struct mount *
453vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp,
454    const char *fspath, struct thread *td)
455{
456	struct mount *mp;
457
458	mp = uma_zalloc(mount_zone, M_WAITOK);
459	bzero(&mp->mnt_startzero,
460	    __rangeof(struct mount, mnt_startzero, mnt_endzero));
461	TAILQ_INIT(&mp->mnt_nvnodelist);
462	mp->mnt_nvnodelistsize = 0;
463	mp->mnt_ref = 0;
464	(void) vfs_busy(mp, LK_NOWAIT, 0, td);
465	mp->mnt_op = vfsp->vfc_vfsops;
466	mp->mnt_vfc = vfsp;
467	vfsp->vfc_refcount++;	/* XXX Unlocked */
468	mp->mnt_stat.f_type = vfsp->vfc_typenum;
469	mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
470	strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
471	mp->mnt_vnodecovered = vp;
472	mp->mnt_cred = crdup(td->td_ucred);
473	mp->mnt_stat.f_owner = td->td_ucred->cr_uid;
474	strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
475	mp->mnt_iosize_max = DFLTPHYS;
476#ifdef MAC
477	mac_init_mount(mp);
478	mac_create_mount(td->td_ucred, mp);
479#endif
480	arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0);
481	return (mp);
482}
483
484/*
485 * Destroy the mount struct previously allocated by vfs_mount_alloc().
486 */
487static void
488vfs_mount_destroy(struct mount *mp, struct thread *td)
489{
490	int i;
491
492	lockmgr(&mp->mnt_lock, LK_RELEASE, NULL, td);
493	MNT_ILOCK(mp);
494	for (i = 0; mp->mnt_ref && i < 3; i++)
495		msleep(mp, MNT_MTX(mp), PVFS, "mntref", hz);
496	/*
497	 * This will always cause a 3 second delay in rebooting due to
498	 * refs on the root mountpoint that never go away.  Most of these
499	 * are held by init which never exits.
500	 */
501	if (i == 3 && (!rebooting || bootverbose))
502		printf("Mount point %s had %d dangling refs\n",
503		    mp->mnt_stat.f_mntonname, mp->mnt_ref);
504	if (mp->mnt_holdcnt != 0) {
505		printf("Waiting for mount point to be unheld\n");
506		while (mp->mnt_holdcnt != 0) {
507			mp->mnt_holdcntwaiters++;
508			msleep(&mp->mnt_holdcnt, MNT_MTX(mp),
509			       PZERO, "mntdestroy", 0);
510			mp->mnt_holdcntwaiters--;
511		}
512		printf("mount point unheld\n");
513	}
514	if (mp->mnt_writeopcount > 0) {
515		printf("Waiting for mount point write ops\n");
516		while (mp->mnt_writeopcount > 0) {
517			mp->mnt_kern_flag |= MNTK_SUSPEND;
518			msleep(&mp->mnt_writeopcount,
519			       MNT_MTX(mp),
520			       PZERO, "mntdestroy2", 0);
521		}
522		printf("mount point write ops completed\n");
523	}
524	if (mp->mnt_secondary_writes > 0) {
525		printf("Waiting for mount point secondary write ops\n");
526		while (mp->mnt_secondary_writes > 0) {
527			mp->mnt_kern_flag |= MNTK_SUSPEND;
528			msleep(&mp->mnt_secondary_writes,
529			       MNT_MTX(mp),
530			       PZERO, "mntdestroy3", 0);
531		}
532		printf("mount point secondary write ops completed\n");
533	}
534	MNT_IUNLOCK(mp);
535	mp->mnt_vfc->vfc_refcount--;
536	if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) {
537		struct vnode *vp;
538
539		TAILQ_FOREACH(vp, &mp->mnt_nvnodelist, v_nmntvnodes)
540			vprint("", vp);
541		panic("unmount: dangling vnode");
542	}
543	MNT_ILOCK(mp);
544	if (mp->mnt_kern_flag & MNTK_MWAIT)
545		wakeup(mp);
546	if (mp->mnt_writeopcount != 0)
547		panic("vfs_mount_destroy: nonzero writeopcount");
548	if (mp->mnt_secondary_writes != 0)
549		panic("vfs_mount_destroy: nonzero secondary_writes");
550	if (mp->mnt_nvnodelistsize != 0)
551		panic("vfs_mount_destroy: nonzero nvnodelistsize");
552	mp->mnt_writeopcount = -1000;
553	mp->mnt_nvnodelistsize = -1000;
554	mp->mnt_secondary_writes = -1000;
555	MNT_IUNLOCK(mp);
556#ifdef MAC
557	mac_destroy_mount(mp);
558#endif
559	if (mp->mnt_opt != NULL)
560		vfs_freeopts(mp->mnt_opt);
561	crfree(mp->mnt_cred);
562	uma_zfree(mount_zone, mp);
563}
564
565static int
566vfs_donmount(struct thread *td, int fsflags, struct uio *fsoptions)
567{
568	struct vfsoptlist *optlist;
569	struct vfsopt *opt;
570	char *fstype, *fspath, *errmsg;
571	int error, fstypelen, fspathlen, errmsg_len, errmsg_pos;
572
573	errmsg_len = 0;
574	errmsg_pos = -1;
575
576	error = vfs_buildopts(fsoptions, &optlist);
577	if (error)
578		return (error);
579
580	if (vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len) == 0)
581		errmsg_pos = vfs_getopt_pos(optlist, "errmsg");
582	else
583		errmsg_len = 0;
584
585	/*
586	 * We need these two options before the others,
587	 * and they are mandatory for any filesystem.
588	 * Ensure they are NUL terminated as well.
589	 */
590	fstypelen = 0;
591	error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
592	if (error || fstype[fstypelen - 1] != '\0') {
593		error = EINVAL;
594		if (errmsg != NULL)
595			strncpy(errmsg, "Invalid fstype", errmsg_len);
596		goto bail;
597	}
598	fspathlen = 0;
599	error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
600	if (error || fspath[fspathlen - 1] != '\0') {
601		error = EINVAL;
602		if (errmsg != NULL)
603			strncpy(errmsg, "Invalid fspath", errmsg_len);
604		goto bail;
605	}
606
607	/*
608	 * We need to see if we have the "update" option
609	 * before we call vfs_domount(), since vfs_domount() has special
610	 * logic based on MNT_UPDATE.  This is very important
611	 * when we want to update the root filesystem.
612	 */
613	TAILQ_FOREACH(opt, optlist, link) {
614		if (strcmp(opt->name, "update") == 0)
615			fsflags |= MNT_UPDATE;
616		else if (strcmp(opt->name, "async") == 0)
617			fsflags |= MNT_ASYNC;
618		else if (strcmp(opt->name, "force") == 0)
619			fsflags |= MNT_FORCE;
620		else if (strcmp(opt->name, "multilabel") == 0)
621			fsflags |= MNT_MULTILABEL;
622		else if (strcmp(opt->name, "noasync") == 0)
623			fsflags &= ~MNT_ASYNC;
624		else if (strcmp(opt->name, "noatime") == 0)
625			fsflags |= MNT_NOATIME;
626		else if (strcmp(opt->name, "noclusterr") == 0)
627			fsflags |= MNT_NOCLUSTERR;
628		else if (strcmp(opt->name, "noclusterw") == 0)
629			fsflags |= MNT_NOCLUSTERW;
630		else if (strcmp(opt->name, "noexec") == 0)
631			fsflags |= MNT_NOEXEC;
632		else if (strcmp(opt->name, "nosuid") == 0)
633			fsflags |= MNT_NOSUID;
634		else if (strcmp(opt->name, "nosymfollow") == 0)
635			fsflags |= MNT_NOSYMFOLLOW;
636		else if (strcmp(opt->name, "noro") == 0 ||
637		    strcmp(opt->name, "rw") == 0)
638			fsflags &= ~MNT_RDONLY;
639		else if (strcmp(opt->name, "ro") == 0 ||
640		    strcmp(opt->name, "rdonly") == 0)
641			fsflags |= MNT_RDONLY;
642		else if (strcmp(opt->name, "snapshot") == 0)
643			fsflags |= MNT_SNAPSHOT;
644		else if (strcmp(opt->name, "suiddir") == 0)
645			fsflags |= MNT_SUIDDIR;
646		else if (strcmp(opt->name, "sync") == 0)
647			fsflags |= MNT_SYNCHRONOUS;
648		else if (strcmp(opt->name, "union") == 0)
649			fsflags |= MNT_UNION;
650	}
651
652	/*
653	 * Be ultra-paranoid about making sure the type and fspath
654	 * variables will fit in our mp buffers, including the
655	 * terminating NUL.
656	 */
657	if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) {
658		error = ENAMETOOLONG;
659		goto bail;
660	}
661
662	mtx_lock(&Giant);
663	error = vfs_domount(td, fstype, fspath, fsflags, optlist);
664	mtx_unlock(&Giant);
665bail:
666	/* copyout the errmsg */
667	if (errmsg_pos != -1 && ((2 * errmsg_pos + 1) < fsoptions->uio_iovcnt)
668	    && errmsg_len > 0 && errmsg != NULL) {
669		if (fsoptions->uio_segflg == UIO_SYSSPACE) {
670			strncpy(fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
671			    errmsg,
672			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len);
673		} else {
674			copystr(errmsg,
675			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_base,
676			    fsoptions->uio_iov[2 * errmsg_pos + 1].iov_len,
677			    NULL);
678		}
679	}
680
681	if (error != 0)
682		vfs_freeopts(optlist);
683	return (error);
684}
685
686/*
687 * ---------------------------------------------------------------------
688 * Old mount API.
689 */
690#ifndef _SYS_SYSPROTO_H_
691struct mount_args {
692	char	*type;
693	char	*path;
694	int	flags;
695	caddr_t	data;
696};
697#endif
698/* ARGSUSED */
699int
700mount(td, uap)
701	struct thread *td;
702	struct mount_args /* {
703		char *type;
704		char *path;
705		int flags;
706		caddr_t data;
707	} */ *uap;
708{
709	char *fstype;
710	struct vfsconf *vfsp = NULL;
711	struct mntarg *ma = NULL;
712	int error;
713
714	/* Kick out MNT_ROOTFS early as it is legal internally */
715	uap->flags &= ~MNT_ROOTFS;
716
717	if (uap->data == NULL)
718		return (EINVAL);
719
720	fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
721	error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
722	if (!error) {
723		mtx_lock(&Giant);	/* XXX ? */
724		vfsp = vfs_byname_kld(fstype, td, &error);
725		mtx_unlock(&Giant);
726	}
727	free(fstype, M_TEMP);
728	if (error)
729		return (error);
730	if (vfsp == NULL)
731		return (ENOENT);
732	if (vfsp->vfc_vfsops->vfs_cmount == NULL)
733		return (EOPNOTSUPP);
734
735	ma = mount_argsu(ma, "fstype", uap->type, MNAMELEN);
736	ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
737	ma = mount_argb(ma, uap->flags & MNT_RDONLY, "noro");
738	ma = mount_argb(ma, !(uap->flags & MNT_NOSUID), "nosuid");
739	ma = mount_argb(ma, !(uap->flags & MNT_NOEXEC), "noexec");
740
741	error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, uap->flags, td);
742	return (error);
743}
744
745
746/*
747 * vfs_domount(): actually attempt a filesystem mount.
748 */
749static int
750vfs_domount(
751	struct thread *td,	/* Flags common to all filesystems. */
752	const char *fstype,	/* Filesystem type. */
753	char *fspath,		/* Mount path. */
754	int fsflags,		/* Flags common to all filesystems. */
755	void *fsdata		/* Options local to the filesystem. */
756	)
757{
758	struct vnode *vp;
759	struct mount *mp;
760	struct vfsconf *vfsp;
761	int error, flag = 0, kern_flag = 0;
762	struct vattr va;
763	struct nameidata nd;
764
765	mtx_assert(&Giant, MA_OWNED);
766	/*
767	 * Be ultra-paranoid about making sure the type and fspath
768	 * variables will fit in our mp buffers, including the
769	 * terminating NUL.
770	 */
771	if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
772		return (ENAMETOOLONG);
773
774	if (jailed(td->td_ucred))
775		return (EPERM);
776	if (usermount == 0) {
777		if ((error = suser(td)) != 0)
778			return (error);
779	}
780
781	/*
782	 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
783	 */
784	if (fsflags & (MNT_EXPORTED | MNT_SUIDDIR)) {
785		if ((error = suser(td)) != 0)
786			return (error);
787	}
788	/*
789	 * Silently enforce MNT_NOSUID and MNT_USER for
790	 * unprivileged users.
791	 */
792	if (suser(td) != 0)
793		fsflags |= MNT_NOSUID | MNT_USER;
794
795	/* Load KLDs before we lock the covered vnode to avoid reversals. */
796	vfsp = NULL;
797	if ((fsflags & MNT_UPDATE) == 0) {
798		/* Don't try to load KLDs if we're mounting the root. */
799		if (fsflags & MNT_ROOTFS)
800			vfsp = vfs_byname(fstype);
801		else
802			vfsp = vfs_byname_kld(fstype, td, &error);
803		if (vfsp == NULL)
804			return (ENODEV);
805	}
806	/*
807	 * Get vnode to be covered
808	 */
809	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td);
810	if ((error = namei(&nd)) != 0)
811		return (error);
812	NDFREE(&nd, NDF_ONLY_PNBUF);
813	vp = nd.ni_vp;
814	if (fsflags & MNT_UPDATE) {
815		if ((vp->v_vflag & VV_ROOT) == 0) {
816			vput(vp);
817			return (EINVAL);
818		}
819		mp = vp->v_mount;
820		flag = mp->mnt_flag;
821		kern_flag = mp->mnt_kern_flag;
822		/*
823		 * We only allow the filesystem to be reloaded if it
824		 * is currently mounted read-only.
825		 */
826		if ((fsflags & MNT_RELOAD) &&
827		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
828			vput(vp);
829			return (EOPNOTSUPP);	/* Needs translation */
830		}
831		/*
832		 * Only privileged root, or (if MNT_USER is set) the user that
833		 * did the original mount is permitted to update it.
834		 */
835		error = vfs_suser(mp, td);
836		if (error) {
837			vput(vp);
838			return (error);
839		}
840		if (vfs_busy(mp, LK_NOWAIT, 0, td)) {
841			vput(vp);
842			return (EBUSY);
843		}
844		VI_LOCK(vp);
845		if ((vp->v_iflag & VI_MOUNT) != 0 ||
846		    vp->v_mountedhere != NULL) {
847			VI_UNLOCK(vp);
848			vfs_unbusy(mp, td);
849			vput(vp);
850			return (EBUSY);
851		}
852		vp->v_iflag |= VI_MOUNT;
853		VI_UNLOCK(vp);
854		mp->mnt_flag |= fsflags &
855		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT | MNT_ROOTFS);
856		VOP_UNLOCK(vp, 0, td);
857		mp->mnt_optnew = fsdata;
858		vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
859	} else {
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, td);
865		if (error) {
866			vput(vp);
867			return (error);
868		}
869		if (va.va_uid != td->td_ucred->cr_uid) {
870			if ((error = suser(td)) != 0) {
871				vput(vp);
872				return (error);
873			}
874		}
875		error = vinvalbuf(vp, V_SAVE, td, 0, 0);
876		if (error != 0) {
877			vput(vp);
878			return (error);
879		}
880		if (vp->v_type != VDIR) {
881			vput(vp);
882			return (ENOTDIR);
883		}
884		VI_LOCK(vp);
885		if ((vp->v_iflag & VI_MOUNT) != 0 ||
886		    vp->v_mountedhere != NULL) {
887			VI_UNLOCK(vp);
888			vput(vp);
889			return (EBUSY);
890		}
891		vp->v_iflag |= VI_MOUNT;
892		VI_UNLOCK(vp);
893
894		/*
895		 * Allocate and initialize the filesystem.
896		 */
897		mp = vfs_mount_alloc(vp, vfsp, fspath, td);
898		VOP_UNLOCK(vp, 0, td);
899
900		/* XXXMAC: pass to vfs_mount_alloc? */
901		mp->mnt_optnew = fsdata;
902	}
903
904	/*
905	 * Set the mount level flags.
906	 */
907	if (fsflags & MNT_RDONLY)
908		mp->mnt_flag |= MNT_RDONLY;
909	mp->mnt_flag &=~ MNT_UPDATEMASK;
910	mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE | MNT_ROOTFS);
911	/*
912	 * Mount the filesystem.
913	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
914	 * get.  No freeing of cn_pnbuf.
915	 */
916        error = VFS_MOUNT(mp, td);
917	if (!error) {
918		if (mp->mnt_opt != NULL)
919			vfs_freeopts(mp->mnt_opt);
920		mp->mnt_opt = mp->mnt_optnew;
921		(void)VFS_STATFS(mp, &mp->mnt_stat, td);
922	}
923	/*
924	 * Prevent external consumers of mount options from reading
925	 * mnt_optnew.
926	*/
927	mp->mnt_optnew = NULL;
928	if (mp->mnt_flag & MNT_UPDATE) {
929		mp->mnt_flag &=
930		    ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT);
931		if (error) {
932			mp->mnt_flag = flag;
933			mp->mnt_kern_flag = kern_flag;
934		}
935		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
936			if (mp->mnt_syncer == NULL)
937				error = vfs_allocate_syncvnode(mp);
938		} else {
939			if (mp->mnt_syncer != NULL)
940				vrele(mp->mnt_syncer);
941			mp->mnt_syncer = NULL;
942		}
943		vfs_unbusy(mp, td);
944		VI_LOCK(vp);
945		vp->v_iflag &= ~VI_MOUNT;
946		VI_UNLOCK(vp);
947		vrele(vp);
948		return (error);
949	}
950	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
951	/*
952	 * Put the new filesystem on the mount list after root.
953	 */
954	cache_purge(vp);
955	if (!error) {
956		struct vnode *newdp;
957
958		VI_LOCK(vp);
959		vp->v_iflag &= ~VI_MOUNT;
960		VI_UNLOCK(vp);
961		vp->v_mountedhere = mp;
962		mtx_lock(&mountlist_mtx);
963		TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
964		mtx_unlock(&mountlist_mtx);
965		vfs_event_signal(NULL, VQ_MOUNT, 0);
966		if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp, td))
967			panic("mount: lost mount");
968		mountcheckdirs(vp, newdp);
969		vput(newdp);
970		VOP_UNLOCK(vp, 0, td);
971		if ((mp->mnt_flag & MNT_RDONLY) == 0)
972			error = vfs_allocate_syncvnode(mp);
973		vfs_unbusy(mp, td);
974		if (error)
975			vrele(vp);
976	} else {
977		VI_LOCK(vp);
978		vp->v_iflag &= ~VI_MOUNT;
979		VI_UNLOCK(vp);
980		vfs_mount_destroy(mp, td);
981		vput(vp);
982	}
983	return (error);
984}
985
986/*
987 * ---------------------------------------------------------------------
988 * Unmount a filesystem.
989 *
990 * Note: unmount takes a path to the vnode mounted on as argument,
991 * not special file (as before).
992 */
993#ifndef _SYS_SYSPROTO_H_
994struct unmount_args {
995	char	*path;
996	int	flags;
997};
998#endif
999/* ARGSUSED */
1000int
1001unmount(td, uap)
1002	struct thread *td;
1003	register struct unmount_args /* {
1004		char *path;
1005		int flags;
1006	} */ *uap;
1007{
1008	struct mount *mp;
1009	char *pathbuf;
1010	int error, id0, id1;
1011
1012	if (jailed(td->td_ucred))
1013		return (EPERM);
1014	if (usermount == 0) {
1015		if ((error = suser(td)) != 0)
1016			return (error);
1017	}
1018
1019	pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1020	error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1021	if (error) {
1022		free(pathbuf, M_TEMP);
1023		return (error);
1024	}
1025	if (uap->flags & MNT_BYFSID) {
1026		/* Decode the filesystem ID. */
1027		if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
1028			free(pathbuf, M_TEMP);
1029			return (EINVAL);
1030		}
1031
1032		mtx_lock(&mountlist_mtx);
1033		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1034			if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1035			    mp->mnt_stat.f_fsid.val[1] == id1)
1036				break;
1037		}
1038		mtx_unlock(&mountlist_mtx);
1039	} else {
1040		mtx_lock(&mountlist_mtx);
1041		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
1042			if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0)
1043				break;
1044		}
1045		mtx_unlock(&mountlist_mtx);
1046	}
1047	free(pathbuf, M_TEMP);
1048	if (mp == NULL) {
1049		/*
1050		 * Previously we returned ENOENT for a nonexistent path and
1051		 * EINVAL for a non-mountpoint.  We cannot tell these apart
1052		 * now, so in the !MNT_BYFSID case return the more likely
1053		 * EINVAL for compatibility.
1054		 */
1055		return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1056	}
1057
1058	/*
1059	 * Only privileged root, or (if MNT_USER is set) the user that did the
1060	 * original mount is permitted to unmount this filesystem.
1061	 */
1062	error = vfs_suser(mp, td);
1063	if (error)
1064		return (error);
1065
1066	/*
1067	 * Don't allow unmounting the root filesystem.
1068	 */
1069	if (mp->mnt_flag & MNT_ROOTFS)
1070		return (EINVAL);
1071	mtx_lock(&Giant);
1072	error = dounmount(mp, uap->flags, td);
1073	mtx_unlock(&Giant);
1074	return (error);
1075}
1076
1077/*
1078 * Do the actual filesystem unmount.
1079 */
1080int
1081dounmount(mp, flags, td)
1082	struct mount *mp;
1083	int flags;
1084	struct thread *td;
1085{
1086	struct vnode *coveredvp, *fsrootvp;
1087	int error;
1088	int async_flag;
1089
1090	mtx_assert(&Giant, MA_OWNED);
1091
1092	if ((coveredvp = mp->mnt_vnodecovered) != NULL)
1093		vn_lock(coveredvp, LK_EXCLUSIVE | LK_RETRY, td);
1094	MNT_ILOCK(mp);
1095	if (mp->mnt_kern_flag & MNTK_UNMOUNT) {
1096		MNT_IUNLOCK(mp);
1097		if (coveredvp)
1098			VOP_UNLOCK(coveredvp, 0, td);
1099		return (EBUSY);
1100	}
1101	mp->mnt_kern_flag |= MNTK_UNMOUNT;
1102	/* Allow filesystems to detect that a forced unmount is in progress. */
1103	if (flags & MNT_FORCE)
1104		mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1105	error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK |
1106	    ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), MNT_MTX(mp), td);
1107	if (error) {
1108		MNT_ILOCK(mp);
1109		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1110		if (mp->mnt_kern_flag & MNTK_MWAIT)
1111			wakeup(mp);
1112		MNT_IUNLOCK(mp);
1113		if (coveredvp)
1114			VOP_UNLOCK(coveredvp, 0, td);
1115		return (error);
1116	}
1117	vn_start_write(NULL, &mp, V_WAIT);
1118
1119	if (mp->mnt_flag & MNT_EXPUBLIC)
1120		vfs_setpublicfs(NULL, NULL, NULL);
1121
1122	vfs_msync(mp, MNT_WAIT);
1123	async_flag = mp->mnt_flag & MNT_ASYNC;
1124	mp->mnt_flag &= ~MNT_ASYNC;
1125	cache_purgevfs(mp);	/* remove cache entries for this file sys */
1126	if (mp->mnt_syncer != NULL)
1127		vrele(mp->mnt_syncer);
1128	/*
1129	 * For forced unmounts, move process cdir/rdir refs on the fs root
1130	 * vnode to the covered vnode.  For non-forced unmounts we want
1131	 * such references to cause an EBUSY error.
1132	 */
1133	if ((flags & MNT_FORCE) &&
1134	    VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp, td) == 0) {
1135		if (mp->mnt_vnodecovered != NULL)
1136			mountcheckdirs(fsrootvp, mp->mnt_vnodecovered);
1137		if (fsrootvp == rootvnode) {
1138			vrele(rootvnode);
1139			rootvnode = NULL;
1140		}
1141		vput(fsrootvp);
1142	}
1143	if (((mp->mnt_flag & MNT_RDONLY) ||
1144	     (error = VFS_SYNC(mp, MNT_WAIT, td)) == 0) ||
1145	    (flags & MNT_FORCE)) {
1146		error = VFS_UNMOUNT(mp, flags, td);
1147	}
1148	vn_finished_write(mp);
1149	if (error) {
1150		/* Undo cdir/rdir and rootvnode changes made above. */
1151		if ((flags & MNT_FORCE) &&
1152		    VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp, td) == 0) {
1153			if (mp->mnt_vnodecovered != NULL)
1154				mountcheckdirs(mp->mnt_vnodecovered, fsrootvp);
1155			if (rootvnode == NULL) {
1156				rootvnode = fsrootvp;
1157				vref(rootvnode);
1158			}
1159			vput(fsrootvp);
1160		}
1161		if ((mp->mnt_flag & MNT_RDONLY) == 0 && mp->mnt_syncer == NULL)
1162			(void) vfs_allocate_syncvnode(mp);
1163		MNT_ILOCK(mp);
1164		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1165		mp->mnt_flag |= async_flag;
1166		lockmgr(&mp->mnt_lock, LK_RELEASE, NULL, td);
1167		if (mp->mnt_kern_flag & MNTK_MWAIT)
1168			wakeup(mp);
1169		MNT_IUNLOCK(mp);
1170		if (coveredvp)
1171			VOP_UNLOCK(coveredvp, 0, td);
1172		return (error);
1173	}
1174	mtx_lock(&mountlist_mtx);
1175	TAILQ_REMOVE(&mountlist, mp, mnt_list);
1176	mtx_unlock(&mountlist_mtx);
1177	if (coveredvp != NULL) {
1178		coveredvp->v_mountedhere = NULL;
1179		vput(coveredvp);
1180	}
1181	vfs_event_signal(NULL, VQ_UNMOUNT, 0);
1182	vfs_mount_destroy(mp, td);
1183	return (0);
1184}
1185
1186/*
1187 * ---------------------------------------------------------------------
1188 * Mounting of root filesystem
1189 *
1190 */
1191
1192struct root_hold_token {
1193	const char 			*who;
1194	LIST_ENTRY(root_hold_token)	list;
1195};
1196
1197static LIST_HEAD(, root_hold_token)	root_holds =
1198    LIST_HEAD_INITIALIZER(&root_holds);
1199
1200struct root_hold_token *
1201root_mount_hold(const char *identifier)
1202{
1203	struct root_hold_token *h;
1204
1205	h = malloc(sizeof *h, M_DEVBUF, M_ZERO | M_WAITOK);
1206	h->who = identifier;
1207	mtx_lock(&mountlist_mtx);
1208	LIST_INSERT_HEAD(&root_holds, h, list);
1209	mtx_unlock(&mountlist_mtx);
1210	return (h);
1211}
1212
1213void
1214root_mount_rel(struct root_hold_token *h)
1215{
1216
1217	mtx_lock(&mountlist_mtx);
1218	LIST_REMOVE(h, list);
1219	wakeup(&root_holds);
1220	mtx_unlock(&mountlist_mtx);
1221	free(h, M_DEVBUF);
1222}
1223
1224static void
1225root_mount_wait(void)
1226{
1227	struct root_hold_token *h;
1228
1229	for (;;) {
1230		DROP_GIANT();
1231		g_waitidle();
1232		PICKUP_GIANT();
1233		mtx_lock(&mountlist_mtx);
1234		if (LIST_EMPTY(&root_holds)) {
1235			mtx_unlock(&mountlist_mtx);
1236			break;
1237		}
1238		printf("Root mount waiting for:");
1239		LIST_FOREACH(h, &root_holds, list)
1240			printf(" %s", h->who);
1241		printf("\n");
1242		msleep(&root_holds, &mountlist_mtx, PZERO | PDROP, "roothold",
1243		    hz);
1244	}
1245}
1246
1247static void
1248set_rootvnode(struct thread *td)
1249{
1250	struct proc *p;
1251
1252	if (VFS_ROOT(TAILQ_FIRST(&mountlist), LK_EXCLUSIVE, &rootvnode, td))
1253		panic("Cannot find root vnode");
1254
1255	p = td->td_proc;
1256	FILEDESC_LOCK(p->p_fd);
1257
1258	if (p->p_fd->fd_cdir != NULL)
1259		vrele(p->p_fd->fd_cdir);
1260	p->p_fd->fd_cdir = rootvnode;
1261	VREF(rootvnode);
1262
1263	if (p->p_fd->fd_rdir != NULL)
1264		vrele(p->p_fd->fd_rdir);
1265	p->p_fd->fd_rdir = rootvnode;
1266	VREF(rootvnode);
1267
1268	FILEDESC_UNLOCK(p->p_fd);
1269
1270	VOP_UNLOCK(rootvnode, 0, td);
1271}
1272
1273/*
1274 * Mount /devfs as our root filesystem, but do not put it on the mountlist
1275 * yet.  Create a /dev -> / symlink so that absolute pathnames will lookup.
1276 */
1277
1278static void
1279devfs_first(void)
1280{
1281	struct thread *td = curthread;
1282	struct vfsoptlist *opts;
1283	struct vfsconf *vfsp;
1284	struct mount *mp = NULL;
1285	int error;
1286
1287	vfsp = vfs_byname("devfs");
1288	KASSERT(vfsp != NULL, ("Could not find devfs by name"));
1289	if (vfsp == NULL)
1290		return;
1291
1292	mp = vfs_mount_alloc(NULLVP, vfsp, "/dev", td);
1293
1294	error = VFS_MOUNT(mp, td);
1295	KASSERT(error == 0, ("VFS_MOUNT(devfs) failed %d", error));
1296	if (error)
1297		return;
1298
1299	opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
1300	TAILQ_INIT(opts);
1301	mp->mnt_opt = opts;
1302
1303	mtx_lock(&mountlist_mtx);
1304	TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list);
1305	mtx_unlock(&mountlist_mtx);
1306
1307	set_rootvnode(td);
1308
1309	error = kern_symlink(td, "/", "dev", UIO_SYSSPACE);
1310	if (error)
1311		printf("kern_symlink /dev -> / returns %d\n", error);
1312}
1313
1314/*
1315 * Surgically move our devfs to be mounted on /dev.
1316 */
1317
1318static void
1319devfs_fixup(struct thread *td)
1320{
1321	struct nameidata nd;
1322	int error;
1323	struct vnode *vp, *dvp;
1324	struct mount *mp;
1325
1326	/* Remove our devfs mount from the mountlist and purge the cache */
1327	mtx_lock(&mountlist_mtx);
1328	mp = TAILQ_FIRST(&mountlist);
1329	TAILQ_REMOVE(&mountlist, mp, mnt_list);
1330	mtx_unlock(&mountlist_mtx);
1331	cache_purgevfs(mp);
1332
1333	VFS_ROOT(mp, LK_EXCLUSIVE, &dvp, td);
1334	VI_LOCK(dvp);
1335	dvp->v_iflag &= ~VI_MOUNT;
1336	dvp->v_mountedhere = NULL;
1337	VI_UNLOCK(dvp);
1338
1339	/* Set up the real rootvnode, and purge the cache */
1340	TAILQ_FIRST(&mountlist)->mnt_vnodecovered = NULL;
1341	set_rootvnode(td);
1342	cache_purgevfs(rootvnode->v_mount);
1343
1344	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, "/dev", td);
1345	error = namei(&nd);
1346	if (error) {
1347		printf("Lookup of /dev for devfs, error: %d\n", error);
1348		return;
1349	}
1350	NDFREE(&nd, NDF_ONLY_PNBUF);
1351	vp = nd.ni_vp;
1352	if (vp->v_type != VDIR) {
1353		vput(vp);
1354	}
1355	error = vinvalbuf(vp, V_SAVE, td, 0, 0);
1356	if (error) {
1357		vput(vp);
1358	}
1359	cache_purge(vp);
1360	mp->mnt_vnodecovered = vp;
1361	vp->v_mountedhere = mp;
1362	mtx_lock(&mountlist_mtx);
1363	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
1364	mtx_unlock(&mountlist_mtx);
1365	VOP_UNLOCK(vp, 0, td);
1366	vput(dvp);
1367	vfs_unbusy(mp, td);
1368
1369	/* Unlink the no longer needed /dev/dev -> / symlink */
1370	kern_unlink(td, "/dev/dev", UIO_SYSSPACE);
1371}
1372
1373/*
1374 * Report errors during filesystem mounting.
1375 */
1376void
1377vfs_mount_error(struct mount *mp, const char *fmt, ...)
1378{
1379	struct vfsoptlist *moptlist = mp->mnt_optnew;
1380	va_list ap;
1381	int error, len;
1382	char *errmsg;
1383
1384	error = vfs_getopt(moptlist, "errmsg", (void **)&errmsg, &len);
1385	if (error || errmsg == NULL || len <= 0)
1386		return;
1387
1388	va_start(ap, fmt);
1389	vsnprintf(errmsg, (size_t)len, fmt, ap);
1390	va_end(ap);
1391}
1392
1393/*
1394 * Find and mount the root filesystem
1395 */
1396void
1397vfs_mountroot(void)
1398{
1399	char *cp;
1400	int error, i, asked = 0;
1401
1402	root_mount_wait();
1403
1404	mount_zone = uma_zcreate("Mountpoints", sizeof(struct mount),
1405	    NULL, NULL, mount_init, mount_fini,
1406	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
1407	devfs_first();
1408
1409	/*
1410	 * We are booted with instructions to prompt for the root filesystem.
1411	 */
1412	if (boothowto & RB_ASKNAME) {
1413		if (!vfs_mountroot_ask())
1414			return;
1415		asked = 1;
1416	}
1417
1418	/*
1419	 * The root filesystem information is compiled in, and we are
1420	 * booted with instructions to use it.
1421	 */
1422	if (ctrootdevname != NULL && (boothowto & RB_DFLTROOT)) {
1423		if (!vfs_mountroot_try(ctrootdevname))
1424			return;
1425		ctrootdevname = NULL;
1426	}
1427
1428	/*
1429	 * We've been given the generic "use CDROM as root" flag.  This is
1430	 * necessary because one media may be used in many different
1431	 * devices, so we need to search for them.
1432	 */
1433	if (boothowto & RB_CDROM) {
1434		for (i = 0; cdrom_rootdevnames[i] != NULL; i++) {
1435			if (!vfs_mountroot_try(cdrom_rootdevnames[i]))
1436				return;
1437		}
1438	}
1439
1440	/*
1441	 * Try to use the value read by the loader from /etc/fstab, or
1442	 * supplied via some other means.  This is the preferred
1443	 * mechanism.
1444	 */
1445	cp = getenv("vfs.root.mountfrom");
1446	if (cp != NULL) {
1447		error = vfs_mountroot_try(cp);
1448		freeenv(cp);
1449		if (!error)
1450			return;
1451	}
1452
1453	/*
1454	 * Try values that may have been computed by code during boot
1455	 */
1456	if (!vfs_mountroot_try(rootdevnames[0]))
1457		return;
1458	if (!vfs_mountroot_try(rootdevnames[1]))
1459		return;
1460
1461	/*
1462	 * If we (still) have a compiled-in default, try it.
1463	 */
1464	if (ctrootdevname != NULL)
1465		if (!vfs_mountroot_try(ctrootdevname))
1466			return;
1467	/*
1468	 * Everything so far has failed, prompt on the console if we haven't
1469	 * already tried that.
1470	 */
1471	if (!asked)
1472		if (!vfs_mountroot_ask())
1473			return;
1474
1475	panic("Root mount failed, startup aborted.");
1476}
1477
1478/*
1479 * Mount (mountfrom) as the root filesystem.
1480 */
1481static int
1482vfs_mountroot_try(const char *mountfrom)
1483{
1484	struct mount	*mp;
1485	char		*vfsname, *path;
1486	time_t		timebase;
1487	int		error;
1488	char		patt[32];
1489
1490	vfsname = NULL;
1491	path    = NULL;
1492	mp      = NULL;
1493	error   = EINVAL;
1494
1495	if (mountfrom == NULL)
1496		return (error);		/* don't complain */
1497	printf("Trying to mount root from %s\n", mountfrom);
1498
1499	/* parse vfs name and path */
1500	vfsname = malloc(MFSNAMELEN, M_MOUNT, M_WAITOK);
1501	path = malloc(MNAMELEN, M_MOUNT, M_WAITOK);
1502	vfsname[0] = path[0] = 0;
1503	sprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN);
1504	if (sscanf(mountfrom, patt, vfsname, path) < 1)
1505		goto out;
1506
1507	if (path[0] == '\0')
1508		strcpy(path, ROOTNAME);
1509
1510	error = kernel_vmount(
1511	    MNT_RDONLY | MNT_ROOTFS,
1512	    "fstype", vfsname,
1513	    "fspath", "/",
1514	    "from", path,
1515	    NULL);
1516	if (error == 0) {
1517		/*
1518		 * We mount devfs prior to mounting the / FS, so the first
1519		 * entry will typically be devfs.
1520		 */
1521		mp = TAILQ_FIRST(&mountlist);
1522		KASSERT(mp != NULL, ("%s: mountlist is empty", __func__));
1523
1524		/*
1525		 * Iterate over all currently mounted file systems and use
1526		 * the time stamp found to check and/or initialize the RTC.
1527		 * Typically devfs has no time stamp and the only other FS
1528		 * is the actual / FS.
1529		 * Call inittodr() only once and pass it the largest of the
1530		 * timestamps we encounter.
1531		 */
1532		timebase = 0;
1533		do {
1534			if (mp->mnt_time > timebase)
1535				timebase = mp->mnt_time;
1536			mp = TAILQ_NEXT(mp, mnt_list);
1537		} while (mp != NULL);
1538		inittodr(timebase);
1539
1540		devfs_fixup(curthread);
1541	}
1542out:
1543	free(path, M_MOUNT);
1544	free(vfsname, M_MOUNT);
1545	return (error);
1546}
1547
1548/*
1549 * ---------------------------------------------------------------------
1550 * Interactive root filesystem selection code.
1551 */
1552
1553static int
1554vfs_mountroot_ask(void)
1555{
1556	char name[128];
1557
1558	for(;;) {
1559		printf("\nManual root filesystem specification:\n");
1560		printf("  <fstype>:<device>  Mount <device> using filesystem <fstype>\n");
1561#if defined(__amd64__) || defined(__i386__) || defined(__ia64__)
1562		printf("                       eg. ufs:da0s1a\n");
1563#else
1564		printf("                       eg. ufs:/dev/da0a\n");
1565#endif
1566		printf("  ?                  List valid disk boot devices\n");
1567		printf("  <empty line>       Abort manual input\n");
1568		printf("\nmountroot> ");
1569		gets(name, sizeof(name), 1);
1570		if (name[0] == '\0')
1571			return (1);
1572		if (name[0] == '?') {
1573			printf("\nList of GEOM managed disk devices:\n  ");
1574			g_dev_print();
1575			continue;
1576		}
1577		if (!vfs_mountroot_try(name))
1578			return (0);
1579	}
1580}
1581
1582/*
1583 * ---------------------------------------------------------------------
1584 * Functions for querying mount options/arguments from filesystems.
1585 */
1586
1587/*
1588 * Check that no unknown options are given
1589 */
1590int
1591vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1592{
1593	struct vfsopt *opt;
1594	const char **t, *p;
1595
1596
1597	TAILQ_FOREACH(opt, opts, link) {
1598		p = opt->name;
1599		if (p[0] == 'n' && p[1] == 'o')
1600			p += 2;
1601		for(t = global_opts; *t != NULL; t++)
1602			if (!strcmp(*t, p))
1603				break;
1604		if (*t != NULL)
1605			continue;
1606		for(t = legal; *t != NULL; t++)
1607			if (!strcmp(*t, p))
1608				break;
1609		if (*t != NULL)
1610			continue;
1611		printf("mount option <%s> is unknown\n", p);
1612		return (EINVAL);
1613	}
1614	return (0);
1615}
1616
1617/*
1618 * Get a mount option by its name.
1619 *
1620 * Return 0 if the option was found, ENOENT otherwise.
1621 * If len is non-NULL it will be filled with the length
1622 * of the option. If buf is non-NULL, it will be filled
1623 * with the address of the option.
1624 */
1625int
1626vfs_getopt(opts, name, buf, len)
1627	struct vfsoptlist *opts;
1628	const char *name;
1629	void **buf;
1630	int *len;
1631{
1632	struct vfsopt *opt;
1633
1634	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1635
1636	TAILQ_FOREACH(opt, opts, link) {
1637		if (strcmp(name, opt->name) == 0) {
1638			if (len != NULL)
1639				*len = opt->len;
1640			if (buf != NULL)
1641				*buf = opt->value;
1642			return (0);
1643		}
1644	}
1645	return (ENOENT);
1646}
1647
1648static int
1649vfs_getopt_pos(struct vfsoptlist *opts, const char *name)
1650{
1651	struct vfsopt *opt;
1652	int i;
1653
1654	if (opts == NULL)
1655		return (-1);
1656
1657	i = 0;
1658	TAILQ_FOREACH(opt, opts, link) {
1659		if (strcmp(name, opt->name) == 0)
1660			return (i);
1661		++i;
1662	}
1663	return (-1);
1664}
1665
1666char *
1667vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
1668{
1669	struct vfsopt *opt;
1670
1671	*error = 0;
1672	TAILQ_FOREACH(opt, opts, link) {
1673		if (strcmp(name, opt->name) != 0)
1674			continue;
1675		if (((char *)opt->value)[opt->len - 1] != '\0') {
1676			*error = EINVAL;
1677			return (NULL);
1678		}
1679		return (opt->value);
1680	}
1681	return (NULL);
1682}
1683
1684int
1685vfs_flagopt(struct vfsoptlist *opts, const char *name, u_int *w, u_int val)
1686{
1687	struct vfsopt *opt;
1688
1689	TAILQ_FOREACH(opt, opts, link) {
1690		if (strcmp(name, opt->name) == 0) {
1691			if (w != NULL)
1692				*w |= val;
1693			return (1);
1694		}
1695	}
1696	if (w != NULL)
1697		*w &= ~val;
1698	return (0);
1699}
1700
1701int
1702vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
1703{
1704	va_list ap;
1705	struct vfsopt *opt;
1706	int ret;
1707
1708	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1709
1710	TAILQ_FOREACH(opt, opts, link) {
1711		if (strcmp(name, opt->name) != 0)
1712			continue;
1713		if (((char *)opt->value)[opt->len - 1] != '\0')
1714			return (0);
1715		va_start(ap, fmt);
1716		ret = vsscanf(opt->value, fmt, ap);
1717		va_end(ap);
1718		return (ret);
1719	}
1720	return (0);
1721}
1722
1723/*
1724 * Find and copy a mount option.
1725 *
1726 * The size of the buffer has to be specified
1727 * in len, if it is not the same length as the
1728 * mount option, EINVAL is returned.
1729 * Returns ENOENT if the option is not found.
1730 */
1731int
1732vfs_copyopt(opts, name, dest, len)
1733	struct vfsoptlist *opts;
1734	const char *name;
1735	void *dest;
1736	int len;
1737{
1738	struct vfsopt *opt;
1739
1740	KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1741
1742	TAILQ_FOREACH(opt, opts, link) {
1743		if (strcmp(name, opt->name) == 0) {
1744			if (len != opt->len)
1745				return (EINVAL);
1746			bcopy(opt->value, dest, opt->len);
1747			return (0);
1748		}
1749	}
1750	return (ENOENT);
1751}
1752
1753/*
1754 * This is a helper function for filesystems to traverse their
1755 * vnodes.  See MNT_VNODE_FOREACH() in sys/mount.h
1756 */
1757
1758struct vnode *
1759__mnt_vnode_next(struct vnode **mvp, struct mount *mp)
1760{
1761	struct vnode *vp;
1762
1763	mtx_assert(MNT_MTX(mp), MA_OWNED);
1764
1765	KASSERT((*mvp)->v_mount == mp, ("marker vnode mount list mismatch"));
1766	vp = TAILQ_NEXT(*mvp, v_nmntvnodes);
1767	while (vp != NULL && vp->v_type == VMARKER)
1768		vp = TAILQ_NEXT(vp, v_nmntvnodes);
1769
1770	/* Check if we are done */
1771	if (vp == NULL) {
1772		__mnt_vnode_markerfree(mvp, mp);
1773		return (NULL);
1774	}
1775	TAILQ_REMOVE(&mp->mnt_nvnodelist, *mvp, v_nmntvnodes);
1776	TAILQ_INSERT_AFTER(&mp->mnt_nvnodelist, vp, *mvp, v_nmntvnodes);
1777	return (vp);
1778}
1779
1780struct vnode *
1781__mnt_vnode_first(struct vnode **mvp, struct mount *mp)
1782{
1783	struct vnode *vp;
1784
1785	mtx_assert(MNT_MTX(mp), MA_OWNED);
1786
1787	vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
1788	while (vp != NULL && vp->v_type == VMARKER)
1789		vp = TAILQ_NEXT(vp, v_nmntvnodes);
1790
1791	/* Check if we are done */
1792	if (vp == NULL) {
1793		*mvp = NULL;
1794		return (NULL);
1795	}
1796	mp->mnt_holdcnt++;
1797	MNT_IUNLOCK(mp);
1798	*mvp = (struct vnode *) malloc(sizeof(struct vnode),
1799				       M_VNODE_MARKER,
1800				       M_WAITOK | M_ZERO);
1801	MNT_ILOCK(mp);
1802	(*mvp)->v_type = VMARKER;
1803
1804	vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
1805	while (vp != NULL && vp->v_type == VMARKER)
1806		vp = TAILQ_NEXT(vp, v_nmntvnodes);
1807
1808	/* Check if we are done */
1809	if (vp == NULL) {
1810		MNT_IUNLOCK(mp);
1811		free(*mvp, M_VNODE_MARKER);
1812		MNT_ILOCK(mp);
1813		*mvp = NULL;
1814		mp->mnt_holdcnt--;
1815		if (mp->mnt_holdcnt == 0 && mp->mnt_holdcntwaiters != 0)
1816			wakeup(&mp->mnt_holdcnt);
1817		return (NULL);
1818	}
1819	mp->mnt_markercnt++;
1820	(*mvp)->v_mount = mp;
1821	TAILQ_INSERT_AFTER(&mp->mnt_nvnodelist, vp, *mvp, v_nmntvnodes);
1822	return (vp);
1823}
1824
1825
1826void
1827__mnt_vnode_markerfree(struct vnode **mvp, struct mount *mp)
1828{
1829
1830	if (*mvp == NULL)
1831		return;
1832
1833	mtx_assert(MNT_MTX(mp), MA_OWNED);
1834
1835	KASSERT((*mvp)->v_mount == mp, ("marker vnode mount list mismatch"));
1836	TAILQ_REMOVE(&mp->mnt_nvnodelist, *mvp, v_nmntvnodes);
1837	MNT_IUNLOCK(mp);
1838	free(*mvp, M_VNODE_MARKER);
1839	MNT_ILOCK(mp);
1840	*mvp = NULL;
1841
1842	mp->mnt_markercnt--;
1843	mp->mnt_holdcnt--;
1844	if (mp->mnt_holdcnt == 0 && mp->mnt_holdcntwaiters != 0)
1845		wakeup(&mp->mnt_holdcnt);
1846}
1847
1848
1849int
1850__vfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
1851{
1852	int error;
1853
1854	error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat, td);
1855	if (sbp != &mp->mnt_stat)
1856		*sbp = mp->mnt_stat;
1857	return (error);
1858}
1859
1860void
1861vfs_mountedfrom(struct mount *mp, const char *from)
1862{
1863
1864	bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
1865	strlcpy(mp->mnt_stat.f_mntfromname, from,
1866	    sizeof mp->mnt_stat.f_mntfromname);
1867}
1868
1869/*
1870 * ---------------------------------------------------------------------
1871 * This is the api for building mount args and mounting filesystems from
1872 * inside the kernel.
1873 *
1874 * The API works by accumulation of individual args.  First error is
1875 * latched.
1876 *
1877 * XXX: should be documented in new manpage kernel_mount(9)
1878 */
1879
1880/* A memory allocation which must be freed when we are done */
1881struct mntaarg {
1882	SLIST_ENTRY(mntaarg)	next;
1883};
1884
1885/* The header for the mount arguments */
1886struct mntarg {
1887	struct iovec *v;
1888	int len;
1889	int error;
1890	SLIST_HEAD(, mntaarg)	list;
1891};
1892
1893/*
1894 * Add a boolean argument.
1895 *
1896 * flag is the boolean value.
1897 * name must start with "no".
1898 */
1899struct mntarg *
1900mount_argb(struct mntarg *ma, int flag, const char *name)
1901{
1902
1903	KASSERT(name[0] == 'n' && name[1] == 'o',
1904	    ("mount_argb(...,%s): name must start with 'no'", name));
1905
1906	return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
1907}
1908
1909/*
1910 * Add an argument printf style
1911 */
1912struct mntarg *
1913mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
1914{
1915	va_list ap;
1916	struct mntaarg *maa;
1917	struct sbuf *sb;
1918	int len;
1919
1920	if (ma == NULL) {
1921		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1922		SLIST_INIT(&ma->list);
1923	}
1924	if (ma->error)
1925		return (ma);
1926
1927	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1928	    M_MOUNT, M_WAITOK);
1929	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1930	ma->v[ma->len].iov_len = strlen(name) + 1;
1931	ma->len++;
1932
1933	sb = sbuf_new(NULL, NULL, 0, SBUF_AUTOEXTEND);
1934	va_start(ap, fmt);
1935	sbuf_vprintf(sb, fmt, ap);
1936	va_end(ap);
1937	sbuf_finish(sb);
1938	len = sbuf_len(sb) + 1;
1939	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1940	SLIST_INSERT_HEAD(&ma->list, maa, next);
1941	bcopy(sbuf_data(sb), maa + 1, len);
1942	sbuf_delete(sb);
1943
1944	ma->v[ma->len].iov_base = maa + 1;
1945	ma->v[ma->len].iov_len = len;
1946	ma->len++;
1947
1948	return (ma);
1949}
1950
1951/*
1952 * Add an argument which is a userland string.
1953 */
1954struct mntarg *
1955mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
1956{
1957	struct mntaarg *maa;
1958	char *tbuf;
1959
1960	if (val == NULL)
1961		return (ma);
1962	if (ma == NULL) {
1963		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1964		SLIST_INIT(&ma->list);
1965	}
1966	if (ma->error)
1967		return (ma);
1968	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1969	SLIST_INSERT_HEAD(&ma->list, maa, next);
1970	tbuf = (void *)(maa + 1);
1971	ma->error = copyinstr(val, tbuf, len, NULL);
1972	return (mount_arg(ma, name, tbuf, -1));
1973}
1974
1975/*
1976 * Plain argument.
1977 *
1978 * If length is -1, use printf.
1979 */
1980struct mntarg *
1981mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
1982{
1983
1984	if (ma == NULL) {
1985		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1986		SLIST_INIT(&ma->list);
1987	}
1988	if (ma->error)
1989		return (ma);
1990
1991	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1992	    M_MOUNT, M_WAITOK);
1993	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1994	ma->v[ma->len].iov_len = strlen(name) + 1;
1995	ma->len++;
1996
1997	ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
1998	if (len < 0)
1999		ma->v[ma->len].iov_len = strlen(val) + 1;
2000	else
2001		ma->v[ma->len].iov_len = len;
2002	ma->len++;
2003	return (ma);
2004}
2005
2006/*
2007 * Free a mntarg structure
2008 */
2009static void
2010free_mntarg(struct mntarg *ma)
2011{
2012	struct mntaarg *maa;
2013
2014	while (!SLIST_EMPTY(&ma->list)) {
2015		maa = SLIST_FIRST(&ma->list);
2016		SLIST_REMOVE_HEAD(&ma->list, next);
2017		free(maa, M_MOUNT);
2018	}
2019	free(ma->v, M_MOUNT);
2020	free(ma, M_MOUNT);
2021}
2022
2023/*
2024 * Mount a filesystem
2025 */
2026int
2027kernel_mount(struct mntarg *ma, int flags)
2028{
2029	struct uio auio;
2030	int error;
2031
2032	KASSERT(ma != NULL, ("kernel_mount NULL ma"));
2033	KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
2034	KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
2035
2036	auio.uio_iov = ma->v;
2037	auio.uio_iovcnt = ma->len;
2038	auio.uio_segflg = UIO_SYSSPACE;
2039
2040	error = ma->error;
2041	if (!error)
2042		error = vfs_donmount(curthread, flags, &auio);
2043	free_mntarg(ma);
2044	return (error);
2045}
2046
2047/*
2048 * A printflike function to mount a filesystem.
2049 */
2050int
2051kernel_vmount(int flags, ...)
2052{
2053	struct mntarg *ma = NULL;
2054	va_list ap;
2055	const char *cp;
2056	const void *vp;
2057	int error;
2058
2059	va_start(ap, flags);
2060	for (;;) {
2061		cp = va_arg(ap, const char *);
2062		if (cp == NULL)
2063			break;
2064		vp = va_arg(ap, const void *);
2065		ma = mount_arg(ma, cp, vp, -1);
2066	}
2067	va_end(ap);
2068
2069	error = kernel_mount(ma, flags);
2070	return (error);
2071}
2072