vfs_mount.c revision 122567
1/*
2 * Copyright (c) 1989, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 * (c) UNIX System Laboratories, Inc.
5 * All or some portions of this file are derived from material licensed
6 * to the University of California by American Telephone and Telegraph
7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
8 * the permission of UNIX System Laboratories, Inc.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 *    notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 *    notice, this list of conditions and the following disclaimer in the
17 *    documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 *    must display the following acknowledgement:
20 *	This product includes software developed by the University of
21 *	California, Berkeley and its contributors.
22 * 4. 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 REGENTS 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 REGENTS 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 * Copyright (c) 1999 Michael Smith
39 * All rights reserved.
40 * Copyright (c) 1999 Poul-Henning Kamp
41 * All rights reserved.
42 *
43 * Redistribution and use in source and binary forms, with or without
44 * modification, are permitted provided that the following conditions
45 * are met:
46 * 1. Redistributions of source code must retain the above copyright
47 *    notice, this list of conditions and the following disclaimer.
48 * 2. Redistributions in binary form must reproduce the above copyright
49 *    notice, this list of conditions and the following disclaimer in the
50 *    documentation and/or other materials provided with the distribution.
51 *
52 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
56 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62 * SUCH DAMAGE.
63 */
64
65#include <sys/cdefs.h>
66__FBSDID("$FreeBSD: head/sys/kern/vfs_mount.c 122567 2003-11-12 17:09:12Z peter $");
67
68#include <sys/param.h>
69#include <sys/conf.h>
70#include <sys/cons.h>
71#include <sys/kernel.h>
72#include <sys/linker.h>
73#include <sys/mac.h>
74#include <sys/malloc.h>
75#include <sys/mount.h>
76#include <sys/mutex.h>
77#include <sys/namei.h>
78#include <sys/proc.h>
79#include <sys/filedesc.h>
80#include <sys/reboot.h>
81#include <sys/sysproto.h>
82#include <sys/sx.h>
83#include <sys/sysctl.h>
84#include <sys/sysent.h>
85#include <sys/systm.h>
86#include <sys/vnode.h>
87
88#include <geom/geom.h>
89
90#include <machine/stdarg.h>
91
92#include "opt_rootdevname.h"
93#include "opt_ddb.h"
94#include "opt_mac.h"
95
96#ifdef DDB
97#include <ddb/ddb.h>
98#endif
99
100#define ROOTNAME	"root_device"
101
102static void	checkdirs(struct vnode *olddp, struct vnode *newdp);
103static int	vfs_nmount(struct thread *td, int, struct uio *);
104static int	vfs_mountroot_try(char *mountfrom);
105static int	vfs_mountroot_ask(void);
106static int	vfs_mount_alloc(struct vnode *, struct vfsconf *,
107		    const char *, struct thread *, struct mount **);
108static int	vfs_domount(struct thread *, const char *, char *,
109    		    int fsflags, void *fsdata, int compat);
110static void	gets(char *cp);
111
112static int	usermount = 0;	/* if 1, non-root can mount fs. */
113SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0, "");
114
115MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
116
117/* List of mounted filesystems. */
118struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
119
120/* For any iteration/modification of mountlist */
121struct mtx mountlist_mtx;
122
123/*
124 * The vnode of the system's root (/ in the filesystem, without chroot
125 * active.)
126 */
127struct vnode	*rootvnode;
128
129/*
130 * The root filesystem is detailed in the kernel environment variable
131 * vfs.root.mountfrom, which is expected to be in the general format
132 *
133 * <vfsname>:[<path>]
134 * vfsname   := the name of a VFS known to the kernel and capable
135 *              of being mounted as root
136 * path      := disk device name or other data used by the filesystem
137 *              to locate its physical store
138 */
139
140/*
141 * The root specifiers we will try if RB_CDROM is specified.
142 */
143static char *cdrom_rootdevnames[] = {
144	"cd9660:cd0",
145	"cd9660:acd0",
146	NULL
147};
148
149/* legacy find-root code */
150char		*rootdevnames[2] = {NULL, NULL};
151static int	setrootbyname(char *name);
152dev_t		rootdev = NODEV;
153
154/*
155 * Has to be dynamic as the value of rootdev can change; however, it can't
156 * change after the root is mounted, so a user process can't access this
157 * sysctl until after the value is unchangeable.
158 */
159static int
160sysctl_rootdev(SYSCTL_HANDLER_ARGS)
161{
162	int error;
163
164	/* _RD prevents this from happening. */
165	KASSERT(req->newptr == NULL, ("Attempt to change root device name"));
166
167	if (rootdev != NODEV)
168		error = sysctl_handle_string(oidp, rootdev->si_name, 0, req);
169	else
170		error = sysctl_handle_string(oidp, "", 0, req);
171
172	return (error);
173}
174
175SYSCTL_PROC(_kern, OID_AUTO, rootdev, CTLTYPE_STRING | CTLFLAG_RD,
176    0, 0, sysctl_rootdev, "A", "Root file system device");
177
178/* Remove one mount option. */
179static void
180vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
181{
182
183	TAILQ_REMOVE(opts, opt, link);
184	free(opt->name, M_MOUNT);
185	if (opt->value != NULL)
186		free(opt->value, M_MOUNT);
187#ifdef INVARIANTS
188	else if (opt->len != 0)
189		panic("%s: mount option with NULL value but length != 0",
190		    __func__);
191#endif
192	free(opt, M_MOUNT);
193}
194
195/* Release all resources related to the mount options. */
196static void
197vfs_freeopts(struct vfsoptlist *opts)
198{
199	struct vfsopt *opt;
200
201	while (!TAILQ_EMPTY(opts)) {
202		opt = TAILQ_FIRST(opts);
203		vfs_freeopt(opts, opt);
204	}
205	free(opts, M_MOUNT);
206}
207
208/*
209 * If a mount option is specified several times,
210 * (with or without the "no" prefix) only keep
211 * the last occurence of it.
212 */
213static void
214vfs_sanitizeopts(struct vfsoptlist *opts)
215{
216	struct vfsopt *opt, *opt2, *tmp;
217	int noopt;
218
219	TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
220		if (strncmp(opt->name, "no", 2) == 0)
221			noopt = 1;
222		else
223			noopt = 0;
224		opt2 = TAILQ_PREV(opt, vfsoptlist, link);
225		while (opt2 != NULL) {
226			if (strcmp(opt2->name, opt->name) == 0 ||
227			    (noopt && strcmp(opt->name + 2, opt2->name) == 0) ||
228			    (!noopt && strncmp(opt2->name, "no", 2) == 0 &&
229			    strcmp(opt2->name + 2, opt->name) == 0)) {
230				tmp = TAILQ_PREV(opt2, vfsoptlist, link);
231				vfs_freeopt(opts, opt2);
232				opt2 = tmp;
233			} else {
234				opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
235			}
236		}
237	}
238}
239
240/*
241 * Build a linked list of mount options from a struct uio.
242 */
243#define VFS_MOUNTARG_SIZE_MAX	(1024*64)
244
245static int
246vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
247{
248	struct vfsoptlist *opts;
249	struct vfsopt *opt;
250	unsigned int i, iovcnt;
251	int error, namelen, optlen;
252	size_t memused = 0;
253
254	iovcnt = auio->uio_iovcnt;
255	opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
256	TAILQ_INIT(opts);
257	for (i = 0; i < iovcnt; i += 2) {
258		opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
259		namelen = auio->uio_iov[i].iov_len;
260		optlen = auio->uio_iov[i + 1].iov_len;
261		opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
262		opt->value = NULL;
263		opt->len = optlen;
264
265		/*
266		 * Do this early, so jumps to "bad" will free the current
267		 * option
268		 */
269		TAILQ_INSERT_TAIL(opts, opt, link);
270		memused += sizeof (struct vfsopt) + optlen + namelen;
271
272		/*
273		 * Avoid consuming too much memory, and attempts to overflow
274		 * memused
275		 */
276		if (memused > VFS_MOUNTARG_SIZE_MAX ||
277		    optlen > VFS_MOUNTARG_SIZE_MAX ||
278		    namelen > VFS_MOUNTARG_SIZE_MAX) {
279			error = EINVAL;
280			goto bad;
281		}
282
283		if (auio->uio_segflg == UIO_SYSSPACE) {
284			bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
285		} else {
286			error = copyin(auio->uio_iov[i].iov_base, opt->name,
287			    namelen);
288			if (error)
289				goto bad;
290		}
291		/* Ensure names are null-terminated strings */
292		if (opt->name[namelen - 1] != '\0') {
293		    error = EINVAL;
294		    goto bad;
295		}
296		if (optlen != 0) {
297			opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
298			if (auio->uio_segflg == UIO_SYSSPACE) {
299				bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
300				    optlen);
301			} else {
302				error = copyin(auio->uio_iov[i + 1].iov_base,
303				    opt->value, optlen);
304				if (error)
305					goto bad;
306			}
307		}
308	}
309	vfs_sanitizeopts(opts);
310	*options = opts;
311	return (0);
312bad:
313	vfs_freeopts(opts);
314	return (error);
315}
316
317/*
318 * Merge the old mount options with the new ones passed
319 * in the MNT_UPDATE case.
320 */
321static void
322vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *opts)
323{
324	struct vfsopt *opt, *opt2, *new;
325
326	TAILQ_FOREACH(opt, opts, link) {
327		/*
328		 * Check that this option hasn't been redefined
329		 * nor cancelled with a "no" mount option.
330		 */
331		opt2 = TAILQ_FIRST(toopts);
332		while (opt2 != NULL) {
333			if (strcmp(opt2->name, opt->name) == 0)
334				goto next;
335			if (strncmp(opt2->name, "no", 2) == 0 &&
336			    strcmp(opt2->name + 2, opt->name) == 0) {
337				vfs_freeopt(toopts, opt2);
338				goto next;
339			}
340			opt2 = TAILQ_NEXT(opt2, link);
341		}
342		/* We want this option, duplicate it. */
343		new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
344		new->name = malloc(strlen(opt->name) + 1, M_MOUNT, M_WAITOK);
345		strcpy(new->name, opt->name);
346		if (opt->len != 0) {
347			new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
348			bcopy(opt->value, new->value, opt->len);
349		} else {
350			new->value = NULL;
351		}
352		new->len = opt->len;
353		TAILQ_INSERT_TAIL(toopts, new, link);
354next:
355		continue;
356	}
357}
358
359/*
360 * New mount API.
361 */
362int
363nmount(td, uap)
364	struct thread *td;
365	struct nmount_args /* {
366		struct iovec *iovp;
367		unsigned int iovcnt;
368		int flags;
369	} */ *uap;
370{
371	struct uio auio;
372	struct iovec *iov, *needfree;
373	struct iovec aiov[UIO_SMALLIOV];
374	unsigned int i;
375	int error;
376	u_int iovlen, iovcnt;
377
378	iovcnt = uap->iovcnt;
379	iovlen = iovcnt * sizeof (struct iovec);
380	/*
381	 * Check that we have an even number of iovec's
382	 * and that we have at least two options.
383	 */
384	if ((iovcnt & 1) || (iovcnt < 4) || (iovcnt > UIO_MAXIOV))
385		return (EINVAL);
386
387	if (iovcnt > UIO_SMALLIOV) {
388		MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK);
389		needfree = iov;
390	} else {
391		iov = aiov;
392		needfree = NULL;
393	}
394	auio.uio_iov = iov;
395	auio.uio_iovcnt = iovcnt;
396	auio.uio_segflg = UIO_USERSPACE;
397	if ((error = copyin(uap->iovp, iov, iovlen)))
398		goto finish;
399
400	for (i = 0; i < iovcnt; i++) {
401		if (iov->iov_len > MMAXOPTIONLEN) {
402			error = EINVAL;
403			goto finish;
404		}
405		iov++;
406	}
407	error = vfs_nmount(td, uap->flags, &auio);
408finish:
409	if (needfree != NULL)
410		free(needfree, M_TEMP);
411	return (error);
412}
413
414int
415kernel_mount(iovp, iovcnt, flags)
416	struct iovec *iovp;
417	unsigned int iovcnt;
418	int flags;
419{
420	struct uio auio;
421	int error;
422
423	/*
424	 * Check that we have an even number of iovec's
425	 * and that we have at least two options.
426	 */
427	if ((iovcnt & 1) || (iovcnt < 4))
428		return (EINVAL);
429
430	auio.uio_iov = iovp;
431	auio.uio_iovcnt = iovcnt;
432	auio.uio_segflg = UIO_SYSSPACE;
433
434	error = vfs_nmount(curthread, flags, &auio);
435	return (error);
436}
437
438int
439kernel_vmount(int flags, ...)
440{
441	struct iovec *iovp;
442	struct uio auio;
443	va_list ap;
444	unsigned int iovcnt, iovlen, len;
445	const char *cp;
446	char *buf, *pos;
447	size_t n;
448	int error, i;
449
450	len = 0;
451	va_start(ap, flags);
452	for (iovcnt = 0; (cp = va_arg(ap, const char *)) != NULL; iovcnt++)
453		len += strlen(cp) + 1;
454	va_end(ap);
455
456	if (iovcnt < 4 || iovcnt & 1)
457		return (EINVAL);
458
459	iovlen = iovcnt * sizeof (struct iovec);
460	MALLOC(iovp, struct iovec *, iovlen, M_MOUNT, M_WAITOK);
461	MALLOC(buf, char *, len, M_MOUNT, M_WAITOK);
462	pos = buf;
463	va_start(ap, flags);
464	for (i = 0; i < iovcnt; i++) {
465		cp = va_arg(ap, const char *);
466		copystr(cp, pos, len - (pos - buf), &n);
467		iovp[i].iov_base = pos;
468		iovp[i].iov_len = n;
469		pos += n;
470	}
471	va_end(ap);
472
473	auio.uio_iov = iovp;
474	auio.uio_iovcnt = iovcnt;
475	auio.uio_segflg = UIO_SYSSPACE;
476
477	error = vfs_nmount(curthread, flags, &auio);
478	FREE(iovp, M_MOUNT);
479	FREE(buf, M_MOUNT);
480	return (error);
481}
482
483/*
484 * Allocate and initialize the mount point struct.
485 */
486static int
487vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp,
488    const char *fspath, struct thread *td, struct mount **mpp)
489{
490	struct mount *mp;
491
492	mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK | M_ZERO);
493	TAILQ_INIT(&mp->mnt_nvnodelist);
494	TAILQ_INIT(&mp->mnt_reservedvnlist);
495	mp->mnt_nvnodelistsize = 0;
496	mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF);
497	lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, LK_NOPAUSE);
498	vfs_busy(mp, LK_NOWAIT, 0, td);
499	mp->mnt_op = vfsp->vfc_vfsops;
500	mp->mnt_vfc = vfsp;
501	vfsp->vfc_refcount++;
502	mp->mnt_stat.f_type = vfsp->vfc_typenum;
503	mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
504	strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
505	mp->mnt_vnodecovered = vp;
506	mp->mnt_cred = crdup(td->td_ucred);
507	mp->mnt_stat.f_owner = td->td_ucred->cr_uid;
508	strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
509	mp->mnt_iosize_max = DFLTPHYS;
510#ifdef MAC
511	mac_init_mount(mp);
512	mac_create_mount(td->td_ucred, mp);
513#endif
514	*mpp = mp;
515	return (0);
516}
517
518void
519vfs_mount_destroy(struct mount *mp, struct thread *td)
520{
521
522	mp->mnt_vfc->vfc_refcount--;
523	if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
524		panic("unmount: dangling vnode");
525	vfs_unbusy(mp,td);
526	lockdestroy(&mp->mnt_lock);
527	mtx_destroy(&mp->mnt_mtx);
528	if (mp->mnt_kern_flag & MNTK_MWAIT)
529		wakeup(mp);
530#ifdef MAC
531	mac_destroy_mount(mp);
532#endif
533	if (mp->mnt_op->vfs_mount == NULL)
534		vfs_freeopts(mp->mnt_opt);
535	crfree(mp->mnt_cred);
536	free(mp, M_MOUNT);
537}
538
539static int
540vfs_nmount(struct thread *td, int fsflags, struct uio *fsoptions)
541{
542	struct vfsoptlist *optlist;
543	char *fstype, *fspath;
544	int fstypelen, fspathlen;
545	int error;
546
547	error = vfs_buildopts(fsoptions, &optlist);
548	if (error)
549		return (error);
550
551	/*
552	 * We need these two options before the others,
553	 * and they are mandatory for any filesystem.
554	 * Ensure they are NUL terminated as well.
555	 */
556	fstypelen = 0;
557	error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
558	if (error || fstype[fstypelen - 1] != '\0') {
559		error = EINVAL;
560		goto bail;
561	}
562	fspathlen = 0;
563	error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
564	if (error || fspath[fspathlen - 1] != '\0') {
565		error = EINVAL;
566		goto bail;
567	}
568
569	/*
570	 * Be ultra-paranoid about making sure the type and fspath
571	 * variables will fit in our mp buffers, including the
572	 * terminating NUL.
573	 */
574	if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) {
575		error = ENAMETOOLONG;
576		goto bail;
577	}
578
579	error = vfs_domount(td, fstype, fspath, fsflags, optlist, 0);
580bail:
581	if (error)
582		vfs_freeopts(optlist);
583	return (error);
584}
585
586/*
587 * Old mount API.
588 */
589#ifndef _SYS_SYSPROTO_H_
590struct mount_args {
591	char	*type;
592	char	*path;
593	int	flags;
594	caddr_t	data;
595};
596#endif
597/* ARGSUSED */
598int
599mount(td, uap)
600	struct thread *td;
601	struct mount_args /* {
602		char *type;
603		char *path;
604		int flags;
605		caddr_t data;
606	} */ *uap;
607{
608	char *fstype;
609	char *fspath;
610	int error;
611
612	fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
613	fspath = malloc(MNAMELEN, M_TEMP, M_WAITOK);
614
615	/*
616	 * vfs_mount() actually takes a kernel string for `type' and
617	 * `path' now, so extract them.
618	 */
619	error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
620	if (error == 0)
621		error = copyinstr(uap->path, fspath, MNAMELEN, NULL);
622	if (error == 0)
623		error = vfs_domount(td, fstype, fspath, uap->flags,
624		    uap->data, 1);
625	free(fstype, M_TEMP);
626	free(fspath, M_TEMP);
627	return (error);
628}
629
630/*
631 * vfs_mount(): actually attempt a filesystem mount.
632 *
633 * This routine is designed to be a "generic" entry point for routines
634 * that wish to mount a filesystem. All parameters except `fsdata' are
635 * pointers into kernel space. `fsdata' is currently still a pointer
636 * into userspace.
637 */
638int
639vfs_mount(td, fstype, fspath, fsflags, fsdata)
640        struct thread *td;
641        const char *fstype;
642        char *fspath;
643        int fsflags;
644        void *fsdata;
645{
646	return (vfs_domount(td,fstype, fspath, fsflags, fsdata, 1));
647}
648
649/*
650 * vfs_domount(): actually attempt a filesystem mount.
651 */
652static int
653vfs_domount(
654	struct thread *td,
655	const char *fstype,
656	char *fspath,
657	int fsflags,
658	void *fsdata,
659	int compat)
660{
661	linker_file_t lf;
662	struct vnode *vp;
663	struct mount *mp;
664	struct vfsconf *vfsp;
665	int error, flag = 0, kern_flag = 0;
666	struct vattr va;
667	struct nameidata nd;
668
669	/*
670	 * Be ultra-paranoid about making sure the type and fspath
671	 * variables will fit in our mp buffers, including the
672	 * terminating NUL.
673	 */
674	if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
675		return (ENAMETOOLONG);
676
677	if (usermount == 0) {
678	       	error = suser(td);
679		if (error)
680			return (error);
681	}
682	/*
683	 * Do not allow NFS export by non-root users.
684	 */
685	if (fsflags & MNT_EXPORTED) {
686		error = suser(td);
687		if (error)
688			return (error);
689	}
690	/*
691	 * Silently enforce MNT_NOSUID and MNT_NODEV for non-root users.
692	 */
693	if (suser(td))
694		fsflags |= MNT_NOSUID | MNT_NODEV;
695	/*
696	 * Get vnode to be covered
697	 */
698	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td);
699	if ((error = namei(&nd)) != 0)
700		return (error);
701	NDFREE(&nd, NDF_ONLY_PNBUF);
702	vp = nd.ni_vp;
703	if (fsflags & MNT_UPDATE) {
704		if ((vp->v_vflag & VV_ROOT) == 0) {
705			vput(vp);
706			return (EINVAL);
707		}
708		mp = vp->v_mount;
709		flag = mp->mnt_flag;
710		kern_flag = mp->mnt_kern_flag;
711		/*
712		 * We only allow the filesystem to be reloaded if it
713		 * is currently mounted read-only.
714		 */
715		if ((fsflags & MNT_RELOAD) &&
716		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
717			vput(vp);
718			return (EOPNOTSUPP);	/* Needs translation */
719		}
720		/*
721		 * Only root, or the user that did the original mount is
722		 * permitted to update it.
723		 */
724		if (mp->mnt_cred->cr_uid != td->td_ucred->cr_uid) {
725			error = suser(td);
726			if (error) {
727				vput(vp);
728				return (error);
729			}
730		}
731		if (vfs_busy(mp, LK_NOWAIT, 0, td)) {
732			vput(vp);
733			return (EBUSY);
734		}
735		VI_LOCK(vp);
736		if ((vp->v_iflag & VI_MOUNT) != 0 ||
737		    vp->v_mountedhere != NULL) {
738			VI_UNLOCK(vp);
739			vfs_unbusy(mp, td);
740			vput(vp);
741			return (EBUSY);
742		}
743		vp->v_iflag |= VI_MOUNT;
744		VI_UNLOCK(vp);
745		mp->mnt_flag |= fsflags &
746		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT);
747		VOP_UNLOCK(vp, 0, td);
748		if (compat == 0) {
749			mp->mnt_optnew = fsdata;
750			vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
751		}
752		goto update;
753	}
754	/*
755	 * If the user is not root, ensure that they own the directory
756	 * onto which we are attempting to mount.
757	 */
758	error = VOP_GETATTR(vp, &va, td->td_ucred, td);
759	if (error) {
760		vput(vp);
761		return (error);
762	}
763	if (va.va_uid != td->td_ucred->cr_uid) {
764		error = suser(td);
765		if (error) {
766			vput(vp);
767			return (error);
768		}
769	}
770	if ((error = vinvalbuf(vp, V_SAVE, td->td_ucred, td, 0, 0)) != 0) {
771		vput(vp);
772		return (error);
773	}
774	if (vp->v_type != VDIR) {
775		vput(vp);
776		return (ENOTDIR);
777	}
778	for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
779		if (!strcmp(vfsp->vfc_name, fstype))
780			break;
781	if (vfsp == NULL) {
782		/* Only load modules for root (very important!). */
783		error = suser(td);
784		if (error) {
785			vput(vp);
786			return (error);
787		}
788		error = securelevel_gt(td->td_ucred, 0);
789		if (error) {
790			vput(vp);
791			return (error);
792		}
793		error = linker_load_module(NULL, fstype, NULL, NULL, &lf);
794		if (error || lf == NULL) {
795			vput(vp);
796			if (lf == NULL)
797				error = ENODEV;
798			return (error);
799		}
800		lf->userrefs++;
801		/* Look up again to see if the VFS was loaded. */
802		for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
803			if (!strcmp(vfsp->vfc_name, fstype))
804				break;
805		if (vfsp == NULL) {
806			lf->userrefs--;
807			linker_file_unload(lf);
808			vput(vp);
809			return (ENODEV);
810		}
811	}
812	VI_LOCK(vp);
813	if ((vp->v_iflag & VI_MOUNT) != 0 ||
814	    vp->v_mountedhere != NULL) {
815		VI_UNLOCK(vp);
816		vput(vp);
817		return (EBUSY);
818	}
819	vp->v_iflag |= VI_MOUNT;
820	VI_UNLOCK(vp);
821
822	/*
823	 * Allocate and initialize the filesystem.
824	 */
825	error = vfs_mount_alloc(vp, vfsp, fspath, td, &mp);
826	if (error) {
827		vput(vp);
828		return (error);
829	}
830	VOP_UNLOCK(vp, 0, td);
831
832	/* XXXMAC: pass to vfs_mount_alloc? */
833	if (compat == 0)
834		mp->mnt_optnew = fsdata;
835update:
836	/*
837	 * Check if the fs implements the type VFS_[N]MOUNT()
838	 * function we are looking for.
839	 */
840	if ((compat == 0) == (mp->mnt_op->vfs_mount != NULL)) {
841		printf("%s doesn't support the %s mount syscall\n",
842		    mp->mnt_vfc->vfc_name, compat? "old" : "new");
843		VI_LOCK(vp);
844		vp->v_iflag &= ~VI_MOUNT;
845		VI_UNLOCK(vp);
846		if (mp->mnt_flag & MNT_UPDATE)
847			vfs_unbusy(mp, td);
848		else
849			vfs_mount_destroy(mp, td);
850		vrele(vp);
851		return (EOPNOTSUPP);
852	}
853
854	/*
855	 * Set the mount level flags.
856	 */
857	if (fsflags & MNT_RDONLY)
858		mp->mnt_flag |= MNT_RDONLY;
859	else if (mp->mnt_flag & MNT_RDONLY)
860		mp->mnt_kern_flag |= MNTK_WANTRDWR;
861	mp->mnt_flag &=~ MNT_UPDATEMASK;
862	mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE);
863	/*
864	 * Mount the filesystem.
865	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
866	 * get.  No freeing of cn_pnbuf.
867	 */
868	error = compat? VFS_MOUNT(mp, fspath, fsdata, &nd, td) :
869	                VFS_NMOUNT(mp, &nd, td);
870	if (!error) {
871		if (mp->mnt_opt != NULL)
872			vfs_freeopts(mp->mnt_opt);
873		mp->mnt_opt = mp->mnt_optnew;
874	}
875	/*
876	 * Prevent external consumers of mount
877	 * options to read mnt_optnew.
878	 */
879	mp->mnt_optnew = NULL;
880	if (mp->mnt_flag & MNT_UPDATE) {
881		if (mp->mnt_kern_flag & MNTK_WANTRDWR)
882			mp->mnt_flag &= ~MNT_RDONLY;
883		mp->mnt_flag &=~
884		    (MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT);
885		mp->mnt_kern_flag &=~ MNTK_WANTRDWR;
886		if (error) {
887			mp->mnt_flag = flag;
888			mp->mnt_kern_flag = kern_flag;
889		}
890		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
891			if (mp->mnt_syncer == NULL)
892				error = vfs_allocate_syncvnode(mp);
893		} else {
894			if (mp->mnt_syncer != NULL)
895				vrele(mp->mnt_syncer);
896			mp->mnt_syncer = NULL;
897		}
898		vfs_unbusy(mp, td);
899		VI_LOCK(vp);
900		vp->v_iflag &= ~VI_MOUNT;
901		VI_UNLOCK(vp);
902		vrele(vp);
903		return (error);
904	}
905	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
906	/*
907	 * Put the new filesystem on the mount list after root.
908	 */
909	cache_purge(vp);
910	if (!error) {
911		struct vnode *newdp;
912
913		VI_LOCK(vp);
914		vp->v_iflag &= ~VI_MOUNT;
915		VI_UNLOCK(vp);
916		vp->v_mountedhere = mp;
917		mtx_lock(&mountlist_mtx);
918		TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
919		mtx_unlock(&mountlist_mtx);
920		if (VFS_ROOT(mp, &newdp))
921			panic("mount: lost mount");
922		checkdirs(vp, newdp);
923		vput(newdp);
924		VOP_UNLOCK(vp, 0, td);
925		if ((mp->mnt_flag & MNT_RDONLY) == 0)
926			error = vfs_allocate_syncvnode(mp);
927		vfs_unbusy(mp, td);
928		if ((error = VFS_START(mp, 0, td)) != 0)
929			vrele(vp);
930	} else {
931		VI_LOCK(vp);
932		vp->v_iflag &= ~VI_MOUNT;
933		VI_UNLOCK(vp);
934
935		vfs_mount_destroy(mp, td);
936		vput(vp);
937	}
938	return (error);
939}
940
941/*
942 * Scan all active processes to see if any of them have a current
943 * or root directory of `olddp'. If so, replace them with the new
944 * mount point.
945 */
946static void
947checkdirs(olddp, newdp)
948	struct vnode *olddp, *newdp;
949{
950	struct filedesc *fdp;
951	struct proc *p;
952	int nrele;
953
954	if (vrefcnt(olddp) == 1)
955		return;
956	sx_slock(&allproc_lock);
957	LIST_FOREACH(p, &allproc, p_list) {
958		mtx_lock(&fdesc_mtx);
959		fdp = p->p_fd;
960		if (fdp == NULL) {
961			mtx_unlock(&fdesc_mtx);
962			continue;
963		}
964		nrele = 0;
965		FILEDESC_LOCK(fdp);
966		if (fdp->fd_cdir == olddp) {
967			VREF(newdp);
968			fdp->fd_cdir = newdp;
969			nrele++;
970		}
971		if (fdp->fd_rdir == olddp) {
972			VREF(newdp);
973			fdp->fd_rdir = newdp;
974			nrele++;
975		}
976		FILEDESC_UNLOCK(fdp);
977		mtx_unlock(&fdesc_mtx);
978		while (nrele--)
979			vrele(olddp);
980	}
981	sx_sunlock(&allproc_lock);
982	if (rootvnode == olddp) {
983		vrele(rootvnode);
984		VREF(newdp);
985		rootvnode = newdp;
986	}
987}
988
989/*
990 * Unmount a filesystem.
991 *
992 * Note: unmount takes a path to the vnode mounted on as argument,
993 * not special file (as before).
994 */
995#ifndef _SYS_SYSPROTO_H_
996struct unmount_args {
997	char	*path;
998	int	flags;
999};
1000#endif
1001/* ARGSUSED */
1002int
1003unmount(td, uap)
1004	struct thread *td;
1005	register struct unmount_args /* {
1006		char *path;
1007		int flags;
1008	} */ *uap;
1009{
1010	struct mount *mp;
1011	char *pathbuf;
1012	int error, id0, id1;
1013
1014	pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1015	error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
1016	if (error) {
1017		free(pathbuf, M_TEMP);
1018		return (error);
1019	}
1020	if (uap->flags & MNT_BYFSID) {
1021		/* Decode the filesystem ID. */
1022		if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
1023			free(pathbuf, M_TEMP);
1024			return (EINVAL);
1025		}
1026
1027		mtx_lock(&mountlist_mtx);
1028		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list)
1029			if (mp->mnt_stat.f_fsid.val[0] == id0 &&
1030			    mp->mnt_stat.f_fsid.val[1] == id1)
1031				break;
1032		mtx_unlock(&mountlist_mtx);
1033	} else {
1034		mtx_lock(&mountlist_mtx);
1035		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list)
1036			if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0)
1037				break;
1038		mtx_unlock(&mountlist_mtx);
1039	}
1040	free(pathbuf, M_TEMP);
1041	if (mp == NULL) {
1042		/*
1043		 * Previously we returned ENOENT for a nonexistent path and
1044		 * EINVAL for a non-mountpoint.  We cannot tell these apart
1045		 * now, so in the !MNT_BYFSID case return the more likely
1046		 * EINVAL for compatibility.
1047		 */
1048		return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
1049	}
1050
1051	/*
1052	 * Only root, or the user that did the original mount is
1053	 * permitted to unmount this filesystem.
1054	 */
1055	if (mp->mnt_cred->cr_uid != td->td_ucred->cr_uid) {
1056		error = suser(td);
1057		if (error)
1058			return (error);
1059	}
1060
1061	/*
1062	 * Don't allow unmounting the root filesystem.
1063	 */
1064	if (mp->mnt_flag & MNT_ROOTFS)
1065		return (EINVAL);
1066	return (dounmount(mp, uap->flags, td));
1067}
1068
1069/*
1070 * Do the actual filesystem unmount.
1071 */
1072int
1073dounmount(mp, flags, td)
1074	struct mount *mp;
1075	int flags;
1076	struct thread *td;
1077{
1078	struct vnode *coveredvp, *fsrootvp;
1079	int error;
1080	int async_flag;
1081
1082	mtx_lock(&mountlist_mtx);
1083	if (mp->mnt_kern_flag & MNTK_UNMOUNT) {
1084		mtx_unlock(&mountlist_mtx);
1085		return (EBUSY);
1086	}
1087	mp->mnt_kern_flag |= MNTK_UNMOUNT;
1088	/* Allow filesystems to detect that a forced unmount is in progress. */
1089	if (flags & MNT_FORCE)
1090		mp->mnt_kern_flag |= MNTK_UNMOUNTF;
1091	error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK |
1092	    ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), &mountlist_mtx, td);
1093	if (error) {
1094		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1095		if (mp->mnt_kern_flag & MNTK_MWAIT)
1096			wakeup(mp);
1097		return (error);
1098	}
1099	vn_start_write(NULL, &mp, V_WAIT);
1100
1101	if (mp->mnt_flag & MNT_EXPUBLIC)
1102		vfs_setpublicfs(NULL, NULL, NULL);
1103
1104	vfs_msync(mp, MNT_WAIT);
1105	async_flag = mp->mnt_flag & MNT_ASYNC;
1106	mp->mnt_flag &=~ MNT_ASYNC;
1107	cache_purgevfs(mp);	/* remove cache entries for this file sys */
1108	if (mp->mnt_syncer != NULL)
1109		vrele(mp->mnt_syncer);
1110	/* Move process cdir/rdir refs on fs root to underlying vnode. */
1111	if (VFS_ROOT(mp, &fsrootvp) == 0) {
1112		if (mp->mnt_vnodecovered != NULL)
1113			checkdirs(fsrootvp, mp->mnt_vnodecovered);
1114		if (fsrootvp == rootvnode) {
1115			vrele(rootvnode);
1116			rootvnode = NULL;
1117		}
1118		vput(fsrootvp);
1119	}
1120	if (((mp->mnt_flag & MNT_RDONLY) ||
1121	     (error = VFS_SYNC(mp, MNT_WAIT, td->td_ucred, td)) == 0) ||
1122	    (flags & MNT_FORCE)) {
1123		error = VFS_UNMOUNT(mp, flags, td);
1124	}
1125	vn_finished_write(mp);
1126	if (error) {
1127		/* Undo cdir/rdir and rootvnode changes made above. */
1128		if (VFS_ROOT(mp, &fsrootvp) == 0) {
1129			if (mp->mnt_vnodecovered != NULL)
1130				checkdirs(mp->mnt_vnodecovered, fsrootvp);
1131			if (rootvnode == NULL) {
1132				rootvnode = fsrootvp;
1133				vref(rootvnode);
1134			}
1135			vput(fsrootvp);
1136		}
1137		if ((mp->mnt_flag & MNT_RDONLY) == 0 && mp->mnt_syncer == NULL)
1138			(void) vfs_allocate_syncvnode(mp);
1139		mtx_lock(&mountlist_mtx);
1140		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
1141		mp->mnt_flag |= async_flag;
1142		lockmgr(&mp->mnt_lock, LK_RELEASE | LK_INTERLOCK,
1143		    &mountlist_mtx, td);
1144		if (mp->mnt_kern_flag & MNTK_MWAIT)
1145			wakeup(mp);
1146		return (error);
1147	}
1148	mtx_lock(&mountlist_mtx);
1149	TAILQ_REMOVE(&mountlist, mp, mnt_list);
1150	if ((coveredvp = mp->mnt_vnodecovered) != NULL)
1151		coveredvp->v_mountedhere = NULL;
1152	mtx_unlock(&mountlist_mtx);
1153	vfs_mount_destroy(mp, td);
1154	if (coveredvp != NULL)
1155		vrele(coveredvp);
1156	return (0);
1157}
1158
1159/*
1160 * Lookup a filesystem type, and if found allocate and initialize
1161 * a mount structure for it.
1162 *
1163 * Devname is usually updated by mount(8) after booting.
1164 */
1165int
1166vfs_rootmountalloc(fstypename, devname, mpp)
1167	char *fstypename;
1168	char *devname;
1169	struct mount **mpp;
1170{
1171	struct thread *td = curthread;	/* XXX */
1172	struct vfsconf *vfsp;
1173	struct mount *mp;
1174	int error;
1175
1176	if (fstypename == NULL)
1177		return (ENODEV);
1178	for (vfsp = vfsconf; vfsp; vfsp = vfsp->vfc_next)
1179		if (!strcmp(vfsp->vfc_name, fstypename))
1180			break;
1181	if (vfsp == NULL)
1182		return (ENODEV);
1183
1184	error = vfs_mount_alloc(NULLVP, vfsp, "/", td, &mp);
1185	if (error)
1186		return (error);
1187
1188	mp->mnt_flag |= MNT_RDONLY | MNT_ROOTFS;
1189	strlcpy(mp->mnt_stat.f_mntfromname, devname, MNAMELEN);
1190
1191	*mpp = mp;
1192	return (0);
1193}
1194
1195/*
1196 * Find and mount the root filesystem
1197 */
1198void
1199vfs_mountroot(void)
1200{
1201	char		*cp;
1202	int		i, error;
1203
1204	g_waitidle();
1205
1206	/*
1207	 * The root filesystem information is compiled in, and we are
1208	 * booted with instructions to use it.
1209	 */
1210#ifdef ROOTDEVNAME
1211	if ((boothowto & RB_DFLTROOT) &&
1212	    !vfs_mountroot_try(ROOTDEVNAME))
1213		return;
1214#endif
1215	/*
1216	 * We are booted with instructions to prompt for the root filesystem,
1217	 * or to use the compiled-in default when it doesn't exist.
1218	 */
1219	if (boothowto & (RB_DFLTROOT | RB_ASKNAME)) {
1220		if (!vfs_mountroot_ask())
1221			return;
1222	}
1223
1224	/*
1225	 * We've been given the generic "use CDROM as root" flag.  This is
1226	 * necessary because one media may be used in many different
1227	 * devices, so we need to search for them.
1228	 */
1229	if (boothowto & RB_CDROM) {
1230		for (i = 0; cdrom_rootdevnames[i] != NULL; i++) {
1231			if (!vfs_mountroot_try(cdrom_rootdevnames[i]))
1232				return;
1233		}
1234	}
1235
1236	/*
1237	 * Try to use the value read by the loader from /etc/fstab, or
1238	 * supplied via some other means.  This is the preferred
1239	 * mechanism.
1240	 */
1241	if ((cp = getenv("vfs.root.mountfrom")) != NULL) {
1242		error = vfs_mountroot_try(cp);
1243		freeenv(cp);
1244		if (!error)
1245			return;
1246	}
1247
1248	/*
1249	 * Try values that may have been computed by the machine-dependant
1250	 * legacy code.
1251	 */
1252	if (!vfs_mountroot_try(rootdevnames[0]))
1253		return;
1254	if (!vfs_mountroot_try(rootdevnames[1]))
1255		return;
1256
1257	/*
1258	 * If we have a compiled-in default, and haven't already tried it, try
1259	 * it now.
1260	 */
1261#ifdef ROOTDEVNAME
1262	if (!(boothowto & RB_DFLTROOT))
1263		if (!vfs_mountroot_try(ROOTDEVNAME))
1264			return;
1265#endif
1266
1267	/*
1268	 * Everything so far has failed, prompt on the console if we haven't
1269	 * already tried that.
1270	 */
1271	if (!(boothowto & (RB_DFLTROOT | RB_ASKNAME)) && !vfs_mountroot_ask())
1272		return;
1273	panic("Root mount failed, startup aborted.");
1274}
1275
1276/*
1277 * Mount (mountfrom) as the root filesystem.
1278 */
1279static int
1280vfs_mountroot_try(char *mountfrom)
1281{
1282        struct mount	*mp;
1283	char		*vfsname, *path;
1284	const char	*devname;
1285	int		error;
1286	char		patt[32];
1287	int		s;
1288
1289	vfsname = NULL;
1290	path    = NULL;
1291	mp      = NULL;
1292	error   = EINVAL;
1293
1294	if (mountfrom == NULL)
1295		return(error);		/* don't complain */
1296
1297	s = splcam();			/* Overkill, but annoying without it */
1298	printf("Mounting root from %s\n", mountfrom);
1299	splx(s);
1300
1301	/* parse vfs name and path */
1302	vfsname = malloc(MFSNAMELEN, M_MOUNT, M_WAITOK);
1303	path = malloc(MNAMELEN, M_MOUNT, M_WAITOK);
1304	vfsname[0] = path[0] = 0;
1305	sprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN);
1306	if (sscanf(mountfrom, patt, vfsname, path) < 1)
1307		goto done;
1308
1309	/* allocate a root mount */
1310	error = vfs_rootmountalloc(vfsname, path[0] != 0 ? path : ROOTNAME,
1311				   &mp);
1312	if (error != 0) {
1313		printf("Can't allocate root mount for filesystem '%s': %d\n",
1314		       vfsname, error);
1315		goto done;
1316	}
1317
1318	/* do our best to set rootdev */
1319	if ((path[0] != 0) && setrootbyname(path))
1320		printf("setrootbyname failed\n");
1321
1322	/* If the root device is a type "memory disk", mount RW */
1323	if (rootdev != NODEV && devsw(rootdev) != NULL) {
1324		devname = devtoname(rootdev);
1325		if (devname[0] == 'm' && devname[1] == 'd')
1326			mp->mnt_flag &= ~MNT_RDONLY;
1327	}
1328
1329	error = VFS_MOUNT(mp, NULL, NULL, NULL, curthread);
1330
1331done:
1332	if (vfsname != NULL)
1333		free(vfsname, M_MOUNT);
1334	if (path != NULL)
1335		free(path, M_MOUNT);
1336	if (error != 0) {
1337		if (mp != NULL)
1338			vfs_mount_destroy(mp, curthread);
1339		printf("Root mount failed: %d\n", error);
1340	} else {
1341
1342		/* register with list of mounted filesystems */
1343		mtx_lock(&mountlist_mtx);
1344		TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list);
1345		mtx_unlock(&mountlist_mtx);
1346
1347		/* sanity check system clock against root fs timestamp */
1348		inittodr(mp->mnt_time);
1349		vfs_unbusy(mp, curthread);
1350		error = VFS_START(mp, 0, curthread);
1351	}
1352	return(error);
1353}
1354
1355/*
1356 * Spin prompting on the console for a suitable root filesystem
1357 */
1358static int
1359vfs_mountroot_ask(void)
1360{
1361	char name[128];
1362
1363	for(;;) {
1364		printf("\nManual root filesystem specification:\n");
1365		printf("  <fstype>:<device>  Mount <device> using filesystem <fstype>\n");
1366#if defined(__i386__) || defined(__ia64__)
1367		printf("                       eg. ufs:da0s1a\n");
1368#else
1369		printf("                       eg. ufs:/dev/da0a\n");
1370#endif
1371		printf("  ?                  List valid disk boot devices\n");
1372		printf("  <empty line>       Abort manual input\n");
1373		printf("\nmountroot> ");
1374		gets(name);
1375		if (name[0] == 0)
1376			return(1);
1377		if (name[0] == '?') {
1378			printf("\nList of GEOM managed disk devices:\n  ");
1379			g_dev_print();
1380			continue;
1381		}
1382		if (!vfs_mountroot_try(name))
1383			return(0);
1384	}
1385}
1386
1387/*
1388 * Local helper function for vfs_mountroot_ask.
1389 */
1390static void
1391gets(char *cp)
1392{
1393	char *lp;
1394	int c;
1395
1396	lp = cp;
1397	for (;;) {
1398		printf("%c", c = cngetc() & 0177);
1399		switch (c) {
1400		case -1:
1401		case '\n':
1402		case '\r':
1403			*lp++ = '\0';
1404			return;
1405		case '\b':
1406		case '\177':
1407			if (lp > cp) {
1408				printf(" \b");
1409				lp--;
1410			}
1411			continue;
1412		case '#':
1413			lp--;
1414			if (lp < cp)
1415				lp = cp;
1416			continue;
1417		case '@':
1418		case 'u' & 037:
1419			lp = cp;
1420			printf("%c", '\n');
1421			continue;
1422		default:
1423			*lp++ = c;
1424		}
1425	}
1426}
1427
1428/*
1429 * Convert a given name to the dev_t of the disk-like device
1430 * it refers to.
1431 */
1432dev_t
1433getdiskbyname(char *name) {
1434	char *cp;
1435	dev_t dev;
1436
1437	cp = name;
1438	if (!bcmp(cp, "/dev/", 5))
1439		cp += 5;
1440
1441	dev = NODEV;
1442	EVENTHANDLER_INVOKE(dev_clone, cp, strlen(cp), &dev);
1443	return (dev);
1444}
1445
1446/*
1447 * Set rootdev to match (name), given that we expect it to
1448 * refer to a disk-like device.
1449 */
1450static int
1451setrootbyname(char *name)
1452{
1453	dev_t diskdev;
1454
1455	diskdev = getdiskbyname(name);
1456	if (diskdev != NODEV) {
1457		rootdev = diskdev;
1458		return (0);
1459	}
1460
1461	return (1);
1462}
1463
1464/* Show the dev_t for a disk specified by name */
1465#ifdef DDB
1466DB_SHOW_COMMAND(disk, db_getdiskbyname)
1467{
1468	dev_t dev;
1469
1470	if (modif[0] == '\0') {
1471		db_error("usage: show disk/devicename");
1472		return;
1473	}
1474	dev = getdiskbyname(modif);
1475	if (dev != NODEV)
1476		db_printf("dev_t = %p\n", dev);
1477	else
1478		db_printf("No disk device matched.\n");
1479}
1480#endif
1481
1482/*
1483 * Get a mount option by its name.
1484 *
1485 * Return 0 if the option was found, ENOENT otherwise.
1486 * If len is non-NULL it will be filled with the length
1487 * of the option. If buf is non-NULL, it will be filled
1488 * with the address of the option.
1489 */
1490int
1491vfs_getopt(opts, name, buf, len)
1492	struct vfsoptlist *opts;
1493	const char *name;
1494	void **buf;
1495	int *len;
1496{
1497	struct vfsopt *opt;
1498
1499	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1500
1501	TAILQ_FOREACH(opt, opts, link) {
1502		if (strcmp(name, opt->name) == 0) {
1503			if (len != NULL)
1504				*len = opt->len;
1505			if (buf != NULL)
1506				*buf = opt->value;
1507			return (0);
1508		}
1509	}
1510	return (ENOENT);
1511}
1512
1513/*
1514 * Find and copy a mount option.
1515 *
1516 * The size of the buffer has to be specified
1517 * in len, if it is not the same length as the
1518 * mount option, EINVAL is returned.
1519 * Returns ENOENT if the option is not found.
1520 */
1521int
1522vfs_copyopt(opts, name, dest, len)
1523	struct vfsoptlist *opts;
1524	const char *name;
1525	void *dest;
1526	int len;
1527{
1528	struct vfsopt *opt;
1529
1530	KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1531
1532	TAILQ_FOREACH(opt, opts, link) {
1533		if (strcmp(name, opt->name) == 0) {
1534			if (len != opt->len)
1535				return (EINVAL);
1536			bcopy(opt->value, dest, opt->len);
1537			return (0);
1538		}
1539	}
1540	return (ENOENT);
1541}
1542