vfs_mountroot.c revision 223919 
1/*-
2 * Copyright (c) 2010 Marcel Moolenaar
3 * Copyright (c) 1999-2004 Poul-Henning Kamp
4 * Copyright (c) 1999 Michael Smith
5 * Copyright (c) 1989, 1993
6 *      The Regents of the University of California.  All rights reserved.
7 * (c) UNIX System Laboratories, Inc.
8 * All or some portions of this file are derived from material licensed
9 * to the University of California by American Telephone and Telegraph
10 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
11 * the permission of UNIX System Laboratories, Inc.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
16 * 1. Redistributions of source code must retain the above copyright
17 *    notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 *    notice, this list of conditions and the following disclaimer in the
20 *    documentation and/or other materials provided with the distribution.
21 * 4. Neither the name of the University nor the names of its contributors
22 *    may be used to endorse or promote products derived from this software
23 *    without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * SUCH DAMAGE.
36 */
37
38#include "opt_rootdevname.h"
39
40#include <sys/cdefs.h>
41__FBSDID("$FreeBSD: head/sys/kern/vfs_mountroot.c 223919 2011-07-11 05:17:46Z ae $");
42
43#include <sys/param.h>
44#include <sys/conf.h>
45#include <sys/fcntl.h>
46#include <sys/jail.h>
47#include <sys/kernel.h>
48#include <sys/libkern.h>
49#include <sys/malloc.h>
50#include <sys/mdioctl.h>
51#include <sys/mount.h>
52#include <sys/mutex.h>
53#include <sys/namei.h>
54#include <sys/priv.h>
55#include <sys/proc.h>
56#include <sys/filedesc.h>
57#include <sys/reboot.h>
58#include <sys/sbuf.h>
59#include <sys/stat.h>
60#include <sys/syscallsubr.h>
61#include <sys/sysproto.h>
62#include <sys/sx.h>
63#include <sys/sysctl.h>
64#include <sys/sysent.h>
65#include <sys/systm.h>
66#include <sys/vnode.h>
67
68#include <geom/geom.h>
69
70/*
71 * The root filesystem is detailed in the kernel environment variable
72 * vfs.root.mountfrom, which is expected to be in the general format
73 *
74 * <vfsname>:[<path>][	<vfsname>:[<path>] ...]
75 * vfsname   := the name of a VFS known to the kernel and capable
76 *              of being mounted as root
77 * path      := disk device name or other data used by the filesystem
78 *              to locate its physical store
79 *
80 * If the environment variable vfs.root.mountfrom is a space separated list,
81 * each list element is tried in turn and the root filesystem will be mounted
82 * from the first one that suceeds.
83 *
84 * The environment variable vfs.root.mountfrom.options is a comma delimited
85 * set of string mount options.  These mount options must be parseable
86 * by nmount() in the kernel.
87 */
88
89static int parse_mount(char **);
90static struct mntarg *parse_mountroot_options(struct mntarg *, const char *);
91
92/*
93 * The vnode of the system's root (/ in the filesystem, without chroot
94 * active.)
95 */
96struct vnode *rootvnode;
97
98char *rootdevnames[2] = {NULL, NULL};
99
100struct root_hold_token {
101	const char			*who;
102	LIST_ENTRY(root_hold_token)	list;
103};
104
105static LIST_HEAD(, root_hold_token)	root_holds =
106    LIST_HEAD_INITIALIZER(root_holds);
107
108enum action {
109	A_CONTINUE,
110	A_PANIC,
111	A_REBOOT,
112	A_RETRY
113};
114
115static enum action root_mount_onfail = A_CONTINUE;
116
117static int root_mount_mddev;
118static int root_mount_complete;
119
120/* By default wait up to 3 seconds for devices to appear. */
121static int root_mount_timeout = 3;
122
123struct root_hold_token *
124root_mount_hold(const char *identifier)
125{
126	struct root_hold_token *h;
127
128	if (root_mounted())
129		return (NULL);
130
131	h = malloc(sizeof *h, M_DEVBUF, M_ZERO | M_WAITOK);
132	h->who = identifier;
133	mtx_lock(&mountlist_mtx);
134	LIST_INSERT_HEAD(&root_holds, h, list);
135	mtx_unlock(&mountlist_mtx);
136	return (h);
137}
138
139void
140root_mount_rel(struct root_hold_token *h)
141{
142
143	if (h == NULL)
144		return;
145	mtx_lock(&mountlist_mtx);
146	LIST_REMOVE(h, list);
147	wakeup(&root_holds);
148	mtx_unlock(&mountlist_mtx);
149	free(h, M_DEVBUF);
150}
151
152int
153root_mounted(void)
154{
155
156	/* No mutex is acquired here because int stores are atomic. */
157	return (root_mount_complete);
158}
159
160void
161root_mount_wait(void)
162{
163
164	/*
165	 * Panic on an obvious deadlock - the function can't be called from
166	 * a thread which is doing the whole SYSINIT stuff.
167	 */
168	KASSERT(curthread->td_proc->p_pid != 0,
169	    ("root_mount_wait: cannot be called from the swapper thread"));
170	mtx_lock(&mountlist_mtx);
171	while (!root_mount_complete) {
172		msleep(&root_mount_complete, &mountlist_mtx, PZERO, "rootwait",
173		    hz);
174	}
175	mtx_unlock(&mountlist_mtx);
176}
177
178static void
179set_rootvnode(void)
180{
181	struct proc *p;
182
183	if (VFS_ROOT(TAILQ_FIRST(&mountlist), LK_EXCLUSIVE, &rootvnode))
184		panic("Cannot find root vnode");
185
186	VOP_UNLOCK(rootvnode, 0);
187
188	p = curthread->td_proc;
189	FILEDESC_XLOCK(p->p_fd);
190
191	if (p->p_fd->fd_cdir != NULL)
192		vrele(p->p_fd->fd_cdir);
193	p->p_fd->fd_cdir = rootvnode;
194	VREF(rootvnode);
195
196	if (p->p_fd->fd_rdir != NULL)
197		vrele(p->p_fd->fd_rdir);
198	p->p_fd->fd_rdir = rootvnode;
199	VREF(rootvnode);
200
201	FILEDESC_XUNLOCK(p->p_fd);
202
203	EVENTHANDLER_INVOKE(mountroot);
204}
205
206static int
207vfs_mountroot_devfs(struct thread *td, struct mount **mpp)
208{
209	struct vfsoptlist *opts;
210	struct vfsconf *vfsp;
211	struct mount *mp;
212	int error;
213
214	*mpp = NULL;
215
216	vfsp = vfs_byname("devfs");
217	KASSERT(vfsp != NULL, ("Could not find devfs by name"));
218	if (vfsp == NULL)
219		return (ENOENT);
220
221	mp = vfs_mount_alloc(NULLVP, vfsp, "/dev", td->td_ucred);
222
223	error = VFS_MOUNT(mp);
224	KASSERT(error == 0, ("VFS_MOUNT(devfs) failed %d", error));
225	if (error)
226		return (error);
227
228	opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
229	TAILQ_INIT(opts);
230	mp->mnt_opt = opts;
231
232	mtx_lock(&mountlist_mtx);
233	TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list);
234	mtx_unlock(&mountlist_mtx);
235
236	*mpp = mp;
237	set_rootvnode();
238
239	error = kern_symlink(td, "/", "dev", UIO_SYSSPACE);
240	if (error)
241		printf("kern_symlink /dev -> / returns %d\n", error);
242
243	return (error);
244}
245
246static int
247vfs_mountroot_shuffle(struct thread *td, struct mount *mpdevfs)
248{
249	struct nameidata nd;
250	struct mount *mporoot, *mpnroot;
251	struct vnode *vp, *vporoot, *vpdevfs;
252	char *fspath;
253	int error;
254
255	mpnroot = TAILQ_NEXT(mpdevfs, mnt_list);
256
257	/* Shuffle the mountlist. */
258	mtx_lock(&mountlist_mtx);
259	mporoot = TAILQ_FIRST(&mountlist);
260	TAILQ_REMOVE(&mountlist, mpdevfs, mnt_list);
261	if (mporoot != mpdevfs) {
262		TAILQ_REMOVE(&mountlist, mpnroot, mnt_list);
263		TAILQ_INSERT_HEAD(&mountlist, mpnroot, mnt_list);
264	}
265	TAILQ_INSERT_TAIL(&mountlist, mpdevfs, mnt_list);
266	mtx_unlock(&mountlist_mtx);
267
268	cache_purgevfs(mporoot);
269	if (mporoot != mpdevfs)
270		cache_purgevfs(mpdevfs);
271
272	VFS_ROOT(mporoot, LK_EXCLUSIVE, &vporoot);
273
274	VI_LOCK(vporoot);
275	vporoot->v_iflag &= ~VI_MOUNT;
276	VI_UNLOCK(vporoot);
277	vporoot->v_mountedhere = NULL;
278	mporoot->mnt_flag &= ~MNT_ROOTFS;
279	mporoot->mnt_vnodecovered = NULL;
280	vput(vporoot);
281
282	/* Set up the new rootvnode, and purge the cache */
283	mpnroot->mnt_vnodecovered = NULL;
284	set_rootvnode();
285	cache_purgevfs(rootvnode->v_mount);
286
287	if (mporoot != mpdevfs) {
288		/* Remount old root under /.mount or /mnt */
289		fspath = "/.mount";
290		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
291		    fspath, td);
292		error = namei(&nd);
293		if (error) {
294			NDFREE(&nd, NDF_ONLY_PNBUF);
295			fspath = "/mnt";
296			NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
297			    fspath, td);
298			error = namei(&nd);
299		}
300		if (!error) {
301			vp = nd.ni_vp;
302			error = (vp->v_type == VDIR) ? 0 : ENOTDIR;
303			if (!error)
304				error = vinvalbuf(vp, V_SAVE, 0, 0);
305			if (!error) {
306				cache_purge(vp);
307				mporoot->mnt_vnodecovered = vp;
308				vp->v_mountedhere = mporoot;
309				strlcpy(mporoot->mnt_stat.f_mntonname,
310				    fspath, MNAMELEN);
311				VOP_UNLOCK(vp, 0);
312			} else
313				vput(vp);
314		}
315		NDFREE(&nd, NDF_ONLY_PNBUF);
316
317		if (error && bootverbose)
318			printf("mountroot: unable to remount previous root "
319			    "under /.mount or /mnt (error %d).\n", error);
320	}
321
322	/* Remount devfs under /dev */
323	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, "/dev", td);
324	error = namei(&nd);
325	if (!error) {
326		vp = nd.ni_vp;
327		error = (vp->v_type == VDIR) ? 0 : ENOTDIR;
328		if (!error)
329			error = vinvalbuf(vp, V_SAVE, 0, 0);
330		if (!error) {
331			vpdevfs = mpdevfs->mnt_vnodecovered;
332			if (vpdevfs != NULL) {
333				cache_purge(vpdevfs);
334				vpdevfs->v_mountedhere = NULL;
335				vrele(vpdevfs);
336			}
337			mpdevfs->mnt_vnodecovered = vp;
338			vp->v_mountedhere = mpdevfs;
339			VOP_UNLOCK(vp, 0);
340		} else
341			vput(vp);
342	}
343	if (error && bootverbose)
344		printf("mountroot: unable to remount devfs under /dev "
345		    "(error %d).\n", error);
346	NDFREE(&nd, NDF_ONLY_PNBUF);
347
348	if (mporoot == mpdevfs) {
349		vfs_unbusy(mpdevfs);
350		/* Unlink the no longer needed /dev/dev -> / symlink */
351		error = kern_unlink(td, "/dev/dev", UIO_SYSSPACE);
352		if (error && bootverbose)
353			printf("mountroot: unable to unlink /dev/dev "
354			    "(error %d)\n", error);
355	}
356
357	return (0);
358}
359
360/*
361 * Configuration parser.
362 */
363
364/* Parser character classes. */
365#define	CC_WHITESPACE		-1
366#define	CC_NONWHITESPACE	-2
367
368/* Parse errors. */
369#define	PE_EOF			-1
370#define	PE_EOL			-2
371
372static __inline int
373parse_peek(char **conf)
374{
375
376	return (**conf);
377}
378
379static __inline void
380parse_poke(char **conf, int c)
381{
382
383	**conf = c;
384}
385
386static __inline void
387parse_advance(char **conf)
388{
389
390	(*conf)++;
391}
392
393static __inline int
394parse_isspace(int c)
395{
396
397	return ((c == ' ' || c == '\t' || c == '\n') ? 1 : 0);
398}
399
400static int
401parse_skipto(char **conf, int mc)
402{
403	int c, match;
404
405	while (1) {
406		c = parse_peek(conf);
407		if (c == 0)
408			return (PE_EOF);
409		switch (mc) {
410		case CC_WHITESPACE:
411			match = (c == ' ' || c == '\t' || c == '\n') ? 1 : 0;
412			break;
413		case CC_NONWHITESPACE:
414			if (c == '\n')
415				return (PE_EOL);
416			match = (c != ' ' && c != '\t') ? 1 : 0;
417			break;
418		default:
419			match = (c == mc) ? 1 : 0;
420			break;
421		}
422		if (match)
423			break;
424		parse_advance(conf);
425	}
426	return (0);
427}
428
429static int
430parse_token(char **conf, char **tok)
431{
432	char *p;
433	size_t len;
434	int error;
435
436	*tok = NULL;
437	error = parse_skipto(conf, CC_NONWHITESPACE);
438	if (error)
439		return (error);
440	p = *conf;
441	error = parse_skipto(conf, CC_WHITESPACE);
442	len = *conf - p;
443	*tok = malloc(len + 1, M_TEMP, M_WAITOK | M_ZERO);
444	bcopy(p, *tok, len);
445	return (0);
446}
447
448static void
449parse_dir_ask_printenv(const char *var)
450{
451	char *val;
452
453	val = getenv(var);
454	if (val != NULL) {
455		printf("  %s=%s\n", var, val);
456		freeenv(val);
457	}
458}
459
460static int
461parse_dir_ask(char **conf)
462{
463	char name[80];
464	char *mnt;
465	int error;
466
467	printf("\nLoader variables:\n");
468	parse_dir_ask_printenv("vfs.root.mountfrom");
469	parse_dir_ask_printenv("vfs.root.mountfrom.options");
470
471	printf("\nManual root filesystem specification:\n");
472	printf("  <fstype>:<device> [options]\n");
473	printf("      Mount <device> using filesystem <fstype>\n");
474	printf("      and with the specified (optional) option list.\n");
475	printf("\n");
476	printf("    eg. ufs:/dev/da0s1a\n");
477	printf("        zfs:tank\n");
478	printf("        cd9660:/dev/acd0 ro\n");
479	printf("          (which is equivalent to: ");
480	printf("mount -t cd9660 -o ro /dev/acd0 /)\n");
481	printf("\n");
482	printf("  ?               List valid disk boot devices\n");
483	printf("  .               Yield 1 second (for background tasks)\n");
484	printf("  <empty line>    Abort manual input\n");
485
486 again:
487	printf("\nmountroot> ");
488	gets(name, sizeof(name), GETS_ECHO);
489	if (name[0] == '\0')
490		return (0);
491	if (name[0] == '?') {
492		printf("\nList of GEOM managed disk devices:\n  ");
493		g_dev_print();
494		goto again;
495	}
496	if (name[0] == '.') {
497		pause("rmask", hz);
498		goto again;
499	}
500	mnt = name;
501	error = parse_mount(&mnt);
502	if (error == -1) {
503		printf("Invalid specification.\n");
504		goto again;
505	}
506	return (error);
507}
508
509static int
510parse_dir_md(char **conf)
511{
512	struct stat sb;
513	struct thread *td;
514	struct md_ioctl *mdio;
515	char *path, *tok;
516	int error, fd, len;
517
518	td = curthread;
519
520	error = parse_token(conf, &tok);
521	if (error)
522		return (error);
523
524	len = strlen(tok);
525	mdio = malloc(sizeof(*mdio) + len + 1, M_TEMP, M_WAITOK | M_ZERO);
526	path = (void *)(mdio + 1);
527	bcopy(tok, path, len);
528	free(tok, M_TEMP);
529
530	/* Get file status. */
531	error = kern_stat(td, path, UIO_SYSSPACE, &sb);
532	if (error)
533		goto out;
534
535	/* Open /dev/mdctl so that we can attach/detach. */
536	error = kern_open(td, "/dev/" MDCTL_NAME, UIO_SYSSPACE, O_RDWR, 0);
537	if (error)
538		goto out;
539
540	fd = td->td_retval[0];
541	mdio->md_version = MDIOVERSION;
542	mdio->md_type = MD_VNODE;
543
544	if (root_mount_mddev != -1) {
545		mdio->md_unit = root_mount_mddev;
546		DROP_GIANT();
547		error = kern_ioctl(td, fd, MDIOCDETACH, (void *)mdio);
548		PICKUP_GIANT();
549		/* Ignore errors. We don't care. */
550		root_mount_mddev = -1;
551	}
552
553	mdio->md_file = (void *)(mdio + 1);
554	mdio->md_options = MD_AUTOUNIT | MD_READONLY;
555	mdio->md_mediasize = sb.st_size;
556	mdio->md_unit = 0;
557	DROP_GIANT();
558	error = kern_ioctl(td, fd, MDIOCATTACH, (void *)mdio);
559	PICKUP_GIANT();
560	if (error)
561		goto out;
562
563	if (mdio->md_unit > 9) {
564		printf("rootmount: too many md units\n");
565		mdio->md_file = NULL;
566		mdio->md_options = 0;
567		mdio->md_mediasize = 0;
568		DROP_GIANT();
569		error = kern_ioctl(td, fd, MDIOCDETACH, (void *)mdio);
570		PICKUP_GIANT();
571		/* Ignore errors. We don't care. */
572		error = ERANGE;
573		goto out;
574	}
575
576	root_mount_mddev = mdio->md_unit;
577	printf(MD_NAME "%u attached to %s\n", root_mount_mddev, mdio->md_file);
578
579	error = kern_close(td, fd);
580
581 out:
582	free(mdio, M_TEMP);
583	return (error);
584}
585
586static int
587parse_dir_onfail(char **conf)
588{
589	char *action;
590	int error;
591
592	error = parse_token(conf, &action);
593	if (error)
594		return (error);
595
596	if (!strcmp(action, "continue"))
597		root_mount_onfail = A_CONTINUE;
598	else if (!strcmp(action, "panic"))
599		root_mount_onfail = A_PANIC;
600	else if (!strcmp(action, "reboot"))
601		root_mount_onfail = A_REBOOT;
602	else if (!strcmp(action, "retry"))
603		root_mount_onfail = A_RETRY;
604	else {
605		printf("rootmount: %s: unknown action\n", action);
606		error = EINVAL;
607	}
608
609	free(action, M_TEMP);
610	return (0);
611}
612
613static int
614parse_dir_timeout(char **conf)
615{
616	char *tok, *endtok;
617	long secs;
618	int error;
619
620	error = parse_token(conf, &tok);
621	if (error)
622		return (error);
623
624	secs = strtol(tok, &endtok, 0);
625	error = (secs < 0 || *endtok != '\0') ? EINVAL : 0;
626	if (!error)
627		root_mount_timeout = secs;
628	free(tok, M_TEMP);
629	return (error);
630}
631
632static int
633parse_directive(char **conf)
634{
635	char *dir;
636	int error;
637
638	error = parse_token(conf, &dir);
639	if (error)
640		return (error);
641
642	if (strcmp(dir, ".ask") == 0)
643		error = parse_dir_ask(conf);
644	else if (strcmp(dir, ".md") == 0)
645		error = parse_dir_md(conf);
646	else if (strcmp(dir, ".onfail") == 0)
647		error = parse_dir_onfail(conf);
648	else if (strcmp(dir, ".timeout") == 0)
649		error = parse_dir_timeout(conf);
650	else {
651		printf("mountroot: invalid directive `%s'\n", dir);
652		/* Ignore the rest of the line. */
653		(void)parse_skipto(conf, '\n');
654		error = EINVAL;
655	}
656	free(dir, M_TEMP);
657	return (error);
658}
659
660static int
661parse_mount_dev_present(const char *dev)
662{
663	struct nameidata nd;
664	int error;
665
666	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, dev, curthread);
667	error = namei(&nd);
668	if (!error)
669		vput(nd.ni_vp);
670	NDFREE(&nd, NDF_ONLY_PNBUF);
671	return (error != 0) ? 0 : 1;
672}
673
674static int
675parse_mount(char **conf)
676{
677	char errmsg[255];
678	struct mntarg *ma;
679	char *dev, *fs, *opts, *tok;
680	int delay, error, timeout;
681
682	error = parse_token(conf, &tok);
683	if (error)
684		return (error);
685	fs = tok;
686	error = parse_skipto(&tok, ':');
687	if (error) {
688		free(fs, M_TEMP);
689		return (error);
690	}
691	parse_poke(&tok, '\0');
692	parse_advance(&tok);
693	dev = tok;
694
695	if (root_mount_mddev != -1) {
696		/* Handle substitution for the md unit number. */
697		tok = strstr(dev, "md#");
698		if (tok != NULL)
699			tok[2] = '0' + root_mount_mddev;
700	}
701
702	/* Parse options. */
703	error = parse_token(conf, &tok);
704	opts = (error == 0) ? tok : NULL;
705
706	printf("Trying to mount root from %s:%s [%s]...\n", fs, dev,
707	    (opts != NULL) ? opts : "");
708
709	bzero(errmsg, sizeof(errmsg));
710
711	if (vfs_byname(fs) == NULL) {
712		strlcpy(errmsg, "unknown file system", sizeof(errmsg));
713		error = ENOENT;
714		goto out;
715	}
716
717	if (strcmp(fs, "zfs") != 0 && dev[0] != '\0' &&
718	    !parse_mount_dev_present(dev)) {
719		printf("mountroot: waiting for device %s ...\n", dev);
720		delay = hz / 10;
721		timeout = root_mount_timeout * hz;
722		do {
723			pause("rmdev", delay);
724			timeout -= delay;
725		} while (timeout > 0 && !parse_mount_dev_present(dev));
726		if (timeout <= 0) {
727			error = ENODEV;
728			goto out;
729		}
730	}
731
732	ma = NULL;
733	ma = mount_arg(ma, "fstype", fs, -1);
734	ma = mount_arg(ma, "fspath", "/", -1);
735	ma = mount_arg(ma, "from", dev, -1);
736	ma = mount_arg(ma, "errmsg", errmsg, sizeof(errmsg));
737	ma = mount_arg(ma, "ro", NULL, 0);
738	ma = parse_mountroot_options(ma, opts);
739	error = kernel_mount(ma, MNT_ROOTFS);
740
741 out:
742	if (error) {
743		printf("Mounting from %s:%s failed with error %d",
744		    fs, dev, error);
745		if (errmsg[0] != '\0')
746			printf(": %s", errmsg);
747		printf(".\n");
748	}
749	free(fs, M_TEMP);
750	if (opts != NULL)
751		free(opts, M_TEMP);
752	/* kernel_mount can return -1 on error. */
753	return ((error < 0) ? EDOOFUS : error);
754}
755
756static int
757vfs_mountroot_parse(struct sbuf *sb, struct mount *mpdevfs)
758{
759	struct mount *mp;
760	char *conf;
761	int error;
762
763	root_mount_mddev = -1;
764
765retry:
766	conf = sbuf_data(sb);
767	mp = TAILQ_NEXT(mpdevfs, mnt_list);
768	error = (mp == NULL) ? 0 : EDOOFUS;
769	root_mount_onfail = A_CONTINUE;
770	while (mp == NULL) {
771		error = parse_skipto(&conf, CC_NONWHITESPACE);
772		if (error == PE_EOL) {
773			parse_advance(&conf);
774			continue;
775		}
776		if (error < 0)
777			break;
778		switch (parse_peek(&conf)) {
779		case '#':
780			error = parse_skipto(&conf, '\n');
781			break;
782		case '.':
783			error = parse_directive(&conf);
784			break;
785		default:
786			error = parse_mount(&conf);
787			break;
788		}
789		if (error < 0)
790			break;
791		/* Ignore any trailing garbage on the line. */
792		if (parse_peek(&conf) != '\n') {
793			printf("mountroot: advancing to next directive...\n");
794			(void)parse_skipto(&conf, '\n');
795		}
796		mp = TAILQ_NEXT(mpdevfs, mnt_list);
797	}
798	if (mp != NULL)
799		return (0);
800
801	/*
802	 * We failed to mount (a new) root.
803	 */
804	switch (root_mount_onfail) {
805	case A_CONTINUE:
806		break;
807	case A_PANIC:
808		panic("mountroot: unable to (re-)mount root.");
809		/* NOTREACHED */
810	case A_RETRY:
811		goto retry;
812	case A_REBOOT:
813		kern_reboot(RB_NOSYNC);
814		/* NOTREACHED */
815	}
816
817	return (error);
818}
819
820static void
821vfs_mountroot_conf0(struct sbuf *sb)
822{
823	char *s, *tok, *mnt, *opt;
824	int error;
825
826	sbuf_printf(sb, ".onfail panic\n");
827	sbuf_printf(sb, ".timeout %d\n", root_mount_timeout);
828	if (boothowto & RB_ASKNAME)
829		sbuf_printf(sb, ".ask\n");
830#ifdef ROOTDEVNAME
831	if (boothowto & RB_DFLTROOT)
832		sbuf_printf(sb, "%s\n", ROOTDEVNAME);
833#endif
834	if (boothowto & RB_CDROM) {
835		sbuf_printf(sb, "cd9660:/dev/cd0 ro\n");
836		sbuf_printf(sb, ".timeout 0\n");
837		sbuf_printf(sb, "cd9660:/dev/acd0 ro\n");
838		sbuf_printf(sb, ".timeout %d\n", root_mount_timeout);
839	}
840	s = getenv("vfs.root.mountfrom");
841	if (s != NULL) {
842		opt = getenv("vfs.root.mountfrom.options");
843		tok = s;
844		error = parse_token(&tok, &mnt);
845		while (!error) {
846			sbuf_printf(sb, "%s %s\n", mnt,
847			    (opt != NULL) ? opt : "");
848			free(mnt, M_TEMP);
849			error = parse_token(&tok, &mnt);
850		}
851		if (opt != NULL)
852			freeenv(opt);
853		freeenv(s);
854	}
855	if (rootdevnames[0] != NULL)
856		sbuf_printf(sb, "%s\n", rootdevnames[0]);
857	if (rootdevnames[1] != NULL)
858		sbuf_printf(sb, "%s\n", rootdevnames[1]);
859#ifdef ROOTDEVNAME
860	if (!(boothowto & RB_DFLTROOT))
861		sbuf_printf(sb, "%s\n", ROOTDEVNAME);
862#endif
863	if (!(boothowto & RB_ASKNAME))
864		sbuf_printf(sb, ".ask\n");
865}
866
867static int
868vfs_mountroot_readconf(struct thread *td, struct sbuf *sb)
869{
870	static char buf[128];
871	struct nameidata nd;
872	off_t ofs;
873	int error, flags;
874	int len, resid;
875	int vfslocked;
876
877	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE, UIO_SYSSPACE,
878	    "/.mount.conf", td);
879	flags = FREAD;
880	error = vn_open(&nd, &flags, 0, NULL);
881	if (error)
882		return (error);
883
884	vfslocked = NDHASGIANT(&nd);
885	NDFREE(&nd, NDF_ONLY_PNBUF);
886	ofs = 0;
887	len = sizeof(buf) - 1;
888	while (1) {
889		error = vn_rdwr(UIO_READ, nd.ni_vp, buf, len, ofs,
890		    UIO_SYSSPACE, IO_NODELOCKED, td->td_ucred,
891		    NOCRED, &resid, td);
892		if (error)
893			break;
894		if (resid == len)
895			break;
896		buf[len - resid] = 0;
897		sbuf_printf(sb, "%s", buf);
898		ofs += len - resid;
899	}
900
901	VOP_UNLOCK(nd.ni_vp, 0);
902	vn_close(nd.ni_vp, FREAD, td->td_ucred, td);
903	VFS_UNLOCK_GIANT(vfslocked);
904	return (error);
905}
906
907static void
908vfs_mountroot_wait(void)
909{
910	struct root_hold_token *h;
911	struct timeval lastfail;
912	int curfail;
913
914	curfail = 0;
915	while (1) {
916		DROP_GIANT();
917		g_waitidle();
918		PICKUP_GIANT();
919		mtx_lock(&mountlist_mtx);
920		if (LIST_EMPTY(&root_holds)) {
921			mtx_unlock(&mountlist_mtx);
922			break;
923		}
924		if (ppsratecheck(&lastfail, &curfail, 1)) {
925			printf("Root mount waiting for:");
926			LIST_FOREACH(h, &root_holds, list)
927				printf(" %s", h->who);
928			printf("\n");
929		}
930		msleep(&root_holds, &mountlist_mtx, PZERO | PDROP, "roothold",
931		    hz);
932	}
933}
934
935void
936vfs_mountroot(void)
937{
938	struct mount *mp;
939	struct sbuf *sb;
940	struct thread *td;
941	time_t timebase;
942	int error;
943
944	td = curthread;
945
946	vfs_mountroot_wait();
947
948	sb = sbuf_new_auto();
949	vfs_mountroot_conf0(sb);
950	sbuf_finish(sb);
951
952	error = vfs_mountroot_devfs(td, &mp);
953	while (!error) {
954		error = vfs_mountroot_parse(sb, mp);
955		if (!error) {
956			error = vfs_mountroot_shuffle(td, mp);
957			if (!error) {
958				sbuf_clear(sb);
959				error = vfs_mountroot_readconf(td, sb);
960				sbuf_finish(sb);
961			}
962		}
963	}
964
965	sbuf_delete(sb);
966
967	/*
968	 * Iterate over all currently mounted file systems and use
969	 * the time stamp found to check and/or initialize the RTC.
970	 * Call inittodr() only once and pass it the largest of the
971	 * timestamps we encounter.
972	 */
973	timebase = 0;
974	mtx_lock(&mountlist_mtx);
975	mp = TAILQ_FIRST(&mountlist);
976	while (mp != NULL) {
977		if (mp->mnt_time > timebase)
978			timebase = mp->mnt_time;
979		mp = TAILQ_NEXT(mp, mnt_list);
980	}
981	mtx_unlock(&mountlist_mtx);
982	inittodr(timebase);
983
984	/* Keep prison0's root in sync with the global rootvnode. */
985	mtx_lock(&prison0.pr_mtx);
986	prison0.pr_root = rootvnode;
987	vref(prison0.pr_root);
988	mtx_unlock(&prison0.pr_mtx);
989
990	mtx_lock(&mountlist_mtx);
991	atomic_store_rel_int(&root_mount_complete, 1);
992	wakeup(&root_mount_complete);
993	mtx_unlock(&mountlist_mtx);
994}
995
996static struct mntarg *
997parse_mountroot_options(struct mntarg *ma, const char *options)
998{
999	char *p;
1000	char *name, *name_arg;
1001	char *val, *val_arg;
1002	char *opts;
1003
1004	if (options == NULL || options[0] == '\0')
1005		return (ma);
1006
1007	p = opts = strdup(options, M_MOUNT);
1008	if (opts == NULL) {
1009		return (ma);
1010	}
1011
1012	while((name = strsep(&p, ",")) != NULL) {
1013		if (name[0] == '\0')
1014			break;
1015
1016		val = strchr(name, '=');
1017		if (val != NULL) {
1018			*val = '\0';
1019			++val;
1020		}
1021		if( strcmp(name, "rw") == 0 ||
1022		    strcmp(name, "noro") == 0) {
1023			/*
1024			 * The first time we mount the root file system,
1025			 * we need to mount 'ro', so We need to ignore
1026			 * 'rw' and 'noro' mount options.
1027			 */
1028			continue;
1029		}
1030		name_arg = strdup(name, M_MOUNT);
1031		val_arg = NULL;
1032		if (val != NULL)
1033			val_arg = strdup(val, M_MOUNT);
1034
1035		ma = mount_arg(ma, name_arg, val_arg,
1036		    (val_arg != NULL ? -1 : 0));
1037	}
1038	free(opts, M_MOUNT);
1039	return (ma);
1040}
1041