vfs_mountroot.c revision 217163 
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 217163 2011-01-08 19:50:13Z nwhitehorn $");
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/stat.h>
59#include <sys/syscallsubr.h>
60#include <sys/sysproto.h>
61#include <sys/sx.h>
62#include <sys/sysctl.h>
63#include <sys/sysent.h>
64#include <sys/systm.h>
65#include <sys/vnode.h>
66
67#include <geom/geom.h>
68
69/*
70 * The root filesystem is detailed in the kernel environment variable
71 * vfs.root.mountfrom, which is expected to be in the general format
72 *
73 * <vfsname>:[<path>][	<vfsname>:[<path>] ...]
74 * vfsname   := the name of a VFS known to the kernel and capable
75 *              of being mounted as root
76 * path      := disk device name or other data used by the filesystem
77 *              to locate its physical store
78 *
79 * If the environment variable vfs.root.mountfrom is a space separated list,
80 * each list element is tried in turn and the root filesystem will be mounted
81 * from the first one that suceeds.
82 *
83 * The environment variable vfs.root.mountfrom.options is a comma delimited
84 * set of string mount options.  These mount options must be parseable
85 * by nmount() in the kernel.
86 */
87
88static int parse_mount(char **);
89static struct mntarg *parse_mountroot_options(struct mntarg *, const char *);
90
91/*
92 * The vnode of the system's root (/ in the filesystem, without chroot
93 * active.)
94 */
95struct vnode *rootvnode;
96
97char *rootdevnames[2] = {NULL, NULL};
98
99struct root_hold_token {
100	const char			*who;
101	LIST_ENTRY(root_hold_token)	list;
102};
103
104static LIST_HEAD(, root_hold_token)	root_holds =
105    LIST_HEAD_INITIALIZER(root_holds);
106
107enum action {
108	A_CONTINUE,
109	A_PANIC,
110	A_REBOOT,
111	A_RETRY
112};
113
114static enum action root_mount_onfail = A_CONTINUE;
115
116static int root_mount_mddev;
117static int root_mount_complete;
118
119/* By default wait up to 3 seconds for devices to appear. */
120static int root_mount_timeout = 3;
121
122struct root_hold_token *
123root_mount_hold(const char *identifier)
124{
125	struct root_hold_token *h;
126
127	if (root_mounted())
128		return (NULL);
129
130	h = malloc(sizeof *h, M_DEVBUF, M_ZERO | M_WAITOK);
131	h->who = identifier;
132	mtx_lock(&mountlist_mtx);
133	LIST_INSERT_HEAD(&root_holds, h, list);
134	mtx_unlock(&mountlist_mtx);
135	return (h);
136}
137
138void
139root_mount_rel(struct root_hold_token *h)
140{
141
142	if (h == NULL)
143		return;
144	mtx_lock(&mountlist_mtx);
145	LIST_REMOVE(h, list);
146	wakeup(&root_holds);
147	mtx_unlock(&mountlist_mtx);
148	free(h, M_DEVBUF);
149}
150
151int
152root_mounted(void)
153{
154
155	/* No mutex is acquired here because int stores are atomic. */
156	return (root_mount_complete);
157}
158
159void
160root_mount_wait(void)
161{
162
163	/*
164	 * Panic on an obvious deadlock - the function can't be called from
165	 * a thread which is doing the whole SYSINIT stuff.
166	 */
167	KASSERT(curthread->td_proc->p_pid != 0,
168	    ("root_mount_wait: cannot be called from the swapper thread"));
169	mtx_lock(&mountlist_mtx);
170	while (!root_mount_complete) {
171		msleep(&root_mount_complete, &mountlist_mtx, PZERO, "rootwait",
172		    hz);
173	}
174	mtx_unlock(&mountlist_mtx);
175}
176
177static void
178set_rootvnode(void)
179{
180	struct proc *p;
181
182	if (VFS_ROOT(TAILQ_FIRST(&mountlist), LK_EXCLUSIVE, &rootvnode))
183		panic("Cannot find root vnode");
184
185	VOP_UNLOCK(rootvnode, 0);
186
187	p = curthread->td_proc;
188	FILEDESC_XLOCK(p->p_fd);
189
190	if (p->p_fd->fd_cdir != NULL)
191		vrele(p->p_fd->fd_cdir);
192	p->p_fd->fd_cdir = rootvnode;
193	VREF(rootvnode);
194
195	if (p->p_fd->fd_rdir != NULL)
196		vrele(p->p_fd->fd_rdir);
197	p->p_fd->fd_rdir = rootvnode;
198	VREF(rootvnode);
199
200	FILEDESC_XUNLOCK(p->p_fd);
201
202	EVENTHANDLER_INVOKE(mountroot);
203}
204
205static int
206vfs_mountroot_devfs(struct thread *td, struct mount **mpp)
207{
208	struct vfsoptlist *opts;
209	struct vfsconf *vfsp;
210	struct mount *mp;
211	int error;
212
213	*mpp = NULL;
214
215	vfsp = vfs_byname("devfs");
216	KASSERT(vfsp != NULL, ("Could not find devfs by name"));
217	if (vfsp == NULL)
218		return (ENOENT);
219
220	mp = vfs_mount_alloc(NULLVP, vfsp, "/dev", td->td_ucred);
221
222	error = VFS_MOUNT(mp);
223	KASSERT(error == 0, ("VFS_MOUNT(devfs) failed %d", error));
224	if (error)
225		return (error);
226
227	opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
228	TAILQ_INIT(opts);
229	mp->mnt_opt = opts;
230
231	mtx_lock(&mountlist_mtx);
232	TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list);
233	mtx_unlock(&mountlist_mtx);
234
235	*mpp = mp;
236	set_rootvnode();
237
238	error = kern_symlink(td, "/", "dev", UIO_SYSSPACE);
239	if (error)
240		printf("kern_symlink /dev -> / returns %d\n", error);
241
242	return (error);
243}
244
245static int
246vfs_mountroot_shuffle(struct thread *td, struct mount *mpdevfs)
247{
248	struct nameidata nd;
249	struct mount *mporoot, *mpnroot;
250	struct vnode *vp, *vporoot, *vpdevfs;
251	char *fspath;
252	int error;
253
254	mpnroot = TAILQ_NEXT(mpdevfs, mnt_list);
255
256	/* Shuffle the mountlist. */
257	mtx_lock(&mountlist_mtx);
258	mporoot = TAILQ_FIRST(&mountlist);
259	TAILQ_REMOVE(&mountlist, mpdevfs, mnt_list);
260	if (mporoot != mpdevfs) {
261		TAILQ_REMOVE(&mountlist, mpnroot, mnt_list);
262		TAILQ_INSERT_HEAD(&mountlist, mpnroot, mnt_list);
263	}
264	TAILQ_INSERT_TAIL(&mountlist, mpdevfs, mnt_list);
265	mtx_unlock(&mountlist_mtx);
266
267	cache_purgevfs(mporoot);
268	if (mporoot != mpdevfs)
269		cache_purgevfs(mpdevfs);
270
271	VFS_ROOT(mporoot, LK_EXCLUSIVE, &vporoot);
272
273	VI_LOCK(vporoot);
274	vporoot->v_iflag &= ~VI_MOUNT;
275	VI_UNLOCK(vporoot);
276	vporoot->v_mountedhere = NULL;
277	mporoot->mnt_flag &= ~MNT_ROOTFS;
278	mporoot->mnt_vnodecovered = NULL;
279	vput(vporoot);
280
281	/* Set up the new rootvnode, and purge the cache */
282	mpnroot->mnt_vnodecovered = NULL;
283	set_rootvnode();
284	cache_purgevfs(rootvnode->v_mount);
285
286	if (mporoot != mpdevfs) {
287		/* Remount old root under /.mount or /mnt */
288		fspath = "/.mount";
289		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
290		    fspath, td);
291		error = namei(&nd);
292		if (error) {
293			NDFREE(&nd, NDF_ONLY_PNBUF);
294			fspath = "/mnt";
295			NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
296			    fspath, td);
297			error = namei(&nd);
298		}
299		if (!error) {
300			vp = nd.ni_vp;
301			error = (vp->v_type == VDIR) ? 0 : ENOTDIR;
302			if (!error)
303				error = vinvalbuf(vp, V_SAVE, 0, 0);
304			if (!error) {
305				cache_purge(vp);
306				mporoot->mnt_vnodecovered = vp;
307				vp->v_mountedhere = mporoot;
308				strlcpy(mporoot->mnt_stat.f_mntonname,
309				    fspath, MNAMELEN);
310				VOP_UNLOCK(vp, 0);
311			} else
312				vput(vp);
313		}
314		NDFREE(&nd, NDF_ONLY_PNBUF);
315
316		if (error && bootverbose)
317			printf("mountroot: unable to remount previous root "
318			    "under /.mount or /mnt (error %d).\n", error);
319	}
320
321	/* Remount devfs under /dev */
322	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, "/dev", td);
323	error = namei(&nd);
324	if (!error) {
325		vp = nd.ni_vp;
326		error = (vp->v_type == VDIR) ? 0 : ENOTDIR;
327		if (!error)
328			error = vinvalbuf(vp, V_SAVE, 0, 0);
329		if (!error) {
330			vpdevfs = mpdevfs->mnt_vnodecovered;
331			if (vpdevfs != NULL) {
332				cache_purge(vpdevfs);
333				vpdevfs->v_mountedhere = NULL;
334				vrele(vpdevfs);
335			}
336			mpdevfs->mnt_vnodecovered = vp;
337			vp->v_mountedhere = mpdevfs;
338			VOP_UNLOCK(vp, 0);
339		} else
340			vput(vp);
341	}
342	if (error && bootverbose)
343		printf("mountroot: unable to remount devfs under /dev "
344		    "(error %d).\n", error);
345	NDFREE(&nd, NDF_ONLY_PNBUF);
346
347	if (mporoot == mpdevfs) {
348		vfs_unbusy(mpdevfs);
349		/* Unlink the no longer needed /dev/dev -> / symlink */
350		error = kern_unlink(td, "/dev/dev", UIO_SYSSPACE);
351		if (error && bootverbose)
352			printf("mountroot: unable to unlink /dev/dev "
353			    "(error %d)\n", error);
354	}
355
356	return (0);
357}
358
359/*
360 * Configuration parser.
361 */
362
363/* Parser character classes. */
364#define	CC_WHITESPACE		-1
365#define	CC_NONWHITESPACE	-2
366
367/* Parse errors. */
368#define	PE_EOF			-1
369#define	PE_EOL			-2
370
371static __inline int
372parse_peek(char **conf)
373{
374
375	return (**conf);
376}
377
378static __inline void
379parse_poke(char **conf, int c)
380{
381
382	**conf = c;
383}
384
385static __inline void
386parse_advance(char **conf)
387{
388
389	(*conf)++;
390}
391
392static __inline int
393parse_isspace(int c)
394{
395
396	return ((c == ' ' || c == '\t' || c == '\n') ? 1 : 0);
397}
398
399static int
400parse_skipto(char **conf, int mc)
401{
402	int c, match;
403
404	while (1) {
405		c = parse_peek(conf);
406		if (c == 0)
407			return (PE_EOF);
408		switch (mc) {
409		case CC_WHITESPACE:
410			match = (c == ' ' || c == '\t' || c == '\n') ? 1 : 0;
411			break;
412		case CC_NONWHITESPACE:
413			if (c == '\n')
414				return (PE_EOL);
415			match = (c != ' ' && c != '\t') ? 1 : 0;
416			break;
417		default:
418			match = (c == mc) ? 1 : 0;
419			break;
420		}
421		if (match)
422			break;
423		parse_advance(conf);
424	}
425	return (0);
426}
427
428static int
429parse_token(char **conf, char **tok)
430{
431	char *p;
432	size_t len;
433	int error;
434
435	*tok = NULL;
436	error = parse_skipto(conf, CC_NONWHITESPACE);
437	if (error)
438		return (error);
439	p = *conf;
440	error = parse_skipto(conf, CC_WHITESPACE);
441	len = *conf - p;
442	*tok = malloc(len + 1, M_TEMP, M_WAITOK | M_ZERO);
443	bcopy(p, *tok, len);
444	return (0);
445}
446
447static void
448parse_dir_ask_printenv(const char *var)
449{
450	char *val;
451
452	val = getenv(var);
453	if (val != NULL) {
454		printf("  %s=%s\n", var, val);
455		freeenv(val);
456	}
457}
458
459static int
460parse_dir_ask(char **conf)
461{
462	char name[80];
463	char *mnt;
464	int error;
465
466	printf("\nLoader variables:\n");
467	parse_dir_ask_printenv("vfs.root.mountfrom");
468	parse_dir_ask_printenv("vfs.root.mountfrom.options");
469
470	printf("\nManual root filesystem specification:\n");
471	printf("  <fstype>:<device> [options]\n");
472	printf("      Mount <device> using filesystem <fstype>\n");
473	printf("      and with the specified (optional) option list.\n");
474	printf("\n");
475	printf("    eg. ufs:/dev/da0s1a\n");
476	printf("        zfs:tank\n");
477	printf("        cd9660:/dev/acd0 ro\n");
478	printf("          (which is equivalent to: ");
479	printf("mount -t cd9660 -o ro /dev/acd0 /)\n");
480	printf("\n");
481	printf("  ?               List valid disk boot devices\n");
482	printf("  .               Yield 1 second (for background tasks)\n");
483	printf("  <empty line>    Abort manual input\n");
484
485 again:
486	printf("\nmountroot> ");
487	gets(name, sizeof(name), GETS_ECHO);
488	if (name[0] == '\0')
489		return (0);
490	if (name[0] == '?') {
491		printf("\nList of GEOM managed disk devices:\n  ");
492		g_dev_print();
493		goto again;
494	}
495	if (name[0] == '.') {
496		pause("rmask", hz);
497		goto again;
498	}
499	mnt = name;
500	error = parse_mount(&mnt);
501	if (error == -1) {
502		printf("Invalid specification.\n");
503		goto again;
504	}
505	return (error);
506}
507
508static int
509parse_dir_md(char **conf)
510{
511	struct stat sb;
512	struct thread *td;
513	struct md_ioctl *mdio;
514	char *path, *tok;
515	int error, fd, len;
516
517	td = curthread;
518
519	error = parse_token(conf, &tok);
520	if (error)
521		return (error);
522
523	len = strlen(tok);
524	mdio = malloc(sizeof(*mdio) + len + 1, M_TEMP, M_WAITOK | M_ZERO);
525	path = (void *)(mdio + 1);
526	bcopy(tok, path, len);
527	free(tok, M_TEMP);
528
529	/* Get file status. */
530	error = kern_stat(td, path, UIO_SYSSPACE, &sb);
531	if (error)
532		goto out;
533
534	/* Open /dev/mdctl so that we can attach/detach. */
535	error = kern_open(td, "/dev/" MDCTL_NAME, UIO_SYSSPACE, O_RDWR, 0);
536	if (error)
537		goto out;
538
539	fd = td->td_retval[0];
540	mdio->md_version = MDIOVERSION;
541	mdio->md_type = MD_VNODE;
542
543	if (root_mount_mddev != -1) {
544		mdio->md_unit = root_mount_mddev;
545		DROP_GIANT();
546		error = kern_ioctl(td, fd, MDIOCDETACH, (void *)mdio);
547		PICKUP_GIANT();
548		/* Ignore errors. We don't care. */
549		root_mount_mddev = -1;
550	}
551
552	mdio->md_file = (void *)(mdio + 1);
553	mdio->md_options = MD_AUTOUNIT | MD_READONLY;
554	mdio->md_mediasize = sb.st_size;
555	mdio->md_unit = 0;
556	DROP_GIANT();
557	error = kern_ioctl(td, fd, MDIOCATTACH, (void *)mdio);
558	PICKUP_GIANT();
559	if (error)
560		goto out;
561
562	if (mdio->md_unit > 9) {
563		printf("rootmount: too many md units\n");
564		mdio->md_file = NULL;
565		mdio->md_options = 0;
566		mdio->md_mediasize = 0;
567		DROP_GIANT();
568		error = kern_ioctl(td, fd, MDIOCDETACH, (void *)mdio);
569		PICKUP_GIANT();
570		/* Ignore errors. We don't care. */
571		error = ERANGE;
572		goto out;
573	}
574
575	root_mount_mddev = mdio->md_unit;
576	printf(MD_NAME "%u attached to %s\n", root_mount_mddev, mdio->md_file);
577
578	error = kern_close(td, fd);
579
580 out:
581	free(mdio, M_TEMP);
582	return (error);
583}
584
585static int
586parse_dir_onfail(char **conf)
587{
588	char *action;
589	int error;
590
591	error = parse_token(conf, &action);
592	if (error)
593		return (error);
594
595	if (!strcmp(action, "continue"))
596		root_mount_onfail = A_CONTINUE;
597	else if (!strcmp(action, "panic"))
598		root_mount_onfail = A_PANIC;
599	else if (!strcmp(action, "reboot"))
600		root_mount_onfail = A_REBOOT;
601	else if (!strcmp(action, "retry"))
602		root_mount_onfail = A_RETRY;
603	else {
604		printf("rootmount: %s: unknown action\n", action);
605		error = EINVAL;
606	}
607
608	free(action, M_TEMP);
609	return (0);
610}
611
612static int
613parse_dir_timeout(char **conf)
614{
615	char *tok, *endtok;
616	long secs;
617	int error;
618
619	error = parse_token(conf, &tok);
620	if (error)
621		return (error);
622
623	secs = strtol(tok, &endtok, 0);
624	error = (secs < 0 || *endtok != '\0') ? EINVAL : 0;
625	if (!error)
626		root_mount_timeout = secs;
627	free(tok, M_TEMP);
628	return (error);
629}
630
631static int
632parse_directive(char **conf)
633{
634	char *dir;
635	int error;
636
637	error = parse_token(conf, &dir);
638	if (error)
639		return (error);
640
641	if (strcmp(dir, ".ask") == 0)
642		error = parse_dir_ask(conf);
643	else if (strcmp(dir, ".md") == 0)
644		error = parse_dir_md(conf);
645	else if (strcmp(dir, ".onfail") == 0)
646		error = parse_dir_onfail(conf);
647	else if (strcmp(dir, ".timeout") == 0)
648		error = parse_dir_timeout(conf);
649	else {
650		printf("mountroot: invalid directive `%s'\n", dir);
651		/* Ignore the rest of the line. */
652		(void)parse_skipto(conf, '\n');
653		error = EINVAL;
654	}
655	free(dir, M_TEMP);
656	return (error);
657}
658
659static int
660parse_mount_dev_present(const char *dev)
661{
662	struct nameidata nd;
663	int error;
664
665	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, dev, curthread);
666	error = namei(&nd);
667	if (!error)
668		vput(nd.ni_vp);
669	NDFREE(&nd, NDF_ONLY_PNBUF);
670	return (error != 0) ? 0 : 1;
671}
672
673static int
674parse_mount(char **conf)
675{
676	char errmsg[255];
677	struct mntarg *ma;
678	char *dev, *fs, *opts, *tok;
679	int delay, error, timeout;
680
681	error = parse_token(conf, &tok);
682	if (error)
683		return (error);
684	fs = tok;
685	error = parse_skipto(&tok, ':');
686	if (error) {
687		free(fs, M_TEMP);
688		return (error);
689	}
690	parse_poke(&tok, '\0');
691	parse_advance(&tok);
692	dev = tok;
693
694	if (root_mount_mddev != -1) {
695		/* Handle substitution for the md unit number. */
696		tok = strstr(dev, "md#");
697		if (tok != NULL)
698			tok[2] = '0' + root_mount_mddev;
699	}
700
701	/* Parse options. */
702	error = parse_token(conf, &tok);
703	opts = (error == 0) ? tok : NULL;
704
705	printf("Trying to mount root from %s:%s [%s]...\n", fs, dev,
706	    (opts != NULL) ? opts : "");
707
708	bzero(errmsg, sizeof(errmsg));
709
710	if (vfs_byname(fs) == NULL) {
711		strlcpy(errmsg, "unknown file system", sizeof(errmsg));
712		error = ENOENT;
713		goto out;
714	}
715
716	if (strcmp(fs, "zfs") != 0 && dev[0] != '\0' &&
717	    !parse_mount_dev_present(dev)) {
718		printf("mountroot: waiting for device %s ...\n", dev);
719		delay = hz / 10;
720		timeout = root_mount_timeout * hz;
721		do {
722			pause("rmdev", delay);
723			timeout -= delay;
724		} while (timeout > 0 && !parse_mount_dev_present(dev));
725		if (timeout <= 0) {
726			error = ENODEV;
727			goto out;
728		}
729	}
730
731	ma = NULL;
732	ma = mount_arg(ma, "fstype", fs, -1);
733	ma = mount_arg(ma, "fspath", "/", -1);
734	ma = mount_arg(ma, "from", dev, -1);
735	ma = mount_arg(ma, "errmsg", errmsg, sizeof(errmsg));
736	ma = mount_arg(ma, "ro", NULL, 0);
737	ma = parse_mountroot_options(ma, opts);
738	error = kernel_mount(ma, MNT_ROOTFS);
739
740 out:
741	if (error) {
742		printf("Mounting from %s:%s failed with error %d",
743		    fs, dev, error);
744		if (errmsg[0] != '\0')
745			printf(": %s", errmsg);
746		printf(".\n");
747	}
748	free(fs, M_TEMP);
749	if (opts != NULL)
750		free(opts, M_TEMP);
751	/* kernel_mount can return -1 on error. */
752	return ((error < 0) ? EDOOFUS : error);
753}
754
755static int
756vfs_mountroot_parse(struct sbuf *sb, struct mount *mpdevfs)
757{
758	struct mount *mp;
759	char *conf;
760	int error;
761
762	root_mount_mddev = -1;
763
764retry:
765	conf = sbuf_data(sb);
766	mp = TAILQ_NEXT(mpdevfs, mnt_list);
767	error = (mp == NULL) ? 0 : EDOOFUS;
768	root_mount_onfail = A_CONTINUE;
769	while (mp == NULL) {
770		error = parse_skipto(&conf, CC_NONWHITESPACE);
771		if (error == PE_EOL) {
772			parse_advance(&conf);
773			continue;
774		}
775		if (error < 0)
776			break;
777		switch (parse_peek(&conf)) {
778		case '#':
779			error = parse_skipto(&conf, '\n');
780			break;
781		case '.':
782			error = parse_directive(&conf);
783			break;
784		default:
785			error = parse_mount(&conf);
786			break;
787		}
788		if (error < 0)
789			break;
790		/* Ignore any trailing garbage on the line. */
791		if (parse_peek(&conf) != '\n') {
792			printf("mountroot: advancing to next directive...\n");
793			(void)parse_skipto(&conf, '\n');
794		}
795		mp = TAILQ_NEXT(mpdevfs, mnt_list);
796	}
797	if (mp != NULL)
798		return (0);
799
800	/*
801	 * We failed to mount (a new) root.
802	 */
803	switch (root_mount_onfail) {
804	case A_CONTINUE:
805		break;
806	case A_PANIC:
807		panic("mountroot: unable to (re-)mount root.");
808		/* NOTREACHED */
809	case A_RETRY:
810		goto retry;
811	case A_REBOOT:
812		kern_reboot(RB_NOSYNC);
813		/* NOTREACHED */
814	}
815
816	return (error);
817}
818
819static void
820vfs_mountroot_conf0(struct sbuf *sb)
821{
822	char *s, *tok, *mnt, *opt;
823	int error;
824
825	sbuf_printf(sb, ".onfail panic\n");
826	sbuf_printf(sb, ".timeout %d\n", root_mount_timeout);
827	if (boothowto & RB_ASKNAME)
828		sbuf_printf(sb, ".ask\n");
829#ifdef ROOTDEVNAME
830	if (boothowto & RB_DFLTROOT)
831		sbuf_printf(sb, "%s\n", ROOTDEVNAME);
832#endif
833	if (boothowto & RB_CDROM) {
834		sbuf_printf(sb, "cd9660:/dev/cd0 ro\n");
835		sbuf_printf(sb, ".timeout 0\n");
836		sbuf_printf(sb, "cd9660:/dev/acd0 ro\n");
837		sbuf_printf(sb, ".timeout %d\n", root_mount_timeout);
838	}
839	s = getenv("vfs.root.mountfrom");
840	if (s != NULL) {
841		opt = getenv("vfs.root.mountfrom.options");
842		tok = s;
843		error = parse_token(&tok, &mnt);
844		while (!error) {
845			sbuf_printf(sb, "%s %s\n", mnt,
846			    (opt != NULL) ? opt : "");
847			free(mnt, M_TEMP);
848			error = parse_token(&tok, &mnt);
849		}
850		if (opt != NULL)
851			freeenv(opt);
852		freeenv(s);
853	}
854	if (rootdevnames[0] != NULL)
855		sbuf_printf(sb, "%s\n", rootdevnames[0]);
856	if (rootdevnames[1] != NULL)
857		sbuf_printf(sb, "%s\n", rootdevnames[1]);
858#ifdef ROOTDEVNAME
859	if (!(boothowto & RB_DFLTROOT))
860		sbuf_printf(sb, "%s\n", ROOTDEVNAME);
861#endif
862	if (!(boothowto & RB_ASKNAME))
863		sbuf_printf(sb, ".ask\n");
864}
865
866static int
867vfs_mountroot_readconf(struct thread *td, struct sbuf *sb)
868{
869	static char buf[128];
870	struct nameidata nd;
871	off_t ofs;
872	int error, flags;
873	int len, resid;
874	int vfslocked;
875
876	NDINIT(&nd, LOOKUP, FOLLOW | MPSAFE, UIO_SYSSPACE,
877	    "/.mount.conf", td);
878	flags = FREAD;
879	error = vn_open(&nd, &flags, 0, NULL);
880	if (error)
881		return (error);
882
883	vfslocked = NDHASGIANT(&nd);
884	NDFREE(&nd, NDF_ONLY_PNBUF);
885	ofs = 0;
886	len = sizeof(buf) - 1;
887	while (1) {
888		error = vn_rdwr(UIO_READ, nd.ni_vp, buf, len, ofs,
889		    UIO_SYSSPACE, IO_NODELOCKED, td->td_ucred,
890		    NOCRED, &resid, td);
891		if (error)
892			break;
893		if (resid == len)
894			break;
895		buf[len - resid] = 0;
896		sbuf_printf(sb, "%s", buf);
897		ofs += len - resid;
898	}
899
900	VOP_UNLOCK(nd.ni_vp, 0);
901	vn_close(nd.ni_vp, FREAD, td->td_ucred, td);
902	VFS_UNLOCK_GIANT(vfslocked);
903	return (error);
904}
905
906static void
907vfs_mountroot_wait(void)
908{
909	struct root_hold_token *h;
910	struct timeval lastfail;
911	int curfail;
912
913	curfail = 0;
914	while (1) {
915		DROP_GIANT();
916		g_waitidle();
917		PICKUP_GIANT();
918		mtx_lock(&mountlist_mtx);
919		if (LIST_EMPTY(&root_holds)) {
920			mtx_unlock(&mountlist_mtx);
921			break;
922		}
923		if (ppsratecheck(&lastfail, &curfail, 1)) {
924			printf("Root mount waiting for:");
925			LIST_FOREACH(h, &root_holds, list)
926				printf(" %s", h->who);
927			printf("\n");
928		}
929		msleep(&root_holds, &mountlist_mtx, PZERO | PDROP, "roothold",
930		    hz);
931	}
932}
933
934void
935vfs_mountroot(void)
936{
937	struct mount *mp;
938	struct sbuf *sb;
939	struct thread *td;
940	time_t timebase;
941	int error;
942
943	td = curthread;
944
945	vfs_mountroot_wait();
946
947	sb = sbuf_new_auto();
948	vfs_mountroot_conf0(sb);
949	sbuf_finish(sb);
950
951	error = vfs_mountroot_devfs(td, &mp);
952	while (!error) {
953		error = vfs_mountroot_parse(sb, mp);
954		if (!error) {
955			error = vfs_mountroot_shuffle(td, mp);
956			if (!error) {
957				sbuf_clear(sb);
958				error = vfs_mountroot_readconf(td, sb);
959				sbuf_finish(sb);
960			}
961		}
962	}
963
964	sbuf_delete(sb);
965
966	/*
967	 * Iterate over all currently mounted file systems and use
968	 * the time stamp found to check and/or initialize the RTC.
969	 * Call inittodr() only once and pass it the largest of the
970	 * timestamps we encounter.
971	 */
972	timebase = 0;
973	mtx_lock(&mountlist_mtx);
974	mp = TAILQ_FIRST(&mountlist);
975	while (mp != NULL) {
976		if (mp->mnt_time > timebase)
977			timebase = mp->mnt_time;
978		mp = TAILQ_NEXT(mp, mnt_list);
979	}
980	mtx_unlock(&mountlist_mtx);
981	inittodr(timebase);
982
983	/* Keep prison0's root in sync with the global rootvnode. */
984	mtx_lock(&prison0.pr_mtx);
985	prison0.pr_root = rootvnode;
986	vref(prison0.pr_root);
987	mtx_unlock(&prison0.pr_mtx);
988
989	mtx_lock(&mountlist_mtx);
990	atomic_store_rel_int(&root_mount_complete, 1);
991	wakeup(&root_mount_complete);
992	mtx_unlock(&mountlist_mtx);
993}
994
995static struct mntarg *
996parse_mountroot_options(struct mntarg *ma, const char *options)
997{
998	char *p;
999	char *name, *name_arg;
1000	char *val, *val_arg;
1001	char *opts;
1002
1003	if (options == NULL || options[0] == '\0')
1004		return (ma);
1005
1006	p = opts = strdup(options, M_MOUNT);
1007	if (opts == NULL) {
1008		return (ma);
1009	}
1010
1011	while((name = strsep(&p, ",")) != NULL) {
1012		if (name[0] == '\0')
1013			break;
1014
1015		val = strchr(name, '=');
1016		if (val != NULL) {
1017			*val = '\0';
1018			++val;
1019		}
1020		if( strcmp(name, "rw") == 0 ||
1021		    strcmp(name, "noro") == 0) {
1022			/*
1023			 * The first time we mount the root file system,
1024			 * we need to mount 'ro', so We need to ignore
1025			 * 'rw' and 'noro' mount options.
1026			 */
1027			continue;
1028		}
1029		name_arg = strdup(name, M_MOUNT);
1030		val_arg = NULL;
1031		if (val != NULL)
1032			val_arg = strdup(val, M_MOUNT);
1033
1034		ma = mount_arg(ma, name_arg, val_arg,
1035		    (val_arg != NULL ? -1 : 0));
1036	}
1037	free(opts, M_MOUNT);
1038	return (ma);
1039}
1040