vfs_mountroot.c revision 273174 
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 273174 2014-10-16 18:04:43Z davide $");
42
43#include <sys/param.h>
44#include <sys/conf.h>
45#include <sys/cons.h>
46#include <sys/fcntl.h>
47#include <sys/jail.h>
48#include <sys/kernel.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 mtx root_holds_mtx;
101MTX_SYSINIT(root_holds, &root_holds_mtx, "root_holds", MTX_DEF);
102
103struct root_hold_token {
104	const char			*who;
105	LIST_ENTRY(root_hold_token)	list;
106};
107
108static LIST_HEAD(, root_hold_token)	root_holds =
109    LIST_HEAD_INITIALIZER(root_holds);
110
111enum action {
112	A_CONTINUE,
113	A_PANIC,
114	A_REBOOT,
115	A_RETRY
116};
117
118static enum action root_mount_onfail = A_CONTINUE;
119
120static int root_mount_mddev;
121static int root_mount_complete;
122
123/* By default wait up to 3 seconds for devices to appear. */
124static int root_mount_timeout = 3;
125TUNABLE_INT("vfs.mountroot.timeout", &root_mount_timeout);
126
127struct root_hold_token *
128root_mount_hold(const char *identifier)
129{
130	struct root_hold_token *h;
131
132	if (root_mounted())
133		return (NULL);
134
135	h = malloc(sizeof *h, M_DEVBUF, M_ZERO | M_WAITOK);
136	h->who = identifier;
137	mtx_lock(&root_holds_mtx);
138	LIST_INSERT_HEAD(&root_holds, h, list);
139	mtx_unlock(&root_holds_mtx);
140	return (h);
141}
142
143void
144root_mount_rel(struct root_hold_token *h)
145{
146
147	if (h == NULL)
148		return;
149	mtx_lock(&root_holds_mtx);
150	LIST_REMOVE(h, list);
151	wakeup(&root_holds);
152	mtx_unlock(&root_holds_mtx);
153	free(h, M_DEVBUF);
154}
155
156int
157root_mounted(void)
158{
159
160	/* No mutex is acquired here because int stores are atomic. */
161	return (root_mount_complete);
162}
163
164void
165root_mount_wait(void)
166{
167
168	/*
169	 * Panic on an obvious deadlock - the function can't be called from
170	 * a thread which is doing the whole SYSINIT stuff.
171	 */
172	KASSERT(curthread->td_proc->p_pid != 0,
173	    ("root_mount_wait: cannot be called from the swapper thread"));
174	mtx_lock(&root_holds_mtx);
175	while (!root_mount_complete) {
176		msleep(&root_mount_complete, &root_holds_mtx, PZERO, "rootwait",
177		    hz);
178	}
179	mtx_unlock(&root_holds_mtx);
180}
181
182static void
183set_rootvnode(void)
184{
185	struct proc *p;
186
187	if (VFS_ROOT(TAILQ_FIRST(&mountlist), LK_EXCLUSIVE, &rootvnode))
188		panic("Cannot find root vnode");
189
190	VOP_UNLOCK(rootvnode, 0);
191
192	p = curthread->td_proc;
193	FILEDESC_XLOCK(p->p_fd);
194
195	if (p->p_fd->fd_cdir != NULL)
196		vrele(p->p_fd->fd_cdir);
197	p->p_fd->fd_cdir = rootvnode;
198	VREF(rootvnode);
199
200	if (p->p_fd->fd_rdir != NULL)
201		vrele(p->p_fd->fd_rdir);
202	p->p_fd->fd_rdir = rootvnode;
203	VREF(rootvnode);
204
205	FILEDESC_XUNLOCK(p->p_fd);
206}
207
208static int
209vfs_mountroot_devfs(struct thread *td, struct mount **mpp)
210{
211	struct vfsoptlist *opts;
212	struct vfsconf *vfsp;
213	struct mount *mp;
214	int error;
215
216	*mpp = NULL;
217
218	vfsp = vfs_byname("devfs");
219	KASSERT(vfsp != NULL, ("Could not find devfs by name"));
220	if (vfsp == NULL)
221		return (ENOENT);
222
223	mp = vfs_mount_alloc(NULLVP, vfsp, "/dev", td->td_ucred);
224
225	error = VFS_MOUNT(mp);
226	KASSERT(error == 0, ("VFS_MOUNT(devfs) failed %d", error));
227	if (error)
228		return (error);
229
230	opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
231	TAILQ_INIT(opts);
232	mp->mnt_opt = opts;
233
234	mtx_lock(&mountlist_mtx);
235	TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list);
236	mtx_unlock(&mountlist_mtx);
237
238	*mpp = mp;
239	set_rootvnode();
240
241	error = kern_symlink(td, "/", "dev", UIO_SYSSPACE);
242	if (error)
243		printf("kern_symlink /dev -> / returns %d\n", error);
244
245	return (error);
246}
247
248static int
249vfs_mountroot_shuffle(struct thread *td, struct mount *mpdevfs)
250{
251	struct nameidata nd;
252	struct mount *mporoot, *mpnroot;
253	struct vnode *vp, *vporoot, *vpdevfs;
254	char *fspath;
255	int error;
256
257	mpnroot = TAILQ_NEXT(mpdevfs, mnt_list);
258
259	/* Shuffle the mountlist. */
260	mtx_lock(&mountlist_mtx);
261	mporoot = TAILQ_FIRST(&mountlist);
262	TAILQ_REMOVE(&mountlist, mpdevfs, mnt_list);
263	if (mporoot != mpdevfs) {
264		TAILQ_REMOVE(&mountlist, mpnroot, mnt_list);
265		TAILQ_INSERT_HEAD(&mountlist, mpnroot, mnt_list);
266	}
267	TAILQ_INSERT_TAIL(&mountlist, mpdevfs, mnt_list);
268	mtx_unlock(&mountlist_mtx);
269
270	cache_purgevfs(mporoot);
271	if (mporoot != mpdevfs)
272		cache_purgevfs(mpdevfs);
273
274	VFS_ROOT(mporoot, LK_EXCLUSIVE, &vporoot);
275
276	VI_LOCK(vporoot);
277	vporoot->v_iflag &= ~VI_MOUNT;
278	VI_UNLOCK(vporoot);
279	vporoot->v_mountedhere = NULL;
280	mporoot->mnt_flag &= ~MNT_ROOTFS;
281	mporoot->mnt_vnodecovered = NULL;
282	vput(vporoot);
283
284	/* Set up the new rootvnode, and purge the cache */
285	mpnroot->mnt_vnodecovered = NULL;
286	set_rootvnode();
287	cache_purgevfs(rootvnode->v_mount);
288
289	if (mporoot != mpdevfs) {
290		/* Remount old root under /.mount or /mnt */
291		fspath = "/.mount";
292		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
293		    fspath, td);
294		error = namei(&nd);
295		if (error) {
296			NDFREE(&nd, NDF_ONLY_PNBUF);
297			fspath = "/mnt";
298			NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
299			    fspath, td);
300			error = namei(&nd);
301		}
302		if (!error) {
303			vp = nd.ni_vp;
304			error = (vp->v_type == VDIR) ? 0 : ENOTDIR;
305			if (!error)
306				error = vinvalbuf(vp, V_SAVE, 0, 0);
307			if (!error) {
308				cache_purge(vp);
309				mporoot->mnt_vnodecovered = vp;
310				vp->v_mountedhere = mporoot;
311				strlcpy(mporoot->mnt_stat.f_mntonname,
312				    fspath, MNAMELEN);
313				VOP_UNLOCK(vp, 0);
314			} else
315				vput(vp);
316		}
317		NDFREE(&nd, NDF_ONLY_PNBUF);
318
319		if (error && bootverbose)
320			printf("mountroot: unable to remount previous root "
321			    "under /.mount or /mnt (error %d).\n", error);
322	}
323
324	/* Remount devfs under /dev */
325	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, "/dev", td);
326	error = namei(&nd);
327	if (!error) {
328		vp = nd.ni_vp;
329		error = (vp->v_type == VDIR) ? 0 : ENOTDIR;
330		if (!error)
331			error = vinvalbuf(vp, V_SAVE, 0, 0);
332		if (!error) {
333			vpdevfs = mpdevfs->mnt_vnodecovered;
334			if (vpdevfs != NULL) {
335				cache_purge(vpdevfs);
336				vpdevfs->v_mountedhere = NULL;
337				vrele(vpdevfs);
338			}
339			mpdevfs->mnt_vnodecovered = vp;
340			vp->v_mountedhere = mpdevfs;
341			VOP_UNLOCK(vp, 0);
342		} else
343			vput(vp);
344	}
345	if (error && bootverbose)
346		printf("mountroot: unable to remount devfs under /dev "
347		    "(error %d).\n", error);
348	NDFREE(&nd, NDF_ONLY_PNBUF);
349
350	if (mporoot == mpdevfs) {
351		vfs_unbusy(mpdevfs);
352		/* Unlink the no longer needed /dev/dev -> / symlink */
353		error = kern_unlink(td, "/dev/dev", UIO_SYSSPACE);
354		if (error && bootverbose)
355			printf("mountroot: unable to unlink /dev/dev "
356			    "(error %d)\n", error);
357	}
358
359	return (0);
360}
361
362/*
363 * Configuration parser.
364 */
365
366/* Parser character classes. */
367#define	CC_WHITESPACE		-1
368#define	CC_NONWHITESPACE	-2
369
370/* Parse errors. */
371#define	PE_EOF			-1
372#define	PE_EOL			-2
373
374static __inline int
375parse_peek(char **conf)
376{
377
378	return (**conf);
379}
380
381static __inline void
382parse_poke(char **conf, int c)
383{
384
385	**conf = c;
386}
387
388static __inline void
389parse_advance(char **conf)
390{
391
392	(*conf)++;
393}
394
395static int
396parse_skipto(char **conf, int mc)
397{
398	int c, match;
399
400	while (1) {
401		c = parse_peek(conf);
402		if (c == 0)
403			return (PE_EOF);
404		switch (mc) {
405		case CC_WHITESPACE:
406			match = (c == ' ' || c == '\t' || c == '\n') ? 1 : 0;
407			break;
408		case CC_NONWHITESPACE:
409			if (c == '\n')
410				return (PE_EOL);
411			match = (c != ' ' && c != '\t') ? 1 : 0;
412			break;
413		default:
414			match = (c == mc) ? 1 : 0;
415			break;
416		}
417		if (match)
418			break;
419		parse_advance(conf);
420	}
421	return (0);
422}
423
424static int
425parse_token(char **conf, char **tok)
426{
427	char *p;
428	size_t len;
429	int error;
430
431	*tok = NULL;
432	error = parse_skipto(conf, CC_NONWHITESPACE);
433	if (error)
434		return (error);
435	p = *conf;
436	error = parse_skipto(conf, CC_WHITESPACE);
437	len = *conf - p;
438	*tok = malloc(len + 1, M_TEMP, M_WAITOK | M_ZERO);
439	bcopy(p, *tok, len);
440	return (0);
441}
442
443static void
444parse_dir_ask_printenv(const char *var)
445{
446	char *val;
447
448	val = kern_getenv(var);
449	if (val != NULL) {
450		printf("  %s=%s\n", var, val);
451		freeenv(val);
452	}
453}
454
455static int
456parse_dir_ask(char **conf)
457{
458	char name[80];
459	char *mnt;
460	int error;
461
462	printf("\nLoader variables:\n");
463	parse_dir_ask_printenv("vfs.root.mountfrom");
464	parse_dir_ask_printenv("vfs.root.mountfrom.options");
465
466	printf("\nManual root filesystem specification:\n");
467	printf("  <fstype>:<device> [options]\n");
468	printf("      Mount <device> using filesystem <fstype>\n");
469	printf("      and with the specified (optional) option list.\n");
470	printf("\n");
471	printf("    eg. ufs:/dev/da0s1a\n");
472	printf("        zfs:tank\n");
473	printf("        cd9660:/dev/acd0 ro\n");
474	printf("          (which is equivalent to: ");
475	printf("mount -t cd9660 -o ro /dev/acd0 /)\n");
476	printf("\n");
477	printf("  ?               List valid disk boot devices\n");
478	printf("  .               Yield 1 second (for background tasks)\n");
479	printf("  <empty line>    Abort manual input\n");
480
481	do {
482		error = EINVAL;
483		printf("\nmountroot> ");
484		cngets(name, sizeof(name), GETS_ECHO);
485		if (name[0] == '\0')
486			break;
487		if (name[0] == '?' && name[1] == '\0') {
488			printf("\nList of GEOM managed disk devices:\n  ");
489			g_dev_print();
490			continue;
491		}
492		if (name[0] == '.' && name[1] == '\0') {
493			pause("rmask", hz);
494			continue;
495		}
496		mnt = name;
497		error = parse_mount(&mnt);
498		if (error == -1)
499			printf("Invalid file system specification.\n");
500	} while (error != 0);
501
502	return (error);
503}
504
505static int
506parse_dir_md(char **conf)
507{
508	struct stat sb;
509	struct thread *td;
510	struct md_ioctl *mdio;
511	char *path, *tok;
512	int error, fd, len;
513
514	td = curthread;
515
516	error = parse_token(conf, &tok);
517	if (error)
518		return (error);
519
520	len = strlen(tok);
521	mdio = malloc(sizeof(*mdio) + len + 1, M_TEMP, M_WAITOK | M_ZERO);
522	path = (void *)(mdio + 1);
523	bcopy(tok, path, len);
524	free(tok, M_TEMP);
525
526	/* Get file status. */
527	error = kern_stat(td, path, UIO_SYSSPACE, &sb);
528	if (error)
529		goto out;
530
531	/* Open /dev/mdctl so that we can attach/detach. */
532	error = kern_open(td, "/dev/" MDCTL_NAME, UIO_SYSSPACE, O_RDWR, 0);
533	if (error)
534		goto out;
535
536	fd = td->td_retval[0];
537	mdio->md_version = MDIOVERSION;
538	mdio->md_type = MD_VNODE;
539
540	if (root_mount_mddev != -1) {
541		mdio->md_unit = root_mount_mddev;
542		DROP_GIANT();
543		error = kern_ioctl(td, fd, MDIOCDETACH, (void *)mdio);
544		PICKUP_GIANT();
545		/* Ignore errors. We don't care. */
546		root_mount_mddev = -1;
547	}
548
549	mdio->md_file = (void *)(mdio + 1);
550	mdio->md_options = MD_AUTOUNIT | MD_READONLY;
551	mdio->md_mediasize = sb.st_size;
552	mdio->md_unit = 0;
553	DROP_GIANT();
554	error = kern_ioctl(td, fd, MDIOCATTACH, (void *)mdio);
555	PICKUP_GIANT();
556	if (error)
557		goto out;
558
559	if (mdio->md_unit > 9) {
560		printf("rootmount: too many md units\n");
561		mdio->md_file = NULL;
562		mdio->md_options = 0;
563		mdio->md_mediasize = 0;
564		DROP_GIANT();
565		error = kern_ioctl(td, fd, MDIOCDETACH, (void *)mdio);
566		PICKUP_GIANT();
567		/* Ignore errors. We don't care. */
568		error = ERANGE;
569		goto out;
570	}
571
572	root_mount_mddev = mdio->md_unit;
573	printf(MD_NAME "%u attached to %s\n", root_mount_mddev, mdio->md_file);
574
575	error = kern_close(td, fd);
576
577 out:
578	free(mdio, M_TEMP);
579	return (error);
580}
581
582static int
583parse_dir_onfail(char **conf)
584{
585	char *action;
586	int error;
587
588	error = parse_token(conf, &action);
589	if (error)
590		return (error);
591
592	if (!strcmp(action, "continue"))
593		root_mount_onfail = A_CONTINUE;
594	else if (!strcmp(action, "panic"))
595		root_mount_onfail = A_PANIC;
596	else if (!strcmp(action, "reboot"))
597		root_mount_onfail = A_REBOOT;
598	else if (!strcmp(action, "retry"))
599		root_mount_onfail = A_RETRY;
600	else {
601		printf("rootmount: %s: unknown action\n", action);
602		error = EINVAL;
603	}
604
605	free(action, M_TEMP);
606	return (0);
607}
608
609static int
610parse_dir_timeout(char **conf)
611{
612	char *tok, *endtok;
613	long secs;
614	int error;
615
616	error = parse_token(conf, &tok);
617	if (error)
618		return (error);
619
620	secs = strtol(tok, &endtok, 0);
621	error = (secs < 0 || *endtok != '\0') ? EINVAL : 0;
622	if (!error)
623		root_mount_timeout = secs;
624	free(tok, M_TEMP);
625	return (error);
626}
627
628static int
629parse_directive(char **conf)
630{
631	char *dir;
632	int error;
633
634	error = parse_token(conf, &dir);
635	if (error)
636		return (error);
637
638	if (strcmp(dir, ".ask") == 0)
639		error = parse_dir_ask(conf);
640	else if (strcmp(dir, ".md") == 0)
641		error = parse_dir_md(conf);
642	else if (strcmp(dir, ".onfail") == 0)
643		error = parse_dir_onfail(conf);
644	else if (strcmp(dir, ".timeout") == 0)
645		error = parse_dir_timeout(conf);
646	else {
647		printf("mountroot: invalid directive `%s'\n", dir);
648		/* Ignore the rest of the line. */
649		(void)parse_skipto(conf, '\n');
650		error = EINVAL;
651	}
652	free(dir, M_TEMP);
653	return (error);
654}
655
656static int
657parse_mount_dev_present(const char *dev)
658{
659	struct nameidata nd;
660	int error;
661
662	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, dev, curthread);
663	error = namei(&nd);
664	if (!error)
665		vput(nd.ni_vp);
666	NDFREE(&nd, NDF_ONLY_PNBUF);
667	return (error != 0) ? 0 : 1;
668}
669
670#define	ERRMSGL	255
671static int
672parse_mount(char **conf)
673{
674	char *errmsg;
675	struct mntarg *ma;
676	char *dev, *fs, *opts, *tok;
677	int delay, error, timeout;
678
679	error = parse_token(conf, &tok);
680	if (error)
681		return (error);
682	fs = tok;
683	error = parse_skipto(&tok, ':');
684	if (error) {
685		free(fs, M_TEMP);
686		return (error);
687	}
688	parse_poke(&tok, '\0');
689	parse_advance(&tok);
690	dev = tok;
691
692	if (root_mount_mddev != -1) {
693		/* Handle substitution for the md unit number. */
694		tok = strstr(dev, "md#");
695		if (tok != NULL)
696			tok[2] = '0' + root_mount_mddev;
697	}
698
699	/* Parse options. */
700	error = parse_token(conf, &tok);
701	opts = (error == 0) ? tok : NULL;
702
703	printf("Trying to mount root from %s:%s [%s]...\n", fs, dev,
704	    (opts != NULL) ? opts : "");
705
706	errmsg = malloc(ERRMSGL, M_TEMP, M_WAITOK | M_ZERO);
707
708	if (vfs_byname(fs) == NULL) {
709		strlcpy(errmsg, "unknown file system", ERRMSGL);
710		error = ENOENT;
711		goto out;
712	}
713
714	if (strcmp(fs, "zfs") != 0 && strstr(fs, "nfs") == NULL &&
715	    dev[0] != '\0' && !parse_mount_dev_present(dev)) {
716		printf("mountroot: waiting for device %s ...\n", dev);
717		delay = hz / 10;
718		timeout = root_mount_timeout * hz;
719		do {
720			pause("rmdev", delay);
721			timeout -= delay;
722		} while (timeout > 0 && !parse_mount_dev_present(dev));
723		if (timeout <= 0) {
724			error = ENODEV;
725			goto out;
726		}
727	}
728
729	ma = NULL;
730	ma = mount_arg(ma, "fstype", fs, -1);
731	ma = mount_arg(ma, "fspath", "/", -1);
732	ma = mount_arg(ma, "from", dev, -1);
733	ma = mount_arg(ma, "errmsg", errmsg, ERRMSGL);
734	ma = mount_arg(ma, "ro", NULL, 0);
735	ma = parse_mountroot_options(ma, opts);
736	error = kernel_mount(ma, MNT_ROOTFS);
737
738 out:
739	if (error) {
740		printf("Mounting from %s:%s failed with error %d",
741		    fs, dev, error);
742		if (errmsg[0] != '\0')
743			printf(": %s", errmsg);
744		printf(".\n");
745	}
746	free(fs, M_TEMP);
747	free(errmsg, M_TEMP);
748	if (opts != NULL)
749		free(opts, M_TEMP);
750	/* kernel_mount can return -1 on error. */
751	return ((error < 0) ? EDOOFUS : error);
752}
753#undef ERRMSGL
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 = kern_getenv("vfs.root.mountfrom");
840	if (s != NULL) {
841		opt = kern_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	ssize_t resid;
873	int error, flags, len;
874
875	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/.mount.conf", td);
876	flags = FREAD;
877	error = vn_open(&nd, &flags, 0, NULL);
878	if (error)
879		return (error);
880
881	NDFREE(&nd, NDF_ONLY_PNBUF);
882	ofs = 0;
883	len = sizeof(buf) - 1;
884	while (1) {
885		error = vn_rdwr(UIO_READ, nd.ni_vp, buf, len, ofs,
886		    UIO_SYSSPACE, IO_NODELOCKED, td->td_ucred,
887		    NOCRED, &resid, td);
888		if (error)
889			break;
890		if (resid == len)
891			break;
892		buf[len - resid] = 0;
893		sbuf_printf(sb, "%s", buf);
894		ofs += len - resid;
895	}
896
897	VOP_UNLOCK(nd.ni_vp, 0);
898	vn_close(nd.ni_vp, FREAD, td->td_ucred, td);
899	return (error);
900}
901
902static void
903vfs_mountroot_wait(void)
904{
905	struct root_hold_token *h;
906	struct timeval lastfail;
907	int curfail;
908
909	curfail = 0;
910	while (1) {
911		DROP_GIANT();
912		g_waitidle();
913		PICKUP_GIANT();
914		mtx_lock(&root_holds_mtx);
915		if (LIST_EMPTY(&root_holds)) {
916			mtx_unlock(&root_holds_mtx);
917			break;
918		}
919		if (ppsratecheck(&lastfail, &curfail, 1)) {
920			printf("Root mount waiting for:");
921			LIST_FOREACH(h, &root_holds, list)
922				printf(" %s", h->who);
923			printf("\n");
924		}
925		msleep(&root_holds, &root_holds_mtx, PZERO | PDROP, "roothold",
926		    hz);
927	}
928}
929
930void
931vfs_mountroot(void)
932{
933	struct mount *mp;
934	struct sbuf *sb;
935	struct thread *td;
936	time_t timebase;
937	int error;
938
939	td = curthread;
940
941	vfs_mountroot_wait();
942
943	sb = sbuf_new_auto();
944	vfs_mountroot_conf0(sb);
945	sbuf_finish(sb);
946
947	error = vfs_mountroot_devfs(td, &mp);
948	while (!error) {
949		error = vfs_mountroot_parse(sb, mp);
950		if (!error) {
951			error = vfs_mountroot_shuffle(td, mp);
952			if (!error) {
953				sbuf_clear(sb);
954				error = vfs_mountroot_readconf(td, sb);
955				sbuf_finish(sb);
956			}
957		}
958	}
959
960	sbuf_delete(sb);
961
962	/*
963	 * Iterate over all currently mounted file systems and use
964	 * the time stamp found to check and/or initialize the RTC.
965	 * Call inittodr() only once and pass it the largest of the
966	 * timestamps we encounter.
967	 */
968	timebase = 0;
969	mtx_lock(&mountlist_mtx);
970	mp = TAILQ_FIRST(&mountlist);
971	while (mp != NULL) {
972		if (mp->mnt_time > timebase)
973			timebase = mp->mnt_time;
974		mp = TAILQ_NEXT(mp, mnt_list);
975	}
976	mtx_unlock(&mountlist_mtx);
977	inittodr(timebase);
978
979	/* Keep prison0's root in sync with the global rootvnode. */
980	mtx_lock(&prison0.pr_mtx);
981	prison0.pr_root = rootvnode;
982	vref(prison0.pr_root);
983	mtx_unlock(&prison0.pr_mtx);
984
985	mtx_lock(&root_holds_mtx);
986	atomic_store_rel_int(&root_mount_complete, 1);
987	wakeup(&root_mount_complete);
988	mtx_unlock(&root_holds_mtx);
989
990	EVENTHANDLER_INVOKE(mountroot);
991}
992
993static struct mntarg *
994parse_mountroot_options(struct mntarg *ma, const char *options)
995{
996	char *p;
997	char *name, *name_arg;
998	char *val, *val_arg;
999	char *opts;
1000
1001	if (options == NULL || options[0] == '\0')
1002		return (ma);
1003
1004	p = opts = strdup(options, M_MOUNT);
1005	if (opts == NULL) {
1006		return (ma);
1007	}
1008
1009	while((name = strsep(&p, ",")) != NULL) {
1010		if (name[0] == '\0')
1011			break;
1012
1013		val = strchr(name, '=');
1014		if (val != NULL) {
1015			*val = '\0';
1016			++val;
1017		}
1018		if( strcmp(name, "rw") == 0 ||
1019		    strcmp(name, "noro") == 0) {
1020			/*
1021			 * The first time we mount the root file system,
1022			 * we need to mount 'ro', so We need to ignore
1023			 * 'rw' and 'noro' mount options.
1024			 */
1025			continue;
1026		}
1027		name_arg = strdup(name, M_MOUNT);
1028		val_arg = NULL;
1029		if (val != NULL)
1030			val_arg = strdup(val, M_MOUNT);
1031
1032		ma = mount_arg(ma, name_arg, val_arg,
1033		    (val_arg != NULL ? -1 : 0));
1034	}
1035	free(opts, M_MOUNT);
1036	return (ma);
1037}
1038