vfs_mountroot.c revision 288400 
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: stable/10/sys/kern/vfs_mountroot.c 288400 2015-09-29 21:54:09Z bdrewery $");
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 void
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
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 int
394parse_skipto(char **conf, int mc)
395{
396	int c, match;
397
398	while (1) {
399		c = parse_peek(conf);
400		if (c == 0)
401			return (PE_EOF);
402		switch (mc) {
403		case CC_WHITESPACE:
404			match = (c == ' ' || c == '\t' || c == '\n') ? 1 : 0;
405			break;
406		case CC_NONWHITESPACE:
407			if (c == '\n')
408				return (PE_EOL);
409			match = (c != ' ' && c != '\t') ? 1 : 0;
410			break;
411		default:
412			match = (c == mc) ? 1 : 0;
413			break;
414		}
415		if (match)
416			break;
417		parse_advance(conf);
418	}
419	return (0);
420}
421
422static int
423parse_token(char **conf, char **tok)
424{
425	char *p;
426	size_t len;
427	int error;
428
429	*tok = NULL;
430	error = parse_skipto(conf, CC_NONWHITESPACE);
431	if (error)
432		return (error);
433	p = *conf;
434	error = parse_skipto(conf, CC_WHITESPACE);
435	len = *conf - p;
436	*tok = malloc(len + 1, M_TEMP, M_WAITOK | M_ZERO);
437	bcopy(p, *tok, len);
438	return (0);
439}
440
441static void
442parse_dir_ask_printenv(const char *var)
443{
444	char *val;
445
446	val = getenv(var);
447	if (val != NULL) {
448		printf("  %s=%s\n", var, val);
449		freeenv(val);
450	}
451}
452
453static int
454parse_dir_ask(char **conf)
455{
456	char name[80];
457	char *mnt;
458	int error;
459
460	printf("\nLoader variables:\n");
461	parse_dir_ask_printenv("vfs.root.mountfrom");
462	parse_dir_ask_printenv("vfs.root.mountfrom.options");
463
464	printf("\nManual root filesystem specification:\n");
465	printf("  <fstype>:<device> [options]\n");
466	printf("      Mount <device> using filesystem <fstype>\n");
467	printf("      and with the specified (optional) option list.\n");
468	printf("\n");
469	printf("    eg. ufs:/dev/da0s1a\n");
470	printf("        zfs:tank\n");
471	printf("        cd9660:/dev/acd0 ro\n");
472	printf("          (which is equivalent to: ");
473	printf("mount -t cd9660 -o ro /dev/acd0 /)\n");
474	printf("\n");
475	printf("  ?               List valid disk boot devices\n");
476	printf("  .               Yield 1 second (for background tasks)\n");
477	printf("  <empty line>    Abort manual input\n");
478
479	do {
480		error = EINVAL;
481		printf("\nmountroot> ");
482		cngets(name, sizeof(name), GETS_ECHO);
483		if (name[0] == '\0')
484			break;
485		if (name[0] == '?' && name[1] == '\0') {
486			printf("\nList of GEOM managed disk devices:\n  ");
487			g_dev_print();
488			continue;
489		}
490		if (name[0] == '.' && name[1] == '\0') {
491			pause("rmask", hz);
492			continue;
493		}
494		mnt = name;
495		error = parse_mount(&mnt);
496		if (error == -1)
497			printf("Invalid file system specification.\n");
498	} while (error != 0);
499
500	return (error);
501}
502
503static int
504parse_dir_md(char **conf)
505{
506	struct stat sb;
507	struct thread *td;
508	struct md_ioctl *mdio;
509	char *path, *tok;
510	int error, fd, len;
511
512	td = curthread;
513
514	error = parse_token(conf, &tok);
515	if (error)
516		return (error);
517
518	len = strlen(tok);
519	mdio = malloc(sizeof(*mdio) + len + 1, M_TEMP, M_WAITOK | M_ZERO);
520	path = (void *)(mdio + 1);
521	bcopy(tok, path, len);
522	free(tok, M_TEMP);
523
524	/* Get file status. */
525	error = kern_stat(td, path, UIO_SYSSPACE, &sb);
526	if (error)
527		goto out;
528
529	/* Open /dev/mdctl so that we can attach/detach. */
530	error = kern_open(td, "/dev/" MDCTL_NAME, UIO_SYSSPACE, O_RDWR, 0);
531	if (error)
532		goto out;
533
534	fd = td->td_retval[0];
535	mdio->md_version = MDIOVERSION;
536	mdio->md_type = MD_VNODE;
537
538	if (root_mount_mddev != -1) {
539		mdio->md_unit = root_mount_mddev;
540		DROP_GIANT();
541		error = kern_ioctl(td, fd, MDIOCDETACH, (void *)mdio);
542		PICKUP_GIANT();
543		/* Ignore errors. We don't care. */
544		root_mount_mddev = -1;
545	}
546
547	mdio->md_file = (void *)(mdio + 1);
548	mdio->md_options = MD_AUTOUNIT | MD_READONLY;
549	mdio->md_mediasize = sb.st_size;
550	mdio->md_unit = 0;
551	DROP_GIANT();
552	error = kern_ioctl(td, fd, MDIOCATTACH, (void *)mdio);
553	PICKUP_GIANT();
554	if (error)
555		goto out;
556
557	if (mdio->md_unit > 9) {
558		printf("rootmount: too many md units\n");
559		mdio->md_file = NULL;
560		mdio->md_options = 0;
561		mdio->md_mediasize = 0;
562		DROP_GIANT();
563		error = kern_ioctl(td, fd, MDIOCDETACH, (void *)mdio);
564		PICKUP_GIANT();
565		/* Ignore errors. We don't care. */
566		error = ERANGE;
567		goto out;
568	}
569
570	root_mount_mddev = mdio->md_unit;
571	printf(MD_NAME "%u attached to %s\n", root_mount_mddev, mdio->md_file);
572
573	error = kern_close(td, fd);
574
575 out:
576	free(mdio, M_TEMP);
577	return (error);
578}
579
580static int
581parse_dir_onfail(char **conf)
582{
583	char *action;
584	int error;
585
586	error = parse_token(conf, &action);
587	if (error)
588		return (error);
589
590	if (!strcmp(action, "continue"))
591		root_mount_onfail = A_CONTINUE;
592	else if (!strcmp(action, "panic"))
593		root_mount_onfail = A_PANIC;
594	else if (!strcmp(action, "reboot"))
595		root_mount_onfail = A_REBOOT;
596	else if (!strcmp(action, "retry"))
597		root_mount_onfail = A_RETRY;
598	else {
599		printf("rootmount: %s: unknown action\n", action);
600		error = EINVAL;
601	}
602
603	free(action, M_TEMP);
604	return (0);
605}
606
607static int
608parse_dir_timeout(char **conf)
609{
610	char *tok, *endtok;
611	long secs;
612	int error;
613
614	error = parse_token(conf, &tok);
615	if (error)
616		return (error);
617
618	secs = strtol(tok, &endtok, 0);
619	error = (secs < 0 || *endtok != '\0') ? EINVAL : 0;
620	if (!error)
621		root_mount_timeout = secs;
622	free(tok, M_TEMP);
623	return (error);
624}
625
626static int
627parse_directive(char **conf)
628{
629	char *dir;
630	int error;
631
632	error = parse_token(conf, &dir);
633	if (error)
634		return (error);
635
636	if (strcmp(dir, ".ask") == 0)
637		error = parse_dir_ask(conf);
638	else if (strcmp(dir, ".md") == 0)
639		error = parse_dir_md(conf);
640	else if (strcmp(dir, ".onfail") == 0)
641		error = parse_dir_onfail(conf);
642	else if (strcmp(dir, ".timeout") == 0)
643		error = parse_dir_timeout(conf);
644	else {
645		printf("mountroot: invalid directive `%s'\n", dir);
646		/* Ignore the rest of the line. */
647		(void)parse_skipto(conf, '\n');
648		error = EINVAL;
649	}
650	free(dir, M_TEMP);
651	return (error);
652}
653
654static int
655parse_mount_dev_present(const char *dev)
656{
657	struct nameidata nd;
658	int error;
659
660	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, dev, curthread);
661	error = namei(&nd);
662	if (!error)
663		vput(nd.ni_vp);
664	NDFREE(&nd, NDF_ONLY_PNBUF);
665	return (error != 0) ? 0 : 1;
666}
667
668#define	ERRMSGL	255
669static int
670parse_mount(char **conf)
671{
672	char *errmsg;
673	struct mntarg *ma;
674	char *dev, *fs, *opts, *tok;
675	int delay, error, timeout;
676
677	error = parse_token(conf, &tok);
678	if (error)
679		return (error);
680	fs = tok;
681	error = parse_skipto(&tok, ':');
682	if (error) {
683		free(fs, M_TEMP);
684		return (error);
685	}
686	parse_poke(&tok, '\0');
687	parse_advance(&tok);
688	dev = tok;
689
690	if (root_mount_mddev != -1) {
691		/* Handle substitution for the md unit number. */
692		tok = strstr(dev, "md#");
693		if (tok != NULL)
694			tok[2] = '0' + root_mount_mddev;
695	}
696
697	/* Parse options. */
698	error = parse_token(conf, &tok);
699	opts = (error == 0) ? tok : NULL;
700
701	printf("Trying to mount root from %s:%s [%s]...\n", fs, dev,
702	    (opts != NULL) ? opts : "");
703
704	errmsg = malloc(ERRMSGL, M_TEMP, M_WAITOK | M_ZERO);
705
706	if (vfs_byname(fs) == NULL) {
707		strlcpy(errmsg, "unknown file system", ERRMSGL);
708		error = ENOENT;
709		goto out;
710	}
711
712	if (strcmp(fs, "zfs") != 0 && strstr(fs, "nfs") == NULL &&
713	    dev[0] != '\0' && !parse_mount_dev_present(dev)) {
714		printf("mountroot: waiting for device %s ...\n", dev);
715		delay = hz / 10;
716		timeout = root_mount_timeout * hz;
717		do {
718			pause("rmdev", delay);
719			timeout -= delay;
720		} while (timeout > 0 && !parse_mount_dev_present(dev));
721		if (timeout <= 0) {
722			error = ENODEV;
723			goto out;
724		}
725	}
726
727	ma = NULL;
728	ma = mount_arg(ma, "fstype", fs, -1);
729	ma = mount_arg(ma, "fspath", "/", -1);
730	ma = mount_arg(ma, "from", dev, -1);
731	ma = mount_arg(ma, "errmsg", errmsg, ERRMSGL);
732	ma = mount_arg(ma, "ro", NULL, 0);
733	ma = parse_mountroot_options(ma, opts);
734	error = kernel_mount(ma, MNT_ROOTFS);
735
736 out:
737	if (error) {
738		printf("Mounting from %s:%s failed with error %d",
739		    fs, dev, error);
740		if (errmsg[0] != '\0')
741			printf(": %s", errmsg);
742		printf(".\n");
743	}
744	free(fs, M_TEMP);
745	free(errmsg, M_TEMP);
746	if (opts != NULL)
747		free(opts, M_TEMP);
748	/* kernel_mount can return -1 on error. */
749	return ((error < 0) ? EDOOFUS : error);
750}
751#undef ERRMSGL
752
753static int
754vfs_mountroot_parse(struct sbuf *sb, struct mount *mpdevfs)
755{
756	struct mount *mp;
757	char *conf;
758	int error;
759
760	root_mount_mddev = -1;
761
762retry:
763	conf = sbuf_data(sb);
764	mp = TAILQ_NEXT(mpdevfs, mnt_list);
765	error = (mp == NULL) ? 0 : EDOOFUS;
766	root_mount_onfail = A_CONTINUE;
767	while (mp == NULL) {
768		error = parse_skipto(&conf, CC_NONWHITESPACE);
769		if (error == PE_EOL) {
770			parse_advance(&conf);
771			continue;
772		}
773		if (error < 0)
774			break;
775		switch (parse_peek(&conf)) {
776		case '#':
777			error = parse_skipto(&conf, '\n');
778			break;
779		case '.':
780			error = parse_directive(&conf);
781			break;
782		default:
783			error = parse_mount(&conf);
784			break;
785		}
786		if (error < 0)
787			break;
788		/* Ignore any trailing garbage on the line. */
789		if (parse_peek(&conf) != '\n') {
790			printf("mountroot: advancing to next directive...\n");
791			(void)parse_skipto(&conf, '\n');
792		}
793		mp = TAILQ_NEXT(mpdevfs, mnt_list);
794	}
795	if (mp != NULL)
796		return (0);
797
798	/*
799	 * We failed to mount (a new) root.
800	 */
801	switch (root_mount_onfail) {
802	case A_CONTINUE:
803		break;
804	case A_PANIC:
805		panic("mountroot: unable to (re-)mount root.");
806		/* NOTREACHED */
807	case A_RETRY:
808		goto retry;
809	case A_REBOOT:
810		kern_reboot(RB_NOSYNC);
811		/* NOTREACHED */
812	}
813
814	return (error);
815}
816
817static void
818vfs_mountroot_conf0(struct sbuf *sb)
819{
820	char *s, *tok, *mnt, *opt;
821	int error;
822
823	sbuf_printf(sb, ".onfail panic\n");
824	sbuf_printf(sb, ".timeout %d\n", root_mount_timeout);
825	if (boothowto & RB_ASKNAME)
826		sbuf_printf(sb, ".ask\n");
827#ifdef ROOTDEVNAME
828	if (boothowto & RB_DFLTROOT)
829		sbuf_printf(sb, "%s\n", ROOTDEVNAME);
830#endif
831	if (boothowto & RB_CDROM) {
832		sbuf_printf(sb, "cd9660:/dev/cd0 ro\n");
833		sbuf_printf(sb, ".timeout 0\n");
834		sbuf_printf(sb, "cd9660:/dev/acd0 ro\n");
835		sbuf_printf(sb, ".timeout %d\n", root_mount_timeout);
836	}
837	s = getenv("vfs.root.mountfrom");
838	if (s != NULL) {
839		opt = getenv("vfs.root.mountfrom.options");
840		tok = s;
841		error = parse_token(&tok, &mnt);
842		while (!error) {
843			sbuf_printf(sb, "%s %s\n", mnt,
844			    (opt != NULL) ? opt : "");
845			free(mnt, M_TEMP);
846			error = parse_token(&tok, &mnt);
847		}
848		if (opt != NULL)
849			freeenv(opt);
850		freeenv(s);
851	}
852	if (rootdevnames[0] != NULL)
853		sbuf_printf(sb, "%s\n", rootdevnames[0]);
854	if (rootdevnames[1] != NULL)
855		sbuf_printf(sb, "%s\n", rootdevnames[1]);
856#ifdef ROOTDEVNAME
857	if (!(boothowto & RB_DFLTROOT))
858		sbuf_printf(sb, "%s\n", ROOTDEVNAME);
859#endif
860	if (!(boothowto & RB_ASKNAME))
861		sbuf_printf(sb, ".ask\n");
862}
863
864static int
865vfs_mountroot_readconf(struct thread *td, struct sbuf *sb)
866{
867	static char buf[128];
868	struct nameidata nd;
869	off_t ofs;
870	ssize_t resid;
871	int error, flags, len;
872
873	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/.mount.conf", td);
874	flags = FREAD;
875	error = vn_open(&nd, &flags, 0, NULL);
876	if (error)
877		return (error);
878
879	NDFREE(&nd, NDF_ONLY_PNBUF);
880	ofs = 0;
881	len = sizeof(buf) - 1;
882	while (1) {
883		error = vn_rdwr(UIO_READ, nd.ni_vp, buf, len, ofs,
884		    UIO_SYSSPACE, IO_NODELOCKED, td->td_ucred,
885		    NOCRED, &resid, td);
886		if (error)
887			break;
888		if (resid == len)
889			break;
890		buf[len - resid] = 0;
891		sbuf_printf(sb, "%s", buf);
892		ofs += len - resid;
893	}
894
895	VOP_UNLOCK(nd.ni_vp, 0);
896	vn_close(nd.ni_vp, FREAD, td->td_ucred, td);
897	return (error);
898}
899
900static void
901vfs_mountroot_wait(void)
902{
903	struct root_hold_token *h;
904	struct timeval lastfail;
905	int curfail;
906
907	curfail = 0;
908	while (1) {
909		DROP_GIANT();
910		g_waitidle();
911		PICKUP_GIANT();
912		mtx_lock(&root_holds_mtx);
913		if (LIST_EMPTY(&root_holds)) {
914			mtx_unlock(&root_holds_mtx);
915			break;
916		}
917		if (ppsratecheck(&lastfail, &curfail, 1)) {
918			printf("Root mount waiting for:");
919			LIST_FOREACH(h, &root_holds, list)
920				printf(" %s", h->who);
921			printf("\n");
922		}
923		msleep(&root_holds, &root_holds_mtx, PZERO | PDROP, "roothold",
924		    hz);
925	}
926}
927
928void
929vfs_mountroot(void)
930{
931	struct mount *mp;
932	struct sbuf *sb;
933	struct thread *td;
934	time_t timebase;
935	int error;
936
937	td = curthread;
938
939	vfs_mountroot_wait();
940
941	sb = sbuf_new_auto();
942	vfs_mountroot_conf0(sb);
943	sbuf_finish(sb);
944
945	error = vfs_mountroot_devfs(td, &mp);
946	while (!error) {
947		error = vfs_mountroot_parse(sb, mp);
948		if (!error) {
949			vfs_mountroot_shuffle(td, mp);
950			sbuf_clear(sb);
951			error = vfs_mountroot_readconf(td, sb);
952			sbuf_finish(sb);
953		}
954	}
955
956	sbuf_delete(sb);
957
958	/*
959	 * Iterate over all currently mounted file systems and use
960	 * the time stamp found to check and/or initialize the RTC.
961	 * Call inittodr() only once and pass it the largest of the
962	 * timestamps we encounter.
963	 */
964	timebase = 0;
965	mtx_lock(&mountlist_mtx);
966	mp = TAILQ_FIRST(&mountlist);
967	while (mp != NULL) {
968		if (mp->mnt_time > timebase)
969			timebase = mp->mnt_time;
970		mp = TAILQ_NEXT(mp, mnt_list);
971	}
972	mtx_unlock(&mountlist_mtx);
973	inittodr(timebase);
974
975	/* Keep prison0's root in sync with the global rootvnode. */
976	mtx_lock(&prison0.pr_mtx);
977	prison0.pr_root = rootvnode;
978	vref(prison0.pr_root);
979	mtx_unlock(&prison0.pr_mtx);
980
981	mtx_lock(&root_holds_mtx);
982	atomic_store_rel_int(&root_mount_complete, 1);
983	wakeup(&root_mount_complete);
984	mtx_unlock(&root_holds_mtx);
985
986	EVENTHANDLER_INVOKE(mountroot);
987}
988
989static struct mntarg *
990parse_mountroot_options(struct mntarg *ma, const char *options)
991{
992	char *p;
993	char *name, *name_arg;
994	char *val, *val_arg;
995	char *opts;
996
997	if (options == NULL || options[0] == '\0')
998		return (ma);
999
1000	p = opts = strdup(options, M_MOUNT);
1001	if (opts == NULL) {
1002		return (ma);
1003	}
1004
1005	while((name = strsep(&p, ",")) != NULL) {
1006		if (name[0] == '\0')
1007			break;
1008
1009		val = strchr(name, '=');
1010		if (val != NULL) {
1011			*val = '\0';
1012			++val;
1013		}
1014		if( strcmp(name, "rw") == 0 ||
1015		    strcmp(name, "noro") == 0) {
1016			/*
1017			 * The first time we mount the root file system,
1018			 * we need to mount 'ro', so We need to ignore
1019			 * 'rw' and 'noro' mount options.
1020			 */
1021			continue;
1022		}
1023		name_arg = strdup(name, M_MOUNT);
1024		val_arg = NULL;
1025		if (val != NULL)
1026			val_arg = strdup(val, M_MOUNT);
1027
1028		ma = mount_arg(ma, name_arg, val_arg,
1029		    (val_arg != NULL ? -1 : 0));
1030	}
1031	free(opts, M_MOUNT);
1032	return (ma);
1033}
1034