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 331276 2018-03-20 22:57:14Z ian $");
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 succeeds.
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
98/*
99 * Mount of the system's /dev.
100 */
101struct mount *rootdevmp;
102
103char *rootdevnames[2] = {NULL, NULL};
104
105struct mtx root_holds_mtx;
106MTX_SYSINIT(root_holds, &root_holds_mtx, "root_holds", MTX_DEF);
107
108struct root_hold_token {
109	const char			*who;
110	LIST_ENTRY(root_hold_token)	list;
111};
112
113static LIST_HEAD(, root_hold_token)	root_holds =
114    LIST_HEAD_INITIALIZER(root_holds);
115
116enum action {
117	A_CONTINUE,
118	A_PANIC,
119	A_REBOOT,
120	A_RETRY
121};
122
123static enum action root_mount_onfail = A_CONTINUE;
124
125static int root_mount_mddev;
126static int root_mount_complete;
127
128/* By default wait up to 3 seconds for devices to appear. */
129static int root_mount_timeout = 3;
130TUNABLE_INT("vfs.mountroot.timeout", &root_mount_timeout);
131
132struct root_hold_token *
133root_mount_hold(const char *identifier)
134{
135	struct root_hold_token *h;
136
137	if (root_mounted())
138		return (NULL);
139
140	h = malloc(sizeof *h, M_DEVBUF, M_ZERO | M_WAITOK);
141	h->who = identifier;
142	mtx_lock(&root_holds_mtx);
143	LIST_INSERT_HEAD(&root_holds, h, list);
144	mtx_unlock(&root_holds_mtx);
145	return (h);
146}
147
148void
149root_mount_rel(struct root_hold_token *h)
150{
151
152	if (h == NULL)
153		return;
154	mtx_lock(&root_holds_mtx);
155	LIST_REMOVE(h, list);
156	wakeup(&root_holds);
157	mtx_unlock(&root_holds_mtx);
158	free(h, M_DEVBUF);
159}
160
161int
162root_mounted(void)
163{
164
165	/* No mutex is acquired here because int stores are atomic. */
166	return (root_mount_complete);
167}
168
169void
170root_mount_wait(void)
171{
172
173	/*
174	 * Panic on an obvious deadlock - the function can't be called from
175	 * a thread which is doing the whole SYSINIT stuff.
176	 */
177	KASSERT(curthread->td_proc->p_pid != 0,
178	    ("root_mount_wait: cannot be called from the swapper thread"));
179	mtx_lock(&root_holds_mtx);
180	while (!root_mount_complete) {
181		msleep(&root_mount_complete, &root_holds_mtx, PZERO, "rootwait",
182		    hz);
183	}
184	mtx_unlock(&root_holds_mtx);
185}
186
187static void
188set_rootvnode(void)
189{
190	struct proc *p;
191
192	if (VFS_ROOT(TAILQ_FIRST(&mountlist), LK_EXCLUSIVE, &rootvnode))
193		panic("Cannot find root vnode");
194
195	VOP_UNLOCK(rootvnode, 0);
196
197	p = curthread->td_proc;
198	FILEDESC_XLOCK(p->p_fd);
199
200	if (p->p_fd->fd_cdir != NULL)
201		vrele(p->p_fd->fd_cdir);
202	p->p_fd->fd_cdir = rootvnode;
203	VREF(rootvnode);
204
205	if (p->p_fd->fd_rdir != NULL)
206		vrele(p->p_fd->fd_rdir);
207	p->p_fd->fd_rdir = rootvnode;
208	VREF(rootvnode);
209
210	FILEDESC_XUNLOCK(p->p_fd);
211}
212
213static int
214vfs_mountroot_devfs(struct thread *td, struct mount **mpp)
215{
216	struct vfsoptlist *opts;
217	struct vfsconf *vfsp;
218	struct mount *mp;
219	int error;
220
221	*mpp = NULL;
222
223	if (rootdevmp != NULL) {
224		/*
225		 * Already have /dev; this happens during rerooting.
226		 */
227		error = vfs_busy(rootdevmp, 0);
228		if (error != 0)
229			return (error);
230		*mpp = rootdevmp;
231	} else {
232		vfsp = vfs_byname("devfs");
233		KASSERT(vfsp != NULL, ("Could not find devfs by name"));
234		if (vfsp == NULL)
235			return (ENOENT);
236
237		mp = vfs_mount_alloc(NULLVP, vfsp, "/dev", td->td_ucred);
238
239		error = VFS_MOUNT(mp);
240		KASSERT(error == 0, ("VFS_MOUNT(devfs) failed %d", error));
241		if (error)
242			return (error);
243
244		opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
245		TAILQ_INIT(opts);
246		mp->mnt_opt = opts;
247
248		mtx_lock(&mountlist_mtx);
249		TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list);
250		mtx_unlock(&mountlist_mtx);
251
252		*mpp = mp;
253		rootdevmp = mp;
254	}
255
256	set_rootvnode();
257
258	error = kern_symlink(td, "/", "dev", UIO_SYSSPACE);
259	if (error)
260		printf("kern_symlink /dev -> / returns %d\n", error);
261
262	return (error);
263}
264
265static void
266vfs_mountroot_shuffle(struct thread *td, struct mount *mpdevfs)
267{
268	struct nameidata nd;
269	struct mount *mporoot, *mpnroot;
270	struct vnode *vp, *vporoot, *vpdevfs;
271	char *fspath;
272	int error;
273
274	mpnroot = TAILQ_NEXT(mpdevfs, mnt_list);
275
276	/* Shuffle the mountlist. */
277	mtx_lock(&mountlist_mtx);
278	mporoot = TAILQ_FIRST(&mountlist);
279	TAILQ_REMOVE(&mountlist, mpdevfs, mnt_list);
280	if (mporoot != mpdevfs) {
281		TAILQ_REMOVE(&mountlist, mpnroot, mnt_list);
282		TAILQ_INSERT_HEAD(&mountlist, mpnroot, mnt_list);
283	}
284	TAILQ_INSERT_TAIL(&mountlist, mpdevfs, mnt_list);
285	mtx_unlock(&mountlist_mtx);
286
287	cache_purgevfs(mporoot);
288	if (mporoot != mpdevfs)
289		cache_purgevfs(mpdevfs);
290
291	VFS_ROOT(mporoot, LK_EXCLUSIVE, &vporoot);
292
293	VI_LOCK(vporoot);
294	vporoot->v_iflag &= ~VI_MOUNT;
295	VI_UNLOCK(vporoot);
296	vporoot->v_mountedhere = NULL;
297	mporoot->mnt_flag &= ~MNT_ROOTFS;
298	mporoot->mnt_vnodecovered = NULL;
299	vput(vporoot);
300
301	/* Set up the new rootvnode, and purge the cache */
302	mpnroot->mnt_vnodecovered = NULL;
303	set_rootvnode();
304	cache_purgevfs(rootvnode->v_mount);
305
306	if (mporoot != mpdevfs) {
307		/* Remount old root under /.mount or /mnt */
308		fspath = "/.mount";
309		NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
310		    fspath, td);
311		error = namei(&nd);
312		if (error) {
313			NDFREE(&nd, NDF_ONLY_PNBUF);
314			fspath = "/mnt";
315			NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE,
316			    fspath, td);
317			error = namei(&nd);
318		}
319		if (!error) {
320			vp = nd.ni_vp;
321			error = (vp->v_type == VDIR) ? 0 : ENOTDIR;
322			if (!error)
323				error = vinvalbuf(vp, V_SAVE, 0, 0);
324			if (!error) {
325				cache_purge(vp);
326				mporoot->mnt_vnodecovered = vp;
327				vp->v_mountedhere = mporoot;
328				strlcpy(mporoot->mnt_stat.f_mntonname,
329				    fspath, MNAMELEN);
330				VOP_UNLOCK(vp, 0);
331			} else
332				vput(vp);
333		}
334		NDFREE(&nd, NDF_ONLY_PNBUF);
335
336		if (error && bootverbose)
337			printf("mountroot: unable to remount previous root "
338			    "under /.mount or /mnt (error %d).\n", error);
339	}
340
341	/* Remount devfs under /dev */
342	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, "/dev", td);
343	error = namei(&nd);
344	if (!error) {
345		vp = nd.ni_vp;
346		error = (vp->v_type == VDIR) ? 0 : ENOTDIR;
347		if (!error)
348			error = vinvalbuf(vp, V_SAVE, 0, 0);
349		if (!error) {
350			vpdevfs = mpdevfs->mnt_vnodecovered;
351			if (vpdevfs != NULL) {
352				cache_purge(vpdevfs);
353				vpdevfs->v_mountedhere = NULL;
354				vrele(vpdevfs);
355			}
356			mpdevfs->mnt_vnodecovered = vp;
357			vp->v_mountedhere = mpdevfs;
358			VOP_UNLOCK(vp, 0);
359		} else
360			vput(vp);
361	}
362	if (error && bootverbose)
363		printf("mountroot: unable to remount devfs under /dev "
364		    "(error %d).\n", error);
365	NDFREE(&nd, NDF_ONLY_PNBUF);
366
367	if (mporoot == mpdevfs) {
368		vfs_unbusy(mpdevfs);
369		/* Unlink the no longer needed /dev/dev -> / symlink */
370		error = kern_unlink(td, "/dev/dev", UIO_SYSSPACE);
371		if (error && bootverbose)
372			printf("mountroot: unable to unlink /dev/dev "
373			    "(error %d)\n", error);
374	}
375}
376
377/*
378 * Configuration parser.
379 */
380
381/* Parser character classes. */
382#define	CC_WHITESPACE		-1
383#define	CC_NONWHITESPACE	-2
384
385/* Parse errors. */
386#define	PE_EOF			-1
387#define	PE_EOL			-2
388
389static __inline int
390parse_peek(char **conf)
391{
392
393	return (**conf);
394}
395
396static __inline void
397parse_poke(char **conf, int c)
398{
399
400	**conf = c;
401}
402
403static __inline void
404parse_advance(char **conf)
405{
406
407	(*conf)++;
408}
409
410static int
411parse_skipto(char **conf, int mc)
412{
413	int c, match;
414
415	while (1) {
416		c = parse_peek(conf);
417		if (c == 0)
418			return (PE_EOF);
419		switch (mc) {
420		case CC_WHITESPACE:
421			match = (c == ' ' || c == '\t' || c == '\n') ? 1 : 0;
422			break;
423		case CC_NONWHITESPACE:
424			if (c == '\n')
425				return (PE_EOL);
426			match = (c != ' ' && c != '\t') ? 1 : 0;
427			break;
428		default:
429			match = (c == mc) ? 1 : 0;
430			break;
431		}
432		if (match)
433			break;
434		parse_advance(conf);
435	}
436	return (0);
437}
438
439static int
440parse_token(char **conf, char **tok)
441{
442	char *p;
443	size_t len;
444	int error;
445
446	*tok = NULL;
447	error = parse_skipto(conf, CC_NONWHITESPACE);
448	if (error)
449		return (error);
450	p = *conf;
451	error = parse_skipto(conf, CC_WHITESPACE);
452	len = *conf - p;
453	*tok = malloc(len + 1, M_TEMP, M_WAITOK | M_ZERO);
454	bcopy(p, *tok, len);
455	return (0);
456}
457
458static void
459parse_dir_ask_printenv(const char *var)
460{
461	char *val;
462
463	val = getenv(var);
464	if (val != NULL) {
465		printf("  %s=%s\n", var, val);
466		freeenv(val);
467	}
468}
469
470static int
471parse_dir_ask(char **conf)
472{
473	char name[80];
474	char *mnt;
475	int error;
476
477	printf("\nLoader variables:\n");
478	parse_dir_ask_printenv("vfs.root.mountfrom");
479	parse_dir_ask_printenv("vfs.root.mountfrom.options");
480
481	printf("\nManual root filesystem specification:\n");
482	printf("  <fstype>:<device> [options]\n");
483	printf("      Mount <device> using filesystem <fstype>\n");
484	printf("      and with the specified (optional) option list.\n");
485	printf("\n");
486	printf("    eg. ufs:/dev/da0s1a\n");
487	printf("        zfs:tank\n");
488	printf("        cd9660:/dev/acd0 ro\n");
489	printf("          (which is equivalent to: ");
490	printf("mount -t cd9660 -o ro /dev/acd0 /)\n");
491	printf("\n");
492	printf("  ?               List valid disk boot devices\n");
493	printf("  .               Yield 1 second (for background tasks)\n");
494	printf("  <empty line>    Abort manual input\n");
495
496	do {
497		error = EINVAL;
498		printf("\nmountroot> ");
499		cngets(name, sizeof(name), GETS_ECHO);
500		if (name[0] == '\0')
501			break;
502		if (name[0] == '?' && name[1] == '\0') {
503			printf("\nList of GEOM managed disk devices:\n  ");
504			g_dev_print();
505			continue;
506		}
507		if (name[0] == '.' && name[1] == '\0') {
508			pause("rmask", hz);
509			continue;
510		}
511		mnt = name;
512		error = parse_mount(&mnt);
513		if (error == -1)
514			printf("Invalid file system specification.\n");
515	} while (error != 0);
516
517	return (error);
518}
519
520static int
521parse_dir_md(char **conf)
522{
523	struct stat sb;
524	struct thread *td;
525	struct md_ioctl *mdio;
526	char *path, *tok;
527	int error, fd, len;
528
529	td = curthread;
530
531	error = parse_token(conf, &tok);
532	if (error)
533		return (error);
534
535	len = strlen(tok);
536	mdio = malloc(sizeof(*mdio) + len + 1, M_TEMP, M_WAITOK | M_ZERO);
537	path = (void *)(mdio + 1);
538	bcopy(tok, path, len);
539	free(tok, M_TEMP);
540
541	/* Get file status. */
542	error = kern_stat(td, path, UIO_SYSSPACE, &sb);
543	if (error)
544		goto out;
545
546	/* Open /dev/mdctl so that we can attach/detach. */
547	error = kern_open(td, "/dev/" MDCTL_NAME, UIO_SYSSPACE, O_RDWR, 0);
548	if (error)
549		goto out;
550
551	fd = td->td_retval[0];
552	mdio->md_version = MDIOVERSION;
553	mdio->md_type = MD_VNODE;
554
555	if (root_mount_mddev != -1) {
556		mdio->md_unit = root_mount_mddev;
557		DROP_GIANT();
558		error = kern_ioctl(td, fd, MDIOCDETACH, (void *)mdio);
559		PICKUP_GIANT();
560		/* Ignore errors. We don't care. */
561		root_mount_mddev = -1;
562	}
563
564	mdio->md_file = (void *)(mdio + 1);
565	mdio->md_options = MD_AUTOUNIT | MD_READONLY;
566	mdio->md_mediasize = sb.st_size;
567	mdio->md_unit = 0;
568	DROP_GIANT();
569	error = kern_ioctl(td, fd, MDIOCATTACH, (void *)mdio);
570	PICKUP_GIANT();
571	if (error)
572		goto out;
573
574	if (mdio->md_unit > 9) {
575		printf("rootmount: too many md units\n");
576		mdio->md_file = NULL;
577		mdio->md_options = 0;
578		mdio->md_mediasize = 0;
579		DROP_GIANT();
580		error = kern_ioctl(td, fd, MDIOCDETACH, (void *)mdio);
581		PICKUP_GIANT();
582		/* Ignore errors. We don't care. */
583		error = ERANGE;
584		goto out;
585	}
586
587	root_mount_mddev = mdio->md_unit;
588	printf(MD_NAME "%u attached to %s\n", root_mount_mddev, mdio->md_file);
589
590	error = kern_close(td, fd);
591
592 out:
593	free(mdio, M_TEMP);
594	return (error);
595}
596
597static int
598parse_dir_onfail(char **conf)
599{
600	char *action;
601	int error;
602
603	error = parse_token(conf, &action);
604	if (error)
605		return (error);
606
607	if (!strcmp(action, "continue"))
608		root_mount_onfail = A_CONTINUE;
609	else if (!strcmp(action, "panic"))
610		root_mount_onfail = A_PANIC;
611	else if (!strcmp(action, "reboot"))
612		root_mount_onfail = A_REBOOT;
613	else if (!strcmp(action, "retry"))
614		root_mount_onfail = A_RETRY;
615	else {
616		printf("rootmount: %s: unknown action\n", action);
617		error = EINVAL;
618	}
619
620	free(action, M_TEMP);
621	return (0);
622}
623
624static int
625parse_dir_timeout(char **conf)
626{
627	char *tok, *endtok;
628	long secs;
629	int error;
630
631	error = parse_token(conf, &tok);
632	if (error)
633		return (error);
634
635	secs = strtol(tok, &endtok, 0);
636	error = (secs < 0 || *endtok != '\0') ? EINVAL : 0;
637	if (!error)
638		root_mount_timeout = secs;
639	free(tok, M_TEMP);
640	return (error);
641}
642
643static int
644parse_directive(char **conf)
645{
646	char *dir;
647	int error;
648
649	error = parse_token(conf, &dir);
650	if (error)
651		return (error);
652
653	if (strcmp(dir, ".ask") == 0)
654		error = parse_dir_ask(conf);
655	else if (strcmp(dir, ".md") == 0)
656		error = parse_dir_md(conf);
657	else if (strcmp(dir, ".onfail") == 0)
658		error = parse_dir_onfail(conf);
659	else if (strcmp(dir, ".timeout") == 0)
660		error = parse_dir_timeout(conf);
661	else {
662		printf("mountroot: invalid directive `%s'\n", dir);
663		/* Ignore the rest of the line. */
664		(void)parse_skipto(conf, '\n');
665		error = EINVAL;
666	}
667	free(dir, M_TEMP);
668	return (error);
669}
670
671static int
672parse_mount_dev_present(const char *dev)
673{
674	struct nameidata nd;
675	int error;
676
677	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, dev, curthread);
678	error = namei(&nd);
679	if (!error)
680		vput(nd.ni_vp);
681	NDFREE(&nd, NDF_ONLY_PNBUF);
682	return (error != 0) ? 0 : 1;
683}
684
685#define	ERRMSGL	255
686static int
687parse_mount(char **conf)
688{
689	char *errmsg;
690	struct mntarg *ma;
691	char *dev, *fs, *opts, *tok;
692	int delay, error, timeout;
693
694	error = parse_token(conf, &tok);
695	if (error)
696		return (error);
697	fs = tok;
698	error = parse_skipto(&tok, ':');
699	if (error) {
700		free(fs, M_TEMP);
701		return (error);
702	}
703	parse_poke(&tok, '\0');
704	parse_advance(&tok);
705	dev = tok;
706
707	if (root_mount_mddev != -1) {
708		/* Handle substitution for the md unit number. */
709		tok = strstr(dev, "md#");
710		if (tok != NULL)
711			tok[2] = '0' + root_mount_mddev;
712	}
713
714	/* Parse options. */
715	error = parse_token(conf, &tok);
716	opts = (error == 0) ? tok : NULL;
717
718	printf("Trying to mount root from %s:%s [%s]...\n", fs, dev,
719	    (opts != NULL) ? opts : "");
720
721	errmsg = malloc(ERRMSGL, M_TEMP, M_WAITOK | M_ZERO);
722
723	if (vfs_byname(fs) == NULL) {
724		strlcpy(errmsg, "unknown file system", ERRMSGL);
725		error = ENOENT;
726		goto out;
727	}
728
729	if (strcmp(fs, "zfs") != 0 && strstr(fs, "nfs") == NULL &&
730	    dev[0] != '\0' && !parse_mount_dev_present(dev)) {
731		printf("mountroot: waiting for device %s ...\n", dev);
732		delay = hz / 10;
733		timeout = root_mount_timeout * hz;
734		do {
735			pause("rmdev", delay);
736			timeout -= delay;
737		} while (timeout > 0 && !parse_mount_dev_present(dev));
738		if (timeout <= 0) {
739			error = ENODEV;
740			goto out;
741		}
742	}
743
744	delay = hz / 10;
745	timeout = root_mount_timeout * hz;
746
747	for (;;) {
748		ma = NULL;
749		ma = mount_arg(ma, "fstype", fs, -1);
750		ma = mount_arg(ma, "fspath", "/", -1);
751		ma = mount_arg(ma, "from", dev, -1);
752		ma = mount_arg(ma, "errmsg", errmsg, ERRMSGL);
753		ma = mount_arg(ma, "ro", NULL, 0);
754		ma = parse_mountroot_options(ma, opts);
755
756		error = kernel_mount(ma, MNT_ROOTFS);
757		if (error == 0 || timeout <= 0)
758			break;
759
760		if (root_mount_timeout * hz == timeout ||
761		    (bootverbose && timeout % hz == 0)) {
762			printf("Mounting from %s:%s failed with error %d; "
763			    "retrying for %d more second%s\n", fs, dev, error,
764			    timeout / hz, (timeout / hz > 1) ? "s" : "");
765		}
766		pause("rmretry", delay);
767		timeout -= delay;
768	}
769 out:
770	if (error) {
771		printf("Mounting from %s:%s failed with error %d",
772		    fs, dev, error);
773		if (errmsg[0] != '\0')
774			printf(": %s", errmsg);
775		printf(".\n");
776	}
777	free(fs, M_TEMP);
778	free(errmsg, M_TEMP);
779	if (opts != NULL)
780		free(opts, M_TEMP);
781	/* kernel_mount can return -1 on error. */
782	return ((error < 0) ? EDOOFUS : error);
783}
784#undef ERRMSGL
785
786static int
787vfs_mountroot_parse(struct sbuf *sb, struct mount *mpdevfs)
788{
789	struct mount *mp;
790	char *conf;
791	int error;
792
793	root_mount_mddev = -1;
794
795retry:
796	conf = sbuf_data(sb);
797	mp = TAILQ_NEXT(mpdevfs, mnt_list);
798	error = (mp == NULL) ? 0 : EDOOFUS;
799	root_mount_onfail = A_CONTINUE;
800	while (mp == NULL) {
801		error = parse_skipto(&conf, CC_NONWHITESPACE);
802		if (error == PE_EOL) {
803			parse_advance(&conf);
804			continue;
805		}
806		if (error < 0)
807			break;
808		switch (parse_peek(&conf)) {
809		case '#':
810			error = parse_skipto(&conf, '\n');
811			break;
812		case '.':
813			error = parse_directive(&conf);
814			break;
815		default:
816			error = parse_mount(&conf);
817			break;
818		}
819		if (error < 0)
820			break;
821		/* Ignore any trailing garbage on the line. */
822		if (parse_peek(&conf) != '\n') {
823			printf("mountroot: advancing to next directive...\n");
824			(void)parse_skipto(&conf, '\n');
825		}
826		mp = TAILQ_NEXT(mpdevfs, mnt_list);
827	}
828	if (mp != NULL)
829		return (0);
830
831	/*
832	 * We failed to mount (a new) root.
833	 */
834	switch (root_mount_onfail) {
835	case A_CONTINUE:
836		break;
837	case A_PANIC:
838		panic("mountroot: unable to (re-)mount root.");
839		/* NOTREACHED */
840	case A_RETRY:
841		goto retry;
842	case A_REBOOT:
843		kern_reboot(RB_NOSYNC);
844		/* NOTREACHED */
845	}
846
847	return (error);
848}
849
850static void
851vfs_mountroot_conf0(struct sbuf *sb)
852{
853	char *s, *tok, *mnt, *opt;
854	int error;
855
856	sbuf_printf(sb, ".onfail panic\n");
857	sbuf_printf(sb, ".timeout %d\n", root_mount_timeout);
858	if (boothowto & RB_ASKNAME)
859		sbuf_printf(sb, ".ask\n");
860#ifdef ROOTDEVNAME
861	if (boothowto & RB_DFLTROOT)
862		sbuf_printf(sb, "%s\n", ROOTDEVNAME);
863#endif
864	if (boothowto & RB_CDROM) {
865		sbuf_printf(sb, "cd9660:/dev/cd0 ro\n");
866		sbuf_printf(sb, ".timeout 0\n");
867		sbuf_printf(sb, "cd9660:/dev/acd0 ro\n");
868		sbuf_printf(sb, ".timeout %d\n", root_mount_timeout);
869	}
870	s = getenv("vfs.root.mountfrom");
871	if (s != NULL) {
872		opt = getenv("vfs.root.mountfrom.options");
873		tok = s;
874		error = parse_token(&tok, &mnt);
875		while (!error) {
876			sbuf_printf(sb, "%s %s\n", mnt,
877			    (opt != NULL) ? opt : "");
878			free(mnt, M_TEMP);
879			error = parse_token(&tok, &mnt);
880		}
881		if (opt != NULL)
882			freeenv(opt);
883		freeenv(s);
884	}
885	if (rootdevnames[0] != NULL)
886		sbuf_printf(sb, "%s\n", rootdevnames[0]);
887	if (rootdevnames[1] != NULL)
888		sbuf_printf(sb, "%s\n", rootdevnames[1]);
889#ifdef ROOTDEVNAME
890	if (!(boothowto & RB_DFLTROOT))
891		sbuf_printf(sb, "%s\n", ROOTDEVNAME);
892#endif
893	if (!(boothowto & RB_ASKNAME))
894		sbuf_printf(sb, ".ask\n");
895}
896
897static int
898vfs_mountroot_readconf(struct thread *td, struct sbuf *sb)
899{
900	static char buf[128];
901	struct nameidata nd;
902	off_t ofs;
903	ssize_t resid;
904	int error, flags, len;
905
906	NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, "/.mount.conf", td);
907	flags = FREAD;
908	error = vn_open(&nd, &flags, 0, NULL);
909	if (error)
910		return (error);
911
912	NDFREE(&nd, NDF_ONLY_PNBUF);
913	ofs = 0;
914	len = sizeof(buf) - 1;
915	while (1) {
916		error = vn_rdwr(UIO_READ, nd.ni_vp, buf, len, ofs,
917		    UIO_SYSSPACE, IO_NODELOCKED, td->td_ucred,
918		    NOCRED, &resid, td);
919		if (error)
920			break;
921		if (resid == len)
922			break;
923		buf[len - resid] = 0;
924		sbuf_printf(sb, "%s", buf);
925		ofs += len - resid;
926	}
927
928	VOP_UNLOCK(nd.ni_vp, 0);
929	vn_close(nd.ni_vp, FREAD, td->td_ucred, td);
930	return (error);
931}
932
933static void
934vfs_mountroot_wait(void)
935{
936	struct root_hold_token *h;
937	struct timeval lastfail;
938	int curfail;
939
940	curfail = 0;
941	while (1) {
942		DROP_GIANT();
943		g_waitidle();
944		PICKUP_GIANT();
945		mtx_lock(&root_holds_mtx);
946		if (LIST_EMPTY(&root_holds)) {
947			mtx_unlock(&root_holds_mtx);
948			break;
949		}
950		if (ppsratecheck(&lastfail, &curfail, 1)) {
951			printf("Root mount waiting for:");
952			LIST_FOREACH(h, &root_holds, list)
953				printf(" %s", h->who);
954			printf("\n");
955		}
956		msleep(&root_holds, &root_holds_mtx, PZERO | PDROP, "roothold",
957		    hz);
958	}
959}
960
961void
962vfs_mountroot(void)
963{
964	struct mount *mp;
965	struct sbuf *sb;
966	struct thread *td;
967	time_t timebase;
968	int error;
969
970	td = curthread;
971
972	vfs_mountroot_wait();
973
974	sb = sbuf_new_auto();
975	vfs_mountroot_conf0(sb);
976	sbuf_finish(sb);
977
978	error = vfs_mountroot_devfs(td, &mp);
979	while (!error) {
980		error = vfs_mountroot_parse(sb, mp);
981		if (!error) {
982			vfs_mountroot_shuffle(td, mp);
983			sbuf_clear(sb);
984			error = vfs_mountroot_readconf(td, sb);
985			sbuf_finish(sb);
986		}
987	}
988
989	sbuf_delete(sb);
990
991	/*
992	 * Iterate over all currently mounted file systems and use
993	 * the time stamp found to check and/or initialize the RTC.
994	 * Call inittodr() only once and pass it the largest of the
995	 * timestamps we encounter.
996	 */
997	timebase = 0;
998	mtx_lock(&mountlist_mtx);
999	mp = TAILQ_FIRST(&mountlist);
1000	while (mp != NULL) {
1001		if (mp->mnt_time > timebase)
1002			timebase = mp->mnt_time;
1003		mp = TAILQ_NEXT(mp, mnt_list);
1004	}
1005	mtx_unlock(&mountlist_mtx);
1006	inittodr(timebase);
1007
1008	/* Keep prison0's root in sync with the global rootvnode. */
1009	mtx_lock(&prison0.pr_mtx);
1010	prison0.pr_root = rootvnode;
1011	vref(prison0.pr_root);
1012	mtx_unlock(&prison0.pr_mtx);
1013
1014	mtx_lock(&root_holds_mtx);
1015	atomic_store_rel_int(&root_mount_complete, 1);
1016	wakeup(&root_mount_complete);
1017	mtx_unlock(&root_holds_mtx);
1018
1019	EVENTHANDLER_INVOKE(mountroot);
1020}
1021
1022static struct mntarg *
1023parse_mountroot_options(struct mntarg *ma, const char *options)
1024{
1025	char *p;
1026	char *name, *name_arg;
1027	char *val, *val_arg;
1028	char *opts;
1029
1030	if (options == NULL || options[0] == '\0')
1031		return (ma);
1032
1033	p = opts = strdup(options, M_MOUNT);
1034	if (opts == NULL) {
1035		return (ma);
1036	}
1037
1038	while((name = strsep(&p, ",")) != NULL) {
1039		if (name[0] == '\0')
1040			break;
1041
1042		val = strchr(name, '=');
1043		if (val != NULL) {
1044			*val = '\0';
1045			++val;
1046		}
1047		if( strcmp(name, "rw") == 0 ||
1048		    strcmp(name, "noro") == 0) {
1049			/*
1050			 * The first time we mount the root file system,
1051			 * we need to mount 'ro', so We need to ignore
1052			 * 'rw' and 'noro' mount options.
1053			 */
1054			continue;
1055		}
1056		name_arg = strdup(name, M_MOUNT);
1057		val_arg = NULL;
1058		if (val != NULL)
1059			val_arg = strdup(val, M_MOUNT);
1060
1061		ma = mount_arg(ma, name_arg, val_arg,
1062		    (val_arg != NULL ? -1 : 0));
1063	}
1064	free(opts, M_MOUNT);
1065	return (ma);
1066}
1067