xref: /openbmc/linux/init/do_mounts.c (revision c8ec3743)
1 #include <linux/module.h>
2 #include <linux/sched.h>
3 #include <linux/ctype.h>
4 #include <linux/fd.h>
5 #include <linux/tty.h>
6 #include <linux/suspend.h>
7 #include <linux/root_dev.h>
8 #include <linux/security.h>
9 #include <linux/delay.h>
10 #include <linux/genhd.h>
11 #include <linux/mount.h>
12 #include <linux/device.h>
13 #include <linux/init.h>
14 #include <linux/fs.h>
15 #include <linux/initrd.h>
16 #include <linux/async.h>
17 #include <linux/fs_struct.h>
18 #include <linux/slab.h>
19 #include <linux/ramfs.h>
20 #include <linux/shmem_fs.h>
21 
22 #include <linux/nfs_fs.h>
23 #include <linux/nfs_fs_sb.h>
24 #include <linux/nfs_mount.h>
25 #include <uapi/linux/mount.h>
26 
27 #include "do_mounts.h"
28 
29 int __initdata rd_doload;	/* 1 = load RAM disk, 0 = don't load */
30 
31 int root_mountflags = MS_RDONLY | MS_SILENT;
32 static char * __initdata root_device_name;
33 static char __initdata saved_root_name[64];
34 static int root_wait;
35 
36 dev_t ROOT_DEV;
37 
38 static int __init load_ramdisk(char *str)
39 {
40 	rd_doload = simple_strtol(str,NULL,0) & 3;
41 	return 1;
42 }
43 __setup("load_ramdisk=", load_ramdisk);
44 
45 static int __init readonly(char *str)
46 {
47 	if (*str)
48 		return 0;
49 	root_mountflags |= MS_RDONLY;
50 	return 1;
51 }
52 
53 static int __init readwrite(char *str)
54 {
55 	if (*str)
56 		return 0;
57 	root_mountflags &= ~MS_RDONLY;
58 	return 1;
59 }
60 
61 __setup("ro", readonly);
62 __setup("rw", readwrite);
63 
64 #ifdef CONFIG_BLOCK
65 struct uuidcmp {
66 	const char *uuid;
67 	int len;
68 };
69 
70 /**
71  * match_dev_by_uuid - callback for finding a partition using its uuid
72  * @dev:	device passed in by the caller
73  * @data:	opaque pointer to the desired struct uuidcmp to match
74  *
75  * Returns 1 if the device matches, and 0 otherwise.
76  */
77 static int match_dev_by_uuid(struct device *dev, const void *data)
78 {
79 	const struct uuidcmp *cmp = data;
80 	struct hd_struct *part = dev_to_part(dev);
81 
82 	if (!part->info)
83 		goto no_match;
84 
85 	if (strncasecmp(cmp->uuid, part->info->uuid, cmp->len))
86 		goto no_match;
87 
88 	return 1;
89 no_match:
90 	return 0;
91 }
92 
93 
94 /**
95  * devt_from_partuuid - looks up the dev_t of a partition by its UUID
96  * @uuid_str:	char array containing ascii UUID
97  *
98  * The function will return the first partition which contains a matching
99  * UUID value in its partition_meta_info struct.  This does not search
100  * by filesystem UUIDs.
101  *
102  * If @uuid_str is followed by a "/PARTNROFF=%d", then the number will be
103  * extracted and used as an offset from the partition identified by the UUID.
104  *
105  * Returns the matching dev_t on success or 0 on failure.
106  */
107 static dev_t devt_from_partuuid(const char *uuid_str)
108 {
109 	dev_t res = 0;
110 	struct uuidcmp cmp;
111 	struct device *dev = NULL;
112 	struct gendisk *disk;
113 	struct hd_struct *part;
114 	int offset = 0;
115 	bool clear_root_wait = false;
116 	char *slash;
117 
118 	cmp.uuid = uuid_str;
119 
120 	slash = strchr(uuid_str, '/');
121 	/* Check for optional partition number offset attributes. */
122 	if (slash) {
123 		char c = 0;
124 		/* Explicitly fail on poor PARTUUID syntax. */
125 		if (sscanf(slash + 1,
126 			   "PARTNROFF=%d%c", &offset, &c) != 1) {
127 			clear_root_wait = true;
128 			goto done;
129 		}
130 		cmp.len = slash - uuid_str;
131 	} else {
132 		cmp.len = strlen(uuid_str);
133 	}
134 
135 	if (!cmp.len) {
136 		clear_root_wait = true;
137 		goto done;
138 	}
139 
140 	dev = class_find_device(&block_class, NULL, &cmp,
141 				&match_dev_by_uuid);
142 	if (!dev)
143 		goto done;
144 
145 	res = dev->devt;
146 
147 	/* Attempt to find the partition by offset. */
148 	if (!offset)
149 		goto no_offset;
150 
151 	res = 0;
152 	disk = part_to_disk(dev_to_part(dev));
153 	part = disk_get_part(disk, dev_to_part(dev)->partno + offset);
154 	if (part) {
155 		res = part_devt(part);
156 		put_device(part_to_dev(part));
157 	}
158 
159 no_offset:
160 	put_device(dev);
161 done:
162 	if (clear_root_wait) {
163 		pr_err("VFS: PARTUUID= is invalid.\n"
164 		       "Expected PARTUUID=<valid-uuid-id>[/PARTNROFF=%%d]\n");
165 		if (root_wait)
166 			pr_err("Disabling rootwait; root= is invalid.\n");
167 		root_wait = 0;
168 	}
169 	return res;
170 }
171 
172 /**
173  * match_dev_by_label - callback for finding a partition using its label
174  * @dev:	device passed in by the caller
175  * @data:	opaque pointer to the label to match
176  *
177  * Returns 1 if the device matches, and 0 otherwise.
178  */
179 static int match_dev_by_label(struct device *dev, const void *data)
180 {
181 	const char *label = data;
182 	struct hd_struct *part = dev_to_part(dev);
183 
184 	if (part->info && !strcmp(label, part->info->volname))
185 		return 1;
186 
187 	return 0;
188 }
189 #endif
190 
191 /*
192  *	Convert a name into device number.  We accept the following variants:
193  *
194  *	1) <hex_major><hex_minor> device number in hexadecimal represents itself
195  *         no leading 0x, for example b302.
196  *	2) /dev/nfs represents Root_NFS (0xff)
197  *	3) /dev/<disk_name> represents the device number of disk
198  *	4) /dev/<disk_name><decimal> represents the device number
199  *         of partition - device number of disk plus the partition number
200  *	5) /dev/<disk_name>p<decimal> - same as the above, that form is
201  *	   used when disk name of partitioned disk ends on a digit.
202  *	6) PARTUUID=00112233-4455-6677-8899-AABBCCDDEEFF representing the
203  *	   unique id of a partition if the partition table provides it.
204  *	   The UUID may be either an EFI/GPT UUID, or refer to an MSDOS
205  *	   partition using the format SSSSSSSS-PP, where SSSSSSSS is a zero-
206  *	   filled hex representation of the 32-bit "NT disk signature", and PP
207  *	   is a zero-filled hex representation of the 1-based partition number.
208  *	7) PARTUUID=<UUID>/PARTNROFF=<int> to select a partition in relation to
209  *	   a partition with a known unique id.
210  *	8) <major>:<minor> major and minor number of the device separated by
211  *	   a colon.
212  *	9) PARTLABEL=<name> with name being the GPT partition label.
213  *	   MSDOS partitions do not support labels!
214  *
215  *	If name doesn't have fall into the categories above, we return (0,0).
216  *	block_class is used to check if something is a disk name. If the disk
217  *	name contains slashes, the device name has them replaced with
218  *	bangs.
219  */
220 
221 dev_t name_to_dev_t(const char *name)
222 {
223 	char s[32];
224 	char *p;
225 	dev_t res = 0;
226 	int part;
227 
228 #ifdef CONFIG_BLOCK
229 	if (strncmp(name, "PARTUUID=", 9) == 0) {
230 		name += 9;
231 		res = devt_from_partuuid(name);
232 		if (!res)
233 			goto fail;
234 		goto done;
235 	} else if (strncmp(name, "PARTLABEL=", 10) == 0) {
236 		struct device *dev;
237 
238 		dev = class_find_device(&block_class, NULL, name + 10,
239 					&match_dev_by_label);
240 		if (!dev)
241 			goto fail;
242 
243 		res = dev->devt;
244 		put_device(dev);
245 		goto done;
246 	}
247 #endif
248 
249 	if (strncmp(name, "/dev/", 5) != 0) {
250 		unsigned maj, min, offset;
251 		char dummy;
252 
253 		if ((sscanf(name, "%u:%u%c", &maj, &min, &dummy) == 2) ||
254 		    (sscanf(name, "%u:%u:%u:%c", &maj, &min, &offset, &dummy) == 3)) {
255 			res = MKDEV(maj, min);
256 			if (maj != MAJOR(res) || min != MINOR(res))
257 				goto fail;
258 		} else {
259 			res = new_decode_dev(simple_strtoul(name, &p, 16));
260 			if (*p)
261 				goto fail;
262 		}
263 		goto done;
264 	}
265 
266 	name += 5;
267 	res = Root_NFS;
268 	if (strcmp(name, "nfs") == 0)
269 		goto done;
270 	res = Root_RAM0;
271 	if (strcmp(name, "ram") == 0)
272 		goto done;
273 
274 	if (strlen(name) > 31)
275 		goto fail;
276 	strcpy(s, name);
277 	for (p = s; *p; p++)
278 		if (*p == '/')
279 			*p = '!';
280 	res = blk_lookup_devt(s, 0);
281 	if (res)
282 		goto done;
283 
284 	/*
285 	 * try non-existent, but valid partition, which may only exist
286 	 * after revalidating the disk, like partitioned md devices
287 	 */
288 	while (p > s && isdigit(p[-1]))
289 		p--;
290 	if (p == s || !*p || *p == '0')
291 		goto fail;
292 
293 	/* try disk name without <part number> */
294 	part = simple_strtoul(p, NULL, 10);
295 	*p = '\0';
296 	res = blk_lookup_devt(s, part);
297 	if (res)
298 		goto done;
299 
300 	/* try disk name without p<part number> */
301 	if (p < s + 2 || !isdigit(p[-2]) || p[-1] != 'p')
302 		goto fail;
303 	p[-1] = '\0';
304 	res = blk_lookup_devt(s, part);
305 	if (res)
306 		goto done;
307 
308 fail:
309 	return 0;
310 done:
311 	return res;
312 }
313 EXPORT_SYMBOL_GPL(name_to_dev_t);
314 
315 static int __init root_dev_setup(char *line)
316 {
317 	strlcpy(saved_root_name, line, sizeof(saved_root_name));
318 	return 1;
319 }
320 
321 __setup("root=", root_dev_setup);
322 
323 static int __init rootwait_setup(char *str)
324 {
325 	if (*str)
326 		return 0;
327 	root_wait = 1;
328 	return 1;
329 }
330 
331 __setup("rootwait", rootwait_setup);
332 
333 static char * __initdata root_mount_data;
334 static int __init root_data_setup(char *str)
335 {
336 	root_mount_data = str;
337 	return 1;
338 }
339 
340 static char * __initdata root_fs_names;
341 static int __init fs_names_setup(char *str)
342 {
343 	root_fs_names = str;
344 	return 1;
345 }
346 
347 static unsigned int __initdata root_delay;
348 static int __init root_delay_setup(char *str)
349 {
350 	root_delay = simple_strtoul(str, NULL, 0);
351 	return 1;
352 }
353 
354 __setup("rootflags=", root_data_setup);
355 __setup("rootfstype=", fs_names_setup);
356 __setup("rootdelay=", root_delay_setup);
357 
358 static void __init get_fs_names(char *page)
359 {
360 	char *s = page;
361 
362 	if (root_fs_names) {
363 		strcpy(page, root_fs_names);
364 		while (*s++) {
365 			if (s[-1] == ',')
366 				s[-1] = '\0';
367 		}
368 	} else {
369 		int len = get_filesystem_list(page);
370 		char *p, *next;
371 
372 		page[len] = '\0';
373 		for (p = page-1; p; p = next) {
374 			next = strchr(++p, '\n');
375 			if (*p++ != '\t')
376 				continue;
377 			while ((*s++ = *p++) != '\n')
378 				;
379 			s[-1] = '\0';
380 		}
381 	}
382 	*s = '\0';
383 }
384 
385 static int __init do_mount_root(char *name, char *fs, int flags, void *data)
386 {
387 	struct super_block *s;
388 	int err = ksys_mount(name, "/root", fs, flags, data);
389 	if (err)
390 		return err;
391 
392 	ksys_chdir("/root");
393 	s = current->fs->pwd.dentry->d_sb;
394 	ROOT_DEV = s->s_dev;
395 	printk(KERN_INFO
396 	       "VFS: Mounted root (%s filesystem)%s on device %u:%u.\n",
397 	       s->s_type->name,
398 	       sb_rdonly(s) ? " readonly" : "",
399 	       MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
400 	return 0;
401 }
402 
403 void __init mount_block_root(char *name, int flags)
404 {
405 	struct page *page = alloc_page(GFP_KERNEL);
406 	char *fs_names = page_address(page);
407 	char *p;
408 #ifdef CONFIG_BLOCK
409 	char b[BDEVNAME_SIZE];
410 #else
411 	const char *b = name;
412 #endif
413 
414 	get_fs_names(fs_names);
415 retry:
416 	for (p = fs_names; *p; p += strlen(p)+1) {
417 		int err = do_mount_root(name, p, flags, root_mount_data);
418 		switch (err) {
419 			case 0:
420 				goto out;
421 			case -EACCES:
422 			case -EINVAL:
423 				continue;
424 		}
425 	        /*
426 		 * Allow the user to distinguish between failed sys_open
427 		 * and bad superblock on root device.
428 		 * and give them a list of the available devices
429 		 */
430 #ifdef CONFIG_BLOCK
431 		__bdevname(ROOT_DEV, b);
432 #endif
433 		printk("VFS: Cannot open root device \"%s\" or %s: error %d\n",
434 				root_device_name, b, err);
435 		printk("Please append a correct \"root=\" boot option; here are the available partitions:\n");
436 
437 		printk_all_partitions();
438 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
439 		printk("DEBUG_BLOCK_EXT_DEVT is enabled, you need to specify "
440 		       "explicit textual name for \"root=\" boot option.\n");
441 #endif
442 		panic("VFS: Unable to mount root fs on %s", b);
443 	}
444 	if (!(flags & SB_RDONLY)) {
445 		flags |= SB_RDONLY;
446 		goto retry;
447 	}
448 
449 	printk("List of all partitions:\n");
450 	printk_all_partitions();
451 	printk("No filesystem could mount root, tried: ");
452 	for (p = fs_names; *p; p += strlen(p)+1)
453 		printk(" %s", p);
454 	printk("\n");
455 #ifdef CONFIG_BLOCK
456 	__bdevname(ROOT_DEV, b);
457 #endif
458 	panic("VFS: Unable to mount root fs on %s", b);
459 out:
460 	put_page(page);
461 }
462 
463 #ifdef CONFIG_ROOT_NFS
464 
465 #define NFSROOT_TIMEOUT_MIN	5
466 #define NFSROOT_TIMEOUT_MAX	30
467 #define NFSROOT_RETRY_MAX	5
468 
469 static int __init mount_nfs_root(void)
470 {
471 	char *root_dev, *root_data;
472 	unsigned int timeout;
473 	int try, err;
474 
475 	err = nfs_root_data(&root_dev, &root_data);
476 	if (err != 0)
477 		return 0;
478 
479 	/*
480 	 * The server or network may not be ready, so try several
481 	 * times.  Stop after a few tries in case the client wants
482 	 * to fall back to other boot methods.
483 	 */
484 	timeout = NFSROOT_TIMEOUT_MIN;
485 	for (try = 1; ; try++) {
486 		err = do_mount_root(root_dev, "nfs",
487 					root_mountflags, root_data);
488 		if (err == 0)
489 			return 1;
490 		if (try > NFSROOT_RETRY_MAX)
491 			break;
492 
493 		/* Wait, in case the server refused us immediately */
494 		ssleep(timeout);
495 		timeout <<= 1;
496 		if (timeout > NFSROOT_TIMEOUT_MAX)
497 			timeout = NFSROOT_TIMEOUT_MAX;
498 	}
499 	return 0;
500 }
501 #endif
502 
503 #if defined(CONFIG_BLK_DEV_RAM) || defined(CONFIG_BLK_DEV_FD)
504 void __init change_floppy(char *fmt, ...)
505 {
506 	struct termios termios;
507 	char buf[80];
508 	char c;
509 	int fd;
510 	va_list args;
511 	va_start(args, fmt);
512 	vsprintf(buf, fmt, args);
513 	va_end(args);
514 	fd = ksys_open("/dev/root", O_RDWR | O_NDELAY, 0);
515 	if (fd >= 0) {
516 		ksys_ioctl(fd, FDEJECT, 0);
517 		ksys_close(fd);
518 	}
519 	printk(KERN_NOTICE "VFS: Insert %s and press ENTER\n", buf);
520 	fd = ksys_open("/dev/console", O_RDWR, 0);
521 	if (fd >= 0) {
522 		ksys_ioctl(fd, TCGETS, (long)&termios);
523 		termios.c_lflag &= ~ICANON;
524 		ksys_ioctl(fd, TCSETSF, (long)&termios);
525 		ksys_read(fd, &c, 1);
526 		termios.c_lflag |= ICANON;
527 		ksys_ioctl(fd, TCSETSF, (long)&termios);
528 		ksys_close(fd);
529 	}
530 }
531 #endif
532 
533 void __init mount_root(void)
534 {
535 #ifdef CONFIG_ROOT_NFS
536 	if (ROOT_DEV == Root_NFS) {
537 		if (mount_nfs_root())
538 			return;
539 
540 		printk(KERN_ERR "VFS: Unable to mount root fs via NFS, trying floppy.\n");
541 		ROOT_DEV = Root_FD0;
542 	}
543 #endif
544 #ifdef CONFIG_BLK_DEV_FD
545 	if (MAJOR(ROOT_DEV) == FLOPPY_MAJOR) {
546 		/* rd_doload is 2 for a dual initrd/ramload setup */
547 		if (rd_doload==2) {
548 			if (rd_load_disk(1)) {
549 				ROOT_DEV = Root_RAM1;
550 				root_device_name = NULL;
551 			}
552 		} else
553 			change_floppy("root floppy");
554 	}
555 #endif
556 #ifdef CONFIG_BLOCK
557 	{
558 		int err = create_dev("/dev/root", ROOT_DEV);
559 
560 		if (err < 0)
561 			pr_emerg("Failed to create /dev/root: %d\n", err);
562 		mount_block_root("/dev/root", root_mountflags);
563 	}
564 #endif
565 }
566 
567 /*
568  * Prepare the namespace - decide what/where to mount, load ramdisks, etc.
569  */
570 void __init prepare_namespace(void)
571 {
572 	int is_floppy;
573 
574 	if (root_delay) {
575 		printk(KERN_INFO "Waiting %d sec before mounting root device...\n",
576 		       root_delay);
577 		ssleep(root_delay);
578 	}
579 
580 	/*
581 	 * wait for the known devices to complete their probing
582 	 *
583 	 * Note: this is a potential source of long boot delays.
584 	 * For example, it is not atypical to wait 5 seconds here
585 	 * for the touchpad of a laptop to initialize.
586 	 */
587 	wait_for_device_probe();
588 
589 	md_run_setup();
590 
591 	if (saved_root_name[0]) {
592 		root_device_name = saved_root_name;
593 		if (!strncmp(root_device_name, "mtd", 3) ||
594 		    !strncmp(root_device_name, "ubi", 3)) {
595 			mount_block_root(root_device_name, root_mountflags);
596 			goto out;
597 		}
598 		ROOT_DEV = name_to_dev_t(root_device_name);
599 		if (strncmp(root_device_name, "/dev/", 5) == 0)
600 			root_device_name += 5;
601 	}
602 
603 	if (initrd_load())
604 		goto out;
605 
606 	/* wait for any asynchronous scanning to complete */
607 	if ((ROOT_DEV == 0) && root_wait) {
608 		printk(KERN_INFO "Waiting for root device %s...\n",
609 			saved_root_name);
610 		while (driver_probe_done() != 0 ||
611 			(ROOT_DEV = name_to_dev_t(saved_root_name)) == 0)
612 			msleep(5);
613 		async_synchronize_full();
614 	}
615 
616 	is_floppy = MAJOR(ROOT_DEV) == FLOPPY_MAJOR;
617 
618 	if (is_floppy && rd_doload && rd_load_disk(0))
619 		ROOT_DEV = Root_RAM0;
620 
621 	mount_root();
622 out:
623 	devtmpfs_mount("dev");
624 	ksys_mount(".", "/", NULL, MS_MOVE, NULL);
625 	ksys_chroot(".");
626 }
627 
628 static bool is_tmpfs;
629 static struct dentry *rootfs_mount(struct file_system_type *fs_type,
630 	int flags, const char *dev_name, void *data)
631 {
632 	static unsigned long once;
633 	void *fill = ramfs_fill_super;
634 
635 	if (test_and_set_bit(0, &once))
636 		return ERR_PTR(-ENODEV);
637 
638 	if (IS_ENABLED(CONFIG_TMPFS) && is_tmpfs)
639 		fill = shmem_fill_super;
640 
641 	return mount_nodev(fs_type, flags, data, fill);
642 }
643 
644 static struct file_system_type rootfs_fs_type = {
645 	.name		= "rootfs",
646 	.mount		= rootfs_mount,
647 	.kill_sb	= kill_litter_super,
648 };
649 
650 int __init init_rootfs(void)
651 {
652 	int err = register_filesystem(&rootfs_fs_type);
653 
654 	if (err)
655 		return err;
656 
657 	if (IS_ENABLED(CONFIG_TMPFS) && !saved_root_name[0] &&
658 		(!root_fs_names || strstr(root_fs_names, "tmpfs"))) {
659 		err = shmem_init();
660 		is_tmpfs = true;
661 	} else {
662 		err = init_ramfs_fs();
663 	}
664 
665 	if (err)
666 		unregister_filesystem(&rootfs_fs_type);
667 
668 	return err;
669 }
670