xref: /openbmc/linux/init/do_mounts.c (revision d0bd7f2a)
1 /*
2  * Many of the syscalls used in this file expect some of the arguments
3  * to be __user pointers not __kernel pointers.  To limit the sparse
4  * noise, turn off sparse checking for this file.
5  */
6 #ifdef __CHECKER__
7 #undef __CHECKER__
8 #warning "Sparse checking disabled for this file"
9 #endif
10 
11 #include <linux/module.h>
12 #include <linux/sched.h>
13 #include <linux/ctype.h>
14 #include <linux/fd.h>
15 #include <linux/tty.h>
16 #include <linux/suspend.h>
17 #include <linux/root_dev.h>
18 #include <linux/security.h>
19 #include <linux/delay.h>
20 #include <linux/genhd.h>
21 #include <linux/mount.h>
22 #include <linux/device.h>
23 #include <linux/init.h>
24 #include <linux/fs.h>
25 #include <linux/initrd.h>
26 #include <linux/async.h>
27 #include <linux/fs_struct.h>
28 #include <linux/slab.h>
29 #include <linux/ramfs.h>
30 #include <linux/shmem_fs.h>
31 
32 #include <linux/nfs_fs.h>
33 #include <linux/nfs_fs_sb.h>
34 #include <linux/nfs_mount.h>
35 
36 #include "do_mounts.h"
37 
38 int __initdata rd_doload;	/* 1 = load RAM disk, 0 = don't load */
39 
40 int root_mountflags = MS_RDONLY | MS_SILENT;
41 static char * __initdata root_device_name;
42 static char __initdata saved_root_name[64];
43 static int root_wait;
44 
45 dev_t ROOT_DEV;
46 
47 static int __init load_ramdisk(char *str)
48 {
49 	rd_doload = simple_strtol(str,NULL,0) & 3;
50 	return 1;
51 }
52 __setup("load_ramdisk=", load_ramdisk);
53 
54 static int __init readonly(char *str)
55 {
56 	if (*str)
57 		return 0;
58 	root_mountflags |= MS_RDONLY;
59 	return 1;
60 }
61 
62 static int __init readwrite(char *str)
63 {
64 	if (*str)
65 		return 0;
66 	root_mountflags &= ~MS_RDONLY;
67 	return 1;
68 }
69 
70 __setup("ro", readonly);
71 __setup("rw", readwrite);
72 
73 #ifdef CONFIG_BLOCK
74 struct uuidcmp {
75 	const char *uuid;
76 	int len;
77 };
78 
79 /**
80  * match_dev_by_uuid - callback for finding a partition using its uuid
81  * @dev:	device passed in by the caller
82  * @data:	opaque pointer to the desired struct uuidcmp to match
83  *
84  * Returns 1 if the device matches, and 0 otherwise.
85  */
86 static int match_dev_by_uuid(struct device *dev, const void *data)
87 {
88 	const struct uuidcmp *cmp = data;
89 	struct hd_struct *part = dev_to_part(dev);
90 
91 	if (!part->info)
92 		goto no_match;
93 
94 	if (strncasecmp(cmp->uuid, part->info->uuid, cmp->len))
95 		goto no_match;
96 
97 	return 1;
98 no_match:
99 	return 0;
100 }
101 
102 
103 /**
104  * devt_from_partuuid - looks up the dev_t of a partition by its UUID
105  * @uuid_str:	char array containing ascii UUID
106  *
107  * The function will return the first partition which contains a matching
108  * UUID value in its partition_meta_info struct.  This does not search
109  * by filesystem UUIDs.
110  *
111  * If @uuid_str is followed by a "/PARTNROFF=%d", then the number will be
112  * extracted and used as an offset from the partition identified by the UUID.
113  *
114  * Returns the matching dev_t on success or 0 on failure.
115  */
116 static dev_t devt_from_partuuid(const char *uuid_str)
117 {
118 	dev_t res = 0;
119 	struct uuidcmp cmp;
120 	struct device *dev = NULL;
121 	struct gendisk *disk;
122 	struct hd_struct *part;
123 	int offset = 0;
124 	bool clear_root_wait = false;
125 	char *slash;
126 
127 	cmp.uuid = uuid_str;
128 
129 	slash = strchr(uuid_str, '/');
130 	/* Check for optional partition number offset attributes. */
131 	if (slash) {
132 		char c = 0;
133 		/* Explicitly fail on poor PARTUUID syntax. */
134 		if (sscanf(slash + 1,
135 			   "PARTNROFF=%d%c", &offset, &c) != 1) {
136 			clear_root_wait = true;
137 			goto done;
138 		}
139 		cmp.len = slash - uuid_str;
140 	} else {
141 		cmp.len = strlen(uuid_str);
142 	}
143 
144 	if (!cmp.len) {
145 		clear_root_wait = true;
146 		goto done;
147 	}
148 
149 	dev = class_find_device(&block_class, NULL, &cmp,
150 				&match_dev_by_uuid);
151 	if (!dev)
152 		goto done;
153 
154 	res = dev->devt;
155 
156 	/* Attempt to find the partition by offset. */
157 	if (!offset)
158 		goto no_offset;
159 
160 	res = 0;
161 	disk = part_to_disk(dev_to_part(dev));
162 	part = disk_get_part(disk, dev_to_part(dev)->partno + offset);
163 	if (part) {
164 		res = part_devt(part);
165 		put_device(part_to_dev(part));
166 	}
167 
168 no_offset:
169 	put_device(dev);
170 done:
171 	if (clear_root_wait) {
172 		pr_err("VFS: PARTUUID= is invalid.\n"
173 		       "Expected PARTUUID=<valid-uuid-id>[/PARTNROFF=%%d]\n");
174 		if (root_wait)
175 			pr_err("Disabling rootwait; root= is invalid.\n");
176 		root_wait = 0;
177 	}
178 	return res;
179 }
180 #endif
181 
182 /*
183  *	Convert a name into device number.  We accept the following variants:
184  *
185  *	1) <hex_major><hex_minor> device number in hexadecimal represents itself
186  *         no leading 0x, for example b302.
187  *	2) /dev/nfs represents Root_NFS (0xff)
188  *	3) /dev/<disk_name> represents the device number of disk
189  *	4) /dev/<disk_name><decimal> represents the device number
190  *         of partition - device number of disk plus the partition number
191  *	5) /dev/<disk_name>p<decimal> - same as the above, that form is
192  *	   used when disk name of partitioned disk ends on a digit.
193  *	6) PARTUUID=00112233-4455-6677-8899-AABBCCDDEEFF representing the
194  *	   unique id of a partition if the partition table provides it.
195  *	   The UUID may be either an EFI/GPT UUID, or refer to an MSDOS
196  *	   partition using the format SSSSSSSS-PP, where SSSSSSSS is a zero-
197  *	   filled hex representation of the 32-bit "NT disk signature", and PP
198  *	   is a zero-filled hex representation of the 1-based partition number.
199  *	7) PARTUUID=<UUID>/PARTNROFF=<int> to select a partition in relation to
200  *	   a partition with a known unique id.
201  *	8) <major>:<minor> major and minor number of the device separated by
202  *	   a colon.
203  *
204  *	If name doesn't have fall into the categories above, we return (0,0).
205  *	block_class is used to check if something is a disk name. If the disk
206  *	name contains slashes, the device name has them replaced with
207  *	bangs.
208  */
209 
210 dev_t name_to_dev_t(const char *name)
211 {
212 	char s[32];
213 	char *p;
214 	dev_t res = 0;
215 	int part;
216 
217 #ifdef CONFIG_BLOCK
218 	if (strncmp(name, "PARTUUID=", 9) == 0) {
219 		name += 9;
220 		res = devt_from_partuuid(name);
221 		if (!res)
222 			goto fail;
223 		goto done;
224 	}
225 #endif
226 
227 	if (strncmp(name, "/dev/", 5) != 0) {
228 		unsigned maj, min, offset;
229 		char dummy;
230 
231 		if ((sscanf(name, "%u:%u%c", &maj, &min, &dummy) == 2) ||
232 		    (sscanf(name, "%u:%u:%u:%c", &maj, &min, &offset, &dummy) == 3)) {
233 			res = MKDEV(maj, min);
234 			if (maj != MAJOR(res) || min != MINOR(res))
235 				goto fail;
236 		} else {
237 			res = new_decode_dev(simple_strtoul(name, &p, 16));
238 			if (*p)
239 				goto fail;
240 		}
241 		goto done;
242 	}
243 
244 	name += 5;
245 	res = Root_NFS;
246 	if (strcmp(name, "nfs") == 0)
247 		goto done;
248 	res = Root_RAM0;
249 	if (strcmp(name, "ram") == 0)
250 		goto done;
251 
252 	if (strlen(name) > 31)
253 		goto fail;
254 	strcpy(s, name);
255 	for (p = s; *p; p++)
256 		if (*p == '/')
257 			*p = '!';
258 	res = blk_lookup_devt(s, 0);
259 	if (res)
260 		goto done;
261 
262 	/*
263 	 * try non-existent, but valid partition, which may only exist
264 	 * after revalidating the disk, like partitioned md devices
265 	 */
266 	while (p > s && isdigit(p[-1]))
267 		p--;
268 	if (p == s || !*p || *p == '0')
269 		goto fail;
270 
271 	/* try disk name without <part number> */
272 	part = simple_strtoul(p, NULL, 10);
273 	*p = '\0';
274 	res = blk_lookup_devt(s, part);
275 	if (res)
276 		goto done;
277 
278 	/* try disk name without p<part number> */
279 	if (p < s + 2 || !isdigit(p[-2]) || p[-1] != 'p')
280 		goto fail;
281 	p[-1] = '\0';
282 	res = blk_lookup_devt(s, part);
283 	if (res)
284 		goto done;
285 
286 fail:
287 	return 0;
288 done:
289 	return res;
290 }
291 EXPORT_SYMBOL_GPL(name_to_dev_t);
292 
293 static int __init root_dev_setup(char *line)
294 {
295 	strlcpy(saved_root_name, line, sizeof(saved_root_name));
296 	return 1;
297 }
298 
299 __setup("root=", root_dev_setup);
300 
301 static int __init rootwait_setup(char *str)
302 {
303 	if (*str)
304 		return 0;
305 	root_wait = 1;
306 	return 1;
307 }
308 
309 __setup("rootwait", rootwait_setup);
310 
311 static char * __initdata root_mount_data;
312 static int __init root_data_setup(char *str)
313 {
314 	root_mount_data = str;
315 	return 1;
316 }
317 
318 static char * __initdata root_fs_names;
319 static int __init fs_names_setup(char *str)
320 {
321 	root_fs_names = str;
322 	return 1;
323 }
324 
325 static unsigned int __initdata root_delay;
326 static int __init root_delay_setup(char *str)
327 {
328 	root_delay = simple_strtoul(str, NULL, 0);
329 	return 1;
330 }
331 
332 __setup("rootflags=", root_data_setup);
333 __setup("rootfstype=", fs_names_setup);
334 __setup("rootdelay=", root_delay_setup);
335 
336 static void __init get_fs_names(char *page)
337 {
338 	char *s = page;
339 
340 	if (root_fs_names) {
341 		strcpy(page, root_fs_names);
342 		while (*s++) {
343 			if (s[-1] == ',')
344 				s[-1] = '\0';
345 		}
346 	} else {
347 		int len = get_filesystem_list(page);
348 		char *p, *next;
349 
350 		page[len] = '\0';
351 		for (p = page-1; p; p = next) {
352 			next = strchr(++p, '\n');
353 			if (*p++ != '\t')
354 				continue;
355 			while ((*s++ = *p++) != '\n')
356 				;
357 			s[-1] = '\0';
358 		}
359 	}
360 	*s = '\0';
361 }
362 
363 static int __init do_mount_root(char *name, char *fs, int flags, void *data)
364 {
365 	struct super_block *s;
366 	int err = sys_mount(name, "/root", fs, flags, data);
367 	if (err)
368 		return err;
369 
370 	sys_chdir("/root");
371 	s = current->fs->pwd.dentry->d_sb;
372 	ROOT_DEV = s->s_dev;
373 	printk(KERN_INFO
374 	       "VFS: Mounted root (%s filesystem)%s on device %u:%u.\n",
375 	       s->s_type->name,
376 	       s->s_flags & MS_RDONLY ?  " readonly" : "",
377 	       MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
378 	return 0;
379 }
380 
381 void __init mount_block_root(char *name, int flags)
382 {
383 	struct page *page = alloc_page(GFP_KERNEL |
384 					__GFP_NOTRACK_FALSE_POSITIVE);
385 	char *fs_names = page_address(page);
386 	char *p;
387 #ifdef CONFIG_BLOCK
388 	char b[BDEVNAME_SIZE];
389 #else
390 	const char *b = name;
391 #endif
392 
393 	get_fs_names(fs_names);
394 retry:
395 	for (p = fs_names; *p; p += strlen(p)+1) {
396 		int err = do_mount_root(name, p, flags, root_mount_data);
397 		switch (err) {
398 			case 0:
399 				goto out;
400 			case -EACCES:
401 			case -EINVAL:
402 				continue;
403 		}
404 	        /*
405 		 * Allow the user to distinguish between failed sys_open
406 		 * and bad superblock on root device.
407 		 * and give them a list of the available devices
408 		 */
409 #ifdef CONFIG_BLOCK
410 		__bdevname(ROOT_DEV, b);
411 #endif
412 		printk("VFS: Cannot open root device \"%s\" or %s: error %d\n",
413 				root_device_name, b, err);
414 		printk("Please append a correct \"root=\" boot option; here are the available partitions:\n");
415 
416 		printk_all_partitions();
417 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
418 		printk("DEBUG_BLOCK_EXT_DEVT is enabled, you need to specify "
419 		       "explicit textual name for \"root=\" boot option.\n");
420 #endif
421 		panic("VFS: Unable to mount root fs on %s", b);
422 	}
423 	if (!(flags & MS_RDONLY)) {
424 		flags |= MS_RDONLY;
425 		goto retry;
426 	}
427 
428 	printk("List of all partitions:\n");
429 	printk_all_partitions();
430 	printk("No filesystem could mount root, tried: ");
431 	for (p = fs_names; *p; p += strlen(p)+1)
432 		printk(" %s", p);
433 	printk("\n");
434 #ifdef CONFIG_BLOCK
435 	__bdevname(ROOT_DEV, b);
436 #endif
437 	panic("VFS: Unable to mount root fs on %s", b);
438 out:
439 	put_page(page);
440 }
441 
442 #ifdef CONFIG_ROOT_NFS
443 
444 #define NFSROOT_TIMEOUT_MIN	5
445 #define NFSROOT_TIMEOUT_MAX	30
446 #define NFSROOT_RETRY_MAX	5
447 
448 static int __init mount_nfs_root(void)
449 {
450 	char *root_dev, *root_data;
451 	unsigned int timeout;
452 	int try, err;
453 
454 	err = nfs_root_data(&root_dev, &root_data);
455 	if (err != 0)
456 		return 0;
457 
458 	/*
459 	 * The server or network may not be ready, so try several
460 	 * times.  Stop after a few tries in case the client wants
461 	 * to fall back to other boot methods.
462 	 */
463 	timeout = NFSROOT_TIMEOUT_MIN;
464 	for (try = 1; ; try++) {
465 		err = do_mount_root(root_dev, "nfs",
466 					root_mountflags, root_data);
467 		if (err == 0)
468 			return 1;
469 		if (try > NFSROOT_RETRY_MAX)
470 			break;
471 
472 		/* Wait, in case the server refused us immediately */
473 		ssleep(timeout);
474 		timeout <<= 1;
475 		if (timeout > NFSROOT_TIMEOUT_MAX)
476 			timeout = NFSROOT_TIMEOUT_MAX;
477 	}
478 	return 0;
479 }
480 #endif
481 
482 #if defined(CONFIG_BLK_DEV_RAM) || defined(CONFIG_BLK_DEV_FD)
483 void __init change_floppy(char *fmt, ...)
484 {
485 	struct termios termios;
486 	char buf[80];
487 	char c;
488 	int fd;
489 	va_list args;
490 	va_start(args, fmt);
491 	vsprintf(buf, fmt, args);
492 	va_end(args);
493 	fd = sys_open("/dev/root", O_RDWR | O_NDELAY, 0);
494 	if (fd >= 0) {
495 		sys_ioctl(fd, FDEJECT, 0);
496 		sys_close(fd);
497 	}
498 	printk(KERN_NOTICE "VFS: Insert %s and press ENTER\n", buf);
499 	fd = sys_open("/dev/console", O_RDWR, 0);
500 	if (fd >= 0) {
501 		sys_ioctl(fd, TCGETS, (long)&termios);
502 		termios.c_lflag &= ~ICANON;
503 		sys_ioctl(fd, TCSETSF, (long)&termios);
504 		sys_read(fd, &c, 1);
505 		termios.c_lflag |= ICANON;
506 		sys_ioctl(fd, TCSETSF, (long)&termios);
507 		sys_close(fd);
508 	}
509 }
510 #endif
511 
512 void __init mount_root(void)
513 {
514 #ifdef CONFIG_ROOT_NFS
515 	if (ROOT_DEV == Root_NFS) {
516 		if (mount_nfs_root())
517 			return;
518 
519 		printk(KERN_ERR "VFS: Unable to mount root fs via NFS, trying floppy.\n");
520 		ROOT_DEV = Root_FD0;
521 	}
522 #endif
523 #ifdef CONFIG_BLK_DEV_FD
524 	if (MAJOR(ROOT_DEV) == FLOPPY_MAJOR) {
525 		/* rd_doload is 2 for a dual initrd/ramload setup */
526 		if (rd_doload==2) {
527 			if (rd_load_disk(1)) {
528 				ROOT_DEV = Root_RAM1;
529 				root_device_name = NULL;
530 			}
531 		} else
532 			change_floppy("root floppy");
533 	}
534 #endif
535 #ifdef CONFIG_BLOCK
536 	create_dev("/dev/root", ROOT_DEV);
537 	mount_block_root("/dev/root", root_mountflags);
538 #endif
539 }
540 
541 /*
542  * Prepare the namespace - decide what/where to mount, load ramdisks, etc.
543  */
544 void __init prepare_namespace(void)
545 {
546 	int is_floppy;
547 
548 	if (root_delay) {
549 		printk(KERN_INFO "Waiting %d sec before mounting root device...\n",
550 		       root_delay);
551 		ssleep(root_delay);
552 	}
553 
554 	/*
555 	 * wait for the known devices to complete their probing
556 	 *
557 	 * Note: this is a potential source of long boot delays.
558 	 * For example, it is not atypical to wait 5 seconds here
559 	 * for the touchpad of a laptop to initialize.
560 	 */
561 	wait_for_device_probe();
562 
563 	md_run_setup();
564 
565 	if (saved_root_name[0]) {
566 		root_device_name = saved_root_name;
567 		if (!strncmp(root_device_name, "mtd", 3) ||
568 		    !strncmp(root_device_name, "ubi", 3)) {
569 			mount_block_root(root_device_name, root_mountflags);
570 			goto out;
571 		}
572 		ROOT_DEV = name_to_dev_t(root_device_name);
573 		if (strncmp(root_device_name, "/dev/", 5) == 0)
574 			root_device_name += 5;
575 	}
576 
577 	if (initrd_load())
578 		goto out;
579 
580 	/* wait for any asynchronous scanning to complete */
581 	if ((ROOT_DEV == 0) && root_wait) {
582 		printk(KERN_INFO "Waiting for root device %s...\n",
583 			saved_root_name);
584 		while (driver_probe_done() != 0 ||
585 			(ROOT_DEV = name_to_dev_t(saved_root_name)) == 0)
586 			msleep(100);
587 		async_synchronize_full();
588 	}
589 
590 	is_floppy = MAJOR(ROOT_DEV) == FLOPPY_MAJOR;
591 
592 	if (is_floppy && rd_doload && rd_load_disk(0))
593 		ROOT_DEV = Root_RAM0;
594 
595 	mount_root();
596 out:
597 	devtmpfs_mount("dev");
598 	sys_mount(".", "/", NULL, MS_MOVE, NULL);
599 	sys_chroot(".");
600 }
601 
602 static bool is_tmpfs;
603 static struct dentry *rootfs_mount(struct file_system_type *fs_type,
604 	int flags, const char *dev_name, void *data)
605 {
606 	static unsigned long once;
607 	void *fill = ramfs_fill_super;
608 
609 	if (test_and_set_bit(0, &once))
610 		return ERR_PTR(-ENODEV);
611 
612 	if (IS_ENABLED(CONFIG_TMPFS) && is_tmpfs)
613 		fill = shmem_fill_super;
614 
615 	return mount_nodev(fs_type, flags, data, fill);
616 }
617 
618 static struct file_system_type rootfs_fs_type = {
619 	.name		= "rootfs",
620 	.mount		= rootfs_mount,
621 	.kill_sb	= kill_litter_super,
622 };
623 
624 int __init init_rootfs(void)
625 {
626 	int err = register_filesystem(&rootfs_fs_type);
627 
628 	if (err)
629 		return err;
630 
631 	if (IS_ENABLED(CONFIG_TMPFS) && !saved_root_name[0] &&
632 		(!root_fs_names || strstr(root_fs_names, "tmpfs"))) {
633 		err = shmem_init();
634 		is_tmpfs = true;
635 	} else {
636 		err = init_ramfs_fs();
637 	}
638 
639 	if (err)
640 		unregister_filesystem(&rootfs_fs_type);
641 
642 	return err;
643 }
644