1 // SPDX-License-Identifier: GPL-2.0-only
2 #include <linux/module.h>
3 #include <linux/sched.h>
4 #include <linux/ctype.h>
5 #include <linux/fd.h>
6 #include <linux/tty.h>
7 #include <linux/suspend.h>
8 #include <linux/root_dev.h>
9 #include <linux/security.h>
10 #include <linux/delay.h>
11 #include <linux/genhd.h>
12 #include <linux/mount.h>
13 #include <linux/device.h>
14 #include <linux/init.h>
15 #include <linux/fs.h>
16 #include <linux/initrd.h>
17 #include <linux/async.h>
18 #include <linux/fs_struct.h>
19 #include <linux/slab.h>
20 #include <linux/ramfs.h>
21 #include <linux/shmem_fs.h>
22
23 #include <linux/nfs_fs.h>
24 #include <linux/nfs_fs_sb.h>
25 #include <linux/nfs_mount.h>
26 #include <linux/raid/detect.h>
27 #include <uapi/linux/mount.h>
28
29 #include "do_mounts.h"
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
load_ramdisk(char * str)38 static int __init load_ramdisk(char *str)
39 {
40 pr_warn("ignoring the deprecated load_ramdisk= option\n");
41 return 1;
42 }
43 __setup("load_ramdisk=", load_ramdisk);
44
readonly(char * str)45 static int __init readonly(char *str)
46 {
47 if (*str)
48 return 0;
49 root_mountflags |= MS_RDONLY;
50 return 1;
51 }
52
readwrite(char * str)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 */
match_dev_by_uuid(struct device * dev,const void * data)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 */
devt_from_partuuid(const char * uuid_str)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 */
match_dev_by_label(struct device * dev,const void * data)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 * 10) /dev/cifs represents Root_CIFS (0xfe)
215 *
216 * If name doesn't have fall into the categories above, we return (0,0).
217 * block_class is used to check if something is a disk name. If the disk
218 * name contains slashes, the device name has them replaced with
219 * bangs.
220 */
221
name_to_dev_t(const char * name)222 dev_t name_to_dev_t(const char *name)
223 {
224 char s[32];
225 char *p;
226 dev_t res = 0;
227 int part;
228
229 #ifdef CONFIG_BLOCK
230 if (strncmp(name, "PARTUUID=", 9) == 0) {
231 name += 9;
232 res = devt_from_partuuid(name);
233 if (!res)
234 goto fail;
235 goto done;
236 } else if (strncmp(name, "PARTLABEL=", 10) == 0) {
237 struct device *dev;
238
239 dev = class_find_device(&block_class, NULL, name + 10,
240 &match_dev_by_label);
241 if (!dev)
242 goto fail;
243
244 res = dev->devt;
245 put_device(dev);
246 goto done;
247 }
248 #endif
249
250 if (strncmp(name, "/dev/", 5) != 0) {
251 unsigned maj, min, offset;
252 char dummy;
253
254 if ((sscanf(name, "%u:%u%c", &maj, &min, &dummy) == 2) ||
255 (sscanf(name, "%u:%u:%u:%c", &maj, &min, &offset, &dummy) == 3)) {
256 res = MKDEV(maj, min);
257 if (maj != MAJOR(res) || min != MINOR(res))
258 goto fail;
259 } else {
260 res = new_decode_dev(simple_strtoul(name, &p, 16));
261 if (*p)
262 goto fail;
263 }
264 goto done;
265 }
266
267 name += 5;
268 res = Root_NFS;
269 if (strcmp(name, "nfs") == 0)
270 goto done;
271 res = Root_CIFS;
272 if (strcmp(name, "cifs") == 0)
273 goto done;
274 res = Root_RAM0;
275 if (strcmp(name, "ram") == 0)
276 goto done;
277
278 if (strlen(name) > 31)
279 goto fail;
280 strcpy(s, name);
281 for (p = s; *p; p++)
282 if (*p == '/')
283 *p = '!';
284 res = blk_lookup_devt(s, 0);
285 if (res)
286 goto done;
287
288 /*
289 * try non-existent, but valid partition, which may only exist
290 * after revalidating the disk, like partitioned md devices
291 */
292 while (p > s && isdigit(p[-1]))
293 p--;
294 if (p == s || !*p || *p == '0')
295 goto fail;
296
297 /* try disk name without <part number> */
298 part = simple_strtoul(p, NULL, 10);
299 *p = '\0';
300 res = blk_lookup_devt(s, part);
301 if (res)
302 goto done;
303
304 /* try disk name without p<part number> */
305 if (p < s + 2 || !isdigit(p[-2]) || p[-1] != 'p')
306 goto fail;
307 p[-1] = '\0';
308 res = blk_lookup_devt(s, part);
309 if (res)
310 goto done;
311
312 fail:
313 return 0;
314 done:
315 return res;
316 }
317 EXPORT_SYMBOL_GPL(name_to_dev_t);
318
root_dev_setup(char * line)319 static int __init root_dev_setup(char *line)
320 {
321 strlcpy(saved_root_name, line, sizeof(saved_root_name));
322 return 1;
323 }
324
325 __setup("root=", root_dev_setup);
326
rootwait_setup(char * str)327 static int __init rootwait_setup(char *str)
328 {
329 if (*str)
330 return 0;
331 root_wait = 1;
332 return 1;
333 }
334
335 __setup("rootwait", rootwait_setup);
336
337 static char * __initdata root_mount_data;
root_data_setup(char * str)338 static int __init root_data_setup(char *str)
339 {
340 root_mount_data = str;
341 return 1;
342 }
343
344 static char * __initdata root_fs_names;
fs_names_setup(char * str)345 static int __init fs_names_setup(char *str)
346 {
347 root_fs_names = str;
348 return 1;
349 }
350
351 static unsigned int __initdata root_delay;
root_delay_setup(char * str)352 static int __init root_delay_setup(char *str)
353 {
354 root_delay = simple_strtoul(str, NULL, 0);
355 return 1;
356 }
357
358 __setup("rootflags=", root_data_setup);
359 __setup("rootfstype=", fs_names_setup);
360 __setup("rootdelay=", root_delay_setup);
361
get_fs_names(char * page)362 static void __init get_fs_names(char *page)
363 {
364 char *s = page;
365
366 if (root_fs_names) {
367 strcpy(page, root_fs_names);
368 while (*s++) {
369 if (s[-1] == ',')
370 s[-1] = '\0';
371 }
372 } else {
373 int len = get_filesystem_list(page);
374 char *p, *next;
375
376 page[len] = '\0';
377 for (p = page-1; p; p = next) {
378 next = strchr(++p, '\n');
379 if (*p++ != '\t')
380 continue;
381 while ((*s++ = *p++) != '\n')
382 ;
383 s[-1] = '\0';
384 }
385 }
386 *s = '\0';
387 }
388
do_mount_root(const char * name,const char * fs,const int flags,const void * data)389 static int __init do_mount_root(const char *name, const char *fs,
390 const int flags, const void *data)
391 {
392 struct super_block *s;
393 struct page *p = NULL;
394 char *data_page = NULL;
395 int ret;
396
397 if (data) {
398 /* init_mount() requires a full page as fifth argument */
399 p = alloc_page(GFP_KERNEL);
400 if (!p)
401 return -ENOMEM;
402 data_page = page_address(p);
403 /* zero-pad. init_mount() will make sure it's terminated */
404 strncpy(data_page, data, PAGE_SIZE);
405 }
406
407 ret = init_mount(name, "/root", fs, flags, data_page);
408 if (ret)
409 goto out;
410
411 init_chdir("/root");
412 s = current->fs->pwd.dentry->d_sb;
413 ROOT_DEV = s->s_dev;
414 printk(KERN_INFO
415 "VFS: Mounted root (%s filesystem)%s on device %u:%u.\n",
416 s->s_type->name,
417 sb_rdonly(s) ? " readonly" : "",
418 MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
419
420 out:
421 if (p)
422 put_page(p);
423 return ret;
424 }
425
mount_block_root(char * name,int flags)426 void __init mount_block_root(char *name, int flags)
427 {
428 struct page *page = alloc_page(GFP_KERNEL);
429 char *fs_names = page_address(page);
430 char *p;
431 char b[BDEVNAME_SIZE];
432
433 scnprintf(b, BDEVNAME_SIZE, "unknown-block(%u,%u)",
434 MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
435 get_fs_names(fs_names);
436 retry:
437 for (p = fs_names; *p; p += strlen(p)+1) {
438 int err = do_mount_root(name, p, flags, root_mount_data);
439 switch (err) {
440 case 0:
441 goto out;
442 case -EACCES:
443 case -EINVAL:
444 continue;
445 }
446 /*
447 * Allow the user to distinguish between failed sys_open
448 * and bad superblock on root device.
449 * and give them a list of the available devices
450 */
451 printk("VFS: Cannot open root device \"%s\" or %s: error %d\n",
452 root_device_name, b, err);
453 printk("Please append a correct \"root=\" boot option; here are the available partitions:\n");
454
455 printk_all_partitions();
456 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
457 printk("DEBUG_BLOCK_EXT_DEVT is enabled, you need to specify "
458 "explicit textual name for \"root=\" boot option.\n");
459 #endif
460 panic("VFS: Unable to mount root fs on %s", b);
461 }
462 if (!(flags & SB_RDONLY)) {
463 flags |= SB_RDONLY;
464 goto retry;
465 }
466
467 printk("List of all partitions:\n");
468 printk_all_partitions();
469 printk("No filesystem could mount root, tried: ");
470 for (p = fs_names; *p; p += strlen(p)+1)
471 printk(" %s", p);
472 printk("\n");
473 panic("VFS: Unable to mount root fs on %s", b);
474 out:
475 put_page(page);
476 }
477
478 #ifdef CONFIG_ROOT_NFS
479
480 #define NFSROOT_TIMEOUT_MIN 5
481 #define NFSROOT_TIMEOUT_MAX 30
482 #define NFSROOT_RETRY_MAX 5
483
mount_nfs_root(void)484 static int __init mount_nfs_root(void)
485 {
486 char *root_dev, *root_data;
487 unsigned int timeout;
488 int try, err;
489
490 err = nfs_root_data(&root_dev, &root_data);
491 if (err != 0)
492 return 0;
493
494 /*
495 * The server or network may not be ready, so try several
496 * times. Stop after a few tries in case the client wants
497 * to fall back to other boot methods.
498 */
499 timeout = NFSROOT_TIMEOUT_MIN;
500 for (try = 1; ; try++) {
501 err = do_mount_root(root_dev, "nfs",
502 root_mountflags, root_data);
503 if (err == 0)
504 return 1;
505 if (try > NFSROOT_RETRY_MAX)
506 break;
507
508 /* Wait, in case the server refused us immediately */
509 ssleep(timeout);
510 timeout <<= 1;
511 if (timeout > NFSROOT_TIMEOUT_MAX)
512 timeout = NFSROOT_TIMEOUT_MAX;
513 }
514 return 0;
515 }
516 #endif
517
518 #ifdef CONFIG_CIFS_ROOT
519
520 extern int cifs_root_data(char **dev, char **opts);
521
522 #define CIFSROOT_TIMEOUT_MIN 5
523 #define CIFSROOT_TIMEOUT_MAX 30
524 #define CIFSROOT_RETRY_MAX 5
525
mount_cifs_root(void)526 static int __init mount_cifs_root(void)
527 {
528 char *root_dev, *root_data;
529 unsigned int timeout;
530 int try, err;
531
532 err = cifs_root_data(&root_dev, &root_data);
533 if (err != 0)
534 return 0;
535
536 timeout = CIFSROOT_TIMEOUT_MIN;
537 for (try = 1; ; try++) {
538 err = do_mount_root(root_dev, "cifs", root_mountflags,
539 root_data);
540 if (err == 0)
541 return 1;
542 if (try > CIFSROOT_RETRY_MAX)
543 break;
544
545 ssleep(timeout);
546 timeout <<= 1;
547 if (timeout > CIFSROOT_TIMEOUT_MAX)
548 timeout = CIFSROOT_TIMEOUT_MAX;
549 }
550 return 0;
551 }
552 #endif
553
mount_root(void)554 void __init mount_root(void)
555 {
556 #ifdef CONFIG_ROOT_NFS
557 if (ROOT_DEV == Root_NFS) {
558 if (!mount_nfs_root())
559 printk(KERN_ERR "VFS: Unable to mount root fs via NFS.\n");
560 return;
561 }
562 #endif
563 #ifdef CONFIG_CIFS_ROOT
564 if (ROOT_DEV == Root_CIFS) {
565 if (!mount_cifs_root())
566 printk(KERN_ERR "VFS: Unable to mount root fs via SMB.\n");
567 return;
568 }
569 #endif
570 #ifdef CONFIG_BLOCK
571 {
572 int err = create_dev("/dev/root", ROOT_DEV);
573
574 if (err < 0)
575 pr_emerg("Failed to create /dev/root: %d\n", err);
576 mount_block_root("/dev/root", root_mountflags);
577 }
578 #endif
579 }
580
581 /*
582 * Prepare the namespace - decide what/where to mount, load ramdisks, etc.
583 */
prepare_namespace(void)584 void __init prepare_namespace(void)
585 {
586 if (root_delay) {
587 printk(KERN_INFO "Waiting %d sec before mounting root device...\n",
588 root_delay);
589 ssleep(root_delay);
590 }
591
592 /*
593 * wait for the known devices to complete their probing
594 *
595 * Note: this is a potential source of long boot delays.
596 * For example, it is not atypical to wait 5 seconds here
597 * for the touchpad of a laptop to initialize.
598 */
599 wait_for_device_probe();
600
601 md_run_setup();
602
603 if (saved_root_name[0]) {
604 root_device_name = saved_root_name;
605 if (!strncmp(root_device_name, "mtd", 3) ||
606 !strncmp(root_device_name, "ubi", 3)) {
607 mount_block_root(root_device_name, root_mountflags);
608 goto out;
609 }
610 ROOT_DEV = name_to_dev_t(root_device_name);
611 if (strncmp(root_device_name, "/dev/", 5) == 0)
612 root_device_name += 5;
613 }
614
615 if (initrd_load())
616 goto out;
617
618 /* wait for any asynchronous scanning to complete */
619 if ((ROOT_DEV == 0) && root_wait) {
620 printk(KERN_INFO "Waiting for root device %s...\n",
621 saved_root_name);
622 while (driver_probe_done() != 0 ||
623 (ROOT_DEV = name_to_dev_t(saved_root_name)) == 0)
624 msleep(5);
625 async_synchronize_full();
626 }
627
628 mount_root();
629 out:
630 devtmpfs_mount();
631 init_mount(".", "/", NULL, MS_MOVE, NULL);
632 init_chroot(".");
633 }
634
635 static bool is_tmpfs;
rootfs_init_fs_context(struct fs_context * fc)636 static int rootfs_init_fs_context(struct fs_context *fc)
637 {
638 if (IS_ENABLED(CONFIG_TMPFS) && is_tmpfs)
639 return shmem_init_fs_context(fc);
640
641 return ramfs_init_fs_context(fc);
642 }
643
644 struct file_system_type rootfs_fs_type = {
645 .name = "rootfs",
646 .init_fs_context = rootfs_init_fs_context,
647 .kill_sb = kill_litter_super,
648 };
649
init_rootfs(void)650 void __init init_rootfs(void)
651 {
652 if (IS_ENABLED(CONFIG_TMPFS)) {
653 if (!saved_root_name[0] && !root_fs_names)
654 is_tmpfs = true;
655 else if (root_fs_names && !!strstr(root_fs_names, "tmpfs"))
656 is_tmpfs = true;
657 }
658 }
659