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 struct block_device *bdev = dev_to_bdev(dev);
80 const struct uuidcmp *cmp = data;
81
82 if (!bdev->bd_meta_info ||
83 strncasecmp(cmp->uuid, bdev->bd_meta_info->uuid, cmp->len))
84 return 0;
85 return 1;
86 }
87
88 /**
89 * devt_from_partuuid - looks up the dev_t of a partition by its UUID
90 * @uuid_str: char array containing ascii UUID
91 *
92 * The function will return the first partition which contains a matching
93 * UUID value in its partition_meta_info struct. This does not search
94 * by filesystem UUIDs.
95 *
96 * If @uuid_str is followed by a "/PARTNROFF=%d", then the number will be
97 * extracted and used as an offset from the partition identified by the UUID.
98 *
99 * Returns the matching dev_t on success or 0 on failure.
100 */
devt_from_partuuid(const char * uuid_str)101 static dev_t devt_from_partuuid(const char *uuid_str)
102 {
103 struct uuidcmp cmp;
104 struct device *dev = NULL;
105 dev_t devt = 0;
106 int offset = 0;
107 char *slash;
108
109 cmp.uuid = uuid_str;
110
111 slash = strchr(uuid_str, '/');
112 /* Check for optional partition number offset attributes. */
113 if (slash) {
114 char c = 0;
115
116 /* Explicitly fail on poor PARTUUID syntax. */
117 if (sscanf(slash + 1, "PARTNROFF=%d%c", &offset, &c) != 1)
118 goto clear_root_wait;
119 cmp.len = slash - uuid_str;
120 } else {
121 cmp.len = strlen(uuid_str);
122 }
123
124 if (!cmp.len)
125 goto clear_root_wait;
126
127 dev = class_find_device(&block_class, NULL, &cmp, &match_dev_by_uuid);
128 if (!dev)
129 return 0;
130
131 if (offset) {
132 /*
133 * Attempt to find the requested partition by adding an offset
134 * to the partition number found by UUID.
135 */
136 struct block_device *part;
137
138 part = bdget_disk(dev_to_disk(dev),
139 dev_to_bdev(dev)->bd_partno + offset);
140 if (part) {
141 devt = part->bd_dev;
142 bdput(part);
143 }
144 } else {
145 devt = dev->devt;
146 }
147
148 put_device(dev);
149 return devt;
150
151 clear_root_wait:
152 pr_err("VFS: PARTUUID= is invalid.\n"
153 "Expected PARTUUID=<valid-uuid-id>[/PARTNROFF=%%d]\n");
154 if (root_wait)
155 pr_err("Disabling rootwait; root= is invalid.\n");
156 root_wait = 0;
157 return 0;
158 }
159
160 /**
161 * match_dev_by_label - callback for finding a partition using its label
162 * @dev: device passed in by the caller
163 * @data: opaque pointer to the label to match
164 *
165 * Returns 1 if the device matches, and 0 otherwise.
166 */
match_dev_by_label(struct device * dev,const void * data)167 static int match_dev_by_label(struct device *dev, const void *data)
168 {
169 struct block_device *bdev = dev_to_bdev(dev);
170 const char *label = data;
171
172 if (!bdev->bd_meta_info || strcmp(label, bdev->bd_meta_info->volname))
173 return 0;
174 return 1;
175 }
176
devt_from_partlabel(const char * label)177 static dev_t devt_from_partlabel(const char *label)
178 {
179 struct device *dev;
180 dev_t devt = 0;
181
182 dev = class_find_device(&block_class, NULL, label, &match_dev_by_label);
183 if (dev) {
184 devt = dev->devt;
185 put_device(dev);
186 }
187
188 return devt;
189 }
190
devt_from_devname(const char * name)191 static dev_t devt_from_devname(const char *name)
192 {
193 dev_t devt = 0;
194 int part;
195 char s[32];
196 char *p;
197
198 if (strlen(name) > 31)
199 return 0;
200 strcpy(s, name);
201 for (p = s; *p; p++) {
202 if (*p == '/')
203 *p = '!';
204 }
205
206 devt = blk_lookup_devt(s, 0);
207 if (devt)
208 return devt;
209
210 /*
211 * Try non-existent, but valid partition, which may only exist after
212 * opening the device, like partitioned md devices.
213 */
214 while (p > s && isdigit(p[-1]))
215 p--;
216 if (p == s || !*p || *p == '0')
217 return 0;
218
219 /* try disk name without <part number> */
220 part = simple_strtoul(p, NULL, 10);
221 *p = '\0';
222 devt = blk_lookup_devt(s, part);
223 if (devt)
224 return devt;
225
226 /* try disk name without p<part number> */
227 if (p < s + 2 || !isdigit(p[-2]) || p[-1] != 'p')
228 return 0;
229 p[-1] = '\0';
230 return blk_lookup_devt(s, part);
231 }
232 #endif /* CONFIG_BLOCK */
233
devt_from_devnum(const char * name)234 static dev_t devt_from_devnum(const char *name)
235 {
236 unsigned maj, min, offset;
237 dev_t devt = 0;
238 char *p, dummy;
239
240 if (sscanf(name, "%u:%u%c", &maj, &min, &dummy) == 2 ||
241 sscanf(name, "%u:%u:%u:%c", &maj, &min, &offset, &dummy) == 3) {
242 devt = MKDEV(maj, min);
243 if (maj != MAJOR(devt) || min != MINOR(devt))
244 return 0;
245 } else {
246 devt = new_decode_dev(simple_strtoul(name, &p, 16));
247 if (*p)
248 return 0;
249 }
250
251 return devt;
252 }
253
254 /*
255 * Convert a name into device number. We accept the following variants:
256 *
257 * 1) <hex_major><hex_minor> device number in hexadecimal represents itself
258 * no leading 0x, for example b302.
259 * 2) /dev/nfs represents Root_NFS (0xff)
260 * 3) /dev/<disk_name> represents the device number of disk
261 * 4) /dev/<disk_name><decimal> represents the device number
262 * of partition - device number of disk plus the partition number
263 * 5) /dev/<disk_name>p<decimal> - same as the above, that form is
264 * used when disk name of partitioned disk ends on a digit.
265 * 6) PARTUUID=00112233-4455-6677-8899-AABBCCDDEEFF representing the
266 * unique id of a partition if the partition table provides it.
267 * The UUID may be either an EFI/GPT UUID, or refer to an MSDOS
268 * partition using the format SSSSSSSS-PP, where SSSSSSSS is a zero-
269 * filled hex representation of the 32-bit "NT disk signature", and PP
270 * is a zero-filled hex representation of the 1-based partition number.
271 * 7) PARTUUID=<UUID>/PARTNROFF=<int> to select a partition in relation to
272 * a partition with a known unique id.
273 * 8) <major>:<minor> major and minor number of the device separated by
274 * a colon.
275 * 9) PARTLABEL=<name> with name being the GPT partition label.
276 * MSDOS partitions do not support labels!
277 * 10) /dev/cifs represents Root_CIFS (0xfe)
278 *
279 * If name doesn't have fall into the categories above, we return (0,0).
280 * block_class is used to check if something is a disk name. If the disk
281 * name contains slashes, the device name has them replaced with
282 * bangs.
283 */
name_to_dev_t(const char * name)284 dev_t name_to_dev_t(const char *name)
285 {
286 if (strcmp(name, "/dev/nfs") == 0)
287 return Root_NFS;
288 if (strcmp(name, "/dev/cifs") == 0)
289 return Root_CIFS;
290 if (strcmp(name, "/dev/ram") == 0)
291 return Root_RAM0;
292 #ifdef CONFIG_BLOCK
293 if (strncmp(name, "PARTUUID=", 9) == 0)
294 return devt_from_partuuid(name + 9);
295 if (strncmp(name, "PARTLABEL=", 10) == 0)
296 return devt_from_partlabel(name + 10);
297 if (strncmp(name, "/dev/", 5) == 0)
298 return devt_from_devname(name + 5);
299 #endif
300 return devt_from_devnum(name);
301 }
302 EXPORT_SYMBOL_GPL(name_to_dev_t);
303
root_dev_setup(char * line)304 static int __init root_dev_setup(char *line)
305 {
306 strlcpy(saved_root_name, line, sizeof(saved_root_name));
307 return 1;
308 }
309
310 __setup("root=", root_dev_setup);
311
rootwait_setup(char * str)312 static int __init rootwait_setup(char *str)
313 {
314 if (*str)
315 return 0;
316 root_wait = 1;
317 return 1;
318 }
319
320 __setup("rootwait", rootwait_setup);
321
322 static char * __initdata root_mount_data;
root_data_setup(char * str)323 static int __init root_data_setup(char *str)
324 {
325 root_mount_data = str;
326 return 1;
327 }
328
329 static char * __initdata root_fs_names;
fs_names_setup(char * str)330 static int __init fs_names_setup(char *str)
331 {
332 root_fs_names = str;
333 return 1;
334 }
335
336 static unsigned int __initdata root_delay;
root_delay_setup(char * str)337 static int __init root_delay_setup(char *str)
338 {
339 root_delay = simple_strtoul(str, NULL, 0);
340 return 1;
341 }
342
343 __setup("rootflags=", root_data_setup);
344 __setup("rootfstype=", fs_names_setup);
345 __setup("rootdelay=", root_delay_setup);
346
get_fs_names(char * page)347 static void __init get_fs_names(char *page)
348 {
349 char *s = page;
350
351 if (root_fs_names) {
352 strcpy(page, root_fs_names);
353 while (*s++) {
354 if (s[-1] == ',')
355 s[-1] = '\0';
356 }
357 } else {
358 int len = get_filesystem_list(page);
359 char *p, *next;
360
361 page[len] = '\0';
362 for (p = page-1; p; p = next) {
363 next = strchr(++p, '\n');
364 if (*p++ != '\t')
365 continue;
366 while ((*s++ = *p++) != '\n')
367 ;
368 s[-1] = '\0';
369 }
370 }
371 *s = '\0';
372 }
373
do_mount_root(const char * name,const char * fs,const int flags,const void * data)374 static int __init do_mount_root(const char *name, const char *fs,
375 const int flags, const void *data)
376 {
377 struct super_block *s;
378 struct page *p = NULL;
379 char *data_page = NULL;
380 int ret;
381
382 if (data) {
383 /* init_mount() requires a full page as fifth argument */
384 p = alloc_page(GFP_KERNEL);
385 if (!p)
386 return -ENOMEM;
387 data_page = page_address(p);
388 /* zero-pad. init_mount() will make sure it's terminated */
389 strncpy(data_page, data, PAGE_SIZE);
390 }
391
392 ret = init_mount(name, "/root", fs, flags, data_page);
393 if (ret)
394 goto out;
395
396 init_chdir("/root");
397 s = current->fs->pwd.dentry->d_sb;
398 ROOT_DEV = s->s_dev;
399 printk(KERN_INFO
400 "VFS: Mounted root (%s filesystem)%s on device %u:%u.\n",
401 s->s_type->name,
402 sb_rdonly(s) ? " readonly" : "",
403 MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
404
405 out:
406 if (p)
407 put_page(p);
408 return ret;
409 }
410
mount_block_root(char * name,int flags)411 void __init mount_block_root(char *name, int flags)
412 {
413 struct page *page = alloc_page(GFP_KERNEL);
414 char *fs_names = page_address(page);
415 char *p;
416 char b[BDEVNAME_SIZE];
417
418 scnprintf(b, BDEVNAME_SIZE, "unknown-block(%u,%u)",
419 MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
420 get_fs_names(fs_names);
421 retry:
422 for (p = fs_names; *p; p += strlen(p)+1) {
423 int err = do_mount_root(name, p, flags, root_mount_data);
424 switch (err) {
425 case 0:
426 goto out;
427 case -EACCES:
428 case -EINVAL:
429 continue;
430 }
431 /*
432 * Allow the user to distinguish between failed sys_open
433 * and bad superblock on root device.
434 * and give them a list of the available devices
435 */
436 printk("VFS: Cannot open root device \"%s\" or %s: error %d\n",
437 root_device_name, b, err);
438 printk("Please append a correct \"root=\" boot option; here are the available partitions:\n");
439
440 printk_all_partitions();
441 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
442 printk("DEBUG_BLOCK_EXT_DEVT is enabled, you need to specify "
443 "explicit textual name for \"root=\" boot option.\n");
444 #endif
445 panic("VFS: Unable to mount root fs on %s", b);
446 }
447 if (!(flags & SB_RDONLY)) {
448 flags |= SB_RDONLY;
449 goto retry;
450 }
451
452 printk("List of all partitions:\n");
453 printk_all_partitions();
454 printk("No filesystem could mount root, tried: ");
455 for (p = fs_names; *p; p += strlen(p)+1)
456 printk(" %s", p);
457 printk("\n");
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
mount_nfs_root(void)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 #ifdef CONFIG_CIFS_ROOT
504
505 extern int cifs_root_data(char **dev, char **opts);
506
507 #define CIFSROOT_TIMEOUT_MIN 5
508 #define CIFSROOT_TIMEOUT_MAX 30
509 #define CIFSROOT_RETRY_MAX 5
510
mount_cifs_root(void)511 static int __init mount_cifs_root(void)
512 {
513 char *root_dev, *root_data;
514 unsigned int timeout;
515 int try, err;
516
517 err = cifs_root_data(&root_dev, &root_data);
518 if (err != 0)
519 return 0;
520
521 timeout = CIFSROOT_TIMEOUT_MIN;
522 for (try = 1; ; try++) {
523 err = do_mount_root(root_dev, "cifs", root_mountflags,
524 root_data);
525 if (err == 0)
526 return 1;
527 if (try > CIFSROOT_RETRY_MAX)
528 break;
529
530 ssleep(timeout);
531 timeout <<= 1;
532 if (timeout > CIFSROOT_TIMEOUT_MAX)
533 timeout = CIFSROOT_TIMEOUT_MAX;
534 }
535 return 0;
536 }
537 #endif
538
mount_root(void)539 void __init mount_root(void)
540 {
541 #ifdef CONFIG_ROOT_NFS
542 if (ROOT_DEV == Root_NFS) {
543 if (!mount_nfs_root())
544 printk(KERN_ERR "VFS: Unable to mount root fs via NFS.\n");
545 return;
546 }
547 #endif
548 #ifdef CONFIG_CIFS_ROOT
549 if (ROOT_DEV == Root_CIFS) {
550 if (!mount_cifs_root())
551 printk(KERN_ERR "VFS: Unable to mount root fs via SMB.\n");
552 return;
553 }
554 #endif
555 #ifdef CONFIG_BLOCK
556 {
557 int err = create_dev("/dev/root", ROOT_DEV);
558
559 if (err < 0)
560 pr_emerg("Failed to create /dev/root: %d\n", err);
561 mount_block_root("/dev/root", root_mountflags);
562 }
563 #endif
564 }
565
566 /*
567 * Prepare the namespace - decide what/where to mount, load ramdisks, etc.
568 */
prepare_namespace(void)569 void __init prepare_namespace(void)
570 {
571 if (root_delay) {
572 printk(KERN_INFO "Waiting %d sec before mounting root device...\n",
573 root_delay);
574 ssleep(root_delay);
575 }
576
577 /*
578 * wait for the known devices to complete their probing
579 *
580 * Note: this is a potential source of long boot delays.
581 * For example, it is not atypical to wait 5 seconds here
582 * for the touchpad of a laptop to initialize.
583 */
584 wait_for_device_probe();
585
586 md_run_setup();
587
588 if (saved_root_name[0]) {
589 root_device_name = saved_root_name;
590 if (!strncmp(root_device_name, "mtd", 3) ||
591 !strncmp(root_device_name, "ubi", 3)) {
592 mount_block_root(root_device_name, root_mountflags);
593 goto out;
594 }
595 ROOT_DEV = name_to_dev_t(root_device_name);
596 if (strncmp(root_device_name, "/dev/", 5) == 0)
597 root_device_name += 5;
598 }
599
600 if (initrd_load())
601 goto out;
602
603 /* wait for any asynchronous scanning to complete */
604 if ((ROOT_DEV == 0) && root_wait) {
605 printk(KERN_INFO "Waiting for root device %s...\n",
606 saved_root_name);
607 while (driver_probe_done() != 0 ||
608 (ROOT_DEV = name_to_dev_t(saved_root_name)) == 0)
609 msleep(5);
610 async_synchronize_full();
611 }
612
613 mount_root();
614 out:
615 devtmpfs_mount();
616 init_mount(".", "/", NULL, MS_MOVE, NULL);
617 init_chroot(".");
618 }
619
620 static bool is_tmpfs;
rootfs_init_fs_context(struct fs_context * fc)621 static int rootfs_init_fs_context(struct fs_context *fc)
622 {
623 if (IS_ENABLED(CONFIG_TMPFS) && is_tmpfs)
624 return shmem_init_fs_context(fc);
625
626 return ramfs_init_fs_context(fc);
627 }
628
629 struct file_system_type rootfs_fs_type = {
630 .name = "rootfs",
631 .init_fs_context = rootfs_init_fs_context,
632 .kill_sb = kill_litter_super,
633 };
634
init_rootfs(void)635 void __init init_rootfs(void)
636 {
637 if (IS_ENABLED(CONFIG_TMPFS) && !saved_root_name[0] &&
638 (!root_fs_names || strstr(root_fs_names, "tmpfs")))
639 is_tmpfs = true;
640 }
641