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/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 <linux/raid/detect.h>
26 #include <uapi/linux/mount.h>
27
28 #include "do_mounts.h"
29
30 int root_mountflags = MS_RDONLY | MS_SILENT;
31 static char * __initdata root_device_name;
32 static char __initdata saved_root_name[64];
33 static int root_wait;
34
35 dev_t ROOT_DEV;
36
load_ramdisk(char * str)37 static int __init load_ramdisk(char *str)
38 {
39 pr_warn("ignoring the deprecated load_ramdisk= option\n");
40 return 1;
41 }
42 __setup("load_ramdisk=", load_ramdisk);
43
readonly(char * str)44 static int __init readonly(char *str)
45 {
46 if (*str)
47 return 0;
48 root_mountflags |= MS_RDONLY;
49 return 1;
50 }
51
readwrite(char * str)52 static int __init readwrite(char *str)
53 {
54 if (*str)
55 return 0;
56 root_mountflags &= ~MS_RDONLY;
57 return 1;
58 }
59
60 __setup("ro", readonly);
61 __setup("rw", readwrite);
62
63 #ifdef CONFIG_BLOCK
64 struct uuidcmp {
65 const char *uuid;
66 int len;
67 };
68
69 /**
70 * match_dev_by_uuid - callback for finding a partition using its uuid
71 * @dev: device passed in by the caller
72 * @data: opaque pointer to the desired struct uuidcmp to match
73 *
74 * Returns 1 if the device matches, and 0 otherwise.
75 */
match_dev_by_uuid(struct device * dev,const void * data)76 static int match_dev_by_uuid(struct device *dev, const void *data)
77 {
78 struct block_device *bdev = dev_to_bdev(dev);
79 const struct uuidcmp *cmp = data;
80
81 if (!bdev->bd_meta_info ||
82 strncasecmp(cmp->uuid, bdev->bd_meta_info->uuid, cmp->len))
83 return 0;
84 return 1;
85 }
86
87 /**
88 * devt_from_partuuid - looks up the dev_t of a partition by its UUID
89 * @uuid_str: char array containing ascii UUID
90 *
91 * The function will return the first partition which contains a matching
92 * UUID value in its partition_meta_info struct. This does not search
93 * by filesystem UUIDs.
94 *
95 * If @uuid_str is followed by a "/PARTNROFF=%d", then the number will be
96 * extracted and used as an offset from the partition identified by the UUID.
97 *
98 * Returns the matching dev_t on success or 0 on failure.
99 */
devt_from_partuuid(const char * uuid_str)100 static dev_t devt_from_partuuid(const char *uuid_str)
101 {
102 struct uuidcmp cmp;
103 struct device *dev = NULL;
104 dev_t devt = 0;
105 int offset = 0;
106 char *slash;
107
108 cmp.uuid = uuid_str;
109
110 slash = strchr(uuid_str, '/');
111 /* Check for optional partition number offset attributes. */
112 if (slash) {
113 char c = 0;
114
115 /* Explicitly fail on poor PARTUUID syntax. */
116 if (sscanf(slash + 1, "PARTNROFF=%d%c", &offset, &c) != 1)
117 goto clear_root_wait;
118 cmp.len = slash - uuid_str;
119 } else {
120 cmp.len = strlen(uuid_str);
121 }
122
123 if (!cmp.len)
124 goto clear_root_wait;
125
126 dev = class_find_device(&block_class, NULL, &cmp, &match_dev_by_uuid);
127 if (!dev)
128 return 0;
129
130 if (offset) {
131 /*
132 * Attempt to find the requested partition by adding an offset
133 * to the partition number found by UUID.
134 */
135 devt = part_devt(dev_to_disk(dev),
136 dev_to_bdev(dev)->bd_partno + offset);
137 } else {
138 devt = dev->devt;
139 }
140
141 put_device(dev);
142 return devt;
143
144 clear_root_wait:
145 pr_err("VFS: PARTUUID= is invalid.\n"
146 "Expected PARTUUID=<valid-uuid-id>[/PARTNROFF=%%d]\n");
147 if (root_wait)
148 pr_err("Disabling rootwait; root= is invalid.\n");
149 root_wait = 0;
150 return 0;
151 }
152
153 /**
154 * match_dev_by_label - callback for finding a partition using its label
155 * @dev: device passed in by the caller
156 * @data: opaque pointer to the label to match
157 *
158 * Returns 1 if the device matches, and 0 otherwise.
159 */
match_dev_by_label(struct device * dev,const void * data)160 static int match_dev_by_label(struct device *dev, const void *data)
161 {
162 struct block_device *bdev = dev_to_bdev(dev);
163 const char *label = data;
164
165 if (!bdev->bd_meta_info || strcmp(label, bdev->bd_meta_info->volname))
166 return 0;
167 return 1;
168 }
169
devt_from_partlabel(const char * label)170 static dev_t devt_from_partlabel(const char *label)
171 {
172 struct device *dev;
173 dev_t devt = 0;
174
175 dev = class_find_device(&block_class, NULL, label, &match_dev_by_label);
176 if (dev) {
177 devt = dev->devt;
178 put_device(dev);
179 }
180
181 return devt;
182 }
183
devt_from_devname(const char * name)184 static dev_t devt_from_devname(const char *name)
185 {
186 dev_t devt = 0;
187 int part;
188 char s[32];
189 char *p;
190
191 if (strlen(name) > 31)
192 return 0;
193 strcpy(s, name);
194 for (p = s; *p; p++) {
195 if (*p == '/')
196 *p = '!';
197 }
198
199 devt = blk_lookup_devt(s, 0);
200 if (devt)
201 return devt;
202
203 /*
204 * Try non-existent, but valid partition, which may only exist after
205 * opening the device, like partitioned md devices.
206 */
207 while (p > s && isdigit(p[-1]))
208 p--;
209 if (p == s || !*p || *p == '0')
210 return 0;
211
212 /* try disk name without <part number> */
213 part = simple_strtoul(p, NULL, 10);
214 *p = '\0';
215 devt = blk_lookup_devt(s, part);
216 if (devt)
217 return devt;
218
219 /* try disk name without p<part number> */
220 if (p < s + 2 || !isdigit(p[-2]) || p[-1] != 'p')
221 return 0;
222 p[-1] = '\0';
223 return blk_lookup_devt(s, part);
224 }
225 #endif /* CONFIG_BLOCK */
226
devt_from_devnum(const char * name)227 static dev_t devt_from_devnum(const char *name)
228 {
229 unsigned maj, min, offset;
230 dev_t devt = 0;
231 char *p, dummy;
232
233 if (sscanf(name, "%u:%u%c", &maj, &min, &dummy) == 2 ||
234 sscanf(name, "%u:%u:%u:%c", &maj, &min, &offset, &dummy) == 3) {
235 devt = MKDEV(maj, min);
236 if (maj != MAJOR(devt) || min != MINOR(devt))
237 return 0;
238 } else {
239 devt = new_decode_dev(simple_strtoul(name, &p, 16));
240 if (*p)
241 return 0;
242 }
243
244 return devt;
245 }
246
247 /*
248 * Convert a name into device number. We accept the following variants:
249 *
250 * 1) <hex_major><hex_minor> device number in hexadecimal represents itself
251 * no leading 0x, for example b302.
252 * 2) /dev/nfs represents Root_NFS (0xff)
253 * 3) /dev/<disk_name> represents the device number of disk
254 * 4) /dev/<disk_name><decimal> represents the device number
255 * of partition - device number of disk plus the partition number
256 * 5) /dev/<disk_name>p<decimal> - same as the above, that form is
257 * used when disk name of partitioned disk ends on a digit.
258 * 6) PARTUUID=00112233-4455-6677-8899-AABBCCDDEEFF representing the
259 * unique id of a partition if the partition table provides it.
260 * The UUID may be either an EFI/GPT UUID, or refer to an MSDOS
261 * partition using the format SSSSSSSS-PP, where SSSSSSSS is a zero-
262 * filled hex representation of the 32-bit "NT disk signature", and PP
263 * is a zero-filled hex representation of the 1-based partition number.
264 * 7) PARTUUID=<UUID>/PARTNROFF=<int> to select a partition in relation to
265 * a partition with a known unique id.
266 * 8) <major>:<minor> major and minor number of the device separated by
267 * a colon.
268 * 9) PARTLABEL=<name> with name being the GPT partition label.
269 * MSDOS partitions do not support labels!
270 * 10) /dev/cifs represents Root_CIFS (0xfe)
271 *
272 * If name doesn't have fall into the categories above, we return (0,0).
273 * block_class is used to check if something is a disk name. If the disk
274 * name contains slashes, the device name has them replaced with
275 * bangs.
276 */
name_to_dev_t(const char * name)277 dev_t name_to_dev_t(const char *name)
278 {
279 if (strcmp(name, "/dev/nfs") == 0)
280 return Root_NFS;
281 if (strcmp(name, "/dev/cifs") == 0)
282 return Root_CIFS;
283 if (strcmp(name, "/dev/ram") == 0)
284 return Root_RAM0;
285 #ifdef CONFIG_BLOCK
286 if (strncmp(name, "PARTUUID=", 9) == 0)
287 return devt_from_partuuid(name + 9);
288 if (strncmp(name, "PARTLABEL=", 10) == 0)
289 return devt_from_partlabel(name + 10);
290 if (strncmp(name, "/dev/", 5) == 0)
291 return devt_from_devname(name + 5);
292 #endif
293 return devt_from_devnum(name);
294 }
295 EXPORT_SYMBOL_GPL(name_to_dev_t);
296
root_dev_setup(char * line)297 static int __init root_dev_setup(char *line)
298 {
299 strscpy(saved_root_name, line, sizeof(saved_root_name));
300 return 1;
301 }
302
303 __setup("root=", root_dev_setup);
304
rootwait_setup(char * str)305 static int __init rootwait_setup(char *str)
306 {
307 if (*str)
308 return 0;
309 root_wait = 1;
310 return 1;
311 }
312
313 __setup("rootwait", rootwait_setup);
314
315 static char * __initdata root_mount_data;
root_data_setup(char * str)316 static int __init root_data_setup(char *str)
317 {
318 root_mount_data = str;
319 return 1;
320 }
321
322 static char * __initdata root_fs_names;
fs_names_setup(char * str)323 static int __init fs_names_setup(char *str)
324 {
325 root_fs_names = str;
326 return 1;
327 }
328
329 static unsigned int __initdata root_delay;
root_delay_setup(char * str)330 static int __init root_delay_setup(char *str)
331 {
332 root_delay = simple_strtoul(str, NULL, 0);
333 return 1;
334 }
335
336 __setup("rootflags=", root_data_setup);
337 __setup("rootfstype=", fs_names_setup);
338 __setup("rootdelay=", root_delay_setup);
339
340 /* This can return zero length strings. Caller should check */
split_fs_names(char * page,size_t size,char * names)341 static int __init split_fs_names(char *page, size_t size, char *names)
342 {
343 int count = 1;
344 char *p = page;
345
346 strscpy(p, root_fs_names, size);
347 while (*p++) {
348 if (p[-1] == ',') {
349 p[-1] = '\0';
350 count++;
351 }
352 }
353
354 return count;
355 }
356
do_mount_root(const char * name,const char * fs,const int flags,const void * data)357 static int __init do_mount_root(const char *name, const char *fs,
358 const int flags, const void *data)
359 {
360 struct super_block *s;
361 struct page *p = NULL;
362 char *data_page = NULL;
363 int ret;
364
365 if (data) {
366 /* init_mount() requires a full page as fifth argument */
367 p = alloc_page(GFP_KERNEL);
368 if (!p)
369 return -ENOMEM;
370 data_page = page_address(p);
371 /* zero-pad. init_mount() will make sure it's terminated */
372 strncpy(data_page, data, PAGE_SIZE);
373 }
374
375 ret = init_mount(name, "/root", fs, flags, data_page);
376 if (ret)
377 goto out;
378
379 init_chdir("/root");
380 s = current->fs->pwd.dentry->d_sb;
381 ROOT_DEV = s->s_dev;
382 printk(KERN_INFO
383 "VFS: Mounted root (%s filesystem)%s on device %u:%u.\n",
384 s->s_type->name,
385 sb_rdonly(s) ? " readonly" : "",
386 MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
387
388 out:
389 if (p)
390 put_page(p);
391 return ret;
392 }
393
mount_block_root(char * name,int flags)394 void __init mount_block_root(char *name, int flags)
395 {
396 struct page *page = alloc_page(GFP_KERNEL);
397 char *fs_names = page_address(page);
398 char *p;
399 char b[BDEVNAME_SIZE];
400 int num_fs, i;
401
402 scnprintf(b, BDEVNAME_SIZE, "unknown-block(%u,%u)",
403 MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
404 if (root_fs_names)
405 num_fs = split_fs_names(fs_names, PAGE_SIZE, root_fs_names);
406 else
407 num_fs = list_bdev_fs_names(fs_names, PAGE_SIZE);
408 retry:
409 for (i = 0, p = fs_names; i < num_fs; i++, p += strlen(p)+1) {
410 int err;
411
412 if (!*p)
413 continue;
414 err = do_mount_root(name, p, flags, root_mount_data);
415 switch (err) {
416 case 0:
417 goto out;
418 case -EACCES:
419 case -EINVAL:
420 continue;
421 }
422 /*
423 * Allow the user to distinguish between failed sys_open
424 * and bad superblock on root device.
425 * and give them a list of the available devices
426 */
427 printk("VFS: Cannot open root device \"%s\" or %s: error %d\n",
428 root_device_name, b, err);
429 printk("Please append a correct \"root=\" boot option; here are the available partitions:\n");
430
431 printk_all_partitions();
432 panic("VFS: Unable to mount root fs on %s", b);
433 }
434 if (!(flags & SB_RDONLY)) {
435 flags |= SB_RDONLY;
436 goto retry;
437 }
438
439 printk("List of all partitions:\n");
440 printk_all_partitions();
441 printk("No filesystem could mount root, tried: ");
442 for (i = 0, p = fs_names; i < num_fs; i++, p += strlen(p)+1)
443 printk(" %s", p);
444 printk("\n");
445 panic("VFS: Unable to mount root fs on %s", b);
446 out:
447 put_page(page);
448 }
449
450 #ifdef CONFIG_ROOT_NFS
451
452 #define NFSROOT_TIMEOUT_MIN 5
453 #define NFSROOT_TIMEOUT_MAX 30
454 #define NFSROOT_RETRY_MAX 5
455
mount_nfs_root(void)456 static int __init mount_nfs_root(void)
457 {
458 char *root_dev, *root_data;
459 unsigned int timeout;
460 int try, err;
461
462 err = nfs_root_data(&root_dev, &root_data);
463 if (err != 0)
464 return 0;
465
466 /*
467 * The server or network may not be ready, so try several
468 * times. Stop after a few tries in case the client wants
469 * to fall back to other boot methods.
470 */
471 timeout = NFSROOT_TIMEOUT_MIN;
472 for (try = 1; ; try++) {
473 err = do_mount_root(root_dev, "nfs",
474 root_mountflags, root_data);
475 if (err == 0)
476 return 1;
477 if (try > NFSROOT_RETRY_MAX)
478 break;
479
480 /* Wait, in case the server refused us immediately */
481 ssleep(timeout);
482 timeout <<= 1;
483 if (timeout > NFSROOT_TIMEOUT_MAX)
484 timeout = NFSROOT_TIMEOUT_MAX;
485 }
486 return 0;
487 }
488 #endif
489
490 #ifdef CONFIG_CIFS_ROOT
491
492 extern int cifs_root_data(char **dev, char **opts);
493
494 #define CIFSROOT_TIMEOUT_MIN 5
495 #define CIFSROOT_TIMEOUT_MAX 30
496 #define CIFSROOT_RETRY_MAX 5
497
mount_cifs_root(void)498 static int __init mount_cifs_root(void)
499 {
500 char *root_dev, *root_data;
501 unsigned int timeout;
502 int try, err;
503
504 err = cifs_root_data(&root_dev, &root_data);
505 if (err != 0)
506 return 0;
507
508 timeout = CIFSROOT_TIMEOUT_MIN;
509 for (try = 1; ; try++) {
510 err = do_mount_root(root_dev, "cifs", root_mountflags,
511 root_data);
512 if (err == 0)
513 return 1;
514 if (try > CIFSROOT_RETRY_MAX)
515 break;
516
517 ssleep(timeout);
518 timeout <<= 1;
519 if (timeout > CIFSROOT_TIMEOUT_MAX)
520 timeout = CIFSROOT_TIMEOUT_MAX;
521 }
522 return 0;
523 }
524 #endif
525
fs_is_nodev(char * fstype)526 static bool __init fs_is_nodev(char *fstype)
527 {
528 struct file_system_type *fs = get_fs_type(fstype);
529 bool ret = false;
530
531 if (fs) {
532 ret = !(fs->fs_flags & FS_REQUIRES_DEV);
533 put_filesystem(fs);
534 }
535
536 return ret;
537 }
538
mount_nodev_root(void)539 static int __init mount_nodev_root(void)
540 {
541 char *fs_names, *fstype;
542 int err = -EINVAL;
543 int num_fs, i;
544
545 fs_names = (void *)__get_free_page(GFP_KERNEL);
546 if (!fs_names)
547 return -EINVAL;
548 num_fs = split_fs_names(fs_names, PAGE_SIZE, root_fs_names);
549
550 for (i = 0, fstype = fs_names; i < num_fs;
551 i++, fstype += strlen(fstype) + 1) {
552 if (!*fstype)
553 continue;
554 if (!fs_is_nodev(fstype))
555 continue;
556 err = do_mount_root(root_device_name, fstype, root_mountflags,
557 root_mount_data);
558 if (!err)
559 break;
560 }
561
562 free_page((unsigned long)fs_names);
563 return err;
564 }
565
mount_root(void)566 void __init mount_root(void)
567 {
568 #ifdef CONFIG_ROOT_NFS
569 if (ROOT_DEV == Root_NFS) {
570 if (!mount_nfs_root())
571 printk(KERN_ERR "VFS: Unable to mount root fs via NFS.\n");
572 return;
573 }
574 #endif
575 #ifdef CONFIG_CIFS_ROOT
576 if (ROOT_DEV == Root_CIFS) {
577 if (!mount_cifs_root())
578 printk(KERN_ERR "VFS: Unable to mount root fs via SMB.\n");
579 return;
580 }
581 #endif
582 if (ROOT_DEV == 0 && root_device_name && root_fs_names) {
583 if (mount_nodev_root() == 0)
584 return;
585 }
586 #ifdef CONFIG_BLOCK
587 {
588 int err = create_dev("/dev/root", ROOT_DEV);
589
590 if (err < 0)
591 pr_emerg("Failed to create /dev/root: %d\n", err);
592 mount_block_root("/dev/root", root_mountflags);
593 }
594 #endif
595 }
596
597 /*
598 * Prepare the namespace - decide what/where to mount, load ramdisks, etc.
599 */
prepare_namespace(void)600 void __init prepare_namespace(void)
601 {
602 if (root_delay) {
603 printk(KERN_INFO "Waiting %d sec before mounting root device...\n",
604 root_delay);
605 ssleep(root_delay);
606 }
607
608 /*
609 * wait for the known devices to complete their probing
610 *
611 * Note: this is a potential source of long boot delays.
612 * For example, it is not atypical to wait 5 seconds here
613 * for the touchpad of a laptop to initialize.
614 */
615 wait_for_device_probe();
616
617 md_run_setup();
618
619 if (saved_root_name[0]) {
620 root_device_name = saved_root_name;
621 if (!strncmp(root_device_name, "mtd", 3) ||
622 !strncmp(root_device_name, "ubi", 3)) {
623 mount_block_root(root_device_name, root_mountflags);
624 goto out;
625 }
626 ROOT_DEV = name_to_dev_t(root_device_name);
627 if (strncmp(root_device_name, "/dev/", 5) == 0)
628 root_device_name += 5;
629 }
630
631 if (initrd_load())
632 goto out;
633
634 /* wait for any asynchronous scanning to complete */
635 if ((ROOT_DEV == 0) && root_wait) {
636 printk(KERN_INFO "Waiting for root device %s...\n",
637 saved_root_name);
638 while (driver_probe_done() != 0 ||
639 (ROOT_DEV = name_to_dev_t(saved_root_name)) == 0)
640 msleep(5);
641 async_synchronize_full();
642 }
643
644 mount_root();
645 out:
646 devtmpfs_mount();
647 init_mount(".", "/", NULL, MS_MOVE, NULL);
648 init_chroot(".");
649 }
650
651 static bool is_tmpfs;
rootfs_init_fs_context(struct fs_context * fc)652 static int rootfs_init_fs_context(struct fs_context *fc)
653 {
654 if (IS_ENABLED(CONFIG_TMPFS) && is_tmpfs)
655 return shmem_init_fs_context(fc);
656
657 return ramfs_init_fs_context(fc);
658 }
659
660 struct file_system_type rootfs_fs_type = {
661 .name = "rootfs",
662 .init_fs_context = rootfs_init_fs_context,
663 .kill_sb = kill_litter_super,
664 };
665
init_rootfs(void)666 void __init init_rootfs(void)
667 {
668 if (IS_ENABLED(CONFIG_TMPFS) && !saved_root_name[0] &&
669 (!root_fs_names || strstr(root_fs_names, "tmpfs")))
670 is_tmpfs = true;
671 }
672