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 devt = part_devt(dev_to_disk(dev),
137 dev_to_bdev(dev)->bd_partno + offset);
138 } else {
139 devt = dev->devt;
140 }
141
142 put_device(dev);
143 return devt;
144
145 clear_root_wait:
146 pr_err("VFS: PARTUUID= is invalid.\n"
147 "Expected PARTUUID=<valid-uuid-id>[/PARTNROFF=%%d]\n");
148 if (root_wait)
149 pr_err("Disabling rootwait; root= is invalid.\n");
150 root_wait = 0;
151 return 0;
152 }
153
154 /**
155 * match_dev_by_label - callback for finding a partition using its label
156 * @dev: device passed in by the caller
157 * @data: opaque pointer to the label to match
158 *
159 * Returns 1 if the device matches, and 0 otherwise.
160 */
match_dev_by_label(struct device * dev,const void * data)161 static int match_dev_by_label(struct device *dev, const void *data)
162 {
163 struct block_device *bdev = dev_to_bdev(dev);
164 const char *label = data;
165
166 if (!bdev->bd_meta_info || strcmp(label, bdev->bd_meta_info->volname))
167 return 0;
168 return 1;
169 }
170
devt_from_partlabel(const char * label)171 static dev_t devt_from_partlabel(const char *label)
172 {
173 struct device *dev;
174 dev_t devt = 0;
175
176 dev = class_find_device(&block_class, NULL, label, &match_dev_by_label);
177 if (dev) {
178 devt = dev->devt;
179 put_device(dev);
180 }
181
182 return devt;
183 }
184
devt_from_devname(const char * name)185 static dev_t devt_from_devname(const char *name)
186 {
187 dev_t devt = 0;
188 int part;
189 char s[32];
190 char *p;
191
192 if (strlen(name) > 31)
193 return 0;
194 strcpy(s, name);
195 for (p = s; *p; p++) {
196 if (*p == '/')
197 *p = '!';
198 }
199
200 devt = blk_lookup_devt(s, 0);
201 if (devt)
202 return devt;
203
204 /*
205 * Try non-existent, but valid partition, which may only exist after
206 * opening the device, like partitioned md devices.
207 */
208 while (p > s && isdigit(p[-1]))
209 p--;
210 if (p == s || !*p || *p == '0')
211 return 0;
212
213 /* try disk name without <part number> */
214 part = simple_strtoul(p, NULL, 10);
215 *p = '\0';
216 devt = blk_lookup_devt(s, part);
217 if (devt)
218 return devt;
219
220 /* try disk name without p<part number> */
221 if (p < s + 2 || !isdigit(p[-2]) || p[-1] != 'p')
222 return 0;
223 p[-1] = '\0';
224 return blk_lookup_devt(s, part);
225 }
226 #endif /* CONFIG_BLOCK */
227
devt_from_devnum(const char * name)228 static dev_t devt_from_devnum(const char *name)
229 {
230 unsigned maj, min, offset;
231 dev_t devt = 0;
232 char *p, dummy;
233
234 if (sscanf(name, "%u:%u%c", &maj, &min, &dummy) == 2 ||
235 sscanf(name, "%u:%u:%u:%c", &maj, &min, &offset, &dummy) == 3) {
236 devt = MKDEV(maj, min);
237 if (maj != MAJOR(devt) || min != MINOR(devt))
238 return 0;
239 } else {
240 devt = new_decode_dev(simple_strtoul(name, &p, 16));
241 if (*p)
242 return 0;
243 }
244
245 return devt;
246 }
247
248 /*
249 * Convert a name into device number. We accept the following variants:
250 *
251 * 1) <hex_major><hex_minor> device number in hexadecimal represents itself
252 * no leading 0x, for example b302.
253 * 2) /dev/nfs represents Root_NFS (0xff)
254 * 3) /dev/<disk_name> represents the device number of disk
255 * 4) /dev/<disk_name><decimal> represents the device number
256 * of partition - device number of disk plus the partition number
257 * 5) /dev/<disk_name>p<decimal> - same as the above, that form is
258 * used when disk name of partitioned disk ends on a digit.
259 * 6) PARTUUID=00112233-4455-6677-8899-AABBCCDDEEFF representing the
260 * unique id of a partition if the partition table provides it.
261 * The UUID may be either an EFI/GPT UUID, or refer to an MSDOS
262 * partition using the format SSSSSSSS-PP, where SSSSSSSS is a zero-
263 * filled hex representation of the 32-bit "NT disk signature", and PP
264 * is a zero-filled hex representation of the 1-based partition number.
265 * 7) PARTUUID=<UUID>/PARTNROFF=<int> to select a partition in relation to
266 * a partition with a known unique id.
267 * 8) <major>:<minor> major and minor number of the device separated by
268 * a colon.
269 * 9) PARTLABEL=<name> with name being the GPT partition label.
270 * MSDOS partitions do not support labels!
271 * 10) /dev/cifs represents Root_CIFS (0xfe)
272 *
273 * If name doesn't have fall into the categories above, we return (0,0).
274 * block_class is used to check if something is a disk name. If the disk
275 * name contains slashes, the device name has them replaced with
276 * bangs.
277 */
name_to_dev_t(const char * name)278 dev_t name_to_dev_t(const char *name)
279 {
280 if (strcmp(name, "/dev/nfs") == 0)
281 return Root_NFS;
282 if (strcmp(name, "/dev/cifs") == 0)
283 return Root_CIFS;
284 if (strcmp(name, "/dev/ram") == 0)
285 return Root_RAM0;
286 #ifdef CONFIG_BLOCK
287 if (strncmp(name, "PARTUUID=", 9) == 0)
288 return devt_from_partuuid(name + 9);
289 if (strncmp(name, "PARTLABEL=", 10) == 0)
290 return devt_from_partlabel(name + 10);
291 if (strncmp(name, "/dev/", 5) == 0)
292 return devt_from_devname(name + 5);
293 #endif
294 return devt_from_devnum(name);
295 }
296 EXPORT_SYMBOL_GPL(name_to_dev_t);
297
root_dev_setup(char * line)298 static int __init root_dev_setup(char *line)
299 {
300 strlcpy(saved_root_name, line, sizeof(saved_root_name));
301 return 1;
302 }
303
304 __setup("root=", root_dev_setup);
305
rootwait_setup(char * str)306 static int __init rootwait_setup(char *str)
307 {
308 if (*str)
309 return 0;
310 root_wait = 1;
311 return 1;
312 }
313
314 __setup("rootwait", rootwait_setup);
315
316 static char * __initdata root_mount_data;
root_data_setup(char * str)317 static int __init root_data_setup(char *str)
318 {
319 root_mount_data = str;
320 return 1;
321 }
322
323 static char * __initdata root_fs_names;
fs_names_setup(char * str)324 static int __init fs_names_setup(char *str)
325 {
326 root_fs_names = str;
327 return 1;
328 }
329
330 static unsigned int __initdata root_delay;
root_delay_setup(char * str)331 static int __init root_delay_setup(char *str)
332 {
333 root_delay = simple_strtoul(str, NULL, 0);
334 return 1;
335 }
336
337 __setup("rootflags=", root_data_setup);
338 __setup("rootfstype=", fs_names_setup);
339 __setup("rootdelay=", root_delay_setup);
340
341 /* This can return zero length strings. Caller should check */
split_fs_names(char * page,size_t size,char * names)342 static int __init split_fs_names(char *page, size_t size, char *names)
343 {
344 int count = 1;
345 char *p = page;
346
347 strlcpy(p, root_fs_names, size);
348 while (*p++) {
349 if (p[-1] == ',') {
350 p[-1] = '\0';
351 count++;
352 }
353 }
354
355 return count;
356 }
357
do_mount_root(const char * name,const char * fs,const int flags,const void * data)358 static int __init do_mount_root(const char *name, const char *fs,
359 const int flags, const void *data)
360 {
361 struct super_block *s;
362 struct page *p = NULL;
363 char *data_page = NULL;
364 int ret;
365
366 if (data) {
367 /* init_mount() requires a full page as fifth argument */
368 p = alloc_page(GFP_KERNEL);
369 if (!p)
370 return -ENOMEM;
371 data_page = page_address(p);
372 /* zero-pad. init_mount() will make sure it's terminated */
373 strncpy(data_page, data, PAGE_SIZE);
374 }
375
376 ret = init_mount(name, "/root", fs, flags, data_page);
377 if (ret)
378 goto out;
379
380 init_chdir("/root");
381 s = current->fs->pwd.dentry->d_sb;
382 ROOT_DEV = s->s_dev;
383 printk(KERN_INFO
384 "VFS: Mounted root (%s filesystem)%s on device %u:%u.\n",
385 s->s_type->name,
386 sb_rdonly(s) ? " readonly" : "",
387 MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
388
389 out:
390 if (p)
391 put_page(p);
392 return ret;
393 }
394
mount_block_root(char * name,int flags)395 void __init mount_block_root(char *name, int flags)
396 {
397 struct page *page = alloc_page(GFP_KERNEL);
398 char *fs_names = page_address(page);
399 char *p;
400 char b[BDEVNAME_SIZE];
401 int num_fs, i;
402
403 scnprintf(b, BDEVNAME_SIZE, "unknown-block(%u,%u)",
404 MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
405 if (root_fs_names)
406 num_fs = split_fs_names(fs_names, PAGE_SIZE, root_fs_names);
407 else
408 num_fs = list_bdev_fs_names(fs_names, PAGE_SIZE);
409 retry:
410 for (i = 0, p = fs_names; i < num_fs; i++, p += strlen(p)+1) {
411 int err;
412
413 if (!*p)
414 continue;
415 err = do_mount_root(name, p, flags, root_mount_data);
416 switch (err) {
417 case 0:
418 goto out;
419 case -EACCES:
420 case -EINVAL:
421 continue;
422 }
423 /*
424 * Allow the user to distinguish between failed sys_open
425 * and bad superblock on root device.
426 * and give them a list of the available devices
427 */
428 printk("VFS: Cannot open root device \"%s\" or %s: error %d\n",
429 root_device_name, b, err);
430 printk("Please append a correct \"root=\" boot option; here are the available partitions:\n");
431
432 printk_all_partitions();
433 panic("VFS: Unable to mount root fs on %s", b);
434 }
435 if (!(flags & SB_RDONLY)) {
436 flags |= SB_RDONLY;
437 goto retry;
438 }
439
440 printk("List of all partitions:\n");
441 printk_all_partitions();
442 printk("No filesystem could mount root, tried: ");
443 for (i = 0, p = fs_names; i < num_fs; i++, p += strlen(p)+1)
444 printk(" %s", p);
445 printk("\n");
446 panic("VFS: Unable to mount root fs on %s", b);
447 out:
448 put_page(page);
449 }
450
451 #ifdef CONFIG_ROOT_NFS
452
453 #define NFSROOT_TIMEOUT_MIN 5
454 #define NFSROOT_TIMEOUT_MAX 30
455 #define NFSROOT_RETRY_MAX 5
456
mount_nfs_root(void)457 static int __init mount_nfs_root(void)
458 {
459 char *root_dev, *root_data;
460 unsigned int timeout;
461 int try, err;
462
463 err = nfs_root_data(&root_dev, &root_data);
464 if (err != 0)
465 return 0;
466
467 /*
468 * The server or network may not be ready, so try several
469 * times. Stop after a few tries in case the client wants
470 * to fall back to other boot methods.
471 */
472 timeout = NFSROOT_TIMEOUT_MIN;
473 for (try = 1; ; try++) {
474 err = do_mount_root(root_dev, "nfs",
475 root_mountflags, root_data);
476 if (err == 0)
477 return 1;
478 if (try > NFSROOT_RETRY_MAX)
479 break;
480
481 /* Wait, in case the server refused us immediately */
482 ssleep(timeout);
483 timeout <<= 1;
484 if (timeout > NFSROOT_TIMEOUT_MAX)
485 timeout = NFSROOT_TIMEOUT_MAX;
486 }
487 return 0;
488 }
489 #endif
490
491 #ifdef CONFIG_CIFS_ROOT
492
493 extern int cifs_root_data(char **dev, char **opts);
494
495 #define CIFSROOT_TIMEOUT_MIN 5
496 #define CIFSROOT_TIMEOUT_MAX 30
497 #define CIFSROOT_RETRY_MAX 5
498
mount_cifs_root(void)499 static int __init mount_cifs_root(void)
500 {
501 char *root_dev, *root_data;
502 unsigned int timeout;
503 int try, err;
504
505 err = cifs_root_data(&root_dev, &root_data);
506 if (err != 0)
507 return 0;
508
509 timeout = CIFSROOT_TIMEOUT_MIN;
510 for (try = 1; ; try++) {
511 err = do_mount_root(root_dev, "cifs", root_mountflags,
512 root_data);
513 if (err == 0)
514 return 1;
515 if (try > CIFSROOT_RETRY_MAX)
516 break;
517
518 ssleep(timeout);
519 timeout <<= 1;
520 if (timeout > CIFSROOT_TIMEOUT_MAX)
521 timeout = CIFSROOT_TIMEOUT_MAX;
522 }
523 return 0;
524 }
525 #endif
526
fs_is_nodev(char * fstype)527 static bool __init fs_is_nodev(char *fstype)
528 {
529 struct file_system_type *fs = get_fs_type(fstype);
530 bool ret = false;
531
532 if (fs) {
533 ret = !(fs->fs_flags & FS_REQUIRES_DEV);
534 put_filesystem(fs);
535 }
536
537 return ret;
538 }
539
mount_nodev_root(void)540 static int __init mount_nodev_root(void)
541 {
542 char *fs_names, *fstype;
543 int err = -EINVAL;
544 int num_fs, i;
545
546 fs_names = (void *)__get_free_page(GFP_KERNEL);
547 if (!fs_names)
548 return -EINVAL;
549 num_fs = split_fs_names(fs_names, PAGE_SIZE, root_fs_names);
550
551 for (i = 0, fstype = fs_names; i < num_fs;
552 i++, fstype += strlen(fstype) + 1) {
553 if (!*fstype)
554 continue;
555 if (!fs_is_nodev(fstype))
556 continue;
557 err = do_mount_root(root_device_name, fstype, root_mountflags,
558 root_mount_data);
559 if (!err)
560 break;
561 }
562
563 free_page((unsigned long)fs_names);
564 return err;
565 }
566
mount_root(void)567 void __init mount_root(void)
568 {
569 #ifdef CONFIG_ROOT_NFS
570 if (ROOT_DEV == Root_NFS) {
571 if (!mount_nfs_root())
572 printk(KERN_ERR "VFS: Unable to mount root fs via NFS.\n");
573 return;
574 }
575 #endif
576 #ifdef CONFIG_CIFS_ROOT
577 if (ROOT_DEV == Root_CIFS) {
578 if (!mount_cifs_root())
579 printk(KERN_ERR "VFS: Unable to mount root fs via SMB.\n");
580 return;
581 }
582 #endif
583 if (ROOT_DEV == 0 && root_device_name && root_fs_names) {
584 if (mount_nodev_root() == 0)
585 return;
586 }
587 #ifdef CONFIG_BLOCK
588 {
589 int err = create_dev("/dev/root", ROOT_DEV);
590
591 if (err < 0)
592 pr_emerg("Failed to create /dev/root: %d\n", err);
593 mount_block_root("/dev/root", root_mountflags);
594 }
595 #endif
596 }
597
598 /*
599 * Prepare the namespace - decide what/where to mount, load ramdisks, etc.
600 */
prepare_namespace(void)601 void __init prepare_namespace(void)
602 {
603 if (root_delay) {
604 printk(KERN_INFO "Waiting %d sec before mounting root device...\n",
605 root_delay);
606 ssleep(root_delay);
607 }
608
609 /*
610 * wait for the known devices to complete their probing
611 *
612 * Note: this is a potential source of long boot delays.
613 * For example, it is not atypical to wait 5 seconds here
614 * for the touchpad of a laptop to initialize.
615 */
616 wait_for_device_probe();
617
618 md_run_setup();
619
620 if (saved_root_name[0]) {
621 root_device_name = saved_root_name;
622 if (!strncmp(root_device_name, "mtd", 3) ||
623 !strncmp(root_device_name, "ubi", 3)) {
624 mount_block_root(root_device_name, root_mountflags);
625 goto out;
626 }
627 ROOT_DEV = name_to_dev_t(root_device_name);
628 if (strncmp(root_device_name, "/dev/", 5) == 0)
629 root_device_name += 5;
630 }
631
632 if (initrd_load())
633 goto out;
634
635 /* wait for any asynchronous scanning to complete */
636 if ((ROOT_DEV == 0) && root_wait) {
637 printk(KERN_INFO "Waiting for root device %s...\n",
638 saved_root_name);
639 while (driver_probe_done() != 0 ||
640 (ROOT_DEV = name_to_dev_t(saved_root_name)) == 0)
641 msleep(5);
642 async_synchronize_full();
643 }
644
645 mount_root();
646 out:
647 devtmpfs_mount();
648 init_mount(".", "/", NULL, MS_MOVE, NULL);
649 init_chroot(".");
650 }
651
652 static bool is_tmpfs;
rootfs_init_fs_context(struct fs_context * fc)653 static int rootfs_init_fs_context(struct fs_context *fc)
654 {
655 if (IS_ENABLED(CONFIG_TMPFS) && is_tmpfs)
656 return shmem_init_fs_context(fc);
657
658 return ramfs_init_fs_context(fc);
659 }
660
661 struct file_system_type rootfs_fs_type = {
662 .name = "rootfs",
663 .init_fs_context = rootfs_init_fs_context,
664 .kill_sb = kill_litter_super,
665 };
666
init_rootfs(void)667 void __init init_rootfs(void)
668 {
669 if (IS_ENABLED(CONFIG_TMPFS) && !saved_root_name[0] &&
670 (!root_fs_names || strstr(root_fs_names, "tmpfs")))
671 is_tmpfs = true;
672 }
673