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
load_ramdisk(char * str)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
readonly(char * str)54 static int __init readonly(char *str)
55 {
56 if (*str)
57 return 0;
58 root_mountflags |= MS_RDONLY;
59 return 1;
60 }
61
readwrite(char * str)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 */
match_dev_by_uuid(struct device * dev,const void * data)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 */
devt_from_partuuid(const char * uuid_str)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
name_to_dev_t(const char * name)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;
229
230 if (sscanf(name, "%u:%u", &maj, &min) == 2) {
231 res = MKDEV(maj, min);
232 if (maj != MAJOR(res) || min != MINOR(res))
233 goto fail;
234 } else {
235 res = new_decode_dev(simple_strtoul(name, &p, 16));
236 if (*p)
237 goto fail;
238 }
239 goto done;
240 }
241
242 name += 5;
243 res = Root_NFS;
244 if (strcmp(name, "nfs") == 0)
245 goto done;
246 res = Root_RAM0;
247 if (strcmp(name, "ram") == 0)
248 goto done;
249
250 if (strlen(name) > 31)
251 goto fail;
252 strcpy(s, name);
253 for (p = s; *p; p++)
254 if (*p == '/')
255 *p = '!';
256 res = blk_lookup_devt(s, 0);
257 if (res)
258 goto done;
259
260 /*
261 * try non-existent, but valid partition, which may only exist
262 * after revalidating the disk, like partitioned md devices
263 */
264 while (p > s && isdigit(p[-1]))
265 p--;
266 if (p == s || !*p || *p == '0')
267 goto fail;
268
269 /* try disk name without <part number> */
270 part = simple_strtoul(p, NULL, 10);
271 *p = '\0';
272 res = blk_lookup_devt(s, part);
273 if (res)
274 goto done;
275
276 /* try disk name without p<part number> */
277 if (p < s + 2 || !isdigit(p[-2]) || p[-1] != 'p')
278 goto fail;
279 p[-1] = '\0';
280 res = blk_lookup_devt(s, part);
281 if (res)
282 goto done;
283
284 fail:
285 return 0;
286 done:
287 return res;
288 }
289 EXPORT_SYMBOL_GPL(name_to_dev_t);
290
root_dev_setup(char * line)291 static int __init root_dev_setup(char *line)
292 {
293 strlcpy(saved_root_name, line, sizeof(saved_root_name));
294 return 1;
295 }
296
297 __setup("root=", root_dev_setup);
298
rootwait_setup(char * str)299 static int __init rootwait_setup(char *str)
300 {
301 if (*str)
302 return 0;
303 root_wait = 1;
304 return 1;
305 }
306
307 __setup("rootwait", rootwait_setup);
308
309 static char * __initdata root_mount_data;
root_data_setup(char * str)310 static int __init root_data_setup(char *str)
311 {
312 root_mount_data = str;
313 return 1;
314 }
315
316 static char * __initdata root_fs_names;
fs_names_setup(char * str)317 static int __init fs_names_setup(char *str)
318 {
319 root_fs_names = str;
320 return 1;
321 }
322
323 static unsigned int __initdata root_delay;
root_delay_setup(char * str)324 static int __init root_delay_setup(char *str)
325 {
326 root_delay = simple_strtoul(str, NULL, 0);
327 return 1;
328 }
329
330 __setup("rootflags=", root_data_setup);
331 __setup("rootfstype=", fs_names_setup);
332 __setup("rootdelay=", root_delay_setup);
333
get_fs_names(char * page)334 static void __init get_fs_names(char *page)
335 {
336 char *s = page;
337
338 if (root_fs_names) {
339 strcpy(page, root_fs_names);
340 while (*s++) {
341 if (s[-1] == ',')
342 s[-1] = '\0';
343 }
344 } else {
345 int len = get_filesystem_list(page);
346 char *p, *next;
347
348 page[len] = '\0';
349 for (p = page-1; p; p = next) {
350 next = strchr(++p, '\n');
351 if (*p++ != '\t')
352 continue;
353 while ((*s++ = *p++) != '\n')
354 ;
355 s[-1] = '\0';
356 }
357 }
358 *s = '\0';
359 }
360
do_mount_root(char * name,char * fs,int flags,void * data)361 static int __init do_mount_root(char *name, char *fs, int flags, void *data)
362 {
363 struct super_block *s;
364 int err = sys_mount(name, "/root", fs, flags, data);
365 if (err)
366 return err;
367
368 sys_chdir("/root");
369 s = current->fs->pwd.dentry->d_sb;
370 ROOT_DEV = s->s_dev;
371 printk(KERN_INFO
372 "VFS: Mounted root (%s filesystem)%s on device %u:%u.\n",
373 s->s_type->name,
374 s->s_flags & MS_RDONLY ? " readonly" : "",
375 MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
376 return 0;
377 }
378
mount_block_root(char * name,int flags)379 void __init mount_block_root(char *name, int flags)
380 {
381 struct page *page = alloc_page(GFP_KERNEL |
382 __GFP_NOTRACK_FALSE_POSITIVE);
383 char *fs_names = page_address(page);
384 char *p;
385 #ifdef CONFIG_BLOCK
386 char b[BDEVNAME_SIZE];
387 #else
388 const char *b = name;
389 #endif
390
391 get_fs_names(fs_names);
392 retry:
393 for (p = fs_names; *p; p += strlen(p)+1) {
394 int err = do_mount_root(name, p, flags, root_mount_data);
395 switch (err) {
396 case 0:
397 goto out;
398 case -EACCES:
399 flags |= MS_RDONLY;
400 goto retry;
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
424 printk("List of all partitions:\n");
425 printk_all_partitions();
426 printk("No filesystem could mount root, tried: ");
427 for (p = fs_names; *p; p += strlen(p)+1)
428 printk(" %s", p);
429 printk("\n");
430 #ifdef CONFIG_BLOCK
431 __bdevname(ROOT_DEV, b);
432 #endif
433 panic("VFS: Unable to mount root fs on %s", b);
434 out:
435 put_page(page);
436 }
437
438 #ifdef CONFIG_ROOT_NFS
439
440 #define NFSROOT_TIMEOUT_MIN 5
441 #define NFSROOT_TIMEOUT_MAX 30
442 #define NFSROOT_RETRY_MAX 5
443
mount_nfs_root(void)444 static int __init mount_nfs_root(void)
445 {
446 char *root_dev, *root_data;
447 unsigned int timeout;
448 int try, err;
449
450 err = nfs_root_data(&root_dev, &root_data);
451 if (err != 0)
452 return 0;
453
454 /*
455 * The server or network may not be ready, so try several
456 * times. Stop after a few tries in case the client wants
457 * to fall back to other boot methods.
458 */
459 timeout = NFSROOT_TIMEOUT_MIN;
460 for (try = 1; ; try++) {
461 err = do_mount_root(root_dev, "nfs",
462 root_mountflags, root_data);
463 if (err == 0)
464 return 1;
465 if (try > NFSROOT_RETRY_MAX)
466 break;
467
468 /* Wait, in case the server refused us immediately */
469 ssleep(timeout);
470 timeout <<= 1;
471 if (timeout > NFSROOT_TIMEOUT_MAX)
472 timeout = NFSROOT_TIMEOUT_MAX;
473 }
474 return 0;
475 }
476 #endif
477
478 #if defined(CONFIG_BLK_DEV_RAM) || defined(CONFIG_BLK_DEV_FD)
change_floppy(char * fmt,...)479 void __init change_floppy(char *fmt, ...)
480 {
481 struct termios termios;
482 char buf[80];
483 char c;
484 int fd;
485 va_list args;
486 va_start(args, fmt);
487 vsprintf(buf, fmt, args);
488 va_end(args);
489 fd = sys_open("/dev/root", O_RDWR | O_NDELAY, 0);
490 if (fd >= 0) {
491 sys_ioctl(fd, FDEJECT, 0);
492 sys_close(fd);
493 }
494 printk(KERN_NOTICE "VFS: Insert %s and press ENTER\n", buf);
495 fd = sys_open("/dev/console", O_RDWR, 0);
496 if (fd >= 0) {
497 sys_ioctl(fd, TCGETS, (long)&termios);
498 termios.c_lflag &= ~ICANON;
499 sys_ioctl(fd, TCSETSF, (long)&termios);
500 sys_read(fd, &c, 1);
501 termios.c_lflag |= ICANON;
502 sys_ioctl(fd, TCSETSF, (long)&termios);
503 sys_close(fd);
504 }
505 }
506 #endif
507
mount_root(void)508 void __init mount_root(void)
509 {
510 #ifdef CONFIG_ROOT_NFS
511 if (ROOT_DEV == Root_NFS) {
512 if (mount_nfs_root())
513 return;
514
515 printk(KERN_ERR "VFS: Unable to mount root fs via NFS, trying floppy.\n");
516 ROOT_DEV = Root_FD0;
517 }
518 #endif
519 #ifdef CONFIG_BLK_DEV_FD
520 if (MAJOR(ROOT_DEV) == FLOPPY_MAJOR) {
521 /* rd_doload is 2 for a dual initrd/ramload setup */
522 if (rd_doload==2) {
523 if (rd_load_disk(1)) {
524 ROOT_DEV = Root_RAM1;
525 root_device_name = NULL;
526 }
527 } else
528 change_floppy("root floppy");
529 }
530 #endif
531 #ifdef CONFIG_BLOCK
532 create_dev("/dev/root", ROOT_DEV);
533 mount_block_root("/dev/root", root_mountflags);
534 #endif
535 }
536
537 /*
538 * Prepare the namespace - decide what/where to mount, load ramdisks, etc.
539 */
prepare_namespace(void)540 void __init prepare_namespace(void)
541 {
542 int is_floppy;
543
544 if (root_delay) {
545 printk(KERN_INFO "Waiting %d sec before mounting root device...\n",
546 root_delay);
547 ssleep(root_delay);
548 }
549
550 /*
551 * wait for the known devices to complete their probing
552 *
553 * Note: this is a potential source of long boot delays.
554 * For example, it is not atypical to wait 5 seconds here
555 * for the touchpad of a laptop to initialize.
556 */
557 wait_for_device_probe();
558
559 md_run_setup();
560 dm_run_setup();
561
562 if (saved_root_name[0]) {
563 root_device_name = saved_root_name;
564 if (!strncmp(root_device_name, "mtd", 3) ||
565 !strncmp(root_device_name, "ubi", 3)) {
566 mount_block_root(root_device_name, root_mountflags);
567 goto out;
568 }
569 ROOT_DEV = name_to_dev_t(root_device_name);
570 if (strncmp(root_device_name, "/dev/", 5) == 0)
571 root_device_name += 5;
572 }
573
574 if (initrd_load())
575 goto out;
576
577 /* wait for any asynchronous scanning to complete */
578 if ((ROOT_DEV == 0) && root_wait) {
579 printk(KERN_INFO "Waiting for root device %s...\n",
580 saved_root_name);
581 while (driver_probe_done() != 0 ||
582 (ROOT_DEV = name_to_dev_t(saved_root_name)) == 0)
583 msleep(100);
584 async_synchronize_full();
585 }
586
587 is_floppy = MAJOR(ROOT_DEV) == FLOPPY_MAJOR;
588
589 if (is_floppy && rd_doload && rd_load_disk(0))
590 ROOT_DEV = Root_RAM0;
591
592 mount_root();
593 out:
594 devtmpfs_mount("dev");
595 sys_mount(".", "/", NULL, MS_MOVE, NULL);
596 sys_chroot(".");
597 }
598
599 static bool is_tmpfs;
rootfs_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)600 static struct dentry *rootfs_mount(struct file_system_type *fs_type,
601 int flags, const char *dev_name, void *data)
602 {
603 static unsigned long once;
604 void *fill = ramfs_fill_super;
605
606 if (test_and_set_bit(0, &once))
607 return ERR_PTR(-ENODEV);
608
609 if (IS_ENABLED(CONFIG_TMPFS) && is_tmpfs)
610 fill = shmem_fill_super;
611
612 return mount_nodev(fs_type, flags, data, fill);
613 }
614
615 static struct file_system_type rootfs_fs_type = {
616 .name = "rootfs",
617 .mount = rootfs_mount,
618 .kill_sb = kill_litter_super,
619 };
620
init_rootfs(void)621 int __init init_rootfs(void)
622 {
623 int err = register_filesystem(&rootfs_fs_type);
624
625 if (err)
626 return err;
627
628 if (IS_ENABLED(CONFIG_TMPFS) && !saved_root_name[0] &&
629 (!root_fs_names || strstr(root_fs_names, "tmpfs"))) {
630 err = shmem_init();
631 is_tmpfs = true;
632 } else {
633 err = init_ramfs_fs();
634 }
635
636 if (err)
637 unregister_filesystem(&rootfs_fs_type);
638
639 return err;
640 }
641