init/do_mounts.c

   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 <uapi/linux/mount.h>
  27
  28 #include "do_mounts.h"
  29
  30 int __initdata rd_doload;       /* 1 = load RAM disk, 0 = don't load */
  31
  32 int root_mountflags = MS_RDONLY | MS_SILENT;
  33 static char * __initdata root_device_name;
  34 static char __initdata saved_root_name[64];
  35 static int root_wait;
  36
  37 dev_t ROOT_DEV;
  38
  39 static int __init load_ramdisk(char *str)
  40 {
  41         rd_doload = simple_strtol(str,NULL,0) & 3;
  42         return 1;
  43 }
  44 __setup("load_ramdisk=", load_ramdisk);
  45
  46 static int __init readonly(char *str)
  47 {
  48         if (*str)
  49                 return 0;
  50         root_mountflags |= MS_RDONLY;
  51         return 1;
  52 }
  53
  54 static int __init readwrite(char *str)
  55 {
  56         if (*str)
  57                 return 0;
  58         root_mountflags &= ~MS_RDONLY;
  59         return 1;
  60 }
  61
  62 __setup("ro", readonly);
  63 __setup("rw", readwrite);
  64
  65 #ifdef CONFIG_BLOCK
  66 struct uuidcmp {
  67         const char *uuid;
  68         int len;
  69 };
  70
  71 /**
  72  * match_dev_by_uuid - callback for finding a partition using its uuid
  73  * @dev:        device passed in by the caller
  74  * @data:       opaque pointer to the desired struct uuidcmp to match
  75  *
  76  * Returns 1 if the device matches, and 0 otherwise.
  77  */
  78 static int match_dev_by_uuid(struct device *dev, const void *data)
  79 {
  80         const struct uuidcmp *cmp = data;
  81         struct hd_struct *part = dev_to_part(dev);
  82
  83         if (!part->info)
  84                 goto no_match;
  85
  86         if (strncasecmp(cmp->uuid, part->info->uuid, cmp->len))
  87                 goto no_match;
  88
  89         return 1;
  90 no_match:
  91         return 0;
  92 }
  93
  94
  95 /**
  96  * devt_from_partuuid - looks up the dev_t of a partition by its UUID
  97  * @uuid_str:   char array containing ascii UUID
  98  *
  99  * The function will return the first partition which contains a matching
 100  * UUID value in its partition_meta_info struct.  This does not search
 101  * by filesystem UUIDs.
 102  *
 103  * If @uuid_str is followed by a "/PARTNROFF=%d", then the number will be
 104  * extracted and used as an offset from the partition identified by the UUID.
 105  *
 106  * Returns the matching dev_t on success or 0 on failure.
 107  */
 108 static dev_t devt_from_partuuid(const char *uuid_str)
 109 {
 110         dev_t res = 0;
 111         struct uuidcmp cmp;
 112         struct device *dev = NULL;
 113         struct gendisk *disk;
 114         struct hd_struct *part;
 115         int offset = 0;
 116         bool clear_root_wait = false;
 117         char *slash;
 118
 119         cmp.uuid = uuid_str;
 120
 121         slash = strchr(uuid_str, '/');
 122         /* Check for optional partition number offset attributes. */
 123         if (slash) {
 124                 char c = 0;
 125                 /* Explicitly fail on poor PARTUUID syntax. */
 126                 if (sscanf(slash + 1,
 127                            "PARTNROFF=%d%c", &offset, &c) != 1) {
 128                         clear_root_wait = true;
 129                         goto done;
 130                 }
 131                 cmp.len = slash - uuid_str;
 132         } else {
 133                 cmp.len = strlen(uuid_str);
 134         }
 135
 136         if (!cmp.len) {
 137                 clear_root_wait = true;
 138                 goto done;
 139         }
 140
 141         dev = class_find_device(&block_class, NULL, &cmp,
 142                                 &match_dev_by_uuid);
 143         if (!dev)
 144                 goto done;
 145
 146         res = dev->devt;
 147
 148         /* Attempt to find the partition by offset. */
 149         if (!offset)
 150                 goto no_offset;
 151
 152         res = 0;
 153         disk = part_to_disk(dev_to_part(dev));
 154         part = disk_get_part(disk, dev_to_part(dev)->partno + offset);
 155         if (part) {
 156                 res = part_devt(part);
 157                 put_device(part_to_dev(part));
 158         }
 159
 160 no_offset:
 161         put_device(dev);
 162 done:
 163         if (clear_root_wait) {
 164                 pr_err("VFS: PARTUUID= is invalid.\n"
 165                        "Expected PARTUUID=<valid-uuid-id>[/PARTNROFF=%%d]\n");
 166                 if (root_wait)
 167                         pr_err("Disabling rootwait; root= is invalid.\n");
 168                 root_wait = 0;
 169         }
 170         return res;
 171 }
 172
 173 /**
 174  * match_dev_by_label - callback for finding a partition using its label
 175  * @dev:        device passed in by the caller
 176  * @data:       opaque pointer to the label to match
 177  *
 178  * Returns 1 if the device matches, and 0 otherwise.
 179  */
 180 static int match_dev_by_label(struct device *dev, const void *data)
 181 {
 182         const char *label = data;
 183         struct hd_struct *part = dev_to_part(dev);
 184
 185         if (part->info && !strcmp(label, part->info->volname))
 186                 return 1;
 187
 188         return 0;
 189 }
 190 #endif
 191
 192 /*
 193  *      Convert a name into device number.  We accept the following variants:
 194  *
 195  *      1) <hex_major><hex_minor> device number in hexadecimal represents itself
 196  *         no leading 0x, for example b302.
 197  *      2) /dev/nfs represents Root_NFS (0xff)
 198  *      3) /dev/<disk_name> represents the device number of disk
 199  *      4) /dev/<disk_name><decimal> represents the device number
 200  *         of partition - device number of disk plus the partition number
 201  *      5) /dev/<disk_name>p<decimal> - same as the above, that form is
 202  *         used when disk name of partitioned disk ends on a digit.
 203  *      6) PARTUUID=00112233-4455-6677-8899-AABBCCDDEEFF representing the
 204  *         unique id of a partition if the partition table provides it.
 205  *         The UUID may be either an EFI/GPT UUID, or refer to an MSDOS
 206  *         partition using the format SSSSSSSS-PP, where SSSSSSSS is a zero-
 207  *         filled hex representation of the 32-bit "NT disk signature", and PP
 208  *         is a zero-filled hex representation of the 1-based partition number.
 209  *      7) PARTUUID=<UUID>/PARTNROFF=<int> to select a partition in relation to
 210  *         a partition with a known unique id.
 211  *      8) <major>:<minor> major and minor number of the device separated by
 212  *         a colon.
 213  *      9) PARTLABEL=<name> with name being the GPT partition label.
 214  *         MSDOS partitions do not support labels!
 215  *      10) /dev/cifs represents Root_CIFS (0xfe)
 216  *
 217  *      If name doesn't have fall into the categories above, we return (0,0).
 218  *      block_class is used to check if something is a disk name. If the disk
 219  *      name contains slashes, the device name has them replaced with
 220  *      bangs.
 221  */
 222
 223 dev_t name_to_dev_t(const char *name)
 224 {
 225         char s[32];
 226         char *p;
 227         dev_t res = 0;
 228         int part;
 229
 230 #ifdef CONFIG_BLOCK
 231         if (strncmp(name, "PARTUUID=", 9) == 0) {
 232                 name += 9;
 233                 res = devt_from_partuuid(name);
 234                 if (!res)
 235                         goto fail;
 236                 goto done;
 237         } else if (strncmp(name, "PARTLABEL=", 10) == 0) {
 238                 struct device *dev;
 239
 240                 dev = class_find_device(&block_class, NULL, name + 10,
 241                                         &match_dev_by_label);
 242                 if (!dev)
 243                         goto fail;
 244
 245                 res = dev->devt;
 246                 put_device(dev);
 247                 goto done;
 248         }
 249 #endif
 250
 251         if (strncmp(name, "/dev/", 5) != 0) {
 252                 unsigned maj, min, offset;
 253                 char dummy;
 254
 255                 if ((sscanf(name, "%u:%u%c", &maj, &min, &dummy) == 2) ||
 256                     (sscanf(name, "%u:%u:%u:%c", &maj, &min, &offset, &dummy) == 3)) {
 257                         res = MKDEV(maj, min);
 258                         if (maj != MAJOR(res) || min != MINOR(res))
 259                                 goto fail;
 260                 } else {
 261                         res = new_decode_dev(simple_strtoul(name, &p, 16));
 262                         if (*p)
 263                                 goto fail;
 264                 }
 265                 goto done;
 266         }
 267
 268         name += 5;
 269         res = Root_NFS;
 270         if (strcmp(name, "nfs") == 0)
 271                 goto done;
 272         res = Root_CIFS;
 273         if (strcmp(name, "cifs") == 0)
 274                 goto done;
 275         res = Root_RAM0;
 276         if (strcmp(name, "ram") == 0)
 277                 goto done;
 278
 279         if (strlen(name) > 31)
 280                 goto fail;
 281         strcpy(s, name);
 282         for (p = s; *p; p++)
 283                 if (*p == '/')
 284                         *p = '!';
 285         res = blk_lookup_devt(s, 0);
 286         if (res)
 287                 goto done;
 288
 289         /*
 290          * try non-existent, but valid partition, which may only exist
 291          * after revalidating the disk, like partitioned md devices
 292          */
 293         while (p > s && isdigit(p[-1]))
 294                 p--;
 295         if (p == s || !*p || *p == '0')
 296                 goto fail;
 297
 298         /* try disk name without <part number> */
 299         part = simple_strtoul(p, NULL, 10);
 300         *p = '\0';
 301         res = blk_lookup_devt(s, part);
 302         if (res)
 303                 goto done;
 304
 305         /* try disk name without p<part number> */
 306         if (p < s + 2 || !isdigit(p[-2]) || p[-1] != 'p')
 307                 goto fail;
 308         p[-1] = '\0';
 309         res = blk_lookup_devt(s, part);
 310         if (res)
 311                 goto done;
 312
 313 fail:
 314         return 0;
 315 done:
 316         return res;
 317 }
 318 EXPORT_SYMBOL_GPL(name_to_dev_t);
 319
 320 static int __init root_dev_setup(char *line)
 321 {
 322         strlcpy(saved_root_name, line, sizeof(saved_root_name));
 323         return 1;
 324 }
 325
 326 __setup("root=", root_dev_setup);
 327
 328 static int __init rootwait_setup(char *str)
 329 {
 330         if (*str)
 331                 return 0;
 332         root_wait = 1;
 333         return 1;
 334 }
 335
 336 __setup("rootwait", rootwait_setup);
 337
 338 static char * __initdata root_mount_data;
 339 static int __init root_data_setup(char *str)
 340 {
 341         root_mount_data = str;
 342         return 1;
 343 }
 344
 345 static char * __initdata root_fs_names;
 346 static int __init fs_names_setup(char *str)
 347 {
 348         root_fs_names = str;
 349         return 1;
 350 }
 351
 352 static unsigned int __initdata root_delay;
 353 static int __init root_delay_setup(char *str)
 354 {
 355         root_delay = simple_strtoul(str, NULL, 0);
 356         return 1;
 357 }
 358
 359 __setup("rootflags=", root_data_setup);
 360 __setup("rootfstype=", fs_names_setup);
 361 __setup("rootdelay=", root_delay_setup);
 362
 363 static void __init get_fs_names(char *page)
 364 {
 365         char *s = page;
 366
 367         if (root_fs_names) {
 368                 strcpy(page, root_fs_names);
 369                 while (*s++) {
 370                         if (s[-1] == ',')
 371                                 s[-1] = '\0';
 372                 }
 373         } else {
 374                 int len = get_filesystem_list(page);
 375                 char *p, *next;
 376
 377                 page[len] = '\0';
 378                 for (p = page-1; p; p = next) {
 379                         next = strchr(++p, '\n');
 380                         if (*p++ != '\t')
 381                                 continue;
 382                         while ((*s++ = *p++) != '\n')
 383                                 ;
 384                         s[-1] = '\0';
 385                 }
 386         }
 387         *s = '\0';
 388 }
 389
 390 static int __init do_mount_root(const char *name, const char *fs,
 391                                  const int flags, const void *data)
 392 {
 393         struct super_block *s;
 394         struct page *p = NULL;
 395         char *data_page = NULL;
 396         int ret;
 397
 398         if (data) {
 399                 /* do_mount() requires a full page as fifth argument */
 400                 p = alloc_page(GFP_KERNEL);
 401                 if (!p)
 402                         return -ENOMEM;
 403                 data_page = page_address(p);
 404                 /* zero-pad. do_mount() will make sure it's terminated */
 405                 strncpy(data_page, data, PAGE_SIZE);
 406         }
 407
 408         ret = do_mount(name, "/root", fs, flags, data_page);
 409         if (ret)
 410                 goto out;
 411
 412         ksys_chdir("/root");
 413         s = current->fs->pwd.dentry->d_sb;
 414         ROOT_DEV = s->s_dev;
 415         printk(KERN_INFO
 416                "VFS: Mounted root (%s filesystem)%s on device %u:%u.\n",
 417                s->s_type->name,
 418                sb_rdonly(s) ? " readonly" : "",
 419                MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
 420
 421 out:
 422         if (p)
 423                 put_page(p);
 424         return ret;
 425 }
 426
 427 void __init mount_block_root(char *name, int flags)
 428 {
 429         struct page *page = alloc_page(GFP_KERNEL);
 430         char *fs_names = page_address(page);
 431         char *p;
 432         char b[BDEVNAME_SIZE];
 433
 434         scnprintf(b, BDEVNAME_SIZE, "unknown-block(%u,%u)",
 435                   MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
 436         get_fs_names(fs_names);
 437 retry:
 438         for (p = fs_names; *p; p += strlen(p)+1) {
 439                 int err = do_mount_root(name, p, flags, root_mount_data);
 440                 switch (err) {
 441                         case 0:
 442                                 goto out;
 443                         case -EACCES:
 444                         case -EINVAL:
 445                                 continue;
 446                 }
 447                 /*
 448                  * Allow the user to distinguish between failed sys_open
 449                  * and bad superblock on root device.
 450                  * and give them a list of the available devices
 451                  */
 452                 printk("VFS: Cannot open root device \"%s\" or %s: error %d\n",
 453                                 root_device_name, b, err);
 454                 printk("Please append a correct \"root=\" boot option; here are the available partitions:\n");
 455
 456                 printk_all_partitions();
 457 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
 458                 printk("DEBUG_BLOCK_EXT_DEVT is enabled, you need to specify "
 459                        "explicit textual name for \"root=\" boot option.\n");
 460 #endif
 461                 panic("VFS: Unable to mount root fs on %s", b);
 462         }
 463         if (!(flags & SB_RDONLY)) {
 464                 flags |= SB_RDONLY;
 465                 goto retry;
 466         }
 467
 468         printk("List of all partitions:\n");
 469         printk_all_partitions();
 470         printk("No filesystem could mount root, tried: ");
 471         for (p = fs_names; *p; p += strlen(p)+1)
 472                 printk(" %s", p);
 473         printk("\n");
 474         panic("VFS: Unable to mount root fs on %s", b);
 475 out:
 476         put_page(page);
 477 }
 478
 479 #ifdef CONFIG_ROOT_NFS
 480
 481 #define NFSROOT_TIMEOUT_MIN     5
 482 #define NFSROOT_TIMEOUT_MAX     30
 483 #define NFSROOT_RETRY_MAX       5
 484
 485 static int __init mount_nfs_root(void)
 486 {
 487         char *root_dev, *root_data;
 488         unsigned int timeout;
 489         int try, err;
 490
 491         err = nfs_root_data(&root_dev, &root_data);
 492         if (err != 0)
 493                 return 0;
 494
 495         /*
 496          * The server or network may not be ready, so try several
 497          * times.  Stop after a few tries in case the client wants
 498          * to fall back to other boot methods.
 499          */
 500         timeout = NFSROOT_TIMEOUT_MIN;
 501         for (try = 1; ; try++) {
 502                 err = do_mount_root(root_dev, "nfs",
 503                                         root_mountflags, root_data);
 504                 if (err == 0)
 505                         return 1;
 506                 if (try > NFSROOT_RETRY_MAX)
 507                         break;
 508
 509                 /* Wait, in case the server refused us immediately */
 510                 ssleep(timeout);
 511                 timeout <<= 1;
 512                 if (timeout > NFSROOT_TIMEOUT_MAX)
 513                         timeout = NFSROOT_TIMEOUT_MAX;
 514         }
 515         return 0;
 516 }
 517 #endif
 518
 519 #ifdef CONFIG_CIFS_ROOT
 520
 521 extern int cifs_root_data(char **dev, char **opts);
 522
 523 #define CIFSROOT_TIMEOUT_MIN    5
 524 #define CIFSROOT_TIMEOUT_MAX    30
 525 #define CIFSROOT_RETRY_MAX      5
 526
 527 static int __init mount_cifs_root(void)
 528 {
 529         char *root_dev, *root_data;
 530         unsigned int timeout;
 531         int try, err;
 532
 533         err = cifs_root_data(&root_dev, &root_data);
 534         if (err != 0)
 535                 return 0;
 536
 537         timeout = CIFSROOT_TIMEOUT_MIN;
 538         for (try = 1; ; try++) {
 539                 err = do_mount_root(root_dev, "cifs", root_mountflags,
 540                                     root_data);
 541                 if (err == 0)
 542                         return 1;
 543                 if (try > CIFSROOT_RETRY_MAX)
 544                         break;
 545
 546                 ssleep(timeout);
 547                 timeout <<= 1;
 548                 if (timeout > CIFSROOT_TIMEOUT_MAX)
 549                         timeout = CIFSROOT_TIMEOUT_MAX;
 550         }
 551         return 0;
 552 }
 553 #endif
 554
 555 #if defined(CONFIG_BLK_DEV_RAM) || defined(CONFIG_BLK_DEV_FD)
 556 void __init change_floppy(char *fmt, ...)
 557 {
 558         struct termios termios;
 559         char buf[80];
 560         char c;
 561         int fd;
 562         va_list args;
 563         va_start(args, fmt);
 564         vsprintf(buf, fmt, args);
 565         va_end(args);
 566         fd = ksys_open("/dev/root", O_RDWR | O_NDELAY, 0);
 567         if (fd >= 0) {
 568                 ksys_ioctl(fd, FDEJECT, 0);
 569                 ksys_close(fd);
 570         }
 571         printk(KERN_NOTICE "VFS: Insert %s and press ENTER\n", buf);
 572         fd = ksys_open("/dev/console", O_RDWR, 0);
 573         if (fd >= 0) {
 574                 ksys_ioctl(fd, TCGETS, (long)&termios);
 575                 termios.c_lflag &= ~ICANON;
 576                 ksys_ioctl(fd, TCSETSF, (long)&termios);
 577                 ksys_read(fd, &c, 1);
 578                 termios.c_lflag |= ICANON;
 579                 ksys_ioctl(fd, TCSETSF, (long)&termios);
 580                 ksys_close(fd);
 581         }
 582 }
 583 #endif
 584
 585 void __init mount_root(void)
 586 {
 587 #ifdef CONFIG_ROOT_NFS
 588         if (ROOT_DEV == Root_NFS) {
 589                 if (mount_nfs_root())
 590                         return;
 591
 592                 printk(KERN_ERR "VFS: Unable to mount root fs via NFS, trying floppy.\n");
 593                 ROOT_DEV = Root_FD0;
 594         }
 595 #endif
 596 #ifdef CONFIG_CIFS_ROOT
 597         if (ROOT_DEV == Root_CIFS) {
 598                 if (mount_cifs_root())
 599                         return;
 600
 601                 printk(KERN_ERR "VFS: Unable to mount root fs via SMB, trying floppy.\n");
 602                 ROOT_DEV = Root_FD0;
 603         }
 604 #endif
 605 #ifdef CONFIG_BLK_DEV_FD
 606         if (MAJOR(ROOT_DEV) == FLOPPY_MAJOR) {
 607                 /* rd_doload is 2 for a dual initrd/ramload setup */
 608                 if (rd_doload==2) {
 609                         if (rd_load_disk(1)) {
 610                                 ROOT_DEV = Root_RAM1;
 611                                 root_device_name = NULL;
 612                         }
 613                 } else
 614                         change_floppy("root floppy");
 615         }
 616 #endif
 617 #ifdef CONFIG_BLOCK
 618         {
 619                 int err = create_dev("/dev/root", ROOT_DEV);
 620
 621                 if (err < 0)
 622                         pr_emerg("Failed to create /dev/root: %d\n", err);
 623                 mount_block_root("/dev/root", root_mountflags);
 624         }
 625 #endif
 626 }
 627
 628 /*
 629  * Prepare the namespace - decide what/where to mount, load ramdisks, etc.
 630  */
 631 void __init prepare_namespace(void)
 632 {
 633         int is_floppy;
 634
 635         if (root_delay) {
 636                 printk(KERN_INFO "Waiting %d sec before mounting root device...\n",
 637                        root_delay);
 638                 ssleep(root_delay);
 639         }
 640
 641         /*
 642          * wait for the known devices to complete their probing
 643          *
 644          * Note: this is a potential source of long boot delays.
 645          * For example, it is not atypical to wait 5 seconds here
 646          * for the touchpad of a laptop to initialize.
 647          */
 648         wait_for_device_probe();
 649
 650         md_run_setup();
 651
 652         if (saved_root_name[0]) {
 653                 root_device_name = saved_root_name;
 654                 if (!strncmp(root_device_name, "mtd", 3) ||
 655                     !strncmp(root_device_name, "ubi", 3)) {
 656                         mount_block_root(root_device_name, root_mountflags);
 657                         goto out;
 658                 }
 659                 ROOT_DEV = name_to_dev_t(root_device_name);
 660                 if (strncmp(root_device_name, "/dev/", 5) == 0)
 661                         root_device_name += 5;
 662         }
 663
 664         if (initrd_load())
 665                 goto out;
 666
 667         /* wait for any asynchronous scanning to complete */
 668         if ((ROOT_DEV == 0) && root_wait) {
 669                 printk(KERN_INFO "Waiting for root device %s...\n",
 670                         saved_root_name);
 671                 while (driver_probe_done() != 0 ||
 672                         (ROOT_DEV = name_to_dev_t(saved_root_name)) == 0)
 673                         msleep(5);
 674                 async_synchronize_full();
 675         }
 676
 677         is_floppy = MAJOR(ROOT_DEV) == FLOPPY_MAJOR;
 678
 679         if (is_floppy && rd_doload && rd_load_disk(0))
 680                 ROOT_DEV = Root_RAM0;
 681
 682         mount_root();
 683 out:
 684         devtmpfs_mount();
 685         do_mount(".", "/", NULL, MS_MOVE, NULL);
 686         ksys_chroot(".");
 687 }
 688
 689 static bool is_tmpfs;
 690 static int rootfs_init_fs_context(struct fs_context *fc)
 691 {
 692         if (IS_ENABLED(CONFIG_TMPFS) && is_tmpfs)
 693                 return shmem_init_fs_context(fc);
 694
 695         return ramfs_init_fs_context(fc);
 696 }
 697
 698 struct file_system_type rootfs_fs_type = {
 699         .name           = "rootfs",
 700         .init_fs_context = rootfs_init_fs_context,
 701         .kill_sb        = kill_litter_super,
 702 };
 703
 704 void __init init_rootfs(void)
 705 {
 706         if (IS_ENABLED(CONFIG_TMPFS) && !saved_root_name[0] &&
 707                 (!root_fs_names || strstr(root_fs_names, "tmpfs")))
 708                 is_tmpfs = true;
 709 }