init/do_mounts.c

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