init/do_mounts.c

   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
  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
  54 static int __init readonly(char *str)
  55 {
  56         if (*str)
  57                 return 0;
  58         root_mountflags |= MS_RDONLY;
  59         return 1;
  60 }
  61
  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  */
  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  */
 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
 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, offset;
 229                 char dummy;
 230
 231                 if ((sscanf(name, "%u:%u%c", &maj, &min, &dummy) == 2) ||
 232                     (sscanf(name, "%u:%u:%u:%c", &maj, &min, &offset, &dummy) == 3)) {
 233                         res = MKDEV(maj, min);
 234                         if (maj != MAJOR(res) || min != MINOR(res))
 235                                 goto fail;
 236                 } else {
 237                         res = new_decode_dev(simple_strtoul(name, &p, 16));
 238                         if (*p)
 239                                 goto fail;
 240                 }
 241                 goto done;
 242         }
 243
 244         name += 5;
 245         res = Root_NFS;
 246         if (strcmp(name, "nfs") == 0)
 247                 goto done;
 248         res = Root_RAM0;
 249         if (strcmp(name, "ram") == 0)
 250                 goto done;
 251
 252         if (strlen(name) > 31)
 253                 goto fail;
 254         strcpy(s, name);
 255         for (p = s; *p; p++)
 256                 if (*p == '/')
 257                         *p = '!';
 258         res = blk_lookup_devt(s, 0);
 259         if (res)
 260                 goto done;
 261
 262         /*
 263          * try non-existent, but valid partition, which may only exist
 264          * after revalidating the disk, like partitioned md devices
 265          */
 266         while (p > s && isdigit(p[-1]))
 267                 p--;
 268         if (p == s || !*p || *p == '0')
 269                 goto fail;
 270
 271         /* try disk name without <part number> */
 272         part = simple_strtoul(p, NULL, 10);
 273         *p = '\0';
 274         res = blk_lookup_devt(s, part);
 275         if (res)
 276                 goto done;
 277
 278         /* try disk name without p<part number> */
 279         if (p < s + 2 || !isdigit(p[-2]) || p[-1] != 'p')
 280                 goto fail;
 281         p[-1] = '\0';
 282         res = blk_lookup_devt(s, part);
 283         if (res)
 284                 goto done;
 285
 286 fail:
 287         return 0;
 288 done:
 289         return res;
 290 }
 291 EXPORT_SYMBOL_GPL(name_to_dev_t);
 292
 293 static int __init root_dev_setup(char *line)
 294 {
 295         strlcpy(saved_root_name, line, sizeof(saved_root_name));
 296         return 1;
 297 }
 298
 299 __setup("root=", root_dev_setup);
 300
 301 static int __init rootwait_setup(char *str)
 302 {
 303         if (*str)
 304                 return 0;
 305         root_wait = 1;
 306         return 1;
 307 }
 308
 309 __setup("rootwait", rootwait_setup);
 310
 311 static char * __initdata root_mount_data;
 312 static int __init root_data_setup(char *str)
 313 {
 314         root_mount_data = str;
 315         return 1;
 316 }
 317
 318 static char * __initdata root_fs_names;
 319 static int __init fs_names_setup(char *str)
 320 {
 321         root_fs_names = str;
 322         return 1;
 323 }
 324
 325 static unsigned int __initdata root_delay;
 326 static int __init root_delay_setup(char *str)
 327 {
 328         root_delay = simple_strtoul(str, NULL, 0);
 329         return 1;
 330 }
 331
 332 __setup("rootflags=", root_data_setup);
 333 __setup("rootfstype=", fs_names_setup);
 334 __setup("rootdelay=", root_delay_setup);
 335
 336 static void __init get_fs_names(char *page)
 337 {
 338         char *s = page;
 339
 340         if (root_fs_names) {
 341                 strcpy(page, root_fs_names);
 342                 while (*s++) {
 343                         if (s[-1] == ',')
 344                                 s[-1] = '\0';
 345                 }
 346         } else {
 347                 int len = get_filesystem_list(page);
 348                 char *p, *next;
 349
 350                 page[len] = '\0';
 351                 for (p = page-1; p; p = next) {
 352                         next = strchr(++p, '\n');
 353                         if (*p++ != '\t')
 354                                 continue;
 355                         while ((*s++ = *p++) != '\n')
 356                                 ;
 357                         s[-1] = '\0';
 358                 }
 359         }
 360         *s = '\0';
 361 }
 362
 363 static int __init do_mount_root(char *name, char *fs, int flags, void *data)
 364 {
 365         struct super_block *s;
 366         int err = sys_mount(name, "/root", fs, flags, data);
 367         if (err)
 368                 return err;
 369
 370         sys_chdir("/root");
 371         s = current->fs->pwd.dentry->d_sb;
 372         ROOT_DEV = s->s_dev;
 373         printk(KERN_INFO
 374                "VFS: Mounted root (%s filesystem)%s on device %u:%u.\n",
 375                s->s_type->name,
 376                sb_rdonly(s) ? " readonly" : "",
 377                MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
 378         return 0;
 379 }
 380
 381 void __init mount_block_root(char *name, int flags)
 382 {
 383         struct page *page = alloc_page(GFP_KERNEL);
 384         char *fs_names = page_address(page);
 385         char *p;
 386 #ifdef CONFIG_BLOCK
 387         char b[BDEVNAME_SIZE];
 388 #else
 389         const char *b = name;
 390 #endif
 391
 392         get_fs_names(fs_names);
 393 retry:
 394         for (p = fs_names; *p; p += strlen(p)+1) {
 395                 int err = do_mount_root(name, p, flags, root_mount_data);
 396                 switch (err) {
 397                         case 0:
 398                                 goto out;
 399                         case -EACCES:
 400                         case -EINVAL:
 401                                 continue;
 402                 }
 403                 /*
 404                  * Allow the user to distinguish between failed sys_open
 405                  * and bad superblock on root device.
 406                  * and give them a list of the available devices
 407                  */
 408 #ifdef CONFIG_BLOCK
 409                 __bdevname(ROOT_DEV, b);
 410 #endif
 411                 printk("VFS: Cannot open root device \"%s\" or %s: error %d\n",
 412                                 root_device_name, b, err);
 413                 printk("Please append a correct \"root=\" boot option; here are the available partitions:\n");
 414
 415                 printk_all_partitions();
 416 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
 417                 printk("DEBUG_BLOCK_EXT_DEVT is enabled, you need to specify "
 418                        "explicit textual name for \"root=\" boot option.\n");
 419 #endif
 420                 panic("VFS: Unable to mount root fs on %s", b);
 421         }
 422         if (!(flags & SB_RDONLY)) {
 423                 flags |= SB_RDONLY;
 424                 goto retry;
 425         }
 426
 427         printk("List of all partitions:\n");
 428         printk_all_partitions();
 429         printk("No filesystem could mount root, tried: ");
 430         for (p = fs_names; *p; p += strlen(p)+1)
 431                 printk(" %s", p);
 432         printk("\n");
 433 #ifdef CONFIG_BLOCK
 434         __bdevname(ROOT_DEV, b);
 435 #endif
 436         panic("VFS: Unable to mount root fs on %s", b);
 437 out:
 438         put_page(page);
 439 }
 440
 441 #ifdef CONFIG_ROOT_NFS
 442
 443 #define NFSROOT_TIMEOUT_MIN     5
 444 #define NFSROOT_TIMEOUT_MAX     30
 445 #define NFSROOT_RETRY_MAX       5
 446
 447 static int __init mount_nfs_root(void)
 448 {
 449         char *root_dev, *root_data;
 450         unsigned int timeout;
 451         int try, err;
 452
 453         err = nfs_root_data(&root_dev, &root_data);
 454         if (err != 0)
 455                 return 0;
 456
 457         /*
 458          * The server or network may not be ready, so try several
 459          * times.  Stop after a few tries in case the client wants
 460          * to fall back to other boot methods.
 461          */
 462         timeout = NFSROOT_TIMEOUT_MIN;
 463         for (try = 1; ; try++) {
 464                 err = do_mount_root(root_dev, "nfs",
 465                                         root_mountflags, root_data);
 466                 if (err == 0)
 467                         return 1;
 468                 if (try > NFSROOT_RETRY_MAX)
 469                         break;
 470
 471                 /* Wait, in case the server refused us immediately */
 472                 ssleep(timeout);
 473                 timeout <<= 1;
 474                 if (timeout > NFSROOT_TIMEOUT_MAX)
 475                         timeout = NFSROOT_TIMEOUT_MAX;
 476         }
 477         return 0;
 478 }
 479 #endif
 480
 481 #if defined(CONFIG_BLK_DEV_RAM) || defined(CONFIG_BLK_DEV_FD)
 482 void __init change_floppy(char *fmt, ...)
 483 {
 484         struct termios termios;
 485         char buf[80];
 486         char c;
 487         int fd;
 488         va_list args;
 489         va_start(args, fmt);
 490         vsprintf(buf, fmt, args);
 491         va_end(args);
 492         fd = sys_open("/dev/root", O_RDWR | O_NDELAY, 0);
 493         if (fd >= 0) {
 494                 sys_ioctl(fd, FDEJECT, 0);
 495                 sys_close(fd);
 496         }
 497         printk(KERN_NOTICE "VFS: Insert %s and press ENTER\n", buf);
 498         fd = sys_open("/dev/console", O_RDWR, 0);
 499         if (fd >= 0) {
 500                 sys_ioctl(fd, TCGETS, (long)&termios);
 501                 termios.c_lflag &= ~ICANON;
 502                 sys_ioctl(fd, TCSETSF, (long)&termios);
 503                 sys_read(fd, &c, 1);
 504                 termios.c_lflag |= ICANON;
 505                 sys_ioctl(fd, TCSETSF, (long)&termios);
 506                 sys_close(fd);
 507         }
 508 }
 509 #endif
 510
 511 void __init mount_root(void)
 512 {
 513 #ifdef CONFIG_ROOT_NFS
 514         if (ROOT_DEV == Root_NFS) {
 515                 if (mount_nfs_root())
 516                         return;
 517
 518                 printk(KERN_ERR "VFS: Unable to mount root fs via NFS, trying floppy.\n");
 519                 ROOT_DEV = Root_FD0;
 520         }
 521 #endif
 522 #ifdef CONFIG_BLK_DEV_FD
 523         if (MAJOR(ROOT_DEV) == FLOPPY_MAJOR) {
 524                 /* rd_doload is 2 for a dual initrd/ramload setup */
 525                 if (rd_doload==2) {
 526                         if (rd_load_disk(1)) {
 527                                 ROOT_DEV = Root_RAM1;
 528                                 root_device_name = NULL;
 529                         }
 530                 } else
 531                         change_floppy("root floppy");
 532         }
 533 #endif
 534 #ifdef CONFIG_BLOCK
 535         {
 536                 int err = create_dev("/dev/root", ROOT_DEV);
 537
 538                 if (err < 0)
 539                         pr_emerg("Failed to create /dev/root: %d\n", err);
 540                 mount_block_root("/dev/root", root_mountflags);
 541         }
 542 #endif
 543 }
 544
 545 /*
 546  * Prepare the namespace - decide what/where to mount, load ramdisks, etc.
 547  */
 548 void __init prepare_namespace(void)
 549 {
 550         int is_floppy;
 551         int err;
 552
 553         if (root_delay) {
 554                 printk(KERN_INFO "Waiting %d sec before mounting root device...\n",
 555                        root_delay);
 556                 ssleep(root_delay);
 557         }
 558
 559         /*
 560          * wait for the known devices to complete their probing
 561          *
 562          * Note: this is a potential source of long boot delays.
 563          * For example, it is not atypical to wait 5 seconds here
 564          * for the touchpad of a laptop to initialize.
 565          */
 566         wait_for_device_probe();
 567
 568         md_run_setup();
 569
 570         if (saved_root_name[0]) {
 571                 root_device_name = saved_root_name;
 572                 if (!strncmp(root_device_name, "mtd", 3) ||
 573                     !strncmp(root_device_name, "ubi", 3)) {
 574                         mount_block_root(root_device_name, root_mountflags);
 575                         goto out;
 576                 }
 577                 ROOT_DEV = name_to_dev_t(root_device_name);
 578                 if (strncmp(root_device_name, "/dev/", 5) == 0)
 579                         root_device_name += 5;
 580         }
 581
 582         if (initrd_load())
 583                 goto out;
 584
 585         /* wait for any asynchronous scanning to complete */
 586         if ((ROOT_DEV == 0) && root_wait) {
 587                 printk(KERN_INFO "Waiting for root device %s...\n",
 588                         saved_root_name);
 589                 while (driver_probe_done() != 0 ||
 590                         (ROOT_DEV = name_to_dev_t(saved_root_name)) == 0)
 591                         msleep(5);
 592                 async_synchronize_full();
 593         }
 594
 595         is_floppy = MAJOR(ROOT_DEV) == FLOPPY_MAJOR;
 596
 597         if (is_floppy && rd_doload && rd_load_disk(0))
 598                 ROOT_DEV = Root_RAM0;
 599
 600         mount_root();
 601 out:
 602         devtmpfs_mount("dev");
 603         sys_mount(".", "/", NULL, MS_MOVE, NULL);
 604         sys_chroot(".");
 605 #ifdef CONFIG_BLOCK
 606         /* recreate the /dev/root */
 607         err = create_dev("/dev/root", ROOT_DEV);
 608
 609         if (err < 0)
 610                 pr_emerg("Failed to create /dev/root: %d\n", err);
 611 #endif
 612 }
 613
 614 static bool is_tmpfs;
 615 static struct dentry *rootfs_mount(struct file_system_type *fs_type,
 616         int flags, const char *dev_name, void *data)
 617 {
 618         static unsigned long once;
 619         void *fill = ramfs_fill_super;
 620
 621         if (test_and_set_bit(0, &once))
 622                 return ERR_PTR(-ENODEV);
 623
 624         if (IS_ENABLED(CONFIG_TMPFS) && is_tmpfs)
 625                 fill = shmem_fill_super;
 626
 627         return mount_nodev(fs_type, flags, data, fill);
 628 }
 629
 630 static struct file_system_type rootfs_fs_type = {
 631         .name           = "rootfs",
 632         .mount          = rootfs_mount,
 633         .kill_sb        = kill_litter_super,
 634 };
 635
 636 int __init init_rootfs(void)
 637 {
 638         int err = register_filesystem(&rootfs_fs_type);
 639
 640         if (err)
 641                 return err;
 642
 643         if (IS_ENABLED(CONFIG_TMPFS) && !saved_root_name[0] &&
 644                 (!root_fs_names || strstr(root_fs_names, "tmpfs"))) {
 645                 err = shmem_init();
 646                 is_tmpfs = true;
 647         } else {
 648                 err = init_ramfs_fs();
 649         }
 650
 651         if (err)
 652                 unregister_filesystem(&rootfs_fs_type);
 653
 654         return err;
 655 }