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:       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 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) device number in hexadecimal represents itself
 186  *      2) /dev/nfs represents Root_NFS (0xff)
 187  *      3) /dev/<disk_name> represents the device number of disk
 188  *      4) /dev/<disk_name><decimal> represents the device number
 189  *         of partition - device number of disk plus the partition number
 190  *      5) /dev/<disk_name>p<decimal> - same as the above, that form is
 191  *         used when disk name of partitioned disk ends on a digit.
 192  *      6) PARTUUID=00112233-4455-6677-8899-AABBCCDDEEFF representing the
 193  *         unique id of a partition if the partition table provides it.
 194  *         The UUID may be either an EFI/GPT UUID, or refer to an MSDOS
 195  *         partition using the format SSSSSSSS-PP, where SSSSSSSS is a zero-
 196  *         filled hex representation of the 32-bit "NT disk signature", and PP
 197  *         is a zero-filled hex representation of the 1-based partition number.
 198  *      7) PARTUUID=<UUID>/PARTNROFF=<int> to select a partition in relation to
 199  *         a partition with a known unique id.
 200  *
 201  *      If name doesn't have fall into the categories above, we return (0,0).
 202  *      block_class is used to check if something is a disk name. If the disk
 203  *      name contains slashes, the device name has them replaced with
 204  *      bangs.
 205  */
 206
 207 dev_t name_to_dev_t(char *name)
 208 {
 209         char s[32];
 210         char *p;
 211         dev_t res = 0;
 212         int part;
 213
 214 #ifdef CONFIG_BLOCK
 215         if (strncmp(name, "PARTUUID=", 9) == 0) {
 216                 name += 9;
 217                 res = devt_from_partuuid(name);
 218                 if (!res)
 219                         goto fail;
 220                 goto done;
 221         }
 222 #endif
 223
 224         if (strncmp(name, "/dev/", 5) != 0) {
 225                 unsigned maj, min;
 226
 227                 if (sscanf(name, "%u:%u", &maj, &min) == 2) {
 228                         res = MKDEV(maj, min);
 229                         if (maj != MAJOR(res) || min != MINOR(res))
 230                                 goto fail;
 231                 } else {
 232                         res = new_decode_dev(simple_strtoul(name, &p, 16));
 233                         if (*p)
 234                                 goto fail;
 235                 }
 236                 goto done;
 237         }
 238
 239         name += 5;
 240         res = Root_NFS;
 241         if (strcmp(name, "nfs") == 0)
 242                 goto done;
 243         res = Root_RAM0;
 244         if (strcmp(name, "ram") == 0)
 245                 goto done;
 246
 247         if (strlen(name) > 31)
 248                 goto fail;
 249         strcpy(s, name);
 250         for (p = s; *p; p++)
 251                 if (*p == '/')
 252                         *p = '!';
 253         res = blk_lookup_devt(s, 0);
 254         if (res)
 255                 goto done;
 256
 257         /*
 258          * try non-existent, but valid partition, which may only exist
 259          * after revalidating the disk, like partitioned md devices
 260          */
 261         while (p > s && isdigit(p[-1]))
 262                 p--;
 263         if (p == s || !*p || *p == '0')
 264                 goto fail;
 265
 266         /* try disk name without <part number> */
 267         part = simple_strtoul(p, NULL, 10);
 268         *p = '\0';
 269         res = blk_lookup_devt(s, part);
 270         if (res)
 271                 goto done;
 272
 273         /* try disk name without p<part number> */
 274         if (p < s + 2 || !isdigit(p[-2]) || p[-1] != 'p')
 275                 goto fail;
 276         p[-1] = '\0';
 277         res = blk_lookup_devt(s, part);
 278         if (res)
 279                 goto done;
 280
 281 fail:
 282         return 0;
 283 done:
 284         return res;
 285 }
 286
 287 static int __init root_dev_setup(char *line)
 288 {
 289         strlcpy(saved_root_name, line, sizeof(saved_root_name));
 290         return 1;
 291 }
 292
 293 __setup("root=", root_dev_setup);
 294
 295 static int __init rootwait_setup(char *str)
 296 {
 297         if (*str)
 298                 return 0;
 299         root_wait = 1;
 300         return 1;
 301 }
 302
 303 __setup("rootwait", rootwait_setup);
 304
 305 static char * __initdata root_mount_data;
 306 static int __init root_data_setup(char *str)
 307 {
 308         root_mount_data = str;
 309         return 1;
 310 }
 311
 312 static char * __initdata root_fs_names;
 313 static int __init fs_names_setup(char *str)
 314 {
 315         root_fs_names = str;
 316         return 1;
 317 }
 318
 319 static unsigned int __initdata root_delay;
 320 static int __init root_delay_setup(char *str)
 321 {
 322         root_delay = simple_strtoul(str, NULL, 0);
 323         return 1;
 324 }
 325
 326 __setup("rootflags=", root_data_setup);
 327 __setup("rootfstype=", fs_names_setup);
 328 __setup("rootdelay=", root_delay_setup);
 329
 330 static void __init get_fs_names(char *page)
 331 {
 332         char *s = page;
 333
 334         if (root_fs_names) {
 335                 strcpy(page, root_fs_names);
 336                 while (*s++) {
 337                         if (s[-1] == ',')
 338                                 s[-1] = '\0';
 339                 }
 340         } else {
 341                 int len = get_filesystem_list(page);
 342                 char *p, *next;
 343
 344                 page[len] = '\0';
 345                 for (p = page-1; p; p = next) {
 346                         next = strchr(++p, '\n');
 347                         if (*p++ != '\t')
 348                                 continue;
 349                         while ((*s++ = *p++) != '\n')
 350                                 ;
 351                         s[-1] = '\0';
 352                 }
 353         }
 354         *s = '\0';
 355 }
 356
 357 static int __init do_mount_root(char *name, char *fs, int flags, void *data)
 358 {
 359         struct super_block *s;
 360         int err = sys_mount(name, "/root", fs, flags, data);
 361         if (err)
 362                 return err;
 363
 364         sys_chdir("/root");
 365         s = current->fs->pwd.dentry->d_sb;
 366         ROOT_DEV = s->s_dev;
 367         printk(KERN_INFO
 368                "VFS: Mounted root (%s filesystem)%s on device %u:%u.\n",
 369                s->s_type->name,
 370                s->s_flags & MS_RDONLY ?  " readonly" : "",
 371                MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
 372         return 0;
 373 }
 374
 375 void __init mount_block_root(char *name, int flags)
 376 {
 377         struct page *page = alloc_page(GFP_KERNEL |
 378                                         __GFP_NOTRACK_FALSE_POSITIVE);
 379         char *fs_names = page_address(page);
 380         char *p;
 381 #ifdef CONFIG_BLOCK
 382         char b[BDEVNAME_SIZE];
 383 #else
 384         const char *b = name;
 385 #endif
 386
 387         get_fs_names(fs_names);
 388 retry:
 389         for (p = fs_names; *p; p += strlen(p)+1) {
 390                 int err = do_mount_root(name, p, flags, root_mount_data);
 391                 switch (err) {
 392                         case 0:
 393                                 goto out;
 394                         case -EACCES:
 395                                 flags |= MS_RDONLY;
 396                                 goto retry;
 397                         case -EINVAL:
 398                                 continue;
 399                 }
 400                 /*
 401                  * Allow the user to distinguish between failed sys_open
 402                  * and bad superblock on root device.
 403                  * and give them a list of the available devices
 404                  */
 405 #ifdef CONFIG_BLOCK
 406                 __bdevname(ROOT_DEV, b);
 407 #endif
 408                 printk("VFS: Cannot open root device \"%s\" or %s: error %d\n",
 409                                 root_device_name, b, err);
 410                 printk("Please append a correct \"root=\" boot option; here are the available partitions:\n");
 411
 412                 printk_all_partitions();
 413 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
 414                 printk("DEBUG_BLOCK_EXT_DEVT is enabled, you need to specify "
 415                        "explicit textual name for \"root=\" boot option.\n");
 416 #endif
 417                 panic("VFS: Unable to mount root fs on %s", b);
 418         }
 419
 420         printk("List of all partitions:\n");
 421         printk_all_partitions();
 422         printk("No filesystem could mount root, tried: ");
 423         for (p = fs_names; *p; p += strlen(p)+1)
 424                 printk(" %s", p);
 425         printk("\n");
 426 #ifdef CONFIG_BLOCK
 427         __bdevname(ROOT_DEV, b);
 428 #endif
 429         panic("VFS: Unable to mount root fs on %s", b);
 430 out:
 431         put_page(page);
 432 }
 433
 434 #ifdef CONFIG_ROOT_NFS
 435
 436 #define NFSROOT_TIMEOUT_MIN     5
 437 #define NFSROOT_TIMEOUT_MAX     30
 438 #define NFSROOT_RETRY_MAX       5
 439
 440 static int __init mount_nfs_root(void)
 441 {
 442         char *root_dev, *root_data;
 443         unsigned int timeout;
 444         int try, err;
 445
 446         err = nfs_root_data(&root_dev, &root_data);
 447         if (err != 0)
 448                 return 0;
 449
 450         /*
 451          * The server or network may not be ready, so try several
 452          * times.  Stop after a few tries in case the client wants
 453          * to fall back to other boot methods.
 454          */
 455         timeout = NFSROOT_TIMEOUT_MIN;
 456         for (try = 1; ; try++) {
 457                 err = do_mount_root(root_dev, "nfs",
 458                                         root_mountflags, root_data);
 459                 if (err == 0)
 460                         return 1;
 461                 if (try > NFSROOT_RETRY_MAX)
 462                         break;
 463
 464                 /* Wait, in case the server refused us immediately */
 465                 ssleep(timeout);
 466                 timeout <<= 1;
 467                 if (timeout > NFSROOT_TIMEOUT_MAX)
 468                         timeout = NFSROOT_TIMEOUT_MAX;
 469         }
 470         return 0;
 471 }
 472 #endif
 473
 474 #if defined(CONFIG_BLK_DEV_RAM) || defined(CONFIG_BLK_DEV_FD)
 475 void __init change_floppy(char *fmt, ...)
 476 {
 477         struct termios termios;
 478         char buf[80];
 479         char c;
 480         int fd;
 481         va_list args;
 482         va_start(args, fmt);
 483         vsprintf(buf, fmt, args);
 484         va_end(args);
 485         fd = sys_open("/dev/root", O_RDWR | O_NDELAY, 0);
 486         if (fd >= 0) {
 487                 sys_ioctl(fd, FDEJECT, 0);
 488                 sys_close(fd);
 489         }
 490         printk(KERN_NOTICE "VFS: Insert %s and press ENTER\n", buf);
 491         fd = sys_open("/dev/console", O_RDWR, 0);
 492         if (fd >= 0) {
 493                 sys_ioctl(fd, TCGETS, (long)&termios);
 494                 termios.c_lflag &= ~ICANON;
 495                 sys_ioctl(fd, TCSETSF, (long)&termios);
 496                 sys_read(fd, &c, 1);
 497                 termios.c_lflag |= ICANON;
 498                 sys_ioctl(fd, TCSETSF, (long)&termios);
 499                 sys_close(fd);
 500         }
 501 }
 502 #endif
 503
 504 void __init mount_root(void)
 505 {
 506 #ifdef CONFIG_ROOT_NFS
 507         if (ROOT_DEV == Root_NFS) {
 508                 if (mount_nfs_root())
 509                         return;
 510
 511                 printk(KERN_ERR "VFS: Unable to mount root fs via NFS, trying floppy.\n");
 512                 ROOT_DEV = Root_FD0;
 513         }
 514 #endif
 515 #ifdef CONFIG_BLK_DEV_FD
 516         if (MAJOR(ROOT_DEV) == FLOPPY_MAJOR) {
 517                 /* rd_doload is 2 for a dual initrd/ramload setup */
 518                 if (rd_doload==2) {
 519                         if (rd_load_disk(1)) {
 520                                 ROOT_DEV = Root_RAM1;
 521                                 root_device_name = NULL;
 522                         }
 523                 } else
 524                         change_floppy("root floppy");
 525         }
 526 #endif
 527 #ifdef CONFIG_BLOCK
 528         create_dev("/dev/root", ROOT_DEV);
 529         mount_block_root("/dev/root", root_mountflags);
 530 #endif
 531 }
 532
 533 /*
 534  * Prepare the namespace - decide what/where to mount, load ramdisks, etc.
 535  */
 536 void __init prepare_namespace(void)
 537 {
 538         int is_floppy;
 539
 540         if (root_delay) {
 541                 printk(KERN_INFO "Waiting %d sec before mounting root device...\n",
 542                        root_delay);
 543                 ssleep(root_delay);
 544         }
 545
 546         /*
 547          * wait for the known devices to complete their probing
 548          *
 549          * Note: this is a potential source of long boot delays.
 550          * For example, it is not atypical to wait 5 seconds here
 551          * for the touchpad of a laptop to initialize.
 552          */
 553         wait_for_device_probe();
 554
 555         md_run_setup();
 556
 557         if (saved_root_name[0]) {
 558                 root_device_name = saved_root_name;
 559                 if (!strncmp(root_device_name, "mtd", 3) ||
 560                     !strncmp(root_device_name, "ubi", 3)) {
 561                         mount_block_root(root_device_name, root_mountflags);
 562                         goto out;
 563                 }
 564                 ROOT_DEV = name_to_dev_t(root_device_name);
 565                 if (strncmp(root_device_name, "/dev/", 5) == 0)
 566                         root_device_name += 5;
 567         }
 568
 569         if (initrd_load())
 570                 goto out;
 571
 572         /* wait for any asynchronous scanning to complete */
 573         if ((ROOT_DEV == 0) && root_wait) {
 574                 printk(KERN_INFO "Waiting for root device %s...\n",
 575                         saved_root_name);
 576                 while (driver_probe_done() != 0 ||
 577                         (ROOT_DEV = name_to_dev_t(saved_root_name)) == 0)
 578                         msleep(100);
 579                 async_synchronize_full();
 580         }
 581
 582         is_floppy = MAJOR(ROOT_DEV) == FLOPPY_MAJOR;
 583
 584         if (is_floppy && rd_doload && rd_load_disk(0))
 585                 ROOT_DEV = Root_RAM0;
 586
 587         mount_root();
 588 out:
 589         devtmpfs_mount("dev");
 590         sys_mount(".", "/", NULL, MS_MOVE, NULL);
 591         sys_chroot(".");
 592 }
 593
 594 static bool is_tmpfs;
 595 static struct dentry *rootfs_mount(struct file_system_type *fs_type,
 596         int flags, const char *dev_name, void *data)
 597 {
 598         static unsigned long once;
 599         void *fill = ramfs_fill_super;
 600
 601         if (test_and_set_bit(0, &once))
 602                 return ERR_PTR(-ENODEV);
 603
 604         if (IS_ENABLED(CONFIG_TMPFS) && is_tmpfs)
 605                 fill = shmem_fill_super;
 606
 607         return mount_nodev(fs_type, flags, data, fill);
 608 }
 609
 610 static struct file_system_type rootfs_fs_type = {
 611         .name           = "rootfs",
 612         .mount          = rootfs_mount,
 613         .kill_sb        = kill_litter_super,
 614 };
 615
 616 int __init init_rootfs(void)
 617 {
 618         int err = register_filesystem(&rootfs_fs_type);
 619
 620         if (err)
 621                 return err;
 622
 623         if (IS_ENABLED(CONFIG_TMPFS) && !saved_root_name[0] &&
 624                 (!root_fs_names || strstr(root_fs_names, "tmpfs"))) {
 625                 err = shmem_init();
 626                 is_tmpfs = true;
 627         } else {
 628                 err = init_ramfs_fs();
 629         }
 630
 631         if (err)
 632                 unregister_filesystem(&rootfs_fs_type);
 633
 634         return err;
 635 }