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  *
 216  *      If name doesn't have fall into the categories above, we return (0,0).
 217  *      block_class is used to check if something is a disk name. If the disk
 218  *      name contains slashes, the device name has them replaced with
 219  *      bangs.
 220  */
 221
 222 dev_t name_to_dev_t(const char *name)
 223 {
 224         char s[32];
 225         char *p;
 226         dev_t res = 0;
 227         int part;
 228
 229 #ifdef CONFIG_BLOCK
 230         if (strncmp(name, "PARTUUID=", 9) == 0) {
 231                 name += 9;
 232                 res = devt_from_partuuid(name);
 233                 if (!res)
 234                         goto fail;
 235                 goto done;
 236         } else if (strncmp(name, "PARTLABEL=", 10) == 0) {
 237                 struct device *dev;
 238
 239                 dev = class_find_device(&block_class, NULL, name + 10,
 240                                         &match_dev_by_label);
 241                 if (!dev)
 242                         goto fail;
 243
 244                 res = dev->devt;
 245                 put_device(dev);
 246                 goto done;
 247         }
 248 #endif
 249
 250         if (strncmp(name, "/dev/", 5) != 0) {
 251                 unsigned maj, min, offset;
 252                 char dummy;
 253
 254                 if ((sscanf(name, "%u:%u%c", &maj, &min, &dummy) == 2) ||
 255                     (sscanf(name, "%u:%u:%u:%c", &maj, &min, &offset, &dummy) == 3)) {
 256                         res = MKDEV(maj, min);
 257                         if (maj != MAJOR(res) || min != MINOR(res))
 258                                 goto fail;
 259                 } else {
 260                         res = new_decode_dev(simple_strtoul(name, &p, 16));
 261                         if (*p)
 262                                 goto fail;
 263                 }
 264                 goto done;
 265         }
 266
 267         name += 5;
 268         res = Root_NFS;
 269         if (strcmp(name, "nfs") == 0)
 270                 goto done;
 271         res = Root_RAM0;
 272         if (strcmp(name, "ram") == 0)
 273                 goto done;
 274
 275         if (strlen(name) > 31)
 276                 goto fail;
 277         strcpy(s, name);
 278         for (p = s; *p; p++)
 279                 if (*p == '/')
 280                         *p = '!';
 281         res = blk_lookup_devt(s, 0);
 282         if (res)
 283                 goto done;
 284
 285         /*
 286          * try non-existent, but valid partition, which may only exist
 287          * after revalidating the disk, like partitioned md devices
 288          */
 289         while (p > s && isdigit(p[-1]))
 290                 p--;
 291         if (p == s || !*p || *p == '0')
 292                 goto fail;
 293
 294         /* try disk name without <part number> */
 295         part = simple_strtoul(p, NULL, 10);
 296         *p = '\0';
 297         res = blk_lookup_devt(s, part);
 298         if (res)
 299                 goto done;
 300
 301         /* try disk name without p<part number> */
 302         if (p < s + 2 || !isdigit(p[-2]) || p[-1] != 'p')
 303                 goto fail;
 304         p[-1] = '\0';
 305         res = blk_lookup_devt(s, part);
 306         if (res)
 307                 goto done;
 308
 309 fail:
 310         return 0;
 311 done:
 312         return res;
 313 }
 314 EXPORT_SYMBOL_GPL(name_to_dev_t);
 315
 316 static int __init root_dev_setup(char *line)
 317 {
 318         strlcpy(saved_root_name, line, sizeof(saved_root_name));
 319         return 1;
 320 }
 321
 322 __setup("root=", root_dev_setup);
 323
 324 static int __init rootwait_setup(char *str)
 325 {
 326         if (*str)
 327                 return 0;
 328         root_wait = 1;
 329         return 1;
 330 }
 331
 332 __setup("rootwait", rootwait_setup);
 333
 334 static char * __initdata root_mount_data;
 335 static int __init root_data_setup(char *str)
 336 {
 337         root_mount_data = str;
 338         return 1;
 339 }
 340
 341 static char * __initdata root_fs_names;
 342 static int __init fs_names_setup(char *str)
 343 {
 344         root_fs_names = str;
 345         return 1;
 346 }
 347
 348 static unsigned int __initdata root_delay;
 349 static int __init root_delay_setup(char *str)
 350 {
 351         root_delay = simple_strtoul(str, NULL, 0);
 352         return 1;
 353 }
 354
 355 __setup("rootflags=", root_data_setup);
 356 __setup("rootfstype=", fs_names_setup);
 357 __setup("rootdelay=", root_delay_setup);
 358
 359 static void __init get_fs_names(char *page)
 360 {
 361         char *s = page;
 362
 363         if (root_fs_names) {
 364                 strcpy(page, root_fs_names);
 365                 while (*s++) {
 366                         if (s[-1] == ',')
 367                                 s[-1] = '\0';
 368                 }
 369         } else {
 370                 int len = get_filesystem_list(page);
 371                 char *p, *next;
 372
 373                 page[len] = '\0';
 374                 for (p = page-1; p; p = next) {
 375                         next = strchr(++p, '\n');
 376                         if (*p++ != '\t')
 377                                 continue;
 378                         while ((*s++ = *p++) != '\n')
 379                                 ;
 380                         s[-1] = '\0';
 381                 }
 382         }
 383         *s = '\0';
 384 }
 385
 386 static int __init do_mount_root(char *name, char *fs, int flags, void *data)
 387 {
 388         struct super_block *s;
 389         int err = ksys_mount(name, "/root", fs, flags, data);
 390         if (err)
 391                 return err;
 392
 393         ksys_chdir("/root");
 394         s = current->fs->pwd.dentry->d_sb;
 395         ROOT_DEV = s->s_dev;
 396         printk(KERN_INFO
 397                "VFS: Mounted root (%s filesystem)%s on device %u:%u.\n",
 398                s->s_type->name,
 399                sb_rdonly(s) ? " readonly" : "",
 400                MAJOR(ROOT_DEV), MINOR(ROOT_DEV));
 401         return 0;
 402 }
 403
 404 void __init mount_block_root(char *name, int flags)
 405 {
 406         struct page *page = alloc_page(GFP_KERNEL);
 407         char *fs_names = page_address(page);
 408         char *p;
 409 #ifdef CONFIG_BLOCK
 410         char b[BDEVNAME_SIZE];
 411 #else
 412         const char *b = name;
 413 #endif
 414
 415         get_fs_names(fs_names);
 416 retry:
 417         for (p = fs_names; *p; p += strlen(p)+1) {
 418                 int err = do_mount_root(name, p, flags, root_mount_data);
 419                 switch (err) {
 420                         case 0:
 421                                 goto out;
 422                         case -EACCES:
 423                         case -EINVAL:
 424                                 continue;
 425                 }
 426                 /*
 427                  * Allow the user to distinguish between failed sys_open
 428                  * and bad superblock on root device.
 429                  * and give them a list of the available devices
 430                  */
 431 #ifdef CONFIG_BLOCK
 432                 __bdevname(ROOT_DEV, b);
 433 #endif
 434                 printk("VFS: Cannot open root device \"%s\" or %s: error %d\n",
 435                                 root_device_name, b, err);
 436                 printk("Please append a correct \"root=\" boot option; here are the available partitions:\n");
 437
 438                 printk_all_partitions();
 439 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
 440                 printk("DEBUG_BLOCK_EXT_DEVT is enabled, you need to specify "
 441                        "explicit textual name for \"root=\" boot option.\n");
 442 #endif
 443                 panic("VFS: Unable to mount root fs on %s", b);
 444         }
 445         if (!(flags & SB_RDONLY)) {
 446                 flags |= SB_RDONLY;
 447                 goto retry;
 448         }
 449
 450         printk("List of all partitions:\n");
 451         printk_all_partitions();
 452         printk("No filesystem could mount root, tried: ");
 453         for (p = fs_names; *p; p += strlen(p)+1)
 454                 printk(" %s", p);
 455         printk("\n");
 456 #ifdef CONFIG_BLOCK
 457         __bdevname(ROOT_DEV, b);
 458 #endif
 459         panic("VFS: Unable to mount root fs on %s", b);
 460 out:
 461         put_page(page);
 462 }
 463
 464 #ifdef CONFIG_ROOT_NFS
 465
 466 #define NFSROOT_TIMEOUT_MIN     5
 467 #define NFSROOT_TIMEOUT_MAX     30
 468 #define NFSROOT_RETRY_MAX       5
 469
 470 static int __init mount_nfs_root(void)
 471 {
 472         char *root_dev, *root_data;
 473         unsigned int timeout;
 474         int try, err;
 475
 476         err = nfs_root_data(&root_dev, &root_data);
 477         if (err != 0)
 478                 return 0;
 479
 480         /*
 481          * The server or network may not be ready, so try several
 482          * times.  Stop after a few tries in case the client wants
 483          * to fall back to other boot methods.
 484          */
 485         timeout = NFSROOT_TIMEOUT_MIN;
 486         for (try = 1; ; try++) {
 487                 err = do_mount_root(root_dev, "nfs",
 488                                         root_mountflags, root_data);
 489                 if (err == 0)
 490                         return 1;
 491                 if (try > NFSROOT_RETRY_MAX)
 492                         break;
 493
 494                 /* Wait, in case the server refused us immediately */
 495                 ssleep(timeout);
 496                 timeout <<= 1;
 497                 if (timeout > NFSROOT_TIMEOUT_MAX)
 498                         timeout = NFSROOT_TIMEOUT_MAX;
 499         }
 500         return 0;
 501 }
 502 #endif
 503
 504 #if defined(CONFIG_BLK_DEV_RAM) || defined(CONFIG_BLK_DEV_FD)
 505 void __init change_floppy(char *fmt, ...)
 506 {
 507         struct termios termios;
 508         char buf[80];
 509         char c;
 510         int fd;
 511         va_list args;
 512         va_start(args, fmt);
 513         vsprintf(buf, fmt, args);
 514         va_end(args);
 515         fd = ksys_open("/dev/root", O_RDWR | O_NDELAY, 0);
 516         if (fd >= 0) {
 517                 ksys_ioctl(fd, FDEJECT, 0);
 518                 ksys_close(fd);
 519         }
 520         printk(KERN_NOTICE "VFS: Insert %s and press ENTER\n", buf);
 521         fd = ksys_open("/dev/console", O_RDWR, 0);
 522         if (fd >= 0) {
 523                 ksys_ioctl(fd, TCGETS, (long)&termios);
 524                 termios.c_lflag &= ~ICANON;
 525                 ksys_ioctl(fd, TCSETSF, (long)&termios);
 526                 ksys_read(fd, &c, 1);
 527                 termios.c_lflag |= ICANON;
 528                 ksys_ioctl(fd, TCSETSF, (long)&termios);
 529                 ksys_close(fd);
 530         }
 531 }
 532 #endif
 533
 534 void __init mount_root(void)
 535 {
 536 #ifdef CONFIG_ROOT_NFS
 537         if (ROOT_DEV == Root_NFS) {
 538                 if (mount_nfs_root())
 539                         return;
 540
 541                 printk(KERN_ERR "VFS: Unable to mount root fs via NFS, trying floppy.\n");
 542                 ROOT_DEV = Root_FD0;
 543         }
 544 #endif
 545 #ifdef CONFIG_BLK_DEV_FD
 546         if (MAJOR(ROOT_DEV) == FLOPPY_MAJOR) {
 547                 /* rd_doload is 2 for a dual initrd/ramload setup */
 548                 if (rd_doload==2) {
 549                         if (rd_load_disk(1)) {
 550                                 ROOT_DEV = Root_RAM1;
 551                                 root_device_name = NULL;
 552                         }
 553                 } else
 554                         change_floppy("root floppy");
 555         }
 556 #endif
 557 #ifdef CONFIG_BLOCK
 558         {
 559                 int err = create_dev("/dev/root", ROOT_DEV);
 560
 561                 if (err < 0)
 562                         pr_emerg("Failed to create /dev/root: %d\n", err);
 563                 mount_block_root("/dev/root", root_mountflags);
 564         }
 565 #endif
 566 }
 567
 568 /*
 569  * Prepare the namespace - decide what/where to mount, load ramdisks, etc.
 570  */
 571 void __init prepare_namespace(void)
 572 {
 573         int is_floppy;
 574
 575         if (root_delay) {
 576                 printk(KERN_INFO "Waiting %d sec before mounting root device...\n",
 577                        root_delay);
 578                 ssleep(root_delay);
 579         }
 580
 581         /*
 582          * wait for the known devices to complete their probing
 583          *
 584          * Note: this is a potential source of long boot delays.
 585          * For example, it is not atypical to wait 5 seconds here
 586          * for the touchpad of a laptop to initialize.
 587          */
 588         wait_for_device_probe();
 589
 590         md_run_setup();
 591
 592         if (saved_root_name[0]) {
 593                 root_device_name = saved_root_name;
 594                 if (!strncmp(root_device_name, "mtd", 3) ||
 595                     !strncmp(root_device_name, "ubi", 3)) {
 596                         mount_block_root(root_device_name, root_mountflags);
 597                         goto out;
 598                 }
 599                 ROOT_DEV = name_to_dev_t(root_device_name);
 600                 if (strncmp(root_device_name, "/dev/", 5) == 0)
 601                         root_device_name += 5;
 602         }
 603
 604         if (initrd_load())
 605                 goto out;
 606
 607         /* wait for any asynchronous scanning to complete */
 608         if ((ROOT_DEV == 0) && root_wait) {
 609                 printk(KERN_INFO "Waiting for root device %s...\n",
 610                         saved_root_name);
 611                 while (driver_probe_done() != 0 ||
 612                         (ROOT_DEV = name_to_dev_t(saved_root_name)) == 0)
 613                         msleep(5);
 614                 async_synchronize_full();
 615         }
 616
 617         is_floppy = MAJOR(ROOT_DEV) == FLOPPY_MAJOR;
 618
 619         if (is_floppy && rd_doload && rd_load_disk(0))
 620                 ROOT_DEV = Root_RAM0;
 621
 622         mount_root();
 623 out:
 624         devtmpfs_mount("dev");
 625         ksys_mount(".", "/", NULL, MS_MOVE, NULL);
 626         ksys_chroot(".");
 627 }
 628
 629 static bool is_tmpfs;
 630 static int rootfs_init_fs_context(struct fs_context *fc)
 631 {
 632         if (IS_ENABLED(CONFIG_TMPFS) && is_tmpfs)
 633                 return shmem_init_fs_context(fc);
 634
 635         return ramfs_init_fs_context(fc);
 636 }
 637
 638 struct file_system_type rootfs_fs_type = {
 639         .name           = "rootfs",
 640         .init_fs_context = rootfs_init_fs_context,
 641         .kill_sb        = kill_litter_super,
 642 };
 643
 644 void __init init_rootfs(void)
 645 {
 646         if (IS_ENABLED(CONFIG_TMPFS) && !saved_root_name[0] &&
 647                 (!root_fs_names || strstr(root_fs_names, "tmpfs")))
 648                 is_tmpfs = true;
 649 }