fs/btrfs/tree-checker.c

/* [<][>][^][v][top][bottom][index][help] */
This source file includes following definitions.
__printf
__printf
file_extent_end
check_extent_data_item
check_csum_item
__printf
check_dir_item
__printf
check_block_group_item
__printf
btrfs_check_chunk_valid
__printf
check_dev_item
check_inode_item
check_root_item
__printf
check_extent_item
check_simple_keyed_refs
check_extent_data_ref
check_leaf_item
check_leaf
btrfs_check_leaf_full
btrfs_check_leaf_relaxed
btrfs_check_node
   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * Copyright (C) Qu Wenruo 2017.  All rights reserved.
   4  */
   5 
   6 /*
   7  * The module is used to catch unexpected/corrupted tree block data.
   8  * Such behavior can be caused either by a fuzzed image or bugs.
   9  *
  10  * The objective is to do leaf/node validation checks when tree block is read
  11  * from disk, and check *every* possible member, so other code won't
  12  * need to checking them again.
  13  *
  14  * Due to the potential and unwanted damage, every checker needs to be
  15  * carefully reviewed otherwise so it does not prevent mount of valid images.
  16  */
  17 
  18 #include <linux/types.h>
  19 #include <linux/stddef.h>
  20 #include <linux/error-injection.h>
  21 #include "ctree.h"
  22 #include "tree-checker.h"
  23 #include "disk-io.h"
  24 #include "compression.h"
  25 #include "volumes.h"
  26 
  27 /*
  28  * Error message should follow the following format:
  29  * corrupt <type>: <identifier>, <reason>[, <bad_value>]
  30  *
  31  * @type:       leaf or node
  32  * @identifier: the necessary info to locate the leaf/node.
  33  *              It's recommended to decode key.objecitd/offset if it's
  34  *              meaningful.
  35  * @reason:     describe the error
  36  * @bad_value:  optional, it's recommended to output bad value and its
  37  *              expected value (range).
  38  *
  39  * Since comma is used to separate the components, only space is allowed
  40  * inside each component.
  41  */
  42 
  43 /*
  44  * Append generic "corrupt leaf/node root=%llu block=%llu slot=%d: " to @fmt.
  45  * Allows callers to customize the output.
  46  */
  47 __printf(3, 4)
  48 __cold
  49 static void generic_err(const struct extent_buffer *eb, int slot,
  50                         const char *fmt, ...)
  51 {
  52         const struct btrfs_fs_info *fs_info = eb->fs_info;
  53         struct va_format vaf;
  54         va_list args;
  55 
  56         va_start(args, fmt);
  57 
  58         vaf.fmt = fmt;
  59         vaf.va = &args;
  60 
  61         btrfs_crit(fs_info,
  62                 "corrupt %s: root=%llu block=%llu slot=%d, %pV",
  63                 btrfs_header_level(eb) == 0 ? "leaf" : "node",
  64                 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, &vaf);
  65         va_end(args);
  66 }
  67 
  68 /*
  69  * Customized reporter for extent data item, since its key objectid and
  70  * offset has its own meaning.
  71  */
  72 __printf(3, 4)
  73 __cold
  74 static void file_extent_err(const struct extent_buffer *eb, int slot,
  75                             const char *fmt, ...)
  76 {
  77         const struct btrfs_fs_info *fs_info = eb->fs_info;
  78         struct btrfs_key key;
  79         struct va_format vaf;
  80         va_list args;
  81 
  82         btrfs_item_key_to_cpu(eb, &key, slot);
  83         va_start(args, fmt);
  84 
  85         vaf.fmt = fmt;
  86         vaf.va = &args;
  87 
  88         btrfs_crit(fs_info,
  89         "corrupt %s: root=%llu block=%llu slot=%d ino=%llu file_offset=%llu, %pV",
  90                 btrfs_header_level(eb) == 0 ? "leaf" : "node",
  91                 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
  92                 key.objectid, key.offset, &vaf);
  93         va_end(args);
  94 }
  95 
  96 /*
  97  * Return 0 if the btrfs_file_extent_##name is aligned to @alignment
  98  * Else return 1
  99  */
 100 #define CHECK_FE_ALIGNED(leaf, slot, fi, name, alignment)                     \
 101 ({                                                                            \
 102         if (!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment))) \
 103                 file_extent_err((leaf), (slot),                               \
 104         "invalid %s for file extent, have %llu, should be aligned to %u",     \
 105                         (#name), btrfs_file_extent_##name((leaf), (fi)),      \
 106                         (alignment));                                         \
 107         (!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment)));   \
 108 })
 109 
 110 static u64 file_extent_end(struct extent_buffer *leaf,
 111                            struct btrfs_key *key,
 112                            struct btrfs_file_extent_item *extent)
 113 {
 114         u64 end;
 115         u64 len;
 116 
 117         if (btrfs_file_extent_type(leaf, extent) == BTRFS_FILE_EXTENT_INLINE) {
 118                 len = btrfs_file_extent_ram_bytes(leaf, extent);
 119                 end = ALIGN(key->offset + len, leaf->fs_info->sectorsize);
 120         } else {
 121                 len = btrfs_file_extent_num_bytes(leaf, extent);
 122                 end = key->offset + len;
 123         }
 124         return end;
 125 }
 126 
 127 static int check_extent_data_item(struct extent_buffer *leaf,
 128                                   struct btrfs_key *key, int slot,
 129                                   struct btrfs_key *prev_key)
 130 {
 131         struct btrfs_fs_info *fs_info = leaf->fs_info;
 132         struct btrfs_file_extent_item *fi;
 133         u32 sectorsize = fs_info->sectorsize;
 134         u32 item_size = btrfs_item_size_nr(leaf, slot);
 135         u64 extent_end;
 136 
 137         if (!IS_ALIGNED(key->offset, sectorsize)) {
 138                 file_extent_err(leaf, slot,
 139 "unaligned file_offset for file extent, have %llu should be aligned to %u",
 140                         key->offset, sectorsize);
 141                 return -EUCLEAN;
 142         }
 143 
 144         fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
 145 
 146         if (btrfs_file_extent_type(leaf, fi) > BTRFS_FILE_EXTENT_TYPES) {
 147                 file_extent_err(leaf, slot,
 148                 "invalid type for file extent, have %u expect range [0, %u]",
 149                         btrfs_file_extent_type(leaf, fi),
 150                         BTRFS_FILE_EXTENT_TYPES);
 151                 return -EUCLEAN;
 152         }
 153 
 154         /*
 155          * Support for new compression/encryption must introduce incompat flag,
 156          * and must be caught in open_ctree().
 157          */
 158         if (btrfs_file_extent_compression(leaf, fi) > BTRFS_COMPRESS_TYPES) {
 159                 file_extent_err(leaf, slot,
 160         "invalid compression for file extent, have %u expect range [0, %u]",
 161                         btrfs_file_extent_compression(leaf, fi),
 162                         BTRFS_COMPRESS_TYPES);
 163                 return -EUCLEAN;
 164         }
 165         if (btrfs_file_extent_encryption(leaf, fi)) {
 166                 file_extent_err(leaf, slot,
 167                         "invalid encryption for file extent, have %u expect 0",
 168                         btrfs_file_extent_encryption(leaf, fi));
 169                 return -EUCLEAN;
 170         }
 171         if (btrfs_file_extent_type(leaf, fi) == BTRFS_FILE_EXTENT_INLINE) {
 172                 /* Inline extent must have 0 as key offset */
 173                 if (key->offset) {
 174                         file_extent_err(leaf, slot,
 175                 "invalid file_offset for inline file extent, have %llu expect 0",
 176                                 key->offset);
 177                         return -EUCLEAN;
 178                 }
 179 
 180                 /* Compressed inline extent has no on-disk size, skip it */
 181                 if (btrfs_file_extent_compression(leaf, fi) !=
 182                     BTRFS_COMPRESS_NONE)
 183                         return 0;
 184 
 185                 /* Uncompressed inline extent size must match item size */
 186                 if (item_size != BTRFS_FILE_EXTENT_INLINE_DATA_START +
 187                     btrfs_file_extent_ram_bytes(leaf, fi)) {
 188                         file_extent_err(leaf, slot,
 189         "invalid ram_bytes for uncompressed inline extent, have %u expect %llu",
 190                                 item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START +
 191                                 btrfs_file_extent_ram_bytes(leaf, fi));
 192                         return -EUCLEAN;
 193                 }
 194                 return 0;
 195         }
 196 
 197         /* Regular or preallocated extent has fixed item size */
 198         if (item_size != sizeof(*fi)) {
 199                 file_extent_err(leaf, slot,
 200         "invalid item size for reg/prealloc file extent, have %u expect %zu",
 201                         item_size, sizeof(*fi));
 202                 return -EUCLEAN;
 203         }
 204         if (CHECK_FE_ALIGNED(leaf, slot, fi, ram_bytes, sectorsize) ||
 205             CHECK_FE_ALIGNED(leaf, slot, fi, disk_bytenr, sectorsize) ||
 206             CHECK_FE_ALIGNED(leaf, slot, fi, disk_num_bytes, sectorsize) ||
 207             CHECK_FE_ALIGNED(leaf, slot, fi, offset, sectorsize) ||
 208             CHECK_FE_ALIGNED(leaf, slot, fi, num_bytes, sectorsize))
 209                 return -EUCLEAN;
 210 
 211         /* Catch extent end overflow */
 212         if (check_add_overflow(btrfs_file_extent_num_bytes(leaf, fi),
 213                                key->offset, &extent_end)) {
 214                 file_extent_err(leaf, slot,
 215         "extent end overflow, have file offset %llu extent num bytes %llu",
 216                                 key->offset,
 217                                 btrfs_file_extent_num_bytes(leaf, fi));
 218                 return -EUCLEAN;
 219         }
 220 
 221         /*
 222          * Check that no two consecutive file extent items, in the same leaf,
 223          * present ranges that overlap each other.
 224          */
 225         if (slot > 0 &&
 226             prev_key->objectid == key->objectid &&
 227             prev_key->type == BTRFS_EXTENT_DATA_KEY) {
 228                 struct btrfs_file_extent_item *prev_fi;
 229                 u64 prev_end;
 230 
 231                 prev_fi = btrfs_item_ptr(leaf, slot - 1,
 232                                          struct btrfs_file_extent_item);
 233                 prev_end = file_extent_end(leaf, prev_key, prev_fi);
 234                 if (prev_end > key->offset) {
 235                         file_extent_err(leaf, slot - 1,
 236 "file extent end range (%llu) goes beyond start offset (%llu) of the next file extent",
 237                                         prev_end, key->offset);
 238                         return -EUCLEAN;
 239                 }
 240         }
 241 
 242         return 0;
 243 }
 244 
 245 static int check_csum_item(struct extent_buffer *leaf, struct btrfs_key *key,
 246                            int slot, struct btrfs_key *prev_key)
 247 {
 248         struct btrfs_fs_info *fs_info = leaf->fs_info;
 249         u32 sectorsize = fs_info->sectorsize;
 250         u32 csumsize = btrfs_super_csum_size(fs_info->super_copy);
 251 
 252         if (key->objectid != BTRFS_EXTENT_CSUM_OBJECTID) {
 253                 generic_err(leaf, slot,
 254                 "invalid key objectid for csum item, have %llu expect %llu",
 255                         key->objectid, BTRFS_EXTENT_CSUM_OBJECTID);
 256                 return -EUCLEAN;
 257         }
 258         if (!IS_ALIGNED(key->offset, sectorsize)) {
 259                 generic_err(leaf, slot,
 260         "unaligned key offset for csum item, have %llu should be aligned to %u",
 261                         key->offset, sectorsize);
 262                 return -EUCLEAN;
 263         }
 264         if (!IS_ALIGNED(btrfs_item_size_nr(leaf, slot), csumsize)) {
 265                 generic_err(leaf, slot,
 266         "unaligned item size for csum item, have %u should be aligned to %u",
 267                         btrfs_item_size_nr(leaf, slot), csumsize);
 268                 return -EUCLEAN;
 269         }
 270         if (slot > 0 && prev_key->type == BTRFS_EXTENT_CSUM_KEY) {
 271                 u64 prev_csum_end;
 272                 u32 prev_item_size;
 273 
 274                 prev_item_size = btrfs_item_size_nr(leaf, slot - 1);
 275                 prev_csum_end = (prev_item_size / csumsize) * sectorsize;
 276                 prev_csum_end += prev_key->offset;
 277                 if (prev_csum_end > key->offset) {
 278                         generic_err(leaf, slot - 1,
 279 "csum end range (%llu) goes beyond the start range (%llu) of the next csum item",
 280                                     prev_csum_end, key->offset);
 281                         return -EUCLEAN;
 282                 }
 283         }
 284         return 0;
 285 }
 286 
 287 /*
 288  * Customized reported for dir_item, only important new info is key->objectid,
 289  * which represents inode number
 290  */
 291 __printf(3, 4)
 292 __cold
 293 static void dir_item_err(const struct extent_buffer *eb, int slot,
 294                          const char *fmt, ...)
 295 {
 296         const struct btrfs_fs_info *fs_info = eb->fs_info;
 297         struct btrfs_key key;
 298         struct va_format vaf;
 299         va_list args;
 300 
 301         btrfs_item_key_to_cpu(eb, &key, slot);
 302         va_start(args, fmt);
 303 
 304         vaf.fmt = fmt;
 305         vaf.va = &args;
 306 
 307         btrfs_crit(fs_info,
 308         "corrupt %s: root=%llu block=%llu slot=%d ino=%llu, %pV",
 309                 btrfs_header_level(eb) == 0 ? "leaf" : "node",
 310                 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
 311                 key.objectid, &vaf);
 312         va_end(args);
 313 }
 314 
 315 static int check_dir_item(struct extent_buffer *leaf,
 316                           struct btrfs_key *key, int slot)
 317 {
 318         struct btrfs_fs_info *fs_info = leaf->fs_info;
 319         struct btrfs_dir_item *di;
 320         u32 item_size = btrfs_item_size_nr(leaf, slot);
 321         u32 cur = 0;
 322 
 323         di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
 324         while (cur < item_size) {
 325                 u32 name_len;
 326                 u32 data_len;
 327                 u32 max_name_len;
 328                 u32 total_size;
 329                 u32 name_hash;
 330                 u8 dir_type;
 331 
 332                 /* header itself should not cross item boundary */
 333                 if (cur + sizeof(*di) > item_size) {
 334                         dir_item_err(leaf, slot,
 335                 "dir item header crosses item boundary, have %zu boundary %u",
 336                                 cur + sizeof(*di), item_size);
 337                         return -EUCLEAN;
 338                 }
 339 
 340                 /* dir type check */
 341                 dir_type = btrfs_dir_type(leaf, di);
 342                 if (dir_type >= BTRFS_FT_MAX) {
 343                         dir_item_err(leaf, slot,
 344                         "invalid dir item type, have %u expect [0, %u)",
 345                                 dir_type, BTRFS_FT_MAX);
 346                         return -EUCLEAN;
 347                 }
 348 
 349                 if (key->type == BTRFS_XATTR_ITEM_KEY &&
 350                     dir_type != BTRFS_FT_XATTR) {
 351                         dir_item_err(leaf, slot,
 352                 "invalid dir item type for XATTR key, have %u expect %u",
 353                                 dir_type, BTRFS_FT_XATTR);
 354                         return -EUCLEAN;
 355                 }
 356                 if (dir_type == BTRFS_FT_XATTR &&
 357                     key->type != BTRFS_XATTR_ITEM_KEY) {
 358                         dir_item_err(leaf, slot,
 359                         "xattr dir type found for non-XATTR key");
 360                         return -EUCLEAN;
 361                 }
 362                 if (dir_type == BTRFS_FT_XATTR)
 363                         max_name_len = XATTR_NAME_MAX;
 364                 else
 365                         max_name_len = BTRFS_NAME_LEN;
 366 
 367                 /* Name/data length check */
 368                 name_len = btrfs_dir_name_len(leaf, di);
 369                 data_len = btrfs_dir_data_len(leaf, di);
 370                 if (name_len > max_name_len) {
 371                         dir_item_err(leaf, slot,
 372                         "dir item name len too long, have %u max %u",
 373                                 name_len, max_name_len);
 374                         return -EUCLEAN;
 375                 }
 376                 if (name_len + data_len > BTRFS_MAX_XATTR_SIZE(fs_info)) {
 377                         dir_item_err(leaf, slot,
 378                         "dir item name and data len too long, have %u max %u",
 379                                 name_len + data_len,
 380                                 BTRFS_MAX_XATTR_SIZE(fs_info));
 381                         return -EUCLEAN;
 382                 }
 383 
 384                 if (data_len && dir_type != BTRFS_FT_XATTR) {
 385                         dir_item_err(leaf, slot,
 386                         "dir item with invalid data len, have %u expect 0",
 387                                 data_len);
 388                         return -EUCLEAN;
 389                 }
 390 
 391                 total_size = sizeof(*di) + name_len + data_len;
 392 
 393                 /* header and name/data should not cross item boundary */
 394                 if (cur + total_size > item_size) {
 395                         dir_item_err(leaf, slot,
 396                 "dir item data crosses item boundary, have %u boundary %u",
 397                                 cur + total_size, item_size);
 398                         return -EUCLEAN;
 399                 }
 400 
 401                 /*
 402                  * Special check for XATTR/DIR_ITEM, as key->offset is name
 403                  * hash, should match its name
 404                  */
 405                 if (key->type == BTRFS_DIR_ITEM_KEY ||
 406                     key->type == BTRFS_XATTR_ITEM_KEY) {
 407                         char namebuf[max(BTRFS_NAME_LEN, XATTR_NAME_MAX)];
 408 
 409                         read_extent_buffer(leaf, namebuf,
 410                                         (unsigned long)(di + 1), name_len);
 411                         name_hash = btrfs_name_hash(namebuf, name_len);
 412                         if (key->offset != name_hash) {
 413                                 dir_item_err(leaf, slot,
 414                 "name hash mismatch with key, have 0x%016x expect 0x%016llx",
 415                                         name_hash, key->offset);
 416                                 return -EUCLEAN;
 417                         }
 418                 }
 419                 cur += total_size;
 420                 di = (struct btrfs_dir_item *)((void *)di + total_size);
 421         }
 422         return 0;
 423 }
 424 
 425 __printf(3, 4)
 426 __cold
 427 static void block_group_err(const struct extent_buffer *eb, int slot,
 428                             const char *fmt, ...)
 429 {
 430         const struct btrfs_fs_info *fs_info = eb->fs_info;
 431         struct btrfs_key key;
 432         struct va_format vaf;
 433         va_list args;
 434 
 435         btrfs_item_key_to_cpu(eb, &key, slot);
 436         va_start(args, fmt);
 437 
 438         vaf.fmt = fmt;
 439         vaf.va = &args;
 440 
 441         btrfs_crit(fs_info,
 442         "corrupt %s: root=%llu block=%llu slot=%d bg_start=%llu bg_len=%llu, %pV",
 443                 btrfs_header_level(eb) == 0 ? "leaf" : "node",
 444                 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
 445                 key.objectid, key.offset, &vaf);
 446         va_end(args);
 447 }
 448 
 449 static int check_block_group_item(struct extent_buffer *leaf,
 450                                   struct btrfs_key *key, int slot)
 451 {
 452         struct btrfs_block_group_item bgi;
 453         u32 item_size = btrfs_item_size_nr(leaf, slot);
 454         u64 flags;
 455         u64 type;
 456 
 457         /*
 458          * Here we don't really care about alignment since extent allocator can
 459          * handle it.  We care more about the size.
 460          */
 461         if (key->offset == 0) {
 462                 block_group_err(leaf, slot,
 463                                 "invalid block group size 0");
 464                 return -EUCLEAN;
 465         }
 466 
 467         if (item_size != sizeof(bgi)) {
 468                 block_group_err(leaf, slot,
 469                         "invalid item size, have %u expect %zu",
 470                                 item_size, sizeof(bgi));
 471                 return -EUCLEAN;
 472         }
 473 
 474         read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot),
 475                            sizeof(bgi));
 476         if (btrfs_block_group_chunk_objectid(&bgi) !=
 477             BTRFS_FIRST_CHUNK_TREE_OBJECTID) {
 478                 block_group_err(leaf, slot,
 479                 "invalid block group chunk objectid, have %llu expect %llu",
 480                                 btrfs_block_group_chunk_objectid(&bgi),
 481                                 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
 482                 return -EUCLEAN;
 483         }
 484 
 485         if (btrfs_block_group_used(&bgi) > key->offset) {
 486                 block_group_err(leaf, slot,
 487                         "invalid block group used, have %llu expect [0, %llu)",
 488                                 btrfs_block_group_used(&bgi), key->offset);
 489                 return -EUCLEAN;
 490         }
 491 
 492         flags = btrfs_block_group_flags(&bgi);
 493         if (hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) > 1) {
 494                 block_group_err(leaf, slot,
 495 "invalid profile flags, have 0x%llx (%lu bits set) expect no more than 1 bit set",
 496                         flags & BTRFS_BLOCK_GROUP_PROFILE_MASK,
 497                         hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK));
 498                 return -EUCLEAN;
 499         }
 500 
 501         type = flags & BTRFS_BLOCK_GROUP_TYPE_MASK;
 502         if (type != BTRFS_BLOCK_GROUP_DATA &&
 503             type != BTRFS_BLOCK_GROUP_METADATA &&
 504             type != BTRFS_BLOCK_GROUP_SYSTEM &&
 505             type != (BTRFS_BLOCK_GROUP_METADATA |
 506                            BTRFS_BLOCK_GROUP_DATA)) {
 507                 block_group_err(leaf, slot,
 508 "invalid type, have 0x%llx (%lu bits set) expect either 0x%llx, 0x%llx, 0x%llx or 0x%llx",
 509                         type, hweight64(type),
 510                         BTRFS_BLOCK_GROUP_DATA, BTRFS_BLOCK_GROUP_METADATA,
 511                         BTRFS_BLOCK_GROUP_SYSTEM,
 512                         BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA);
 513                 return -EUCLEAN;
 514         }
 515         return 0;
 516 }
 517 
 518 __printf(4, 5)
 519 __cold
 520 static void chunk_err(const struct extent_buffer *leaf,
 521                       const struct btrfs_chunk *chunk, u64 logical,
 522                       const char *fmt, ...)
 523 {
 524         const struct btrfs_fs_info *fs_info = leaf->fs_info;
 525         bool is_sb;
 526         struct va_format vaf;
 527         va_list args;
 528         int i;
 529         int slot = -1;
 530 
 531         /* Only superblock eb is able to have such small offset */
 532         is_sb = (leaf->start == BTRFS_SUPER_INFO_OFFSET);
 533 
 534         if (!is_sb) {
 535                 /*
 536                  * Get the slot number by iterating through all slots, this
 537                  * would provide better readability.
 538                  */
 539                 for (i = 0; i < btrfs_header_nritems(leaf); i++) {
 540                         if (btrfs_item_ptr_offset(leaf, i) ==
 541                                         (unsigned long)chunk) {
 542                                 slot = i;
 543                                 break;
 544                         }
 545                 }
 546         }
 547         va_start(args, fmt);
 548         vaf.fmt = fmt;
 549         vaf.va = &args;
 550 
 551         if (is_sb)
 552                 btrfs_crit(fs_info,
 553                 "corrupt superblock syschunk array: chunk_start=%llu, %pV",
 554                            logical, &vaf);
 555         else
 556                 btrfs_crit(fs_info,
 557         "corrupt leaf: root=%llu block=%llu slot=%d chunk_start=%llu, %pV",
 558                            BTRFS_CHUNK_TREE_OBJECTID, leaf->start, slot,
 559                            logical, &vaf);
 560         va_end(args);
 561 }
 562 
 563 /*
 564  * The common chunk check which could also work on super block sys chunk array.
 565  *
 566  * Return -EUCLEAN if anything is corrupted.
 567  * Return 0 if everything is OK.
 568  */
 569 int btrfs_check_chunk_valid(struct extent_buffer *leaf,
 570                             struct btrfs_chunk *chunk, u64 logical)
 571 {
 572         struct btrfs_fs_info *fs_info = leaf->fs_info;
 573         u64 length;
 574         u64 stripe_len;
 575         u16 num_stripes;
 576         u16 sub_stripes;
 577         u64 type;
 578         u64 features;
 579         bool mixed = false;
 580 
 581         length = btrfs_chunk_length(leaf, chunk);
 582         stripe_len = btrfs_chunk_stripe_len(leaf, chunk);
 583         num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
 584         sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk);
 585         type = btrfs_chunk_type(leaf, chunk);
 586 
 587         if (!num_stripes) {
 588                 chunk_err(leaf, chunk, logical,
 589                           "invalid chunk num_stripes, have %u", num_stripes);
 590                 return -EUCLEAN;
 591         }
 592         if (!IS_ALIGNED(logical, fs_info->sectorsize)) {
 593                 chunk_err(leaf, chunk, logical,
 594                 "invalid chunk logical, have %llu should aligned to %u",
 595                           logical, fs_info->sectorsize);
 596                 return -EUCLEAN;
 597         }
 598         if (btrfs_chunk_sector_size(leaf, chunk) != fs_info->sectorsize) {
 599                 chunk_err(leaf, chunk, logical,
 600                           "invalid chunk sectorsize, have %u expect %u",
 601                           btrfs_chunk_sector_size(leaf, chunk),
 602                           fs_info->sectorsize);
 603                 return -EUCLEAN;
 604         }
 605         if (!length || !IS_ALIGNED(length, fs_info->sectorsize)) {
 606                 chunk_err(leaf, chunk, logical,
 607                           "invalid chunk length, have %llu", length);
 608                 return -EUCLEAN;
 609         }
 610         if (!is_power_of_2(stripe_len) || stripe_len != BTRFS_STRIPE_LEN) {
 611                 chunk_err(leaf, chunk, logical,
 612                           "invalid chunk stripe length: %llu",
 613                           stripe_len);
 614                 return -EUCLEAN;
 615         }
 616         if (~(BTRFS_BLOCK_GROUP_TYPE_MASK | BTRFS_BLOCK_GROUP_PROFILE_MASK) &
 617             type) {
 618                 chunk_err(leaf, chunk, logical,
 619                           "unrecognized chunk type: 0x%llx",
 620                           ~(BTRFS_BLOCK_GROUP_TYPE_MASK |
 621                             BTRFS_BLOCK_GROUP_PROFILE_MASK) &
 622                           btrfs_chunk_type(leaf, chunk));
 623                 return -EUCLEAN;
 624         }
 625 
 626         if (!is_power_of_2(type & BTRFS_BLOCK_GROUP_PROFILE_MASK) &&
 627             (type & BTRFS_BLOCK_GROUP_PROFILE_MASK) != 0) {
 628                 chunk_err(leaf, chunk, logical,
 629                 "invalid chunk profile flag: 0x%llx, expect 0 or 1 bit set",
 630                           type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
 631                 return -EUCLEAN;
 632         }
 633         if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == 0) {
 634                 chunk_err(leaf, chunk, logical,
 635         "missing chunk type flag, have 0x%llx one bit must be set in 0x%llx",
 636                           type, BTRFS_BLOCK_GROUP_TYPE_MASK);
 637                 return -EUCLEAN;
 638         }
 639 
 640         if ((type & BTRFS_BLOCK_GROUP_SYSTEM) &&
 641             (type & (BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA))) {
 642                 chunk_err(leaf, chunk, logical,
 643                           "system chunk with data or metadata type: 0x%llx",
 644                           type);
 645                 return -EUCLEAN;
 646         }
 647 
 648         features = btrfs_super_incompat_flags(fs_info->super_copy);
 649         if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS)
 650                 mixed = true;
 651 
 652         if (!mixed) {
 653                 if ((type & BTRFS_BLOCK_GROUP_METADATA) &&
 654                     (type & BTRFS_BLOCK_GROUP_DATA)) {
 655                         chunk_err(leaf, chunk, logical,
 656                         "mixed chunk type in non-mixed mode: 0x%llx", type);
 657                         return -EUCLEAN;
 658                 }
 659         }
 660 
 661         if ((type & BTRFS_BLOCK_GROUP_RAID10 && sub_stripes != 2) ||
 662             (type & BTRFS_BLOCK_GROUP_RAID1 && num_stripes != 2) ||
 663             (type & BTRFS_BLOCK_GROUP_RAID5 && num_stripes < 2) ||
 664             (type & BTRFS_BLOCK_GROUP_RAID6 && num_stripes < 3) ||
 665             (type & BTRFS_BLOCK_GROUP_DUP && num_stripes != 2) ||
 666             ((type & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 && num_stripes != 1)) {
 667                 chunk_err(leaf, chunk, logical,
 668                         "invalid num_stripes:sub_stripes %u:%u for profile %llu",
 669                         num_stripes, sub_stripes,
 670                         type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
 671                 return -EUCLEAN;
 672         }
 673 
 674         return 0;
 675 }
 676 
 677 __printf(3, 4)
 678 __cold
 679 static void dev_item_err(const struct extent_buffer *eb, int slot,
 680                          const char *fmt, ...)
 681 {
 682         struct btrfs_key key;
 683         struct va_format vaf;
 684         va_list args;
 685 
 686         btrfs_item_key_to_cpu(eb, &key, slot);
 687         va_start(args, fmt);
 688 
 689         vaf.fmt = fmt;
 690         vaf.va = &args;
 691 
 692         btrfs_crit(eb->fs_info,
 693         "corrupt %s: root=%llu block=%llu slot=%d devid=%llu %pV",
 694                 btrfs_header_level(eb) == 0 ? "leaf" : "node",
 695                 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
 696                 key.objectid, &vaf);
 697         va_end(args);
 698 }
 699 
 700 static int check_dev_item(struct extent_buffer *leaf,
 701                           struct btrfs_key *key, int slot)
 702 {
 703         struct btrfs_dev_item *ditem;
 704 
 705         if (key->objectid != BTRFS_DEV_ITEMS_OBJECTID) {
 706                 dev_item_err(leaf, slot,
 707                              "invalid objectid: has=%llu expect=%llu",
 708                              key->objectid, BTRFS_DEV_ITEMS_OBJECTID);
 709                 return -EUCLEAN;
 710         }
 711         ditem = btrfs_item_ptr(leaf, slot, struct btrfs_dev_item);
 712         if (btrfs_device_id(leaf, ditem) != key->offset) {
 713                 dev_item_err(leaf, slot,
 714                              "devid mismatch: key has=%llu item has=%llu",
 715                              key->offset, btrfs_device_id(leaf, ditem));
 716                 return -EUCLEAN;
 717         }
 718 
 719         /*
 720          * For device total_bytes, we don't have reliable way to check it, as
 721          * it can be 0 for device removal. Device size check can only be done
 722          * by dev extents check.
 723          */
 724         if (btrfs_device_bytes_used(leaf, ditem) >
 725             btrfs_device_total_bytes(leaf, ditem)) {
 726                 dev_item_err(leaf, slot,
 727                              "invalid bytes used: have %llu expect [0, %llu]",
 728                              btrfs_device_bytes_used(leaf, ditem),
 729                              btrfs_device_total_bytes(leaf, ditem));
 730                 return -EUCLEAN;
 731         }
 732         /*
 733          * Remaining members like io_align/type/gen/dev_group aren't really
 734          * utilized.  Skip them to make later usage of them easier.
 735          */
 736         return 0;
 737 }
 738 
 739 /* Inode item error output has the same format as dir_item_err() */
 740 #define inode_item_err(fs_info, eb, slot, fmt, ...)                     \
 741         dir_item_err(eb, slot, fmt, __VA_ARGS__)
 742 
 743 static int check_inode_item(struct extent_buffer *leaf,
 744                             struct btrfs_key *key, int slot)
 745 {
 746         struct btrfs_fs_info *fs_info = leaf->fs_info;
 747         struct btrfs_inode_item *iitem;
 748         u64 super_gen = btrfs_super_generation(fs_info->super_copy);
 749         u32 valid_mask = (S_IFMT | S_ISUID | S_ISGID | S_ISVTX | 0777);
 750         u32 mode;
 751 
 752         if ((key->objectid < BTRFS_FIRST_FREE_OBJECTID ||
 753              key->objectid > BTRFS_LAST_FREE_OBJECTID) &&
 754             key->objectid != BTRFS_ROOT_TREE_DIR_OBJECTID &&
 755             key->objectid != BTRFS_FREE_INO_OBJECTID) {
 756                 generic_err(leaf, slot,
 757         "invalid key objectid: has %llu expect %llu or [%llu, %llu] or %llu",
 758                             key->objectid, BTRFS_ROOT_TREE_DIR_OBJECTID,
 759                             BTRFS_FIRST_FREE_OBJECTID,
 760                             BTRFS_LAST_FREE_OBJECTID,
 761                             BTRFS_FREE_INO_OBJECTID);
 762                 return -EUCLEAN;
 763         }
 764         if (key->offset != 0) {
 765                 inode_item_err(fs_info, leaf, slot,
 766                         "invalid key offset: has %llu expect 0",
 767                         key->offset);
 768                 return -EUCLEAN;
 769         }
 770         iitem = btrfs_item_ptr(leaf, slot, struct btrfs_inode_item);
 771 
 772         /* Here we use super block generation + 1 to handle log tree */
 773         if (btrfs_inode_generation(leaf, iitem) > super_gen + 1) {
 774                 inode_item_err(fs_info, leaf, slot,
 775                         "invalid inode generation: has %llu expect (0, %llu]",
 776                                btrfs_inode_generation(leaf, iitem),
 777                                super_gen + 1);
 778                 return -EUCLEAN;
 779         }
 780         /* Note for ROOT_TREE_DIR_ITEM, mkfs could set its transid 0 */
 781         if (btrfs_inode_transid(leaf, iitem) > super_gen + 1) {
 782                 inode_item_err(fs_info, leaf, slot,
 783                         "invalid inode generation: has %llu expect [0, %llu]",
 784                                btrfs_inode_transid(leaf, iitem), super_gen + 1);
 785                 return -EUCLEAN;
 786         }
 787 
 788         /*
 789          * For size and nbytes it's better not to be too strict, as for dir
 790          * item its size/nbytes can easily get wrong, but doesn't affect
 791          * anything in the fs. So here we skip the check.
 792          */
 793         mode = btrfs_inode_mode(leaf, iitem);
 794         if (mode & ~valid_mask) {
 795                 inode_item_err(fs_info, leaf, slot,
 796                                "unknown mode bit detected: 0x%x",
 797                                mode & ~valid_mask);
 798                 return -EUCLEAN;
 799         }
 800 
 801         /*
 802          * S_IFMT is not bit mapped so we can't completely rely on is_power_of_2,
 803          * but is_power_of_2() can save us from checking FIFO/CHR/DIR/REG.
 804          * Only needs to check BLK, LNK and SOCKS
 805          */
 806         if (!is_power_of_2(mode & S_IFMT)) {
 807                 if (!S_ISLNK(mode) && !S_ISBLK(mode) && !S_ISSOCK(mode)) {
 808                         inode_item_err(fs_info, leaf, slot,
 809                         "invalid mode: has 0%o expect valid S_IF* bit(s)",
 810                                        mode & S_IFMT);
 811                         return -EUCLEAN;
 812                 }
 813         }
 814         if (S_ISDIR(mode) && btrfs_inode_nlink(leaf, iitem) > 1) {
 815                 inode_item_err(fs_info, leaf, slot,
 816                        "invalid nlink: has %u expect no more than 1 for dir",
 817                         btrfs_inode_nlink(leaf, iitem));
 818                 return -EUCLEAN;
 819         }
 820         if (btrfs_inode_flags(leaf, iitem) & ~BTRFS_INODE_FLAG_MASK) {
 821                 inode_item_err(fs_info, leaf, slot,
 822                                "unknown flags detected: 0x%llx",
 823                                btrfs_inode_flags(leaf, iitem) &
 824                                ~BTRFS_INODE_FLAG_MASK);
 825                 return -EUCLEAN;
 826         }
 827         return 0;
 828 }
 829 
 830 static int check_root_item(struct extent_buffer *leaf, struct btrfs_key *key,
 831                            int slot)
 832 {
 833         struct btrfs_fs_info *fs_info = leaf->fs_info;
 834         struct btrfs_root_item ri;
 835         const u64 valid_root_flags = BTRFS_ROOT_SUBVOL_RDONLY |
 836                                      BTRFS_ROOT_SUBVOL_DEAD;
 837 
 838         /* No such tree id */
 839         if (key->objectid == 0) {
 840                 generic_err(leaf, slot, "invalid root id 0");
 841                 return -EUCLEAN;
 842         }
 843 
 844         /*
 845          * Some older kernel may create ROOT_ITEM with non-zero offset, so here
 846          * we only check offset for reloc tree whose key->offset must be a
 847          * valid tree.
 848          */
 849         if (key->objectid == BTRFS_TREE_RELOC_OBJECTID && key->offset == 0) {
 850                 generic_err(leaf, slot, "invalid root id 0 for reloc tree");
 851                 return -EUCLEAN;
 852         }
 853 
 854         if (btrfs_item_size_nr(leaf, slot) != sizeof(ri)) {
 855                 generic_err(leaf, slot,
 856                             "invalid root item size, have %u expect %zu",
 857                             btrfs_item_size_nr(leaf, slot), sizeof(ri));
 858         }
 859 
 860         read_extent_buffer(leaf, &ri, btrfs_item_ptr_offset(leaf, slot),
 861                            sizeof(ri));
 862 
 863         /* Generation related */
 864         if (btrfs_root_generation(&ri) >
 865             btrfs_super_generation(fs_info->super_copy) + 1) {
 866                 generic_err(leaf, slot,
 867                         "invalid root generation, have %llu expect (0, %llu]",
 868                             btrfs_root_generation(&ri),
 869                             btrfs_super_generation(fs_info->super_copy) + 1);
 870                 return -EUCLEAN;
 871         }
 872         if (btrfs_root_generation_v2(&ri) >
 873             btrfs_super_generation(fs_info->super_copy) + 1) {
 874                 generic_err(leaf, slot,
 875                 "invalid root v2 generation, have %llu expect (0, %llu]",
 876                             btrfs_root_generation_v2(&ri),
 877                             btrfs_super_generation(fs_info->super_copy) + 1);
 878                 return -EUCLEAN;
 879         }
 880         if (btrfs_root_last_snapshot(&ri) >
 881             btrfs_super_generation(fs_info->super_copy) + 1) {
 882                 generic_err(leaf, slot,
 883                 "invalid root last_snapshot, have %llu expect (0, %llu]",
 884                             btrfs_root_last_snapshot(&ri),
 885                             btrfs_super_generation(fs_info->super_copy) + 1);
 886                 return -EUCLEAN;
 887         }
 888 
 889         /* Alignment and level check */
 890         if (!IS_ALIGNED(btrfs_root_bytenr(&ri), fs_info->sectorsize)) {
 891                 generic_err(leaf, slot,
 892                 "invalid root bytenr, have %llu expect to be aligned to %u",
 893                             btrfs_root_bytenr(&ri), fs_info->sectorsize);
 894                 return -EUCLEAN;
 895         }
 896         if (btrfs_root_level(&ri) >= BTRFS_MAX_LEVEL) {
 897                 generic_err(leaf, slot,
 898                             "invalid root level, have %u expect [0, %u]",
 899                             btrfs_root_level(&ri), BTRFS_MAX_LEVEL - 1);
 900                 return -EUCLEAN;
 901         }
 902         if (ri.drop_level >= BTRFS_MAX_LEVEL) {
 903                 generic_err(leaf, slot,
 904                             "invalid root level, have %u expect [0, %u]",
 905                             ri.drop_level, BTRFS_MAX_LEVEL - 1);
 906                 return -EUCLEAN;
 907         }
 908 
 909         /* Flags check */
 910         if (btrfs_root_flags(&ri) & ~valid_root_flags) {
 911                 generic_err(leaf, slot,
 912                             "invalid root flags, have 0x%llx expect mask 0x%llx",
 913                             btrfs_root_flags(&ri), valid_root_flags);
 914                 return -EUCLEAN;
 915         }
 916         return 0;
 917 }
 918 
 919 __printf(3,4)
 920 __cold
 921 static void extent_err(const struct extent_buffer *eb, int slot,
 922                        const char *fmt, ...)
 923 {
 924         struct btrfs_key key;
 925         struct va_format vaf;
 926         va_list args;
 927         u64 bytenr;
 928         u64 len;
 929 
 930         btrfs_item_key_to_cpu(eb, &key, slot);
 931         bytenr = key.objectid;
 932         if (key.type == BTRFS_METADATA_ITEM_KEY ||
 933             key.type == BTRFS_TREE_BLOCK_REF_KEY ||
 934             key.type == BTRFS_SHARED_BLOCK_REF_KEY)
 935                 len = eb->fs_info->nodesize;
 936         else
 937                 len = key.offset;
 938         va_start(args, fmt);
 939 
 940         vaf.fmt = fmt;
 941         vaf.va = &args;
 942 
 943         btrfs_crit(eb->fs_info,
 944         "corrupt %s: block=%llu slot=%d extent bytenr=%llu len=%llu %pV",
 945                 btrfs_header_level(eb) == 0 ? "leaf" : "node",
 946                 eb->start, slot, bytenr, len, &vaf);
 947         va_end(args);
 948 }
 949 
 950 static int check_extent_item(struct extent_buffer *leaf,
 951                              struct btrfs_key *key, int slot)
 952 {
 953         struct btrfs_fs_info *fs_info = leaf->fs_info;
 954         struct btrfs_extent_item *ei;
 955         bool is_tree_block = false;
 956         unsigned long ptr;      /* Current pointer inside inline refs */
 957         unsigned long end;      /* Extent item end */
 958         const u32 item_size = btrfs_item_size_nr(leaf, slot);
 959         u64 flags;
 960         u64 generation;
 961         u64 total_refs;         /* Total refs in btrfs_extent_item */
 962         u64 inline_refs = 0;    /* found total inline refs */
 963 
 964         if (key->type == BTRFS_METADATA_ITEM_KEY &&
 965             !btrfs_fs_incompat(fs_info, SKINNY_METADATA)) {
 966                 generic_err(leaf, slot,
 967 "invalid key type, METADATA_ITEM type invalid when SKINNY_METADATA feature disabled");
 968                 return -EUCLEAN;
 969         }
 970         /* key->objectid is the bytenr for both key types */
 971         if (!IS_ALIGNED(key->objectid, fs_info->sectorsize)) {
 972                 generic_err(leaf, slot,
 973                 "invalid key objectid, have %llu expect to be aligned to %u",
 974                            key->objectid, fs_info->sectorsize);
 975                 return -EUCLEAN;
 976         }
 977 
 978         /* key->offset is tree level for METADATA_ITEM_KEY */
 979         if (key->type == BTRFS_METADATA_ITEM_KEY &&
 980             key->offset >= BTRFS_MAX_LEVEL) {
 981                 extent_err(leaf, slot,
 982                            "invalid tree level, have %llu expect [0, %u]",
 983                            key->offset, BTRFS_MAX_LEVEL - 1);
 984                 return -EUCLEAN;
 985         }
 986 
 987         /*
 988          * EXTENT/METADATA_ITEM consists of:
 989          * 1) One btrfs_extent_item
 990          *    Records the total refs, type and generation of the extent.
 991          *
 992          * 2) One btrfs_tree_block_info (for EXTENT_ITEM and tree backref only)
 993          *    Records the first key and level of the tree block.
 994          *
 995          * 2) Zero or more btrfs_extent_inline_ref(s)
 996          *    Each inline ref has one btrfs_extent_inline_ref shows:
 997          *    2.1) The ref type, one of the 4
 998          *         TREE_BLOCK_REF       Tree block only
 999          *         SHARED_BLOCK_REF     Tree block only
1000          *         EXTENT_DATA_REF      Data only
1001          *         SHARED_DATA_REF      Data only
1002          *    2.2) Ref type specific data
1003          *         Either using btrfs_extent_inline_ref::offset, or specific
1004          *         data structure.
1005          */
1006         if (item_size < sizeof(*ei)) {
1007                 extent_err(leaf, slot,
1008                            "invalid item size, have %u expect [%zu, %u)",
1009                            item_size, sizeof(*ei),
1010                            BTRFS_LEAF_DATA_SIZE(fs_info));
1011                 return -EUCLEAN;
1012         }
1013         end = item_size + btrfs_item_ptr_offset(leaf, slot);
1014 
1015         /* Checks against extent_item */
1016         ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
1017         flags = btrfs_extent_flags(leaf, ei);
1018         total_refs = btrfs_extent_refs(leaf, ei);
1019         generation = btrfs_extent_generation(leaf, ei);
1020         if (generation > btrfs_super_generation(fs_info->super_copy) + 1) {
1021                 extent_err(leaf, slot,
1022                            "invalid generation, have %llu expect (0, %llu]",
1023                            generation,
1024                            btrfs_super_generation(fs_info->super_copy) + 1);
1025                 return -EUCLEAN;
1026         }
1027         if (!is_power_of_2(flags & (BTRFS_EXTENT_FLAG_DATA |
1028                                     BTRFS_EXTENT_FLAG_TREE_BLOCK))) {
1029                 extent_err(leaf, slot,
1030                 "invalid extent flag, have 0x%llx expect 1 bit set in 0x%llx",
1031                         flags, BTRFS_EXTENT_FLAG_DATA |
1032                         BTRFS_EXTENT_FLAG_TREE_BLOCK);
1033                 return -EUCLEAN;
1034         }
1035         is_tree_block = !!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK);
1036         if (is_tree_block) {
1037                 if (key->type == BTRFS_EXTENT_ITEM_KEY &&
1038                     key->offset != fs_info->nodesize) {
1039                         extent_err(leaf, slot,
1040                                    "invalid extent length, have %llu expect %u",
1041                                    key->offset, fs_info->nodesize);
1042                         return -EUCLEAN;
1043                 }
1044         } else {
1045                 if (key->type != BTRFS_EXTENT_ITEM_KEY) {
1046                         extent_err(leaf, slot,
1047                         "invalid key type, have %u expect %u for data backref",
1048                                    key->type, BTRFS_EXTENT_ITEM_KEY);
1049                         return -EUCLEAN;
1050                 }
1051                 if (!IS_ALIGNED(key->offset, fs_info->sectorsize)) {
1052                         extent_err(leaf, slot,
1053                         "invalid extent length, have %llu expect aligned to %u",
1054                                    key->offset, fs_info->sectorsize);
1055                         return -EUCLEAN;
1056                 }
1057         }
1058         ptr = (unsigned long)(struct btrfs_extent_item *)(ei + 1);
1059 
1060         /* Check the special case of btrfs_tree_block_info */
1061         if (is_tree_block && key->type != BTRFS_METADATA_ITEM_KEY) {
1062                 struct btrfs_tree_block_info *info;
1063 
1064                 info = (struct btrfs_tree_block_info *)ptr;
1065                 if (btrfs_tree_block_level(leaf, info) >= BTRFS_MAX_LEVEL) {
1066                         extent_err(leaf, slot,
1067                         "invalid tree block info level, have %u expect [0, %u]",
1068                                    btrfs_tree_block_level(leaf, info),
1069                                    BTRFS_MAX_LEVEL - 1);
1070                         return -EUCLEAN;
1071                 }
1072                 ptr = (unsigned long)(struct btrfs_tree_block_info *)(info + 1);
1073         }
1074 
1075         /* Check inline refs */
1076         while (ptr < end) {
1077                 struct btrfs_extent_inline_ref *iref;
1078                 struct btrfs_extent_data_ref *dref;
1079                 struct btrfs_shared_data_ref *sref;
1080                 u64 dref_offset;
1081                 u64 inline_offset;
1082                 u8 inline_type;
1083 
1084                 if (ptr + sizeof(*iref) > end) {
1085                         extent_err(leaf, slot,
1086 "inline ref item overflows extent item, ptr %lu iref size %zu end %lu",
1087                                    ptr, sizeof(*iref), end);
1088                         return -EUCLEAN;
1089                 }
1090                 iref = (struct btrfs_extent_inline_ref *)ptr;
1091                 inline_type = btrfs_extent_inline_ref_type(leaf, iref);
1092                 inline_offset = btrfs_extent_inline_ref_offset(leaf, iref);
1093                 if (ptr + btrfs_extent_inline_ref_size(inline_type) > end) {
1094                         extent_err(leaf, slot,
1095 "inline ref item overflows extent item, ptr %lu iref size %u end %lu",
1096                                    ptr, inline_type, end);
1097                         return -EUCLEAN;
1098                 }
1099 
1100                 switch (inline_type) {
1101                 /* inline_offset is subvolid of the owner, no need to check */
1102                 case BTRFS_TREE_BLOCK_REF_KEY:
1103                         inline_refs++;
1104                         break;
1105                 /* Contains parent bytenr */
1106                 case BTRFS_SHARED_BLOCK_REF_KEY:
1107                         if (!IS_ALIGNED(inline_offset, fs_info->sectorsize)) {
1108                                 extent_err(leaf, slot,
1109                 "invalid tree parent bytenr, have %llu expect aligned to %u",
1110                                            inline_offset, fs_info->sectorsize);
1111                                 return -EUCLEAN;
1112                         }
1113                         inline_refs++;
1114                         break;
1115                 /*
1116                  * Contains owner subvolid, owner key objectid, adjusted offset.
1117                  * The only obvious corruption can happen in that offset.
1118                  */
1119                 case BTRFS_EXTENT_DATA_REF_KEY:
1120                         dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1121                         dref_offset = btrfs_extent_data_ref_offset(leaf, dref);
1122                         if (!IS_ALIGNED(dref_offset, fs_info->sectorsize)) {
1123                                 extent_err(leaf, slot,
1124                 "invalid data ref offset, have %llu expect aligned to %u",
1125                                            dref_offset, fs_info->sectorsize);
1126                                 return -EUCLEAN;
1127                         }
1128                         inline_refs += btrfs_extent_data_ref_count(leaf, dref);
1129                         break;
1130                 /* Contains parent bytenr and ref count */
1131                 case BTRFS_SHARED_DATA_REF_KEY:
1132                         sref = (struct btrfs_shared_data_ref *)(iref + 1);
1133                         if (!IS_ALIGNED(inline_offset, fs_info->sectorsize)) {
1134                                 extent_err(leaf, slot,
1135                 "invalid data parent bytenr, have %llu expect aligned to %u",
1136                                            inline_offset, fs_info->sectorsize);
1137                                 return -EUCLEAN;
1138                         }
1139                         inline_refs += btrfs_shared_data_ref_count(leaf, sref);
1140                         break;
1141                 default:
1142                         extent_err(leaf, slot, "unknown inline ref type: %u",
1143                                    inline_type);
1144                         return -EUCLEAN;
1145                 }
1146                 ptr += btrfs_extent_inline_ref_size(inline_type);
1147         }
1148         /* No padding is allowed */
1149         if (ptr != end) {
1150                 extent_err(leaf, slot,
1151                            "invalid extent item size, padding bytes found");
1152                 return -EUCLEAN;
1153         }
1154 
1155         /* Finally, check the inline refs against total refs */
1156         if (inline_refs > total_refs) {
1157                 extent_err(leaf, slot,
1158                         "invalid extent refs, have %llu expect >= inline %llu",
1159                            total_refs, inline_refs);
1160                 return -EUCLEAN;
1161         }
1162         return 0;
1163 }
1164 
1165 static int check_simple_keyed_refs(struct extent_buffer *leaf,
1166                                    struct btrfs_key *key, int slot)
1167 {
1168         u32 expect_item_size = 0;
1169 
1170         if (key->type == BTRFS_SHARED_DATA_REF_KEY)
1171                 expect_item_size = sizeof(struct btrfs_shared_data_ref);
1172 
1173         if (btrfs_item_size_nr(leaf, slot) != expect_item_size) {
1174                 generic_err(leaf, slot,
1175                 "invalid item size, have %u expect %u for key type %u",
1176                             btrfs_item_size_nr(leaf, slot),
1177                             expect_item_size, key->type);
1178                 return -EUCLEAN;
1179         }
1180         if (!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize)) {
1181                 generic_err(leaf, slot,
1182 "invalid key objectid for shared block ref, have %llu expect aligned to %u",
1183                             key->objectid, leaf->fs_info->sectorsize);
1184                 return -EUCLEAN;
1185         }
1186         if (key->type != BTRFS_TREE_BLOCK_REF_KEY &&
1187             !IS_ALIGNED(key->offset, leaf->fs_info->sectorsize)) {
1188                 extent_err(leaf, slot,
1189                 "invalid tree parent bytenr, have %llu expect aligned to %u",
1190                            key->offset, leaf->fs_info->sectorsize);
1191                 return -EUCLEAN;
1192         }
1193         return 0;
1194 }
1195 
1196 static int check_extent_data_ref(struct extent_buffer *leaf,
1197                                  struct btrfs_key *key, int slot)
1198 {
1199         struct btrfs_extent_data_ref *dref;
1200         unsigned long ptr = btrfs_item_ptr_offset(leaf, slot);
1201         const unsigned long end = ptr + btrfs_item_size_nr(leaf, slot);
1202 
1203         if (btrfs_item_size_nr(leaf, slot) % sizeof(*dref) != 0) {
1204                 generic_err(leaf, slot,
1205         "invalid item size, have %u expect aligned to %zu for key type %u",
1206                             btrfs_item_size_nr(leaf, slot),
1207                             sizeof(*dref), key->type);
1208         }
1209         if (!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize)) {
1210                 generic_err(leaf, slot,
1211 "invalid key objectid for shared block ref, have %llu expect aligned to %u",
1212                             key->objectid, leaf->fs_info->sectorsize);
1213                 return -EUCLEAN;
1214         }
1215         for (; ptr < end; ptr += sizeof(*dref)) {
1216                 u64 root_objectid;
1217                 u64 owner;
1218                 u64 offset;
1219                 u64 hash;
1220 
1221                 dref = (struct btrfs_extent_data_ref *)ptr;
1222                 root_objectid = btrfs_extent_data_ref_root(leaf, dref);
1223                 owner = btrfs_extent_data_ref_objectid(leaf, dref);
1224                 offset = btrfs_extent_data_ref_offset(leaf, dref);
1225                 hash = hash_extent_data_ref(root_objectid, owner, offset);
1226                 if (hash != key->offset) {
1227                         extent_err(leaf, slot,
1228         "invalid extent data ref hash, item has 0x%016llx key has 0x%016llx",
1229                                    hash, key->offset);
1230                         return -EUCLEAN;
1231                 }
1232                 if (!IS_ALIGNED(offset, leaf->fs_info->sectorsize)) {
1233                         extent_err(leaf, slot,
1234         "invalid extent data backref offset, have %llu expect aligned to %u",
1235                                    offset, leaf->fs_info->sectorsize);
1236                 }
1237         }
1238         return 0;
1239 }
1240 
1241 /*
1242  * Common point to switch the item-specific validation.
1243  */
1244 static int check_leaf_item(struct extent_buffer *leaf,
1245                            struct btrfs_key *key, int slot,
1246                            struct btrfs_key *prev_key)
1247 {
1248         int ret = 0;
1249         struct btrfs_chunk *chunk;
1250 
1251         switch (key->type) {
1252         case BTRFS_EXTENT_DATA_KEY:
1253                 ret = check_extent_data_item(leaf, key, slot, prev_key);
1254                 break;
1255         case BTRFS_EXTENT_CSUM_KEY:
1256                 ret = check_csum_item(leaf, key, slot, prev_key);
1257                 break;
1258         case BTRFS_DIR_ITEM_KEY:
1259         case BTRFS_DIR_INDEX_KEY:
1260         case BTRFS_XATTR_ITEM_KEY:
1261                 ret = check_dir_item(leaf, key, slot);
1262                 break;
1263         case BTRFS_BLOCK_GROUP_ITEM_KEY:
1264                 ret = check_block_group_item(leaf, key, slot);
1265                 break;
1266         case BTRFS_CHUNK_ITEM_KEY:
1267                 chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
1268                 ret = btrfs_check_chunk_valid(leaf, chunk, key->offset);
1269                 break;
1270         case BTRFS_DEV_ITEM_KEY:
1271                 ret = check_dev_item(leaf, key, slot);
1272                 break;
1273         case BTRFS_INODE_ITEM_KEY:
1274                 ret = check_inode_item(leaf, key, slot);
1275                 break;
1276         case BTRFS_ROOT_ITEM_KEY:
1277                 ret = check_root_item(leaf, key, slot);
1278                 break;
1279         case BTRFS_EXTENT_ITEM_KEY:
1280         case BTRFS_METADATA_ITEM_KEY:
1281                 ret = check_extent_item(leaf, key, slot);
1282                 break;
1283         case BTRFS_TREE_BLOCK_REF_KEY:
1284         case BTRFS_SHARED_DATA_REF_KEY:
1285         case BTRFS_SHARED_BLOCK_REF_KEY:
1286                 ret = check_simple_keyed_refs(leaf, key, slot);
1287                 break;
1288         case BTRFS_EXTENT_DATA_REF_KEY:
1289                 ret = check_extent_data_ref(leaf, key, slot);
1290                 break;
1291         }
1292         return ret;
1293 }
1294 
1295 static int check_leaf(struct extent_buffer *leaf, bool check_item_data)
1296 {
1297         struct btrfs_fs_info *fs_info = leaf->fs_info;
1298         /* No valid key type is 0, so all key should be larger than this key */
1299         struct btrfs_key prev_key = {0, 0, 0};
1300         struct btrfs_key key;
1301         u32 nritems = btrfs_header_nritems(leaf);
1302         int slot;
1303 
1304         if (btrfs_header_level(leaf) != 0) {
1305                 generic_err(leaf, 0,
1306                         "invalid level for leaf, have %d expect 0",
1307                         btrfs_header_level(leaf));
1308                 return -EUCLEAN;
1309         }
1310 
1311         /*
1312          * Extent buffers from a relocation tree have a owner field that
1313          * corresponds to the subvolume tree they are based on. So just from an
1314          * extent buffer alone we can not find out what is the id of the
1315          * corresponding subvolume tree, so we can not figure out if the extent
1316          * buffer corresponds to the root of the relocation tree or not. So
1317          * skip this check for relocation trees.
1318          */
1319         if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) {
1320                 u64 owner = btrfs_header_owner(leaf);
1321 
1322                 /* These trees must never be empty */
1323                 if (owner == BTRFS_ROOT_TREE_OBJECTID ||
1324                     owner == BTRFS_CHUNK_TREE_OBJECTID ||
1325                     owner == BTRFS_EXTENT_TREE_OBJECTID ||
1326                     owner == BTRFS_DEV_TREE_OBJECTID ||
1327                     owner == BTRFS_FS_TREE_OBJECTID ||
1328                     owner == BTRFS_DATA_RELOC_TREE_OBJECTID) {
1329                         generic_err(leaf, 0,
1330                         "invalid root, root %llu must never be empty",
1331                                     owner);
1332                         return -EUCLEAN;
1333                 }
1334                 /* Unknown tree */
1335                 if (owner == 0) {
1336                         generic_err(leaf, 0,
1337                                 "invalid owner, root 0 is not defined");
1338                         return -EUCLEAN;
1339                 }
1340                 return 0;
1341         }
1342 
1343         if (nritems == 0)
1344                 return 0;
1345 
1346         /*
1347          * Check the following things to make sure this is a good leaf, and
1348          * leaf users won't need to bother with similar sanity checks:
1349          *
1350          * 1) key ordering
1351          * 2) item offset and size
1352          *    No overlap, no hole, all inside the leaf.
1353          * 3) item content
1354          *    If possible, do comprehensive sanity check.
1355          *    NOTE: All checks must only rely on the item data itself.
1356          */
1357         for (slot = 0; slot < nritems; slot++) {
1358                 u32 item_end_expected;
1359                 int ret;
1360 
1361                 btrfs_item_key_to_cpu(leaf, &key, slot);
1362 
1363                 /* Make sure the keys are in the right order */
1364                 if (btrfs_comp_cpu_keys(&prev_key, &key) >= 0) {
1365                         generic_err(leaf, slot,
1366         "bad key order, prev (%llu %u %llu) current (%llu %u %llu)",
1367                                 prev_key.objectid, prev_key.type,
1368                                 prev_key.offset, key.objectid, key.type,
1369                                 key.offset);
1370                         return -EUCLEAN;
1371                 }
1372 
1373                 /*
1374                  * Make sure the offset and ends are right, remember that the
1375                  * item data starts at the end of the leaf and grows towards the
1376                  * front.
1377                  */
1378                 if (slot == 0)
1379                         item_end_expected = BTRFS_LEAF_DATA_SIZE(fs_info);
1380                 else
1381                         item_end_expected = btrfs_item_offset_nr(leaf,
1382                                                                  slot - 1);
1383                 if (btrfs_item_end_nr(leaf, slot) != item_end_expected) {
1384                         generic_err(leaf, slot,
1385                                 "unexpected item end, have %u expect %u",
1386                                 btrfs_item_end_nr(leaf, slot),
1387                                 item_end_expected);
1388                         return -EUCLEAN;
1389                 }
1390 
1391                 /*
1392                  * Check to make sure that we don't point outside of the leaf,
1393                  * just in case all the items are consistent to each other, but
1394                  * all point outside of the leaf.
1395                  */
1396                 if (btrfs_item_end_nr(leaf, slot) >
1397                     BTRFS_LEAF_DATA_SIZE(fs_info)) {
1398                         generic_err(leaf, slot,
1399                         "slot end outside of leaf, have %u expect range [0, %u]",
1400                                 btrfs_item_end_nr(leaf, slot),
1401                                 BTRFS_LEAF_DATA_SIZE(fs_info));
1402                         return -EUCLEAN;
1403                 }
1404 
1405                 /* Also check if the item pointer overlaps with btrfs item. */
1406                 if (btrfs_item_nr_offset(slot) + sizeof(struct btrfs_item) >
1407                     btrfs_item_ptr_offset(leaf, slot)) {
1408                         generic_err(leaf, slot,
1409                 "slot overlaps with its data, item end %lu data start %lu",
1410                                 btrfs_item_nr_offset(slot) +
1411                                 sizeof(struct btrfs_item),
1412                                 btrfs_item_ptr_offset(leaf, slot));
1413                         return -EUCLEAN;
1414                 }
1415 
1416                 if (check_item_data) {
1417                         /*
1418                          * Check if the item size and content meet other
1419                          * criteria
1420                          */
1421                         ret = check_leaf_item(leaf, &key, slot, &prev_key);
1422                         if (ret < 0)
1423                                 return ret;
1424                 }
1425 
1426                 prev_key.objectid = key.objectid;
1427                 prev_key.type = key.type;
1428                 prev_key.offset = key.offset;
1429         }
1430 
1431         return 0;
1432 }
1433 
1434 int btrfs_check_leaf_full(struct extent_buffer *leaf)
1435 {
1436         return check_leaf(leaf, true);
1437 }
1438 ALLOW_ERROR_INJECTION(btrfs_check_leaf_full, ERRNO);
1439 
1440 int btrfs_check_leaf_relaxed(struct extent_buffer *leaf)
1441 {
1442         return check_leaf(leaf, false);
1443 }
1444 
1445 int btrfs_check_node(struct extent_buffer *node)
1446 {
1447         struct btrfs_fs_info *fs_info = node->fs_info;
1448         unsigned long nr = btrfs_header_nritems(node);
1449         struct btrfs_key key, next_key;
1450         int slot;
1451         int level = btrfs_header_level(node);
1452         u64 bytenr;
1453         int ret = 0;
1454 
1455         if (level <= 0 || level >= BTRFS_MAX_LEVEL) {
1456                 generic_err(node, 0,
1457                         "invalid level for node, have %d expect [1, %d]",
1458                         level, BTRFS_MAX_LEVEL - 1);
1459                 return -EUCLEAN;
1460         }
1461         if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(fs_info)) {
1462                 btrfs_crit(fs_info,
1463 "corrupt node: root=%llu block=%llu, nritems too %s, have %lu expect range [1,%u]",
1464                            btrfs_header_owner(node), node->start,
1465                            nr == 0 ? "small" : "large", nr,
1466                            BTRFS_NODEPTRS_PER_BLOCK(fs_info));
1467                 return -EUCLEAN;
1468         }
1469 
1470         for (slot = 0; slot < nr - 1; slot++) {
1471                 bytenr = btrfs_node_blockptr(node, slot);
1472                 btrfs_node_key_to_cpu(node, &key, slot);
1473                 btrfs_node_key_to_cpu(node, &next_key, slot + 1);
1474 
1475                 if (!bytenr) {
1476                         generic_err(node, slot,
1477                                 "invalid NULL node pointer");
1478                         ret = -EUCLEAN;
1479                         goto out;
1480                 }
1481                 if (!IS_ALIGNED(bytenr, fs_info->sectorsize)) {
1482                         generic_err(node, slot,
1483                         "unaligned pointer, have %llu should be aligned to %u",
1484                                 bytenr, fs_info->sectorsize);
1485                         ret = -EUCLEAN;
1486                         goto out;
1487                 }
1488 
1489                 if (btrfs_comp_cpu_keys(&key, &next_key) >= 0) {
1490                         generic_err(node, slot,
1491         "bad key order, current (%llu %u %llu) next (%llu %u %llu)",
1492                                 key.objectid, key.type, key.offset,
1493                                 next_key.objectid, next_key.type,
1494                                 next_key.offset);
1495                         ret = -EUCLEAN;
1496                         goto out;
1497                 }
1498         }
1499 out:
1500         return ret;
1501 }
1502 ALLOW_ERROR_INJECTION(btrfs_check_node, ERRNO);
/* [<][>][^][v][top][bottom][index][help] */
root/fs/btrfs/tree-checker.c

DEFINITIONS
