fs/btrfs/tree-checker.c

   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 "ctree.h"
  19 #include "tree-checker.h"
  20 #include "disk-io.h"
  21 #include "compression.h"
  22 #include "volumes.h"
  23
  24 /*
  25  * Error message should follow the following format:
  26  * corrupt <type>: <identifier>, <reason>[, <bad_value>]
  27  *
  28  * @type:       leaf or node
  29  * @identifier: the necessary info to locate the leaf/node.
  30  *              It's recommened to decode key.objecitd/offset if it's
  31  *              meaningful.
  32  * @reason:     describe the error
  33  * @bad_value:  optional, it's recommened to output bad value and its
  34  *              expected value (range).
  35  *
  36  * Since comma is used to separate the components, only space is allowed
  37  * inside each component.
  38  */
  39
  40 /*
  41  * Append generic "corrupt leaf/node root=%llu block=%llu slot=%d: " to @fmt.
  42  * Allows callers to customize the output.
  43  */
  44 __printf(4, 5)
  45 __cold
  46 static void generic_err(const struct btrfs_fs_info *fs_info,
  47                         const struct extent_buffer *eb, int slot,
  48                         const char *fmt, ...)
  49 {
  50         struct va_format vaf;
  51         va_list args;
  52
  53         va_start(args, fmt);
  54
  55         vaf.fmt = fmt;
  56         vaf.va = &args;
  57
  58         btrfs_crit(fs_info,
  59                 "corrupt %s: root=%llu block=%llu slot=%d, %pV",
  60                 btrfs_header_level(eb) == 0 ? "leaf" : "node",
  61                 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, &vaf);
  62         va_end(args);
  63 }
  64
  65 /*
  66  * Customized reporter for extent data item, since its key objectid and
  67  * offset has its own meaning.
  68  */
  69 __printf(4, 5)
  70 __cold
  71 static void file_extent_err(const struct btrfs_fs_info *fs_info,
  72                             const struct extent_buffer *eb, int slot,
  73                             const char *fmt, ...)
  74 {
  75         struct btrfs_key key;
  76         struct va_format vaf;
  77         va_list args;
  78
  79         btrfs_item_key_to_cpu(eb, &key, slot);
  80         va_start(args, fmt);
  81
  82         vaf.fmt = fmt;
  83         vaf.va = &args;
  84
  85         btrfs_crit(fs_info,
  86         "corrupt %s: root=%llu block=%llu slot=%d ino=%llu file_offset=%llu, %pV",
  87                 btrfs_header_level(eb) == 0 ? "leaf" : "node",
  88                 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
  89                 key.objectid, key.offset, &vaf);
  90         va_end(args);
  91 }
  92
  93 /*
  94  * Return 0 if the btrfs_file_extent_##name is aligned to @alignment
  95  * Else return 1
  96  */
  97 #define CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, name, alignment)            \
  98 ({                                                                            \
  99         if (!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment))) \
 100                 file_extent_err((fs_info), (leaf), (slot),                    \
 101         "invalid %s for file extent, have %llu, should be aligned to %u",     \
 102                         (#name), btrfs_file_extent_##name((leaf), (fi)),      \
 103                         (alignment));                                         \
 104         (!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment)));   \
 105 })
 106
 107 static int check_extent_data_item(struct btrfs_fs_info *fs_info,
 108                                   struct extent_buffer *leaf,
 109                                   struct btrfs_key *key, int slot)
 110 {
 111         struct btrfs_file_extent_item *fi;
 112         u32 sectorsize = fs_info->sectorsize;
 113         u32 item_size = btrfs_item_size_nr(leaf, slot);
 114
 115         if (!IS_ALIGNED(key->offset, sectorsize)) {
 116                 file_extent_err(fs_info, leaf, slot,
 117 "unaligned file_offset for file extent, have %llu should be aligned to %u",
 118                         key->offset, sectorsize);
 119                 return -EUCLEAN;
 120         }
 121
 122         fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
 123
 124         if (btrfs_file_extent_type(leaf, fi) > BTRFS_FILE_EXTENT_TYPES) {
 125                 file_extent_err(fs_info, leaf, slot,
 126                 "invalid type for file extent, have %u expect range [0, %u]",
 127                         btrfs_file_extent_type(leaf, fi),
 128                         BTRFS_FILE_EXTENT_TYPES);
 129                 return -EUCLEAN;
 130         }
 131
 132         /*
 133          * Support for new compression/encrption must introduce incompat flag,
 134          * and must be caught in open_ctree().
 135          */
 136         if (btrfs_file_extent_compression(leaf, fi) > BTRFS_COMPRESS_TYPES) {
 137                 file_extent_err(fs_info, leaf, slot,
 138         "invalid compression for file extent, have %u expect range [0, %u]",
 139                         btrfs_file_extent_compression(leaf, fi),
 140                         BTRFS_COMPRESS_TYPES);
 141                 return -EUCLEAN;
 142         }
 143         if (btrfs_file_extent_encryption(leaf, fi)) {
 144                 file_extent_err(fs_info, leaf, slot,
 145                         "invalid encryption for file extent, have %u expect 0",
 146                         btrfs_file_extent_encryption(leaf, fi));
 147                 return -EUCLEAN;
 148         }
 149         if (btrfs_file_extent_type(leaf, fi) == BTRFS_FILE_EXTENT_INLINE) {
 150                 /* Inline extent must have 0 as key offset */
 151                 if (key->offset) {
 152                         file_extent_err(fs_info, leaf, slot,
 153                 "invalid file_offset for inline file extent, have %llu expect 0",
 154                                 key->offset);
 155                         return -EUCLEAN;
 156                 }
 157
 158                 /* Compressed inline extent has no on-disk size, skip it */
 159                 if (btrfs_file_extent_compression(leaf, fi) !=
 160                     BTRFS_COMPRESS_NONE)
 161                         return 0;
 162
 163                 /* Uncompressed inline extent size must match item size */
 164                 if (item_size != BTRFS_FILE_EXTENT_INLINE_DATA_START +
 165                     btrfs_file_extent_ram_bytes(leaf, fi)) {
 166                         file_extent_err(fs_info, leaf, slot,
 167         "invalid ram_bytes for uncompressed inline extent, have %u expect %llu",
 168                                 item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START +
 169                                 btrfs_file_extent_ram_bytes(leaf, fi));
 170                         return -EUCLEAN;
 171                 }
 172                 return 0;
 173         }
 174
 175         /* Regular or preallocated extent has fixed item size */
 176         if (item_size != sizeof(*fi)) {
 177                 file_extent_err(fs_info, leaf, slot,
 178         "invalid item size for reg/prealloc file extent, have %u expect %zu",
 179                         item_size, sizeof(*fi));
 180                 return -EUCLEAN;
 181         }
 182         if (CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, ram_bytes, sectorsize) ||
 183             CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, disk_bytenr, sectorsize) ||
 184             CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, disk_num_bytes, sectorsize) ||
 185             CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, offset, sectorsize) ||
 186             CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, num_bytes, sectorsize))
 187                 return -EUCLEAN;
 188         return 0;
 189 }
 190
 191 static int check_csum_item(struct btrfs_fs_info *fs_info,
 192                            struct extent_buffer *leaf, struct btrfs_key *key,
 193                            int slot)
 194 {
 195         u32 sectorsize = fs_info->sectorsize;
 196         u32 csumsize = btrfs_super_csum_size(fs_info->super_copy);
 197
 198         if (key->objectid != BTRFS_EXTENT_CSUM_OBJECTID) {
 199                 generic_err(fs_info, leaf, slot,
 200                 "invalid key objectid for csum item, have %llu expect %llu",
 201                         key->objectid, BTRFS_EXTENT_CSUM_OBJECTID);
 202                 return -EUCLEAN;
 203         }
 204         if (!IS_ALIGNED(key->offset, sectorsize)) {
 205                 generic_err(fs_info, leaf, slot,
 206         "unaligned key offset for csum item, have %llu should be aligned to %u",
 207                         key->offset, sectorsize);
 208                 return -EUCLEAN;
 209         }
 210         if (!IS_ALIGNED(btrfs_item_size_nr(leaf, slot), csumsize)) {
 211                 generic_err(fs_info, leaf, slot,
 212         "unaligned item size for csum item, have %u should be aligned to %u",
 213                         btrfs_item_size_nr(leaf, slot), csumsize);
 214                 return -EUCLEAN;
 215         }
 216         return 0;
 217 }
 218
 219 /*
 220  * Customized reported for dir_item, only important new info is key->objectid,
 221  * which represents inode number
 222  */
 223 __printf(4, 5)
 224 __cold
 225 static void dir_item_err(const struct btrfs_fs_info *fs_info,
 226                          const struct extent_buffer *eb, int slot,
 227                          const char *fmt, ...)
 228 {
 229         struct btrfs_key key;
 230         struct va_format vaf;
 231         va_list args;
 232
 233         btrfs_item_key_to_cpu(eb, &key, slot);
 234         va_start(args, fmt);
 235
 236         vaf.fmt = fmt;
 237         vaf.va = &args;
 238
 239         btrfs_crit(fs_info,
 240         "corrupt %s: root=%llu block=%llu slot=%d ino=%llu, %pV",
 241                 btrfs_header_level(eb) == 0 ? "leaf" : "node",
 242                 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
 243                 key.objectid, &vaf);
 244         va_end(args);
 245 }
 246
 247 static int check_dir_item(struct btrfs_fs_info *fs_info,
 248                           struct extent_buffer *leaf,
 249                           struct btrfs_key *key, int slot)
 250 {
 251         struct btrfs_dir_item *di;
 252         u32 item_size = btrfs_item_size_nr(leaf, slot);
 253         u32 cur = 0;
 254
 255         di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
 256         while (cur < item_size) {
 257                 u32 name_len;
 258                 u32 data_len;
 259                 u32 max_name_len;
 260                 u32 total_size;
 261                 u32 name_hash;
 262                 u8 dir_type;
 263
 264                 /* header itself should not cross item boundary */
 265                 if (cur + sizeof(*di) > item_size) {
 266                         dir_item_err(fs_info, leaf, slot,
 267                 "dir item header crosses item boundary, have %zu boundary %u",
 268                                 cur + sizeof(*di), item_size);
 269                         return -EUCLEAN;
 270                 }
 271
 272                 /* dir type check */
 273                 dir_type = btrfs_dir_type(leaf, di);
 274                 if (dir_type >= BTRFS_FT_MAX) {
 275                         dir_item_err(fs_info, leaf, slot,
 276                         "invalid dir item type, have %u expect [0, %u)",
 277                                 dir_type, BTRFS_FT_MAX);
 278                         return -EUCLEAN;
 279                 }
 280
 281                 if (key->type == BTRFS_XATTR_ITEM_KEY &&
 282                     dir_type != BTRFS_FT_XATTR) {
 283                         dir_item_err(fs_info, leaf, slot,
 284                 "invalid dir item type for XATTR key, have %u expect %u",
 285                                 dir_type, BTRFS_FT_XATTR);
 286                         return -EUCLEAN;
 287                 }
 288                 if (dir_type == BTRFS_FT_XATTR &&
 289                     key->type != BTRFS_XATTR_ITEM_KEY) {
 290                         dir_item_err(fs_info, leaf, slot,
 291                         "xattr dir type found for non-XATTR key");
 292                         return -EUCLEAN;
 293                 }
 294                 if (dir_type == BTRFS_FT_XATTR)
 295                         max_name_len = XATTR_NAME_MAX;
 296                 else
 297                         max_name_len = BTRFS_NAME_LEN;
 298
 299                 /* Name/data length check */
 300                 name_len = btrfs_dir_name_len(leaf, di);
 301                 data_len = btrfs_dir_data_len(leaf, di);
 302                 if (name_len > max_name_len) {
 303                         dir_item_err(fs_info, leaf, slot,
 304                         "dir item name len too long, have %u max %u",
 305                                 name_len, max_name_len);
 306                         return -EUCLEAN;
 307                 }
 308                 if (name_len + data_len > BTRFS_MAX_XATTR_SIZE(fs_info)) {
 309                         dir_item_err(fs_info, leaf, slot,
 310                         "dir item name and data len too long, have %u max %u",
 311                                 name_len + data_len,
 312                                 BTRFS_MAX_XATTR_SIZE(fs_info));
 313                         return -EUCLEAN;
 314                 }
 315
 316                 if (data_len && dir_type != BTRFS_FT_XATTR) {
 317                         dir_item_err(fs_info, leaf, slot,
 318                         "dir item with invalid data len, have %u expect 0",
 319                                 data_len);
 320                         return -EUCLEAN;
 321                 }
 322
 323                 total_size = sizeof(*di) + name_len + data_len;
 324
 325                 /* header and name/data should not cross item boundary */
 326                 if (cur + total_size > item_size) {
 327                         dir_item_err(fs_info, leaf, slot,
 328                 "dir item data crosses item boundary, have %u boundary %u",
 329                                 cur + total_size, item_size);
 330                         return -EUCLEAN;
 331                 }
 332
 333                 /*
 334                  * Special check for XATTR/DIR_ITEM, as key->offset is name
 335                  * hash, should match its name
 336                  */
 337                 if (key->type == BTRFS_DIR_ITEM_KEY ||
 338                     key->type == BTRFS_XATTR_ITEM_KEY) {
 339                         char namebuf[max(BTRFS_NAME_LEN, XATTR_NAME_MAX)];
 340
 341                         read_extent_buffer(leaf, namebuf,
 342                                         (unsigned long)(di + 1), name_len);
 343                         name_hash = btrfs_name_hash(namebuf, name_len);
 344                         if (key->offset != name_hash) {
 345                                 dir_item_err(fs_info, leaf, slot,
 346                 "name hash mismatch with key, have 0x%016x expect 0x%016llx",
 347                                         name_hash, key->offset);
 348                                 return -EUCLEAN;
 349                         }
 350                 }
 351                 cur += total_size;
 352                 di = (struct btrfs_dir_item *)((void *)di + total_size);
 353         }
 354         return 0;
 355 }
 356
 357 __printf(4, 5)
 358 __cold
 359 static void block_group_err(const struct btrfs_fs_info *fs_info,
 360                             const struct extent_buffer *eb, int slot,
 361                             const char *fmt, ...)
 362 {
 363         struct btrfs_key key;
 364         struct va_format vaf;
 365         va_list args;
 366
 367         btrfs_item_key_to_cpu(eb, &key, slot);
 368         va_start(args, fmt);
 369
 370         vaf.fmt = fmt;
 371         vaf.va = &args;
 372
 373         btrfs_crit(fs_info,
 374         "corrupt %s: root=%llu block=%llu slot=%d bg_start=%llu bg_len=%llu, %pV",
 375                 btrfs_header_level(eb) == 0 ? "leaf" : "node",
 376                 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
 377                 key.objectid, key.offset, &vaf);
 378         va_end(args);
 379 }
 380
 381 static int check_block_group_item(struct btrfs_fs_info *fs_info,
 382                                   struct extent_buffer *leaf,
 383                                   struct btrfs_key *key, int slot)
 384 {
 385         struct btrfs_block_group_item bgi;
 386         u32 item_size = btrfs_item_size_nr(leaf, slot);
 387         u64 flags;
 388         u64 type;
 389
 390         /*
 391          * Here we don't really care about alignment since extent allocator can
 392          * handle it.  We care more about the size.
 393          */
 394         if (key->offset == 0) {
 395                 block_group_err(fs_info, leaf, slot,
 396                                 "invalid block group size 0");
 397                 return -EUCLEAN;
 398         }
 399
 400         if (item_size != sizeof(bgi)) {
 401                 block_group_err(fs_info, leaf, slot,
 402                         "invalid item size, have %u expect %zu",
 403                                 item_size, sizeof(bgi));
 404                 return -EUCLEAN;
 405         }
 406
 407         read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot),
 408                            sizeof(bgi));
 409         if (btrfs_block_group_chunk_objectid(&bgi) !=
 410             BTRFS_FIRST_CHUNK_TREE_OBJECTID) {
 411                 block_group_err(fs_info, leaf, slot,
 412                 "invalid block group chunk objectid, have %llu expect %llu",
 413                                 btrfs_block_group_chunk_objectid(&bgi),
 414                                 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
 415                 return -EUCLEAN;
 416         }
 417
 418         if (btrfs_block_group_used(&bgi) > key->offset) {
 419                 block_group_err(fs_info, leaf, slot,
 420                         "invalid block group used, have %llu expect [0, %llu)",
 421                                 btrfs_block_group_used(&bgi), key->offset);
 422                 return -EUCLEAN;
 423         }
 424
 425         flags = btrfs_block_group_flags(&bgi);
 426         if (hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) > 1) {
 427                 block_group_err(fs_info, leaf, slot,
 428 "invalid profile flags, have 0x%llx (%lu bits set) expect no more than 1 bit set",
 429                         flags & BTRFS_BLOCK_GROUP_PROFILE_MASK,
 430                         hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK));
 431                 return -EUCLEAN;
 432         }
 433
 434         type = flags & BTRFS_BLOCK_GROUP_TYPE_MASK;
 435         if (type != BTRFS_BLOCK_GROUP_DATA &&
 436             type != BTRFS_BLOCK_GROUP_METADATA &&
 437             type != BTRFS_BLOCK_GROUP_SYSTEM &&
 438             type != (BTRFS_BLOCK_GROUP_METADATA |
 439                            BTRFS_BLOCK_GROUP_DATA)) {
 440                 block_group_err(fs_info, leaf, slot,
 441 "invalid type, have 0x%llx (%lu bits set) expect either 0x%llx, 0x%llx, 0x%llx or 0x%llx",
 442                         type, hweight64(type),
 443                         BTRFS_BLOCK_GROUP_DATA, BTRFS_BLOCK_GROUP_METADATA,
 444                         BTRFS_BLOCK_GROUP_SYSTEM,
 445                         BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA);
 446                 return -EUCLEAN;
 447         }
 448         return 0;
 449 }
 450
 451 /*
 452  * Common point to switch the item-specific validation.
 453  */
 454 static int check_leaf_item(struct btrfs_fs_info *fs_info,
 455                            struct extent_buffer *leaf,
 456                            struct btrfs_key *key, int slot)
 457 {
 458         int ret = 0;
 459
 460         switch (key->type) {
 461         case BTRFS_EXTENT_DATA_KEY:
 462                 ret = check_extent_data_item(fs_info, leaf, key, slot);
 463                 break;
 464         case BTRFS_EXTENT_CSUM_KEY:
 465                 ret = check_csum_item(fs_info, leaf, key, slot);
 466                 break;
 467         case BTRFS_DIR_ITEM_KEY:
 468         case BTRFS_DIR_INDEX_KEY:
 469         case BTRFS_XATTR_ITEM_KEY:
 470                 ret = check_dir_item(fs_info, leaf, key, slot);
 471                 break;
 472         case BTRFS_BLOCK_GROUP_ITEM_KEY:
 473                 ret = check_block_group_item(fs_info, leaf, key, slot);
 474                 break;
 475         }
 476         return ret;
 477 }
 478
 479 static int check_leaf(struct btrfs_fs_info *fs_info, struct extent_buffer *leaf,
 480                       bool check_item_data)
 481 {
 482         /* No valid key type is 0, so all key should be larger than this key */
 483         struct btrfs_key prev_key = {0, 0, 0};
 484         struct btrfs_key key;
 485         u32 nritems = btrfs_header_nritems(leaf);
 486         int slot;
 487
 488         if (btrfs_header_level(leaf) != 0) {
 489                 generic_err(fs_info, leaf, 0,
 490                         "invalid level for leaf, have %d expect 0",
 491                         btrfs_header_level(leaf));
 492                 return -EUCLEAN;
 493         }
 494
 495         /*
 496          * Extent buffers from a relocation tree have a owner field that
 497          * corresponds to the subvolume tree they are based on. So just from an
 498          * extent buffer alone we can not find out what is the id of the
 499          * corresponding subvolume tree, so we can not figure out if the extent
 500          * buffer corresponds to the root of the relocation tree or not. So
 501          * skip this check for relocation trees.
 502          */
 503         if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) {
 504                 u64 owner = btrfs_header_owner(leaf);
 505                 struct btrfs_root *check_root;
 506
 507                 /* These trees must never be empty */
 508                 if (owner == BTRFS_ROOT_TREE_OBJECTID ||
 509                     owner == BTRFS_CHUNK_TREE_OBJECTID ||
 510                     owner == BTRFS_EXTENT_TREE_OBJECTID ||
 511                     owner == BTRFS_DEV_TREE_OBJECTID ||
 512                     owner == BTRFS_FS_TREE_OBJECTID ||
 513                     owner == BTRFS_DATA_RELOC_TREE_OBJECTID) {
 514                         generic_err(fs_info, leaf, 0,
 515                         "invalid root, root %llu must never be empty",
 516                                     owner);
 517                         return -EUCLEAN;
 518                 }
 519                 /* Unknown tree */
 520                 if (owner == 0) {
 521                         generic_err(fs_info, leaf, 0,
 522                                 "invalid owner, root 0 is not defined");
 523                         return -EUCLEAN;
 524                 }
 525                 key.objectid = owner;
 526                 key.type = BTRFS_ROOT_ITEM_KEY;
 527                 key.offset = (u64)-1;
 528
 529                 check_root = btrfs_get_fs_root(fs_info, &key, false);
 530                 /*
 531                  * The only reason we also check NULL here is that during
 532                  * open_ctree() some roots has not yet been set up.
 533                  */
 534                 if (!IS_ERR_OR_NULL(check_root)) {
 535                         struct extent_buffer *eb;
 536
 537                         eb = btrfs_root_node(check_root);
 538                         /* if leaf is the root, then it's fine */
 539                         if (leaf != eb) {
 540                                 generic_err(fs_info, leaf, 0,
 541                 "invalid nritems, have %u should not be 0 for non-root leaf",
 542                                         nritems);
 543                                 free_extent_buffer(eb);
 544                                 return -EUCLEAN;
 545                         }
 546                         free_extent_buffer(eb);
 547                 }
 548                 return 0;
 549         }
 550
 551         if (nritems == 0)
 552                 return 0;
 553
 554         /*
 555          * Check the following things to make sure this is a good leaf, and
 556          * leaf users won't need to bother with similar sanity checks:
 557          *
 558          * 1) key ordering
 559          * 2) item offset and size
 560          *    No overlap, no hole, all inside the leaf.
 561          * 3) item content
 562          *    If possible, do comprehensive sanity check.
 563          *    NOTE: All checks must only rely on the item data itself.
 564          */
 565         for (slot = 0; slot < nritems; slot++) {
 566                 u32 item_end_expected;
 567                 int ret;
 568
 569                 btrfs_item_key_to_cpu(leaf, &key, slot);
 570
 571                 /* Make sure the keys are in the right order */
 572                 if (btrfs_comp_cpu_keys(&prev_key, &key) >= 0) {
 573                         generic_err(fs_info, leaf, slot,
 574         "bad key order, prev (%llu %u %llu) current (%llu %u %llu)",
 575                                 prev_key.objectid, prev_key.type,
 576                                 prev_key.offset, key.objectid, key.type,
 577                                 key.offset);
 578                         return -EUCLEAN;
 579                 }
 580
 581                 /*
 582                  * Make sure the offset and ends are right, remember that the
 583                  * item data starts at the end of the leaf and grows towards the
 584                  * front.
 585                  */
 586                 if (slot == 0)
 587                         item_end_expected = BTRFS_LEAF_DATA_SIZE(fs_info);
 588                 else
 589                         item_end_expected = btrfs_item_offset_nr(leaf,
 590                                                                  slot - 1);
 591                 if (btrfs_item_end_nr(leaf, slot) != item_end_expected) {
 592                         generic_err(fs_info, leaf, slot,
 593                                 "unexpected item end, have %u expect %u",
 594                                 btrfs_item_end_nr(leaf, slot),
 595                                 item_end_expected);
 596                         return -EUCLEAN;
 597                 }
 598
 599                 /*
 600                  * Check to make sure that we don't point outside of the leaf,
 601                  * just in case all the items are consistent to each other, but
 602                  * all point outside of the leaf.
 603                  */
 604                 if (btrfs_item_end_nr(leaf, slot) >
 605                     BTRFS_LEAF_DATA_SIZE(fs_info)) {
 606                         generic_err(fs_info, leaf, slot,
 607                         "slot end outside of leaf, have %u expect range [0, %u]",
 608                                 btrfs_item_end_nr(leaf, slot),
 609                                 BTRFS_LEAF_DATA_SIZE(fs_info));
 610                         return -EUCLEAN;
 611                 }
 612
 613                 /* Also check if the item pointer overlaps with btrfs item. */
 614                 if (btrfs_item_nr_offset(slot) + sizeof(struct btrfs_item) >
 615                     btrfs_item_ptr_offset(leaf, slot)) {
 616                         generic_err(fs_info, leaf, slot,
 617                 "slot overlaps with its data, item end %lu data start %lu",
 618                                 btrfs_item_nr_offset(slot) +
 619                                 sizeof(struct btrfs_item),
 620                                 btrfs_item_ptr_offset(leaf, slot));
 621                         return -EUCLEAN;
 622                 }
 623
 624                 if (check_item_data) {
 625                         /*
 626                          * Check if the item size and content meet other
 627                          * criteria
 628                          */
 629                         ret = check_leaf_item(fs_info, leaf, &key, slot);
 630                         if (ret < 0)
 631                                 return ret;
 632                 }
 633
 634                 prev_key.objectid = key.objectid;
 635                 prev_key.type = key.type;
 636                 prev_key.offset = key.offset;
 637         }
 638
 639         return 0;
 640 }
 641
 642 int btrfs_check_leaf_full(struct btrfs_fs_info *fs_info,
 643                           struct extent_buffer *leaf)
 644 {
 645         return check_leaf(fs_info, leaf, true);
 646 }
 647
 648 int btrfs_check_leaf_relaxed(struct btrfs_fs_info *fs_info,
 649                              struct extent_buffer *leaf)
 650 {
 651         return check_leaf(fs_info, leaf, false);
 652 }
 653
 654 int btrfs_check_node(struct btrfs_fs_info *fs_info, struct extent_buffer *node)
 655 {
 656         unsigned long nr = btrfs_header_nritems(node);
 657         struct btrfs_key key, next_key;
 658         int slot;
 659         int level = btrfs_header_level(node);
 660         u64 bytenr;
 661         int ret = 0;
 662
 663         if (level <= 0 || level >= BTRFS_MAX_LEVEL) {
 664                 generic_err(fs_info, node, 0,
 665                         "invalid level for node, have %d expect [1, %d]",
 666                         level, BTRFS_MAX_LEVEL - 1);
 667                 return -EUCLEAN;
 668         }
 669         if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(fs_info)) {
 670                 btrfs_crit(fs_info,
 671 "corrupt node: root=%llu block=%llu, nritems too %s, have %lu expect range [1,%u]",
 672                            btrfs_header_owner(node), node->start,
 673                            nr == 0 ? "small" : "large", nr,
 674                            BTRFS_NODEPTRS_PER_BLOCK(fs_info));
 675                 return -EUCLEAN;
 676         }
 677
 678         for (slot = 0; slot < nr - 1; slot++) {
 679                 bytenr = btrfs_node_blockptr(node, slot);
 680                 btrfs_node_key_to_cpu(node, &key, slot);
 681                 btrfs_node_key_to_cpu(node, &next_key, slot + 1);
 682
 683                 if (!bytenr) {
 684                         generic_err(fs_info, node, slot,
 685                                 "invalid NULL node pointer");
 686                         ret = -EUCLEAN;
 687                         goto out;
 688                 }
 689                 if (!IS_ALIGNED(bytenr, fs_info->sectorsize)) {
 690                         generic_err(fs_info, node, slot,
 691                         "unaligned pointer, have %llu should be aligned to %u",
 692                                 bytenr, fs_info->sectorsize);
 693                         ret = -EUCLEAN;
 694                         goto out;
 695                 }
 696
 697                 if (btrfs_comp_cpu_keys(&key, &next_key) >= 0) {
 698                         generic_err(fs_info, node, slot,
 699         "bad key order, current (%llu %u %llu) next (%llu %u %llu)",
 700                                 key.objectid, key.type, key.offset,
 701                                 next_key.objectid, next_key.type,
 702                                 next_key.offset);
 703                         ret = -EUCLEAN;
 704                         goto out;
 705                 }
 706         }
 707 out:
 708         return ret;
 709 }