1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) Qu Wenruo 2017. All rights reserved.
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.
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.
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.
18 #include <linux/types.h>
19 #include <linux/stddef.h>
20 #include <linux/error-injection.h>
22 #include "tree-checker.h"
24 #include "compression.h"
29 * Error message should follow the following format:
30 * corrupt <type>: <identifier>, <reason>[, <bad_value>]
33 * @identifier: the necessary info to locate the leaf/node.
34 * It's recommended to decode key.objecitd/offset if it's
36 * @reason: describe the error
37 * @bad_value: optional, it's recommended to output bad value and its
38 * expected value (range).
40 * Since comma is used to separate the components, only space is allowed
41 * inside each component.
45 * Append generic "corrupt leaf/node root=%llu block=%llu slot=%d: " to @fmt.
46 * Allows callers to customize the output.
50 static void generic_err(const struct extent_buffer *eb, int slot,
53 const struct btrfs_fs_info *fs_info = eb->fs_info;
63 "corrupt %s: root=%llu block=%llu slot=%d, %pV",
64 btrfs_header_level(eb) == 0 ? "leaf" : "node",
65 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, &vaf);
70 * Customized reporter for extent data item, since its key objectid and
71 * offset has its own meaning.
75 static void file_extent_err(const struct extent_buffer *eb, int slot,
78 const struct btrfs_fs_info *fs_info = eb->fs_info;
83 btrfs_item_key_to_cpu(eb, &key, slot);
90 "corrupt %s: root=%llu block=%llu slot=%d ino=%llu file_offset=%llu, %pV",
91 btrfs_header_level(eb) == 0 ? "leaf" : "node",
92 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
93 key.objectid, key.offset, &vaf);
98 * Return 0 if the btrfs_file_extent_##name is aligned to @alignment
101 #define CHECK_FE_ALIGNED(leaf, slot, fi, name, alignment) \
103 if (!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment))) \
104 file_extent_err((leaf), (slot), \
105 "invalid %s for file extent, have %llu, should be aligned to %u", \
106 (#name), btrfs_file_extent_##name((leaf), (fi)), \
108 (!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment))); \
111 static u64 file_extent_end(struct extent_buffer *leaf,
112 struct btrfs_key *key,
113 struct btrfs_file_extent_item *extent)
118 if (btrfs_file_extent_type(leaf, extent) == BTRFS_FILE_EXTENT_INLINE) {
119 len = btrfs_file_extent_ram_bytes(leaf, extent);
120 end = ALIGN(key->offset + len, leaf->fs_info->sectorsize);
122 len = btrfs_file_extent_num_bytes(leaf, extent);
123 end = key->offset + len;
129 * Customized report for dir_item, the only new important information is
130 * key->objectid, which represents inode number
134 static void dir_item_err(const struct extent_buffer *eb, int slot,
135 const char *fmt, ...)
137 const struct btrfs_fs_info *fs_info = eb->fs_info;
138 struct btrfs_key key;
139 struct va_format vaf;
142 btrfs_item_key_to_cpu(eb, &key, slot);
149 "corrupt %s: root=%llu block=%llu slot=%d ino=%llu, %pV",
150 btrfs_header_level(eb) == 0 ? "leaf" : "node",
151 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
157 * This functions checks prev_key->objectid, to ensure current key and prev_key
158 * share the same objectid as inode number.
160 * This is to detect missing INODE_ITEM in subvolume trees.
162 * Return true if everything is OK or we don't need to check.
163 * Return false if anything is wrong.
165 static bool check_prev_ino(struct extent_buffer *leaf,
166 struct btrfs_key *key, int slot,
167 struct btrfs_key *prev_key)
169 /* No prev key, skip check */
173 /* Only these key->types needs to be checked */
174 ASSERT(key->type == BTRFS_XATTR_ITEM_KEY ||
175 key->type == BTRFS_INODE_REF_KEY ||
176 key->type == BTRFS_DIR_INDEX_KEY ||
177 key->type == BTRFS_DIR_ITEM_KEY ||
178 key->type == BTRFS_EXTENT_DATA_KEY);
181 * Only subvolume trees along with their reloc trees need this check.
182 * Things like log tree doesn't follow this ino requirement.
184 if (!is_fstree(btrfs_header_owner(leaf)))
187 if (key->objectid == prev_key->objectid)
191 dir_item_err(leaf, slot,
192 "invalid previous key objectid, have %llu expect %llu",
193 prev_key->objectid, key->objectid);
196 static int check_extent_data_item(struct extent_buffer *leaf,
197 struct btrfs_key *key, int slot,
198 struct btrfs_key *prev_key)
200 struct btrfs_fs_info *fs_info = leaf->fs_info;
201 struct btrfs_file_extent_item *fi;
202 u32 sectorsize = fs_info->sectorsize;
203 u32 item_size = btrfs_item_size_nr(leaf, slot);
206 if (!IS_ALIGNED(key->offset, sectorsize)) {
207 file_extent_err(leaf, slot,
208 "unaligned file_offset for file extent, have %llu should be aligned to %u",
209 key->offset, sectorsize);
214 * Previous key must have the same key->objectid (ino).
215 * It can be XATTR_ITEM, INODE_ITEM or just another EXTENT_DATA.
216 * But if objectids mismatch, it means we have a missing
219 if (!check_prev_ino(leaf, key, slot, prev_key))
222 fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
225 * Make sure the item contains at least inline header, so the file
226 * extent type is not some garbage.
228 if (item_size < BTRFS_FILE_EXTENT_INLINE_DATA_START) {
229 file_extent_err(leaf, slot,
230 "invalid item size, have %u expect [%zu, %u)",
231 item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START,
235 if (btrfs_file_extent_type(leaf, fi) >= BTRFS_NR_FILE_EXTENT_TYPES) {
236 file_extent_err(leaf, slot,
237 "invalid type for file extent, have %u expect range [0, %u]",
238 btrfs_file_extent_type(leaf, fi),
239 BTRFS_NR_FILE_EXTENT_TYPES - 1);
244 * Support for new compression/encryption must introduce incompat flag,
245 * and must be caught in open_ctree().
247 if (btrfs_file_extent_compression(leaf, fi) >= BTRFS_NR_COMPRESS_TYPES) {
248 file_extent_err(leaf, slot,
249 "invalid compression for file extent, have %u expect range [0, %u]",
250 btrfs_file_extent_compression(leaf, fi),
251 BTRFS_NR_COMPRESS_TYPES - 1);
254 if (btrfs_file_extent_encryption(leaf, fi)) {
255 file_extent_err(leaf, slot,
256 "invalid encryption for file extent, have %u expect 0",
257 btrfs_file_extent_encryption(leaf, fi));
260 if (btrfs_file_extent_type(leaf, fi) == BTRFS_FILE_EXTENT_INLINE) {
261 /* Inline extent must have 0 as key offset */
263 file_extent_err(leaf, slot,
264 "invalid file_offset for inline file extent, have %llu expect 0",
269 /* Compressed inline extent has no on-disk size, skip it */
270 if (btrfs_file_extent_compression(leaf, fi) !=
274 /* Uncompressed inline extent size must match item size */
275 if (item_size != BTRFS_FILE_EXTENT_INLINE_DATA_START +
276 btrfs_file_extent_ram_bytes(leaf, fi)) {
277 file_extent_err(leaf, slot,
278 "invalid ram_bytes for uncompressed inline extent, have %u expect %llu",
279 item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START +
280 btrfs_file_extent_ram_bytes(leaf, fi));
286 /* Regular or preallocated extent has fixed item size */
287 if (item_size != sizeof(*fi)) {
288 file_extent_err(leaf, slot,
289 "invalid item size for reg/prealloc file extent, have %u expect %zu",
290 item_size, sizeof(*fi));
293 if (CHECK_FE_ALIGNED(leaf, slot, fi, ram_bytes, sectorsize) ||
294 CHECK_FE_ALIGNED(leaf, slot, fi, disk_bytenr, sectorsize) ||
295 CHECK_FE_ALIGNED(leaf, slot, fi, disk_num_bytes, sectorsize) ||
296 CHECK_FE_ALIGNED(leaf, slot, fi, offset, sectorsize) ||
297 CHECK_FE_ALIGNED(leaf, slot, fi, num_bytes, sectorsize))
300 /* Catch extent end overflow */
301 if (check_add_overflow(btrfs_file_extent_num_bytes(leaf, fi),
302 key->offset, &extent_end)) {
303 file_extent_err(leaf, slot,
304 "extent end overflow, have file offset %llu extent num bytes %llu",
306 btrfs_file_extent_num_bytes(leaf, fi));
311 * Check that no two consecutive file extent items, in the same leaf,
312 * present ranges that overlap each other.
315 prev_key->objectid == key->objectid &&
316 prev_key->type == BTRFS_EXTENT_DATA_KEY) {
317 struct btrfs_file_extent_item *prev_fi;
320 prev_fi = btrfs_item_ptr(leaf, slot - 1,
321 struct btrfs_file_extent_item);
322 prev_end = file_extent_end(leaf, prev_key, prev_fi);
323 if (prev_end > key->offset) {
324 file_extent_err(leaf, slot - 1,
325 "file extent end range (%llu) goes beyond start offset (%llu) of the next file extent",
326 prev_end, key->offset);
334 static int check_csum_item(struct extent_buffer *leaf, struct btrfs_key *key,
335 int slot, struct btrfs_key *prev_key)
337 struct btrfs_fs_info *fs_info = leaf->fs_info;
338 u32 sectorsize = fs_info->sectorsize;
339 u32 csumsize = btrfs_super_csum_size(fs_info->super_copy);
341 if (key->objectid != BTRFS_EXTENT_CSUM_OBJECTID) {
342 generic_err(leaf, slot,
343 "invalid key objectid for csum item, have %llu expect %llu",
344 key->objectid, BTRFS_EXTENT_CSUM_OBJECTID);
347 if (!IS_ALIGNED(key->offset, sectorsize)) {
348 generic_err(leaf, slot,
349 "unaligned key offset for csum item, have %llu should be aligned to %u",
350 key->offset, sectorsize);
353 if (!IS_ALIGNED(btrfs_item_size_nr(leaf, slot), csumsize)) {
354 generic_err(leaf, slot,
355 "unaligned item size for csum item, have %u should be aligned to %u",
356 btrfs_item_size_nr(leaf, slot), csumsize);
359 if (slot > 0 && prev_key->type == BTRFS_EXTENT_CSUM_KEY) {
363 prev_item_size = btrfs_item_size_nr(leaf, slot - 1);
364 prev_csum_end = (prev_item_size / csumsize) * sectorsize;
365 prev_csum_end += prev_key->offset;
366 if (prev_csum_end > key->offset) {
367 generic_err(leaf, slot - 1,
368 "csum end range (%llu) goes beyond the start range (%llu) of the next csum item",
369 prev_csum_end, key->offset);
376 /* Inode item error output has the same format as dir_item_err() */
377 #define inode_item_err(eb, slot, fmt, ...) \
378 dir_item_err(eb, slot, fmt, __VA_ARGS__)
380 static int check_inode_key(struct extent_buffer *leaf, struct btrfs_key *key,
383 struct btrfs_key item_key;
386 btrfs_item_key_to_cpu(leaf, &item_key, slot);
387 is_inode_item = (item_key.type == BTRFS_INODE_ITEM_KEY);
389 /* For XATTR_ITEM, location key should be all 0 */
390 if (item_key.type == BTRFS_XATTR_ITEM_KEY) {
391 if (key->type != 0 || key->objectid != 0 || key->offset != 0)
396 if ((key->objectid < BTRFS_FIRST_FREE_OBJECTID ||
397 key->objectid > BTRFS_LAST_FREE_OBJECTID) &&
398 key->objectid != BTRFS_ROOT_TREE_DIR_OBJECTID &&
399 key->objectid != BTRFS_FREE_INO_OBJECTID) {
401 generic_err(leaf, slot,
402 "invalid key objectid: has %llu expect %llu or [%llu, %llu] or %llu",
403 key->objectid, BTRFS_ROOT_TREE_DIR_OBJECTID,
404 BTRFS_FIRST_FREE_OBJECTID,
405 BTRFS_LAST_FREE_OBJECTID,
406 BTRFS_FREE_INO_OBJECTID);
408 dir_item_err(leaf, slot,
409 "invalid location key objectid: has %llu expect %llu or [%llu, %llu] or %llu",
410 key->objectid, BTRFS_ROOT_TREE_DIR_OBJECTID,
411 BTRFS_FIRST_FREE_OBJECTID,
412 BTRFS_LAST_FREE_OBJECTID,
413 BTRFS_FREE_INO_OBJECTID);
417 if (key->offset != 0) {
419 inode_item_err(leaf, slot,
420 "invalid key offset: has %llu expect 0",
423 dir_item_err(leaf, slot,
424 "invalid location key offset:has %llu expect 0",
431 static int check_root_key(struct extent_buffer *leaf, struct btrfs_key *key,
434 struct btrfs_key item_key;
437 btrfs_item_key_to_cpu(leaf, &item_key, slot);
438 is_root_item = (item_key.type == BTRFS_ROOT_ITEM_KEY);
440 /* No such tree id */
441 if (key->objectid == 0) {
443 generic_err(leaf, slot, "invalid root id 0");
445 dir_item_err(leaf, slot,
446 "invalid location key root id 0");
450 /* DIR_ITEM/INDEX/INODE_REF is not allowed to point to non-fs trees */
451 if (!is_fstree(key->objectid) && !is_root_item) {
452 dir_item_err(leaf, slot,
453 "invalid location key objectid, have %llu expect [%llu, %llu]",
454 key->objectid, BTRFS_FIRST_FREE_OBJECTID,
455 BTRFS_LAST_FREE_OBJECTID);
460 * ROOT_ITEM with non-zero offset means this is a snapshot, created at
462 * Furthermore, for location key in DIR_ITEM, its offset is always -1.
464 * So here we only check offset for reloc tree whose key->offset must
467 if (key->objectid == BTRFS_TREE_RELOC_OBJECTID && key->offset == 0) {
468 generic_err(leaf, slot, "invalid root id 0 for reloc tree");
474 static int check_dir_item(struct extent_buffer *leaf,
475 struct btrfs_key *key, struct btrfs_key *prev_key,
478 struct btrfs_fs_info *fs_info = leaf->fs_info;
479 struct btrfs_dir_item *di;
480 u32 item_size = btrfs_item_size_nr(leaf, slot);
483 if (!check_prev_ino(leaf, key, slot, prev_key))
485 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
486 while (cur < item_size) {
487 struct btrfs_key location_key;
496 /* header itself should not cross item boundary */
497 if (cur + sizeof(*di) > item_size) {
498 dir_item_err(leaf, slot,
499 "dir item header crosses item boundary, have %zu boundary %u",
500 cur + sizeof(*di), item_size);
504 /* Location key check */
505 btrfs_dir_item_key_to_cpu(leaf, di, &location_key);
506 if (location_key.type == BTRFS_ROOT_ITEM_KEY) {
507 ret = check_root_key(leaf, &location_key, slot);
510 } else if (location_key.type == BTRFS_INODE_ITEM_KEY ||
511 location_key.type == 0) {
512 ret = check_inode_key(leaf, &location_key, slot);
516 dir_item_err(leaf, slot,
517 "invalid location key type, have %u, expect %u or %u",
518 location_key.type, BTRFS_ROOT_ITEM_KEY,
519 BTRFS_INODE_ITEM_KEY);
524 dir_type = btrfs_dir_type(leaf, di);
525 if (dir_type >= BTRFS_FT_MAX) {
526 dir_item_err(leaf, slot,
527 "invalid dir item type, have %u expect [0, %u)",
528 dir_type, BTRFS_FT_MAX);
532 if (key->type == BTRFS_XATTR_ITEM_KEY &&
533 dir_type != BTRFS_FT_XATTR) {
534 dir_item_err(leaf, slot,
535 "invalid dir item type for XATTR key, have %u expect %u",
536 dir_type, BTRFS_FT_XATTR);
539 if (dir_type == BTRFS_FT_XATTR &&
540 key->type != BTRFS_XATTR_ITEM_KEY) {
541 dir_item_err(leaf, slot,
542 "xattr dir type found for non-XATTR key");
545 if (dir_type == BTRFS_FT_XATTR)
546 max_name_len = XATTR_NAME_MAX;
548 max_name_len = BTRFS_NAME_LEN;
550 /* Name/data length check */
551 name_len = btrfs_dir_name_len(leaf, di);
552 data_len = btrfs_dir_data_len(leaf, di);
553 if (name_len > max_name_len) {
554 dir_item_err(leaf, slot,
555 "dir item name len too long, have %u max %u",
556 name_len, max_name_len);
559 if (name_len + data_len > BTRFS_MAX_XATTR_SIZE(fs_info)) {
560 dir_item_err(leaf, slot,
561 "dir item name and data len too long, have %u max %u",
563 BTRFS_MAX_XATTR_SIZE(fs_info));
567 if (data_len && dir_type != BTRFS_FT_XATTR) {
568 dir_item_err(leaf, slot,
569 "dir item with invalid data len, have %u expect 0",
574 total_size = sizeof(*di) + name_len + data_len;
576 /* header and name/data should not cross item boundary */
577 if (cur + total_size > item_size) {
578 dir_item_err(leaf, slot,
579 "dir item data crosses item boundary, have %u boundary %u",
580 cur + total_size, item_size);
585 * Special check for XATTR/DIR_ITEM, as key->offset is name
586 * hash, should match its name
588 if (key->type == BTRFS_DIR_ITEM_KEY ||
589 key->type == BTRFS_XATTR_ITEM_KEY) {
590 char namebuf[max(BTRFS_NAME_LEN, XATTR_NAME_MAX)];
592 read_extent_buffer(leaf, namebuf,
593 (unsigned long)(di + 1), name_len);
594 name_hash = btrfs_name_hash(namebuf, name_len);
595 if (key->offset != name_hash) {
596 dir_item_err(leaf, slot,
597 "name hash mismatch with key, have 0x%016x expect 0x%016llx",
598 name_hash, key->offset);
603 di = (struct btrfs_dir_item *)((void *)di + total_size);
610 static void block_group_err(const struct extent_buffer *eb, int slot,
611 const char *fmt, ...)
613 const struct btrfs_fs_info *fs_info = eb->fs_info;
614 struct btrfs_key key;
615 struct va_format vaf;
618 btrfs_item_key_to_cpu(eb, &key, slot);
625 "corrupt %s: root=%llu block=%llu slot=%d bg_start=%llu bg_len=%llu, %pV",
626 btrfs_header_level(eb) == 0 ? "leaf" : "node",
627 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
628 key.objectid, key.offset, &vaf);
632 static int check_block_group_item(struct extent_buffer *leaf,
633 struct btrfs_key *key, int slot)
635 struct btrfs_block_group_item bgi;
636 u32 item_size = btrfs_item_size_nr(leaf, slot);
641 * Here we don't really care about alignment since extent allocator can
642 * handle it. We care more about the size.
644 if (key->offset == 0) {
645 block_group_err(leaf, slot,
646 "invalid block group size 0");
650 if (item_size != sizeof(bgi)) {
651 block_group_err(leaf, slot,
652 "invalid item size, have %u expect %zu",
653 item_size, sizeof(bgi));
657 read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot),
659 if (btrfs_stack_block_group_chunk_objectid(&bgi) !=
660 BTRFS_FIRST_CHUNK_TREE_OBJECTID) {
661 block_group_err(leaf, slot,
662 "invalid block group chunk objectid, have %llu expect %llu",
663 btrfs_stack_block_group_chunk_objectid(&bgi),
664 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
668 if (btrfs_stack_block_group_used(&bgi) > key->offset) {
669 block_group_err(leaf, slot,
670 "invalid block group used, have %llu expect [0, %llu)",
671 btrfs_stack_block_group_used(&bgi), key->offset);
675 flags = btrfs_stack_block_group_flags(&bgi);
676 if (hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) > 1) {
677 block_group_err(leaf, slot,
678 "invalid profile flags, have 0x%llx (%lu bits set) expect no more than 1 bit set",
679 flags & BTRFS_BLOCK_GROUP_PROFILE_MASK,
680 hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK));
684 type = flags & BTRFS_BLOCK_GROUP_TYPE_MASK;
685 if (type != BTRFS_BLOCK_GROUP_DATA &&
686 type != BTRFS_BLOCK_GROUP_METADATA &&
687 type != BTRFS_BLOCK_GROUP_SYSTEM &&
688 type != (BTRFS_BLOCK_GROUP_METADATA |
689 BTRFS_BLOCK_GROUP_DATA)) {
690 block_group_err(leaf, slot,
691 "invalid type, have 0x%llx (%lu bits set) expect either 0x%llx, 0x%llx, 0x%llx or 0x%llx",
692 type, hweight64(type),
693 BTRFS_BLOCK_GROUP_DATA, BTRFS_BLOCK_GROUP_METADATA,
694 BTRFS_BLOCK_GROUP_SYSTEM,
695 BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA);
703 static void chunk_err(const struct extent_buffer *leaf,
704 const struct btrfs_chunk *chunk, u64 logical,
705 const char *fmt, ...)
707 const struct btrfs_fs_info *fs_info = leaf->fs_info;
709 struct va_format vaf;
714 /* Only superblock eb is able to have such small offset */
715 is_sb = (leaf->start == BTRFS_SUPER_INFO_OFFSET);
719 * Get the slot number by iterating through all slots, this
720 * would provide better readability.
722 for (i = 0; i < btrfs_header_nritems(leaf); i++) {
723 if (btrfs_item_ptr_offset(leaf, i) ==
724 (unsigned long)chunk) {
736 "corrupt superblock syschunk array: chunk_start=%llu, %pV",
740 "corrupt leaf: root=%llu block=%llu slot=%d chunk_start=%llu, %pV",
741 BTRFS_CHUNK_TREE_OBJECTID, leaf->start, slot,
747 * The common chunk check which could also work on super block sys chunk array.
749 * Return -EUCLEAN if anything is corrupted.
750 * Return 0 if everything is OK.
752 int btrfs_check_chunk_valid(struct extent_buffer *leaf,
753 struct btrfs_chunk *chunk, u64 logical)
755 struct btrfs_fs_info *fs_info = leaf->fs_info;
764 length = btrfs_chunk_length(leaf, chunk);
765 stripe_len = btrfs_chunk_stripe_len(leaf, chunk);
766 num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
767 sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk);
768 type = btrfs_chunk_type(leaf, chunk);
771 chunk_err(leaf, chunk, logical,
772 "invalid chunk num_stripes, have %u", num_stripes);
775 if (!IS_ALIGNED(logical, fs_info->sectorsize)) {
776 chunk_err(leaf, chunk, logical,
777 "invalid chunk logical, have %llu should aligned to %u",
778 logical, fs_info->sectorsize);
781 if (btrfs_chunk_sector_size(leaf, chunk) != fs_info->sectorsize) {
782 chunk_err(leaf, chunk, logical,
783 "invalid chunk sectorsize, have %u expect %u",
784 btrfs_chunk_sector_size(leaf, chunk),
785 fs_info->sectorsize);
788 if (!length || !IS_ALIGNED(length, fs_info->sectorsize)) {
789 chunk_err(leaf, chunk, logical,
790 "invalid chunk length, have %llu", length);
793 if (!is_power_of_2(stripe_len) || stripe_len != BTRFS_STRIPE_LEN) {
794 chunk_err(leaf, chunk, logical,
795 "invalid chunk stripe length: %llu",
799 if (~(BTRFS_BLOCK_GROUP_TYPE_MASK | BTRFS_BLOCK_GROUP_PROFILE_MASK) &
801 chunk_err(leaf, chunk, logical,
802 "unrecognized chunk type: 0x%llx",
803 ~(BTRFS_BLOCK_GROUP_TYPE_MASK |
804 BTRFS_BLOCK_GROUP_PROFILE_MASK) &
805 btrfs_chunk_type(leaf, chunk));
809 if (!has_single_bit_set(type & BTRFS_BLOCK_GROUP_PROFILE_MASK) &&
810 (type & BTRFS_BLOCK_GROUP_PROFILE_MASK) != 0) {
811 chunk_err(leaf, chunk, logical,
812 "invalid chunk profile flag: 0x%llx, expect 0 or 1 bit set",
813 type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
816 if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == 0) {
817 chunk_err(leaf, chunk, logical,
818 "missing chunk type flag, have 0x%llx one bit must be set in 0x%llx",
819 type, BTRFS_BLOCK_GROUP_TYPE_MASK);
823 if ((type & BTRFS_BLOCK_GROUP_SYSTEM) &&
824 (type & (BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA))) {
825 chunk_err(leaf, chunk, logical,
826 "system chunk with data or metadata type: 0x%llx",
831 features = btrfs_super_incompat_flags(fs_info->super_copy);
832 if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS)
836 if ((type & BTRFS_BLOCK_GROUP_METADATA) &&
837 (type & BTRFS_BLOCK_GROUP_DATA)) {
838 chunk_err(leaf, chunk, logical,
839 "mixed chunk type in non-mixed mode: 0x%llx", type);
844 if ((type & BTRFS_BLOCK_GROUP_RAID10 && sub_stripes != 2) ||
845 (type & BTRFS_BLOCK_GROUP_RAID1 && num_stripes != 2) ||
846 (type & BTRFS_BLOCK_GROUP_RAID5 && num_stripes < 2) ||
847 (type & BTRFS_BLOCK_GROUP_RAID6 && num_stripes < 3) ||
848 (type & BTRFS_BLOCK_GROUP_DUP && num_stripes != 2) ||
849 ((type & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 && num_stripes != 1)) {
850 chunk_err(leaf, chunk, logical,
851 "invalid num_stripes:sub_stripes %u:%u for profile %llu",
852 num_stripes, sub_stripes,
853 type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
861 * Enhanced version of chunk item checker.
863 * The common btrfs_check_chunk_valid() doesn't check item size since it needs
864 * to work on super block sys_chunk_array which doesn't have full item ptr.
866 static int check_leaf_chunk_item(struct extent_buffer *leaf,
867 struct btrfs_chunk *chunk,
868 struct btrfs_key *key, int slot)
872 if (btrfs_item_size_nr(leaf, slot) < sizeof(struct btrfs_chunk)) {
873 chunk_err(leaf, chunk, key->offset,
874 "invalid chunk item size: have %u expect [%zu, %u)",
875 btrfs_item_size_nr(leaf, slot),
876 sizeof(struct btrfs_chunk),
877 BTRFS_LEAF_DATA_SIZE(leaf->fs_info));
881 num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
882 /* Let btrfs_check_chunk_valid() handle this error type */
883 if (num_stripes == 0)
886 if (btrfs_chunk_item_size(num_stripes) !=
887 btrfs_item_size_nr(leaf, slot)) {
888 chunk_err(leaf, chunk, key->offset,
889 "invalid chunk item size: have %u expect %lu",
890 btrfs_item_size_nr(leaf, slot),
891 btrfs_chunk_item_size(num_stripes));
895 return btrfs_check_chunk_valid(leaf, chunk, key->offset);
900 static void dev_item_err(const struct extent_buffer *eb, int slot,
901 const char *fmt, ...)
903 struct btrfs_key key;
904 struct va_format vaf;
907 btrfs_item_key_to_cpu(eb, &key, slot);
913 btrfs_crit(eb->fs_info,
914 "corrupt %s: root=%llu block=%llu slot=%d devid=%llu %pV",
915 btrfs_header_level(eb) == 0 ? "leaf" : "node",
916 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
921 static int check_dev_item(struct extent_buffer *leaf,
922 struct btrfs_key *key, int slot)
924 struct btrfs_dev_item *ditem;
926 if (key->objectid != BTRFS_DEV_ITEMS_OBJECTID) {
927 dev_item_err(leaf, slot,
928 "invalid objectid: has=%llu expect=%llu",
929 key->objectid, BTRFS_DEV_ITEMS_OBJECTID);
932 ditem = btrfs_item_ptr(leaf, slot, struct btrfs_dev_item);
933 if (btrfs_device_id(leaf, ditem) != key->offset) {
934 dev_item_err(leaf, slot,
935 "devid mismatch: key has=%llu item has=%llu",
936 key->offset, btrfs_device_id(leaf, ditem));
941 * For device total_bytes, we don't have reliable way to check it, as
942 * it can be 0 for device removal. Device size check can only be done
943 * by dev extents check.
945 if (btrfs_device_bytes_used(leaf, ditem) >
946 btrfs_device_total_bytes(leaf, ditem)) {
947 dev_item_err(leaf, slot,
948 "invalid bytes used: have %llu expect [0, %llu]",
949 btrfs_device_bytes_used(leaf, ditem),
950 btrfs_device_total_bytes(leaf, ditem));
954 * Remaining members like io_align/type/gen/dev_group aren't really
955 * utilized. Skip them to make later usage of them easier.
960 static int check_inode_item(struct extent_buffer *leaf,
961 struct btrfs_key *key, int slot)
963 struct btrfs_fs_info *fs_info = leaf->fs_info;
964 struct btrfs_inode_item *iitem;
965 u64 super_gen = btrfs_super_generation(fs_info->super_copy);
966 u32 valid_mask = (S_IFMT | S_ISUID | S_ISGID | S_ISVTX | 0777);
970 ret = check_inode_key(leaf, key, slot);
974 iitem = btrfs_item_ptr(leaf, slot, struct btrfs_inode_item);
976 /* Here we use super block generation + 1 to handle log tree */
977 if (btrfs_inode_generation(leaf, iitem) > super_gen + 1) {
978 inode_item_err(leaf, slot,
979 "invalid inode generation: has %llu expect (0, %llu]",
980 btrfs_inode_generation(leaf, iitem),
984 /* Note for ROOT_TREE_DIR_ITEM, mkfs could set its transid 0 */
985 if (btrfs_inode_transid(leaf, iitem) > super_gen + 1) {
986 inode_item_err(leaf, slot,
987 "invalid inode transid: has %llu expect [0, %llu]",
988 btrfs_inode_transid(leaf, iitem), super_gen + 1);
993 * For size and nbytes it's better not to be too strict, as for dir
994 * item its size/nbytes can easily get wrong, but doesn't affect
995 * anything in the fs. So here we skip the check.
997 mode = btrfs_inode_mode(leaf, iitem);
998 if (mode & ~valid_mask) {
999 inode_item_err(leaf, slot,
1000 "unknown mode bit detected: 0x%x",
1001 mode & ~valid_mask);
1006 * S_IFMT is not bit mapped so we can't completely rely on
1007 * is_power_of_2/has_single_bit_set, but it can save us from checking
1008 * FIFO/CHR/DIR/REG. Only needs to check BLK, LNK and SOCKS
1010 if (!has_single_bit_set(mode & S_IFMT)) {
1011 if (!S_ISLNK(mode) && !S_ISBLK(mode) && !S_ISSOCK(mode)) {
1012 inode_item_err(leaf, slot,
1013 "invalid mode: has 0%o expect valid S_IF* bit(s)",
1018 if (S_ISDIR(mode) && btrfs_inode_nlink(leaf, iitem) > 1) {
1019 inode_item_err(leaf, slot,
1020 "invalid nlink: has %u expect no more than 1 for dir",
1021 btrfs_inode_nlink(leaf, iitem));
1024 if (btrfs_inode_flags(leaf, iitem) & ~BTRFS_INODE_FLAG_MASK) {
1025 inode_item_err(leaf, slot,
1026 "unknown flags detected: 0x%llx",
1027 btrfs_inode_flags(leaf, iitem) &
1028 ~BTRFS_INODE_FLAG_MASK);
1034 static int check_root_item(struct extent_buffer *leaf, struct btrfs_key *key,
1037 struct btrfs_fs_info *fs_info = leaf->fs_info;
1038 struct btrfs_root_item ri = { 0 };
1039 const u64 valid_root_flags = BTRFS_ROOT_SUBVOL_RDONLY |
1040 BTRFS_ROOT_SUBVOL_DEAD;
1043 ret = check_root_key(leaf, key, slot);
1047 if (btrfs_item_size_nr(leaf, slot) != sizeof(ri) &&
1048 btrfs_item_size_nr(leaf, slot) != btrfs_legacy_root_item_size()) {
1049 generic_err(leaf, slot,
1050 "invalid root item size, have %u expect %zu or %u",
1051 btrfs_item_size_nr(leaf, slot), sizeof(ri),
1052 btrfs_legacy_root_item_size());
1056 * For legacy root item, the members starting at generation_v2 will be
1057 * all filled with 0.
1058 * And since we allow geneartion_v2 as 0, it will still pass the check.
1060 read_extent_buffer(leaf, &ri, btrfs_item_ptr_offset(leaf, slot),
1061 btrfs_item_size_nr(leaf, slot));
1063 /* Generation related */
1064 if (btrfs_root_generation(&ri) >
1065 btrfs_super_generation(fs_info->super_copy) + 1) {
1066 generic_err(leaf, slot,
1067 "invalid root generation, have %llu expect (0, %llu]",
1068 btrfs_root_generation(&ri),
1069 btrfs_super_generation(fs_info->super_copy) + 1);
1072 if (btrfs_root_generation_v2(&ri) >
1073 btrfs_super_generation(fs_info->super_copy) + 1) {
1074 generic_err(leaf, slot,
1075 "invalid root v2 generation, have %llu expect (0, %llu]",
1076 btrfs_root_generation_v2(&ri),
1077 btrfs_super_generation(fs_info->super_copy) + 1);
1080 if (btrfs_root_last_snapshot(&ri) >
1081 btrfs_super_generation(fs_info->super_copy) + 1) {
1082 generic_err(leaf, slot,
1083 "invalid root last_snapshot, have %llu expect (0, %llu]",
1084 btrfs_root_last_snapshot(&ri),
1085 btrfs_super_generation(fs_info->super_copy) + 1);
1089 /* Alignment and level check */
1090 if (!IS_ALIGNED(btrfs_root_bytenr(&ri), fs_info->sectorsize)) {
1091 generic_err(leaf, slot,
1092 "invalid root bytenr, have %llu expect to be aligned to %u",
1093 btrfs_root_bytenr(&ri), fs_info->sectorsize);
1096 if (btrfs_root_level(&ri) >= BTRFS_MAX_LEVEL) {
1097 generic_err(leaf, slot,
1098 "invalid root level, have %u expect [0, %u]",
1099 btrfs_root_level(&ri), BTRFS_MAX_LEVEL - 1);
1102 if (ri.drop_level >= BTRFS_MAX_LEVEL) {
1103 generic_err(leaf, slot,
1104 "invalid root level, have %u expect [0, %u]",
1105 ri.drop_level, BTRFS_MAX_LEVEL - 1);
1110 if (btrfs_root_flags(&ri) & ~valid_root_flags) {
1111 generic_err(leaf, slot,
1112 "invalid root flags, have 0x%llx expect mask 0x%llx",
1113 btrfs_root_flags(&ri), valid_root_flags);
1121 static void extent_err(const struct extent_buffer *eb, int slot,
1122 const char *fmt, ...)
1124 struct btrfs_key key;
1125 struct va_format vaf;
1130 btrfs_item_key_to_cpu(eb, &key, slot);
1131 bytenr = key.objectid;
1132 if (key.type == BTRFS_METADATA_ITEM_KEY ||
1133 key.type == BTRFS_TREE_BLOCK_REF_KEY ||
1134 key.type == BTRFS_SHARED_BLOCK_REF_KEY)
1135 len = eb->fs_info->nodesize;
1138 va_start(args, fmt);
1143 btrfs_crit(eb->fs_info,
1144 "corrupt %s: block=%llu slot=%d extent bytenr=%llu len=%llu %pV",
1145 btrfs_header_level(eb) == 0 ? "leaf" : "node",
1146 eb->start, slot, bytenr, len, &vaf);
1150 static int check_extent_item(struct extent_buffer *leaf,
1151 struct btrfs_key *key, int slot)
1153 struct btrfs_fs_info *fs_info = leaf->fs_info;
1154 struct btrfs_extent_item *ei;
1155 bool is_tree_block = false;
1156 unsigned long ptr; /* Current pointer inside inline refs */
1157 unsigned long end; /* Extent item end */
1158 const u32 item_size = btrfs_item_size_nr(leaf, slot);
1161 u64 total_refs; /* Total refs in btrfs_extent_item */
1162 u64 inline_refs = 0; /* found total inline refs */
1164 if (key->type == BTRFS_METADATA_ITEM_KEY &&
1165 !btrfs_fs_incompat(fs_info, SKINNY_METADATA)) {
1166 generic_err(leaf, slot,
1167 "invalid key type, METADATA_ITEM type invalid when SKINNY_METADATA feature disabled");
1170 /* key->objectid is the bytenr for both key types */
1171 if (!IS_ALIGNED(key->objectid, fs_info->sectorsize)) {
1172 generic_err(leaf, slot,
1173 "invalid key objectid, have %llu expect to be aligned to %u",
1174 key->objectid, fs_info->sectorsize);
1178 /* key->offset is tree level for METADATA_ITEM_KEY */
1179 if (key->type == BTRFS_METADATA_ITEM_KEY &&
1180 key->offset >= BTRFS_MAX_LEVEL) {
1181 extent_err(leaf, slot,
1182 "invalid tree level, have %llu expect [0, %u]",
1183 key->offset, BTRFS_MAX_LEVEL - 1);
1188 * EXTENT/METADATA_ITEM consists of:
1189 * 1) One btrfs_extent_item
1190 * Records the total refs, type and generation of the extent.
1192 * 2) One btrfs_tree_block_info (for EXTENT_ITEM and tree backref only)
1193 * Records the first key and level of the tree block.
1195 * 2) Zero or more btrfs_extent_inline_ref(s)
1196 * Each inline ref has one btrfs_extent_inline_ref shows:
1197 * 2.1) The ref type, one of the 4
1198 * TREE_BLOCK_REF Tree block only
1199 * SHARED_BLOCK_REF Tree block only
1200 * EXTENT_DATA_REF Data only
1201 * SHARED_DATA_REF Data only
1202 * 2.2) Ref type specific data
1203 * Either using btrfs_extent_inline_ref::offset, or specific
1206 if (item_size < sizeof(*ei)) {
1207 extent_err(leaf, slot,
1208 "invalid item size, have %u expect [%zu, %u)",
1209 item_size, sizeof(*ei),
1210 BTRFS_LEAF_DATA_SIZE(fs_info));
1213 end = item_size + btrfs_item_ptr_offset(leaf, slot);
1215 /* Checks against extent_item */
1216 ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
1217 flags = btrfs_extent_flags(leaf, ei);
1218 total_refs = btrfs_extent_refs(leaf, ei);
1219 generation = btrfs_extent_generation(leaf, ei);
1220 if (generation > btrfs_super_generation(fs_info->super_copy) + 1) {
1221 extent_err(leaf, slot,
1222 "invalid generation, have %llu expect (0, %llu]",
1224 btrfs_super_generation(fs_info->super_copy) + 1);
1227 if (!has_single_bit_set(flags & (BTRFS_EXTENT_FLAG_DATA |
1228 BTRFS_EXTENT_FLAG_TREE_BLOCK))) {
1229 extent_err(leaf, slot,
1230 "invalid extent flag, have 0x%llx expect 1 bit set in 0x%llx",
1231 flags, BTRFS_EXTENT_FLAG_DATA |
1232 BTRFS_EXTENT_FLAG_TREE_BLOCK);
1235 is_tree_block = !!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK);
1236 if (is_tree_block) {
1237 if (key->type == BTRFS_EXTENT_ITEM_KEY &&
1238 key->offset != fs_info->nodesize) {
1239 extent_err(leaf, slot,
1240 "invalid extent length, have %llu expect %u",
1241 key->offset, fs_info->nodesize);
1245 if (key->type != BTRFS_EXTENT_ITEM_KEY) {
1246 extent_err(leaf, slot,
1247 "invalid key type, have %u expect %u for data backref",
1248 key->type, BTRFS_EXTENT_ITEM_KEY);
1251 if (!IS_ALIGNED(key->offset, fs_info->sectorsize)) {
1252 extent_err(leaf, slot,
1253 "invalid extent length, have %llu expect aligned to %u",
1254 key->offset, fs_info->sectorsize);
1258 ptr = (unsigned long)(struct btrfs_extent_item *)(ei + 1);
1260 /* Check the special case of btrfs_tree_block_info */
1261 if (is_tree_block && key->type != BTRFS_METADATA_ITEM_KEY) {
1262 struct btrfs_tree_block_info *info;
1264 info = (struct btrfs_tree_block_info *)ptr;
1265 if (btrfs_tree_block_level(leaf, info) >= BTRFS_MAX_LEVEL) {
1266 extent_err(leaf, slot,
1267 "invalid tree block info level, have %u expect [0, %u]",
1268 btrfs_tree_block_level(leaf, info),
1269 BTRFS_MAX_LEVEL - 1);
1272 ptr = (unsigned long)(struct btrfs_tree_block_info *)(info + 1);
1275 /* Check inline refs */
1277 struct btrfs_extent_inline_ref *iref;
1278 struct btrfs_extent_data_ref *dref;
1279 struct btrfs_shared_data_ref *sref;
1284 if (ptr + sizeof(*iref) > end) {
1285 extent_err(leaf, slot,
1286 "inline ref item overflows extent item, ptr %lu iref size %zu end %lu",
1287 ptr, sizeof(*iref), end);
1290 iref = (struct btrfs_extent_inline_ref *)ptr;
1291 inline_type = btrfs_extent_inline_ref_type(leaf, iref);
1292 inline_offset = btrfs_extent_inline_ref_offset(leaf, iref);
1293 if (ptr + btrfs_extent_inline_ref_size(inline_type) > end) {
1294 extent_err(leaf, slot,
1295 "inline ref item overflows extent item, ptr %lu iref size %u end %lu",
1296 ptr, inline_type, end);
1300 switch (inline_type) {
1301 /* inline_offset is subvolid of the owner, no need to check */
1302 case BTRFS_TREE_BLOCK_REF_KEY:
1305 /* Contains parent bytenr */
1306 case BTRFS_SHARED_BLOCK_REF_KEY:
1307 if (!IS_ALIGNED(inline_offset, fs_info->sectorsize)) {
1308 extent_err(leaf, slot,
1309 "invalid tree parent bytenr, have %llu expect aligned to %u",
1310 inline_offset, fs_info->sectorsize);
1316 * Contains owner subvolid, owner key objectid, adjusted offset.
1317 * The only obvious corruption can happen in that offset.
1319 case BTRFS_EXTENT_DATA_REF_KEY:
1320 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1321 dref_offset = btrfs_extent_data_ref_offset(leaf, dref);
1322 if (!IS_ALIGNED(dref_offset, fs_info->sectorsize)) {
1323 extent_err(leaf, slot,
1324 "invalid data ref offset, have %llu expect aligned to %u",
1325 dref_offset, fs_info->sectorsize);
1328 inline_refs += btrfs_extent_data_ref_count(leaf, dref);
1330 /* Contains parent bytenr and ref count */
1331 case BTRFS_SHARED_DATA_REF_KEY:
1332 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1333 if (!IS_ALIGNED(inline_offset, fs_info->sectorsize)) {
1334 extent_err(leaf, slot,
1335 "invalid data parent bytenr, have %llu expect aligned to %u",
1336 inline_offset, fs_info->sectorsize);
1339 inline_refs += btrfs_shared_data_ref_count(leaf, sref);
1342 extent_err(leaf, slot, "unknown inline ref type: %u",
1346 ptr += btrfs_extent_inline_ref_size(inline_type);
1348 /* No padding is allowed */
1350 extent_err(leaf, slot,
1351 "invalid extent item size, padding bytes found");
1355 /* Finally, check the inline refs against total refs */
1356 if (inline_refs > total_refs) {
1357 extent_err(leaf, slot,
1358 "invalid extent refs, have %llu expect >= inline %llu",
1359 total_refs, inline_refs);
1365 static int check_simple_keyed_refs(struct extent_buffer *leaf,
1366 struct btrfs_key *key, int slot)
1368 u32 expect_item_size = 0;
1370 if (key->type == BTRFS_SHARED_DATA_REF_KEY)
1371 expect_item_size = sizeof(struct btrfs_shared_data_ref);
1373 if (btrfs_item_size_nr(leaf, slot) != expect_item_size) {
1374 generic_err(leaf, slot,
1375 "invalid item size, have %u expect %u for key type %u",
1376 btrfs_item_size_nr(leaf, slot),
1377 expect_item_size, key->type);
1380 if (!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize)) {
1381 generic_err(leaf, slot,
1382 "invalid key objectid for shared block ref, have %llu expect aligned to %u",
1383 key->objectid, leaf->fs_info->sectorsize);
1386 if (key->type != BTRFS_TREE_BLOCK_REF_KEY &&
1387 !IS_ALIGNED(key->offset, leaf->fs_info->sectorsize)) {
1388 extent_err(leaf, slot,
1389 "invalid tree parent bytenr, have %llu expect aligned to %u",
1390 key->offset, leaf->fs_info->sectorsize);
1396 static int check_extent_data_ref(struct extent_buffer *leaf,
1397 struct btrfs_key *key, int slot)
1399 struct btrfs_extent_data_ref *dref;
1400 unsigned long ptr = btrfs_item_ptr_offset(leaf, slot);
1401 const unsigned long end = ptr + btrfs_item_size_nr(leaf, slot);
1403 if (btrfs_item_size_nr(leaf, slot) % sizeof(*dref) != 0) {
1404 generic_err(leaf, slot,
1405 "invalid item size, have %u expect aligned to %zu for key type %u",
1406 btrfs_item_size_nr(leaf, slot),
1407 sizeof(*dref), key->type);
1409 if (!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize)) {
1410 generic_err(leaf, slot,
1411 "invalid key objectid for shared block ref, have %llu expect aligned to %u",
1412 key->objectid, leaf->fs_info->sectorsize);
1415 for (; ptr < end; ptr += sizeof(*dref)) {
1421 dref = (struct btrfs_extent_data_ref *)ptr;
1422 root_objectid = btrfs_extent_data_ref_root(leaf, dref);
1423 owner = btrfs_extent_data_ref_objectid(leaf, dref);
1424 offset = btrfs_extent_data_ref_offset(leaf, dref);
1425 hash = hash_extent_data_ref(root_objectid, owner, offset);
1426 if (hash != key->offset) {
1427 extent_err(leaf, slot,
1428 "invalid extent data ref hash, item has 0x%016llx key has 0x%016llx",
1432 if (!IS_ALIGNED(offset, leaf->fs_info->sectorsize)) {
1433 extent_err(leaf, slot,
1434 "invalid extent data backref offset, have %llu expect aligned to %u",
1435 offset, leaf->fs_info->sectorsize);
1441 #define inode_ref_err(eb, slot, fmt, args...) \
1442 inode_item_err(eb, slot, fmt, ##args)
1443 static int check_inode_ref(struct extent_buffer *leaf,
1444 struct btrfs_key *key, struct btrfs_key *prev_key,
1447 struct btrfs_inode_ref *iref;
1451 if (!check_prev_ino(leaf, key, slot, prev_key))
1453 /* namelen can't be 0, so item_size == sizeof() is also invalid */
1454 if (btrfs_item_size_nr(leaf, slot) <= sizeof(*iref)) {
1455 inode_ref_err(leaf, slot,
1456 "invalid item size, have %u expect (%zu, %u)",
1457 btrfs_item_size_nr(leaf, slot),
1458 sizeof(*iref), BTRFS_LEAF_DATA_SIZE(leaf->fs_info));
1462 ptr = btrfs_item_ptr_offset(leaf, slot);
1463 end = ptr + btrfs_item_size_nr(leaf, slot);
1467 if (ptr + sizeof(iref) > end) {
1468 inode_ref_err(leaf, slot,
1469 "inode ref overflow, ptr %lu end %lu inode_ref_size %zu",
1470 ptr, end, sizeof(iref));
1474 iref = (struct btrfs_inode_ref *)ptr;
1475 namelen = btrfs_inode_ref_name_len(leaf, iref);
1476 if (ptr + sizeof(*iref) + namelen > end) {
1477 inode_ref_err(leaf, slot,
1478 "inode ref overflow, ptr %lu end %lu namelen %u",
1484 * NOTE: In theory we should record all found index numbers
1485 * to find any duplicated indexes, but that will be too time
1486 * consuming for inodes with too many hard links.
1488 ptr += sizeof(*iref) + namelen;
1494 * Common point to switch the item-specific validation.
1496 static int check_leaf_item(struct extent_buffer *leaf,
1497 struct btrfs_key *key, int slot,
1498 struct btrfs_key *prev_key)
1501 struct btrfs_chunk *chunk;
1503 switch (key->type) {
1504 case BTRFS_EXTENT_DATA_KEY:
1505 ret = check_extent_data_item(leaf, key, slot, prev_key);
1507 case BTRFS_EXTENT_CSUM_KEY:
1508 ret = check_csum_item(leaf, key, slot, prev_key);
1510 case BTRFS_DIR_ITEM_KEY:
1511 case BTRFS_DIR_INDEX_KEY:
1512 case BTRFS_XATTR_ITEM_KEY:
1513 ret = check_dir_item(leaf, key, prev_key, slot);
1515 case BTRFS_INODE_REF_KEY:
1516 ret = check_inode_ref(leaf, key, prev_key, slot);
1518 case BTRFS_BLOCK_GROUP_ITEM_KEY:
1519 ret = check_block_group_item(leaf, key, slot);
1521 case BTRFS_CHUNK_ITEM_KEY:
1522 chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
1523 ret = check_leaf_chunk_item(leaf, chunk, key, slot);
1525 case BTRFS_DEV_ITEM_KEY:
1526 ret = check_dev_item(leaf, key, slot);
1528 case BTRFS_INODE_ITEM_KEY:
1529 ret = check_inode_item(leaf, key, slot);
1531 case BTRFS_ROOT_ITEM_KEY:
1532 ret = check_root_item(leaf, key, slot);
1534 case BTRFS_EXTENT_ITEM_KEY:
1535 case BTRFS_METADATA_ITEM_KEY:
1536 ret = check_extent_item(leaf, key, slot);
1538 case BTRFS_TREE_BLOCK_REF_KEY:
1539 case BTRFS_SHARED_DATA_REF_KEY:
1540 case BTRFS_SHARED_BLOCK_REF_KEY:
1541 ret = check_simple_keyed_refs(leaf, key, slot);
1543 case BTRFS_EXTENT_DATA_REF_KEY:
1544 ret = check_extent_data_ref(leaf, key, slot);
1550 static int check_leaf(struct extent_buffer *leaf, bool check_item_data)
1552 struct btrfs_fs_info *fs_info = leaf->fs_info;
1553 /* No valid key type is 0, so all key should be larger than this key */
1554 struct btrfs_key prev_key = {0, 0, 0};
1555 struct btrfs_key key;
1556 u32 nritems = btrfs_header_nritems(leaf);
1559 if (btrfs_header_level(leaf) != 0) {
1560 generic_err(leaf, 0,
1561 "invalid level for leaf, have %d expect 0",
1562 btrfs_header_level(leaf));
1567 * Extent buffers from a relocation tree have a owner field that
1568 * corresponds to the subvolume tree they are based on. So just from an
1569 * extent buffer alone we can not find out what is the id of the
1570 * corresponding subvolume tree, so we can not figure out if the extent
1571 * buffer corresponds to the root of the relocation tree or not. So
1572 * skip this check for relocation trees.
1574 if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) {
1575 u64 owner = btrfs_header_owner(leaf);
1577 /* These trees must never be empty */
1578 if (owner == BTRFS_ROOT_TREE_OBJECTID ||
1579 owner == BTRFS_CHUNK_TREE_OBJECTID ||
1580 owner == BTRFS_EXTENT_TREE_OBJECTID ||
1581 owner == BTRFS_DEV_TREE_OBJECTID ||
1582 owner == BTRFS_FS_TREE_OBJECTID ||
1583 owner == BTRFS_DATA_RELOC_TREE_OBJECTID) {
1584 generic_err(leaf, 0,
1585 "invalid root, root %llu must never be empty",
1591 generic_err(leaf, 0,
1592 "invalid owner, root 0 is not defined");
1602 * Check the following things to make sure this is a good leaf, and
1603 * leaf users won't need to bother with similar sanity checks:
1606 * 2) item offset and size
1607 * No overlap, no hole, all inside the leaf.
1609 * If possible, do comprehensive sanity check.
1610 * NOTE: All checks must only rely on the item data itself.
1612 for (slot = 0; slot < nritems; slot++) {
1613 u32 item_end_expected;
1616 btrfs_item_key_to_cpu(leaf, &key, slot);
1618 /* Make sure the keys are in the right order */
1619 if (btrfs_comp_cpu_keys(&prev_key, &key) >= 0) {
1620 generic_err(leaf, slot,
1621 "bad key order, prev (%llu %u %llu) current (%llu %u %llu)",
1622 prev_key.objectid, prev_key.type,
1623 prev_key.offset, key.objectid, key.type,
1629 * Make sure the offset and ends are right, remember that the
1630 * item data starts at the end of the leaf and grows towards the
1634 item_end_expected = BTRFS_LEAF_DATA_SIZE(fs_info);
1636 item_end_expected = btrfs_item_offset_nr(leaf,
1638 if (btrfs_item_end_nr(leaf, slot) != item_end_expected) {
1639 generic_err(leaf, slot,
1640 "unexpected item end, have %u expect %u",
1641 btrfs_item_end_nr(leaf, slot),
1647 * Check to make sure that we don't point outside of the leaf,
1648 * just in case all the items are consistent to each other, but
1649 * all point outside of the leaf.
1651 if (btrfs_item_end_nr(leaf, slot) >
1652 BTRFS_LEAF_DATA_SIZE(fs_info)) {
1653 generic_err(leaf, slot,
1654 "slot end outside of leaf, have %u expect range [0, %u]",
1655 btrfs_item_end_nr(leaf, slot),
1656 BTRFS_LEAF_DATA_SIZE(fs_info));
1660 /* Also check if the item pointer overlaps with btrfs item. */
1661 if (btrfs_item_nr_offset(slot) + sizeof(struct btrfs_item) >
1662 btrfs_item_ptr_offset(leaf, slot)) {
1663 generic_err(leaf, slot,
1664 "slot overlaps with its data, item end %lu data start %lu",
1665 btrfs_item_nr_offset(slot) +
1666 sizeof(struct btrfs_item),
1667 btrfs_item_ptr_offset(leaf, slot));
1671 if (check_item_data) {
1673 * Check if the item size and content meet other
1676 ret = check_leaf_item(leaf, &key, slot, &prev_key);
1681 prev_key.objectid = key.objectid;
1682 prev_key.type = key.type;
1683 prev_key.offset = key.offset;
1689 int btrfs_check_leaf_full(struct extent_buffer *leaf)
1691 return check_leaf(leaf, true);
1693 ALLOW_ERROR_INJECTION(btrfs_check_leaf_full, ERRNO);
1695 int btrfs_check_leaf_relaxed(struct extent_buffer *leaf)
1697 return check_leaf(leaf, false);
1700 int btrfs_check_node(struct extent_buffer *node)
1702 struct btrfs_fs_info *fs_info = node->fs_info;
1703 unsigned long nr = btrfs_header_nritems(node);
1704 struct btrfs_key key, next_key;
1706 int level = btrfs_header_level(node);
1710 if (level <= 0 || level >= BTRFS_MAX_LEVEL) {
1711 generic_err(node, 0,
1712 "invalid level for node, have %d expect [1, %d]",
1713 level, BTRFS_MAX_LEVEL - 1);
1716 if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(fs_info)) {
1718 "corrupt node: root=%llu block=%llu, nritems too %s, have %lu expect range [1,%u]",
1719 btrfs_header_owner(node), node->start,
1720 nr == 0 ? "small" : "large", nr,
1721 BTRFS_NODEPTRS_PER_BLOCK(fs_info));
1725 for (slot = 0; slot < nr - 1; slot++) {
1726 bytenr = btrfs_node_blockptr(node, slot);
1727 btrfs_node_key_to_cpu(node, &key, slot);
1728 btrfs_node_key_to_cpu(node, &next_key, slot + 1);
1731 generic_err(node, slot,
1732 "invalid NULL node pointer");
1736 if (!IS_ALIGNED(bytenr, fs_info->sectorsize)) {
1737 generic_err(node, slot,
1738 "unaligned pointer, have %llu should be aligned to %u",
1739 bytenr, fs_info->sectorsize);
1744 if (btrfs_comp_cpu_keys(&key, &next_key) >= 0) {
1745 generic_err(node, slot,
1746 "bad key order, current (%llu %u %llu) next (%llu %u %llu)",
1747 key.objectid, key.type, key.offset,
1748 next_key.objectid, next_key.type,
1757 ALLOW_ERROR_INJECTION(btrfs_check_node, ERRNO);