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 static int check_dir_item(struct extent_buffer *leaf,
377 struct btrfs_key *key, struct btrfs_key *prev_key,
380 struct btrfs_fs_info *fs_info = leaf->fs_info;
381 struct btrfs_dir_item *di;
382 u32 item_size = btrfs_item_size_nr(leaf, slot);
385 if (!check_prev_ino(leaf, key, slot, prev_key))
387 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
388 while (cur < item_size) {
396 /* header itself should not cross item boundary */
397 if (cur + sizeof(*di) > item_size) {
398 dir_item_err(leaf, slot,
399 "dir item header crosses item boundary, have %zu boundary %u",
400 cur + sizeof(*di), item_size);
405 dir_type = btrfs_dir_type(leaf, di);
406 if (dir_type >= BTRFS_FT_MAX) {
407 dir_item_err(leaf, slot,
408 "invalid dir item type, have %u expect [0, %u)",
409 dir_type, BTRFS_FT_MAX);
413 if (key->type == BTRFS_XATTR_ITEM_KEY &&
414 dir_type != BTRFS_FT_XATTR) {
415 dir_item_err(leaf, slot,
416 "invalid dir item type for XATTR key, have %u expect %u",
417 dir_type, BTRFS_FT_XATTR);
420 if (dir_type == BTRFS_FT_XATTR &&
421 key->type != BTRFS_XATTR_ITEM_KEY) {
422 dir_item_err(leaf, slot,
423 "xattr dir type found for non-XATTR key");
426 if (dir_type == BTRFS_FT_XATTR)
427 max_name_len = XATTR_NAME_MAX;
429 max_name_len = BTRFS_NAME_LEN;
431 /* Name/data length check */
432 name_len = btrfs_dir_name_len(leaf, di);
433 data_len = btrfs_dir_data_len(leaf, di);
434 if (name_len > max_name_len) {
435 dir_item_err(leaf, slot,
436 "dir item name len too long, have %u max %u",
437 name_len, max_name_len);
440 if (name_len + data_len > BTRFS_MAX_XATTR_SIZE(fs_info)) {
441 dir_item_err(leaf, slot,
442 "dir item name and data len too long, have %u max %u",
444 BTRFS_MAX_XATTR_SIZE(fs_info));
448 if (data_len && dir_type != BTRFS_FT_XATTR) {
449 dir_item_err(leaf, slot,
450 "dir item with invalid data len, have %u expect 0",
455 total_size = sizeof(*di) + name_len + data_len;
457 /* header and name/data should not cross item boundary */
458 if (cur + total_size > item_size) {
459 dir_item_err(leaf, slot,
460 "dir item data crosses item boundary, have %u boundary %u",
461 cur + total_size, item_size);
466 * Special check for XATTR/DIR_ITEM, as key->offset is name
467 * hash, should match its name
469 if (key->type == BTRFS_DIR_ITEM_KEY ||
470 key->type == BTRFS_XATTR_ITEM_KEY) {
471 char namebuf[max(BTRFS_NAME_LEN, XATTR_NAME_MAX)];
473 read_extent_buffer(leaf, namebuf,
474 (unsigned long)(di + 1), name_len);
475 name_hash = btrfs_name_hash(namebuf, name_len);
476 if (key->offset != name_hash) {
477 dir_item_err(leaf, slot,
478 "name hash mismatch with key, have 0x%016x expect 0x%016llx",
479 name_hash, key->offset);
484 di = (struct btrfs_dir_item *)((void *)di + total_size);
491 static void block_group_err(const struct extent_buffer *eb, int slot,
492 const char *fmt, ...)
494 const struct btrfs_fs_info *fs_info = eb->fs_info;
495 struct btrfs_key key;
496 struct va_format vaf;
499 btrfs_item_key_to_cpu(eb, &key, slot);
506 "corrupt %s: root=%llu block=%llu slot=%d bg_start=%llu bg_len=%llu, %pV",
507 btrfs_header_level(eb) == 0 ? "leaf" : "node",
508 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
509 key.objectid, key.offset, &vaf);
513 static int check_block_group_item(struct extent_buffer *leaf,
514 struct btrfs_key *key, int slot)
516 struct btrfs_block_group_item bgi;
517 u32 item_size = btrfs_item_size_nr(leaf, slot);
522 * Here we don't really care about alignment since extent allocator can
523 * handle it. We care more about the size.
525 if (key->offset == 0) {
526 block_group_err(leaf, slot,
527 "invalid block group size 0");
531 if (item_size != sizeof(bgi)) {
532 block_group_err(leaf, slot,
533 "invalid item size, have %u expect %zu",
534 item_size, sizeof(bgi));
538 read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot),
540 if (btrfs_stack_block_group_chunk_objectid(&bgi) !=
541 BTRFS_FIRST_CHUNK_TREE_OBJECTID) {
542 block_group_err(leaf, slot,
543 "invalid block group chunk objectid, have %llu expect %llu",
544 btrfs_stack_block_group_chunk_objectid(&bgi),
545 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
549 if (btrfs_stack_block_group_used(&bgi) > key->offset) {
550 block_group_err(leaf, slot,
551 "invalid block group used, have %llu expect [0, %llu)",
552 btrfs_stack_block_group_used(&bgi), key->offset);
556 flags = btrfs_stack_block_group_flags(&bgi);
557 if (hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) > 1) {
558 block_group_err(leaf, slot,
559 "invalid profile flags, have 0x%llx (%lu bits set) expect no more than 1 bit set",
560 flags & BTRFS_BLOCK_GROUP_PROFILE_MASK,
561 hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK));
565 type = flags & BTRFS_BLOCK_GROUP_TYPE_MASK;
566 if (type != BTRFS_BLOCK_GROUP_DATA &&
567 type != BTRFS_BLOCK_GROUP_METADATA &&
568 type != BTRFS_BLOCK_GROUP_SYSTEM &&
569 type != (BTRFS_BLOCK_GROUP_METADATA |
570 BTRFS_BLOCK_GROUP_DATA)) {
571 block_group_err(leaf, slot,
572 "invalid type, have 0x%llx (%lu bits set) expect either 0x%llx, 0x%llx, 0x%llx or 0x%llx",
573 type, hweight64(type),
574 BTRFS_BLOCK_GROUP_DATA, BTRFS_BLOCK_GROUP_METADATA,
575 BTRFS_BLOCK_GROUP_SYSTEM,
576 BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA);
584 static void chunk_err(const struct extent_buffer *leaf,
585 const struct btrfs_chunk *chunk, u64 logical,
586 const char *fmt, ...)
588 const struct btrfs_fs_info *fs_info = leaf->fs_info;
590 struct va_format vaf;
595 /* Only superblock eb is able to have such small offset */
596 is_sb = (leaf->start == BTRFS_SUPER_INFO_OFFSET);
600 * Get the slot number by iterating through all slots, this
601 * would provide better readability.
603 for (i = 0; i < btrfs_header_nritems(leaf); i++) {
604 if (btrfs_item_ptr_offset(leaf, i) ==
605 (unsigned long)chunk) {
617 "corrupt superblock syschunk array: chunk_start=%llu, %pV",
621 "corrupt leaf: root=%llu block=%llu slot=%d chunk_start=%llu, %pV",
622 BTRFS_CHUNK_TREE_OBJECTID, leaf->start, slot,
628 * The common chunk check which could also work on super block sys chunk array.
630 * Return -EUCLEAN if anything is corrupted.
631 * Return 0 if everything is OK.
633 int btrfs_check_chunk_valid(struct extent_buffer *leaf,
634 struct btrfs_chunk *chunk, u64 logical)
636 struct btrfs_fs_info *fs_info = leaf->fs_info;
645 length = btrfs_chunk_length(leaf, chunk);
646 stripe_len = btrfs_chunk_stripe_len(leaf, chunk);
647 num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
648 sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk);
649 type = btrfs_chunk_type(leaf, chunk);
652 chunk_err(leaf, chunk, logical,
653 "invalid chunk num_stripes, have %u", num_stripes);
656 if (!IS_ALIGNED(logical, fs_info->sectorsize)) {
657 chunk_err(leaf, chunk, logical,
658 "invalid chunk logical, have %llu should aligned to %u",
659 logical, fs_info->sectorsize);
662 if (btrfs_chunk_sector_size(leaf, chunk) != fs_info->sectorsize) {
663 chunk_err(leaf, chunk, logical,
664 "invalid chunk sectorsize, have %u expect %u",
665 btrfs_chunk_sector_size(leaf, chunk),
666 fs_info->sectorsize);
669 if (!length || !IS_ALIGNED(length, fs_info->sectorsize)) {
670 chunk_err(leaf, chunk, logical,
671 "invalid chunk length, have %llu", length);
674 if (!is_power_of_2(stripe_len) || stripe_len != BTRFS_STRIPE_LEN) {
675 chunk_err(leaf, chunk, logical,
676 "invalid chunk stripe length: %llu",
680 if (~(BTRFS_BLOCK_GROUP_TYPE_MASK | BTRFS_BLOCK_GROUP_PROFILE_MASK) &
682 chunk_err(leaf, chunk, logical,
683 "unrecognized chunk type: 0x%llx",
684 ~(BTRFS_BLOCK_GROUP_TYPE_MASK |
685 BTRFS_BLOCK_GROUP_PROFILE_MASK) &
686 btrfs_chunk_type(leaf, chunk));
690 if (!has_single_bit_set(type & BTRFS_BLOCK_GROUP_PROFILE_MASK) &&
691 (type & BTRFS_BLOCK_GROUP_PROFILE_MASK) != 0) {
692 chunk_err(leaf, chunk, logical,
693 "invalid chunk profile flag: 0x%llx, expect 0 or 1 bit set",
694 type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
697 if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == 0) {
698 chunk_err(leaf, chunk, logical,
699 "missing chunk type flag, have 0x%llx one bit must be set in 0x%llx",
700 type, BTRFS_BLOCK_GROUP_TYPE_MASK);
704 if ((type & BTRFS_BLOCK_GROUP_SYSTEM) &&
705 (type & (BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA))) {
706 chunk_err(leaf, chunk, logical,
707 "system chunk with data or metadata type: 0x%llx",
712 features = btrfs_super_incompat_flags(fs_info->super_copy);
713 if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS)
717 if ((type & BTRFS_BLOCK_GROUP_METADATA) &&
718 (type & BTRFS_BLOCK_GROUP_DATA)) {
719 chunk_err(leaf, chunk, logical,
720 "mixed chunk type in non-mixed mode: 0x%llx", type);
725 if ((type & BTRFS_BLOCK_GROUP_RAID10 && sub_stripes != 2) ||
726 (type & BTRFS_BLOCK_GROUP_RAID1 && num_stripes != 2) ||
727 (type & BTRFS_BLOCK_GROUP_RAID5 && num_stripes < 2) ||
728 (type & BTRFS_BLOCK_GROUP_RAID6 && num_stripes < 3) ||
729 (type & BTRFS_BLOCK_GROUP_DUP && num_stripes != 2) ||
730 ((type & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 && num_stripes != 1)) {
731 chunk_err(leaf, chunk, logical,
732 "invalid num_stripes:sub_stripes %u:%u for profile %llu",
733 num_stripes, sub_stripes,
734 type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
742 * Enhanced version of chunk item checker.
744 * The common btrfs_check_chunk_valid() doesn't check item size since it needs
745 * to work on super block sys_chunk_array which doesn't have full item ptr.
747 static int check_leaf_chunk_item(struct extent_buffer *leaf,
748 struct btrfs_chunk *chunk,
749 struct btrfs_key *key, int slot)
753 if (btrfs_item_size_nr(leaf, slot) < sizeof(struct btrfs_chunk)) {
754 chunk_err(leaf, chunk, key->offset,
755 "invalid chunk item size: have %u expect [%zu, %u)",
756 btrfs_item_size_nr(leaf, slot),
757 sizeof(struct btrfs_chunk),
758 BTRFS_LEAF_DATA_SIZE(leaf->fs_info));
762 num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
763 /* Let btrfs_check_chunk_valid() handle this error type */
764 if (num_stripes == 0)
767 if (btrfs_chunk_item_size(num_stripes) !=
768 btrfs_item_size_nr(leaf, slot)) {
769 chunk_err(leaf, chunk, key->offset,
770 "invalid chunk item size: have %u expect %lu",
771 btrfs_item_size_nr(leaf, slot),
772 btrfs_chunk_item_size(num_stripes));
776 return btrfs_check_chunk_valid(leaf, chunk, key->offset);
781 static void dev_item_err(const struct extent_buffer *eb, int slot,
782 const char *fmt, ...)
784 struct btrfs_key key;
785 struct va_format vaf;
788 btrfs_item_key_to_cpu(eb, &key, slot);
794 btrfs_crit(eb->fs_info,
795 "corrupt %s: root=%llu block=%llu slot=%d devid=%llu %pV",
796 btrfs_header_level(eb) == 0 ? "leaf" : "node",
797 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
802 static int check_dev_item(struct extent_buffer *leaf,
803 struct btrfs_key *key, int slot)
805 struct btrfs_dev_item *ditem;
807 if (key->objectid != BTRFS_DEV_ITEMS_OBJECTID) {
808 dev_item_err(leaf, slot,
809 "invalid objectid: has=%llu expect=%llu",
810 key->objectid, BTRFS_DEV_ITEMS_OBJECTID);
813 ditem = btrfs_item_ptr(leaf, slot, struct btrfs_dev_item);
814 if (btrfs_device_id(leaf, ditem) != key->offset) {
815 dev_item_err(leaf, slot,
816 "devid mismatch: key has=%llu item has=%llu",
817 key->offset, btrfs_device_id(leaf, ditem));
822 * For device total_bytes, we don't have reliable way to check it, as
823 * it can be 0 for device removal. Device size check can only be done
824 * by dev extents check.
826 if (btrfs_device_bytes_used(leaf, ditem) >
827 btrfs_device_total_bytes(leaf, ditem)) {
828 dev_item_err(leaf, slot,
829 "invalid bytes used: have %llu expect [0, %llu]",
830 btrfs_device_bytes_used(leaf, ditem),
831 btrfs_device_total_bytes(leaf, ditem));
835 * Remaining members like io_align/type/gen/dev_group aren't really
836 * utilized. Skip them to make later usage of them easier.
841 /* Inode item error output has the same format as dir_item_err() */
842 #define inode_item_err(fs_info, eb, slot, fmt, ...) \
843 dir_item_err(eb, slot, fmt, __VA_ARGS__)
845 static int check_inode_item(struct extent_buffer *leaf,
846 struct btrfs_key *key, int slot)
848 struct btrfs_fs_info *fs_info = leaf->fs_info;
849 struct btrfs_inode_item *iitem;
850 u64 super_gen = btrfs_super_generation(fs_info->super_copy);
851 u32 valid_mask = (S_IFMT | S_ISUID | S_ISGID | S_ISVTX | 0777);
854 if ((key->objectid < BTRFS_FIRST_FREE_OBJECTID ||
855 key->objectid > BTRFS_LAST_FREE_OBJECTID) &&
856 key->objectid != BTRFS_ROOT_TREE_DIR_OBJECTID &&
857 key->objectid != BTRFS_FREE_INO_OBJECTID) {
858 generic_err(leaf, slot,
859 "invalid key objectid: has %llu expect %llu or [%llu, %llu] or %llu",
860 key->objectid, BTRFS_ROOT_TREE_DIR_OBJECTID,
861 BTRFS_FIRST_FREE_OBJECTID,
862 BTRFS_LAST_FREE_OBJECTID,
863 BTRFS_FREE_INO_OBJECTID);
866 if (key->offset != 0) {
867 inode_item_err(fs_info, leaf, slot,
868 "invalid key offset: has %llu expect 0",
872 iitem = btrfs_item_ptr(leaf, slot, struct btrfs_inode_item);
874 /* Here we use super block generation + 1 to handle log tree */
875 if (btrfs_inode_generation(leaf, iitem) > super_gen + 1) {
876 inode_item_err(fs_info, leaf, slot,
877 "invalid inode generation: has %llu expect (0, %llu]",
878 btrfs_inode_generation(leaf, iitem),
882 /* Note for ROOT_TREE_DIR_ITEM, mkfs could set its transid 0 */
883 if (btrfs_inode_transid(leaf, iitem) > super_gen + 1) {
884 inode_item_err(fs_info, leaf, slot,
885 "invalid inode generation: has %llu expect [0, %llu]",
886 btrfs_inode_transid(leaf, iitem), super_gen + 1);
891 * For size and nbytes it's better not to be too strict, as for dir
892 * item its size/nbytes can easily get wrong, but doesn't affect
893 * anything in the fs. So here we skip the check.
895 mode = btrfs_inode_mode(leaf, iitem);
896 if (mode & ~valid_mask) {
897 inode_item_err(fs_info, leaf, slot,
898 "unknown mode bit detected: 0x%x",
904 * S_IFMT is not bit mapped so we can't completely rely on
905 * is_power_of_2/has_single_bit_set, but it can save us from checking
906 * FIFO/CHR/DIR/REG. Only needs to check BLK, LNK and SOCKS
908 if (!has_single_bit_set(mode & S_IFMT)) {
909 if (!S_ISLNK(mode) && !S_ISBLK(mode) && !S_ISSOCK(mode)) {
910 inode_item_err(fs_info, leaf, slot,
911 "invalid mode: has 0%o expect valid S_IF* bit(s)",
916 if (S_ISDIR(mode) && btrfs_inode_nlink(leaf, iitem) > 1) {
917 inode_item_err(fs_info, leaf, slot,
918 "invalid nlink: has %u expect no more than 1 for dir",
919 btrfs_inode_nlink(leaf, iitem));
922 if (btrfs_inode_flags(leaf, iitem) & ~BTRFS_INODE_FLAG_MASK) {
923 inode_item_err(fs_info, leaf, slot,
924 "unknown flags detected: 0x%llx",
925 btrfs_inode_flags(leaf, iitem) &
926 ~BTRFS_INODE_FLAG_MASK);
932 static int check_root_item(struct extent_buffer *leaf, struct btrfs_key *key,
935 struct btrfs_fs_info *fs_info = leaf->fs_info;
936 struct btrfs_root_item ri;
937 const u64 valid_root_flags = BTRFS_ROOT_SUBVOL_RDONLY |
938 BTRFS_ROOT_SUBVOL_DEAD;
940 /* No such tree id */
941 if (key->objectid == 0) {
942 generic_err(leaf, slot, "invalid root id 0");
947 * Some older kernel may create ROOT_ITEM with non-zero offset, so here
948 * we only check offset for reloc tree whose key->offset must be a
951 if (key->objectid == BTRFS_TREE_RELOC_OBJECTID && key->offset == 0) {
952 generic_err(leaf, slot, "invalid root id 0 for reloc tree");
956 if (btrfs_item_size_nr(leaf, slot) != sizeof(ri)) {
957 generic_err(leaf, slot,
958 "invalid root item size, have %u expect %zu",
959 btrfs_item_size_nr(leaf, slot), sizeof(ri));
962 read_extent_buffer(leaf, &ri, btrfs_item_ptr_offset(leaf, slot),
965 /* Generation related */
966 if (btrfs_root_generation(&ri) >
967 btrfs_super_generation(fs_info->super_copy) + 1) {
968 generic_err(leaf, slot,
969 "invalid root generation, have %llu expect (0, %llu]",
970 btrfs_root_generation(&ri),
971 btrfs_super_generation(fs_info->super_copy) + 1);
974 if (btrfs_root_generation_v2(&ri) >
975 btrfs_super_generation(fs_info->super_copy) + 1) {
976 generic_err(leaf, slot,
977 "invalid root v2 generation, have %llu expect (0, %llu]",
978 btrfs_root_generation_v2(&ri),
979 btrfs_super_generation(fs_info->super_copy) + 1);
982 if (btrfs_root_last_snapshot(&ri) >
983 btrfs_super_generation(fs_info->super_copy) + 1) {
984 generic_err(leaf, slot,
985 "invalid root last_snapshot, have %llu expect (0, %llu]",
986 btrfs_root_last_snapshot(&ri),
987 btrfs_super_generation(fs_info->super_copy) + 1);
991 /* Alignment and level check */
992 if (!IS_ALIGNED(btrfs_root_bytenr(&ri), fs_info->sectorsize)) {
993 generic_err(leaf, slot,
994 "invalid root bytenr, have %llu expect to be aligned to %u",
995 btrfs_root_bytenr(&ri), fs_info->sectorsize);
998 if (btrfs_root_level(&ri) >= BTRFS_MAX_LEVEL) {
999 generic_err(leaf, slot,
1000 "invalid root level, have %u expect [0, %u]",
1001 btrfs_root_level(&ri), BTRFS_MAX_LEVEL - 1);
1004 if (ri.drop_level >= BTRFS_MAX_LEVEL) {
1005 generic_err(leaf, slot,
1006 "invalid root level, have %u expect [0, %u]",
1007 ri.drop_level, BTRFS_MAX_LEVEL - 1);
1012 if (btrfs_root_flags(&ri) & ~valid_root_flags) {
1013 generic_err(leaf, slot,
1014 "invalid root flags, have 0x%llx expect mask 0x%llx",
1015 btrfs_root_flags(&ri), valid_root_flags);
1023 static void extent_err(const struct extent_buffer *eb, int slot,
1024 const char *fmt, ...)
1026 struct btrfs_key key;
1027 struct va_format vaf;
1032 btrfs_item_key_to_cpu(eb, &key, slot);
1033 bytenr = key.objectid;
1034 if (key.type == BTRFS_METADATA_ITEM_KEY ||
1035 key.type == BTRFS_TREE_BLOCK_REF_KEY ||
1036 key.type == BTRFS_SHARED_BLOCK_REF_KEY)
1037 len = eb->fs_info->nodesize;
1040 va_start(args, fmt);
1045 btrfs_crit(eb->fs_info,
1046 "corrupt %s: block=%llu slot=%d extent bytenr=%llu len=%llu %pV",
1047 btrfs_header_level(eb) == 0 ? "leaf" : "node",
1048 eb->start, slot, bytenr, len, &vaf);
1052 static int check_extent_item(struct extent_buffer *leaf,
1053 struct btrfs_key *key, int slot)
1055 struct btrfs_fs_info *fs_info = leaf->fs_info;
1056 struct btrfs_extent_item *ei;
1057 bool is_tree_block = false;
1058 unsigned long ptr; /* Current pointer inside inline refs */
1059 unsigned long end; /* Extent item end */
1060 const u32 item_size = btrfs_item_size_nr(leaf, slot);
1063 u64 total_refs; /* Total refs in btrfs_extent_item */
1064 u64 inline_refs = 0; /* found total inline refs */
1066 if (key->type == BTRFS_METADATA_ITEM_KEY &&
1067 !btrfs_fs_incompat(fs_info, SKINNY_METADATA)) {
1068 generic_err(leaf, slot,
1069 "invalid key type, METADATA_ITEM type invalid when SKINNY_METADATA feature disabled");
1072 /* key->objectid is the bytenr for both key types */
1073 if (!IS_ALIGNED(key->objectid, fs_info->sectorsize)) {
1074 generic_err(leaf, slot,
1075 "invalid key objectid, have %llu expect to be aligned to %u",
1076 key->objectid, fs_info->sectorsize);
1080 /* key->offset is tree level for METADATA_ITEM_KEY */
1081 if (key->type == BTRFS_METADATA_ITEM_KEY &&
1082 key->offset >= BTRFS_MAX_LEVEL) {
1083 extent_err(leaf, slot,
1084 "invalid tree level, have %llu expect [0, %u]",
1085 key->offset, BTRFS_MAX_LEVEL - 1);
1090 * EXTENT/METADATA_ITEM consists of:
1091 * 1) One btrfs_extent_item
1092 * Records the total refs, type and generation of the extent.
1094 * 2) One btrfs_tree_block_info (for EXTENT_ITEM and tree backref only)
1095 * Records the first key and level of the tree block.
1097 * 2) Zero or more btrfs_extent_inline_ref(s)
1098 * Each inline ref has one btrfs_extent_inline_ref shows:
1099 * 2.1) The ref type, one of the 4
1100 * TREE_BLOCK_REF Tree block only
1101 * SHARED_BLOCK_REF Tree block only
1102 * EXTENT_DATA_REF Data only
1103 * SHARED_DATA_REF Data only
1104 * 2.2) Ref type specific data
1105 * Either using btrfs_extent_inline_ref::offset, or specific
1108 if (item_size < sizeof(*ei)) {
1109 extent_err(leaf, slot,
1110 "invalid item size, have %u expect [%zu, %u)",
1111 item_size, sizeof(*ei),
1112 BTRFS_LEAF_DATA_SIZE(fs_info));
1115 end = item_size + btrfs_item_ptr_offset(leaf, slot);
1117 /* Checks against extent_item */
1118 ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
1119 flags = btrfs_extent_flags(leaf, ei);
1120 total_refs = btrfs_extent_refs(leaf, ei);
1121 generation = btrfs_extent_generation(leaf, ei);
1122 if (generation > btrfs_super_generation(fs_info->super_copy) + 1) {
1123 extent_err(leaf, slot,
1124 "invalid generation, have %llu expect (0, %llu]",
1126 btrfs_super_generation(fs_info->super_copy) + 1);
1129 if (!has_single_bit_set(flags & (BTRFS_EXTENT_FLAG_DATA |
1130 BTRFS_EXTENT_FLAG_TREE_BLOCK))) {
1131 extent_err(leaf, slot,
1132 "invalid extent flag, have 0x%llx expect 1 bit set in 0x%llx",
1133 flags, BTRFS_EXTENT_FLAG_DATA |
1134 BTRFS_EXTENT_FLAG_TREE_BLOCK);
1137 is_tree_block = !!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK);
1138 if (is_tree_block) {
1139 if (key->type == BTRFS_EXTENT_ITEM_KEY &&
1140 key->offset != fs_info->nodesize) {
1141 extent_err(leaf, slot,
1142 "invalid extent length, have %llu expect %u",
1143 key->offset, fs_info->nodesize);
1147 if (key->type != BTRFS_EXTENT_ITEM_KEY) {
1148 extent_err(leaf, slot,
1149 "invalid key type, have %u expect %u for data backref",
1150 key->type, BTRFS_EXTENT_ITEM_KEY);
1153 if (!IS_ALIGNED(key->offset, fs_info->sectorsize)) {
1154 extent_err(leaf, slot,
1155 "invalid extent length, have %llu expect aligned to %u",
1156 key->offset, fs_info->sectorsize);
1160 ptr = (unsigned long)(struct btrfs_extent_item *)(ei + 1);
1162 /* Check the special case of btrfs_tree_block_info */
1163 if (is_tree_block && key->type != BTRFS_METADATA_ITEM_KEY) {
1164 struct btrfs_tree_block_info *info;
1166 info = (struct btrfs_tree_block_info *)ptr;
1167 if (btrfs_tree_block_level(leaf, info) >= BTRFS_MAX_LEVEL) {
1168 extent_err(leaf, slot,
1169 "invalid tree block info level, have %u expect [0, %u]",
1170 btrfs_tree_block_level(leaf, info),
1171 BTRFS_MAX_LEVEL - 1);
1174 ptr = (unsigned long)(struct btrfs_tree_block_info *)(info + 1);
1177 /* Check inline refs */
1179 struct btrfs_extent_inline_ref *iref;
1180 struct btrfs_extent_data_ref *dref;
1181 struct btrfs_shared_data_ref *sref;
1186 if (ptr + sizeof(*iref) > end) {
1187 extent_err(leaf, slot,
1188 "inline ref item overflows extent item, ptr %lu iref size %zu end %lu",
1189 ptr, sizeof(*iref), end);
1192 iref = (struct btrfs_extent_inline_ref *)ptr;
1193 inline_type = btrfs_extent_inline_ref_type(leaf, iref);
1194 inline_offset = btrfs_extent_inline_ref_offset(leaf, iref);
1195 if (ptr + btrfs_extent_inline_ref_size(inline_type) > end) {
1196 extent_err(leaf, slot,
1197 "inline ref item overflows extent item, ptr %lu iref size %u end %lu",
1198 ptr, inline_type, end);
1202 switch (inline_type) {
1203 /* inline_offset is subvolid of the owner, no need to check */
1204 case BTRFS_TREE_BLOCK_REF_KEY:
1207 /* Contains parent bytenr */
1208 case BTRFS_SHARED_BLOCK_REF_KEY:
1209 if (!IS_ALIGNED(inline_offset, fs_info->sectorsize)) {
1210 extent_err(leaf, slot,
1211 "invalid tree parent bytenr, have %llu expect aligned to %u",
1212 inline_offset, fs_info->sectorsize);
1218 * Contains owner subvolid, owner key objectid, adjusted offset.
1219 * The only obvious corruption can happen in that offset.
1221 case BTRFS_EXTENT_DATA_REF_KEY:
1222 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1223 dref_offset = btrfs_extent_data_ref_offset(leaf, dref);
1224 if (!IS_ALIGNED(dref_offset, fs_info->sectorsize)) {
1225 extent_err(leaf, slot,
1226 "invalid data ref offset, have %llu expect aligned to %u",
1227 dref_offset, fs_info->sectorsize);
1230 inline_refs += btrfs_extent_data_ref_count(leaf, dref);
1232 /* Contains parent bytenr and ref count */
1233 case BTRFS_SHARED_DATA_REF_KEY:
1234 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1235 if (!IS_ALIGNED(inline_offset, fs_info->sectorsize)) {
1236 extent_err(leaf, slot,
1237 "invalid data parent bytenr, have %llu expect aligned to %u",
1238 inline_offset, fs_info->sectorsize);
1241 inline_refs += btrfs_shared_data_ref_count(leaf, sref);
1244 extent_err(leaf, slot, "unknown inline ref type: %u",
1248 ptr += btrfs_extent_inline_ref_size(inline_type);
1250 /* No padding is allowed */
1252 extent_err(leaf, slot,
1253 "invalid extent item size, padding bytes found");
1257 /* Finally, check the inline refs against total refs */
1258 if (inline_refs > total_refs) {
1259 extent_err(leaf, slot,
1260 "invalid extent refs, have %llu expect >= inline %llu",
1261 total_refs, inline_refs);
1267 static int check_simple_keyed_refs(struct extent_buffer *leaf,
1268 struct btrfs_key *key, int slot)
1270 u32 expect_item_size = 0;
1272 if (key->type == BTRFS_SHARED_DATA_REF_KEY)
1273 expect_item_size = sizeof(struct btrfs_shared_data_ref);
1275 if (btrfs_item_size_nr(leaf, slot) != expect_item_size) {
1276 generic_err(leaf, slot,
1277 "invalid item size, have %u expect %u for key type %u",
1278 btrfs_item_size_nr(leaf, slot),
1279 expect_item_size, key->type);
1282 if (!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize)) {
1283 generic_err(leaf, slot,
1284 "invalid key objectid for shared block ref, have %llu expect aligned to %u",
1285 key->objectid, leaf->fs_info->sectorsize);
1288 if (key->type != BTRFS_TREE_BLOCK_REF_KEY &&
1289 !IS_ALIGNED(key->offset, leaf->fs_info->sectorsize)) {
1290 extent_err(leaf, slot,
1291 "invalid tree parent bytenr, have %llu expect aligned to %u",
1292 key->offset, leaf->fs_info->sectorsize);
1298 static int check_extent_data_ref(struct extent_buffer *leaf,
1299 struct btrfs_key *key, int slot)
1301 struct btrfs_extent_data_ref *dref;
1302 unsigned long ptr = btrfs_item_ptr_offset(leaf, slot);
1303 const unsigned long end = ptr + btrfs_item_size_nr(leaf, slot);
1305 if (btrfs_item_size_nr(leaf, slot) % sizeof(*dref) != 0) {
1306 generic_err(leaf, slot,
1307 "invalid item size, have %u expect aligned to %zu for key type %u",
1308 btrfs_item_size_nr(leaf, slot),
1309 sizeof(*dref), key->type);
1311 if (!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize)) {
1312 generic_err(leaf, slot,
1313 "invalid key objectid for shared block ref, have %llu expect aligned to %u",
1314 key->objectid, leaf->fs_info->sectorsize);
1317 for (; ptr < end; ptr += sizeof(*dref)) {
1323 dref = (struct btrfs_extent_data_ref *)ptr;
1324 root_objectid = btrfs_extent_data_ref_root(leaf, dref);
1325 owner = btrfs_extent_data_ref_objectid(leaf, dref);
1326 offset = btrfs_extent_data_ref_offset(leaf, dref);
1327 hash = hash_extent_data_ref(root_objectid, owner, offset);
1328 if (hash != key->offset) {
1329 extent_err(leaf, slot,
1330 "invalid extent data ref hash, item has 0x%016llx key has 0x%016llx",
1334 if (!IS_ALIGNED(offset, leaf->fs_info->sectorsize)) {
1335 extent_err(leaf, slot,
1336 "invalid extent data backref offset, have %llu expect aligned to %u",
1337 offset, leaf->fs_info->sectorsize);
1343 #define inode_ref_err(fs_info, eb, slot, fmt, args...) \
1344 inode_item_err(fs_info, eb, slot, fmt, ##args)
1345 static int check_inode_ref(struct extent_buffer *leaf,
1346 struct btrfs_key *key, struct btrfs_key *prev_key,
1349 struct btrfs_inode_ref *iref;
1353 if (!check_prev_ino(leaf, key, slot, prev_key))
1355 /* namelen can't be 0, so item_size == sizeof() is also invalid */
1356 if (btrfs_item_size_nr(leaf, slot) <= sizeof(*iref)) {
1357 inode_ref_err(fs_info, leaf, slot,
1358 "invalid item size, have %u expect (%zu, %u)",
1359 btrfs_item_size_nr(leaf, slot),
1360 sizeof(*iref), BTRFS_LEAF_DATA_SIZE(leaf->fs_info));
1364 ptr = btrfs_item_ptr_offset(leaf, slot);
1365 end = ptr + btrfs_item_size_nr(leaf, slot);
1369 if (ptr + sizeof(iref) > end) {
1370 inode_ref_err(fs_info, leaf, slot,
1371 "inode ref overflow, ptr %lu end %lu inode_ref_size %zu",
1372 ptr, end, sizeof(iref));
1376 iref = (struct btrfs_inode_ref *)ptr;
1377 namelen = btrfs_inode_ref_name_len(leaf, iref);
1378 if (ptr + sizeof(*iref) + namelen > end) {
1379 inode_ref_err(fs_info, leaf, slot,
1380 "inode ref overflow, ptr %lu end %lu namelen %u",
1386 * NOTE: In theory we should record all found index numbers
1387 * to find any duplicated indexes, but that will be too time
1388 * consuming for inodes with too many hard links.
1390 ptr += sizeof(*iref) + namelen;
1396 * Common point to switch the item-specific validation.
1398 static int check_leaf_item(struct extent_buffer *leaf,
1399 struct btrfs_key *key, int slot,
1400 struct btrfs_key *prev_key)
1403 struct btrfs_chunk *chunk;
1405 switch (key->type) {
1406 case BTRFS_EXTENT_DATA_KEY:
1407 ret = check_extent_data_item(leaf, key, slot, prev_key);
1409 case BTRFS_EXTENT_CSUM_KEY:
1410 ret = check_csum_item(leaf, key, slot, prev_key);
1412 case BTRFS_DIR_ITEM_KEY:
1413 case BTRFS_DIR_INDEX_KEY:
1414 case BTRFS_XATTR_ITEM_KEY:
1415 ret = check_dir_item(leaf, key, prev_key, slot);
1417 case BTRFS_INODE_REF_KEY:
1418 ret = check_inode_ref(leaf, key, prev_key, slot);
1420 case BTRFS_BLOCK_GROUP_ITEM_KEY:
1421 ret = check_block_group_item(leaf, key, slot);
1423 case BTRFS_CHUNK_ITEM_KEY:
1424 chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
1425 ret = check_leaf_chunk_item(leaf, chunk, key, slot);
1427 case BTRFS_DEV_ITEM_KEY:
1428 ret = check_dev_item(leaf, key, slot);
1430 case BTRFS_INODE_ITEM_KEY:
1431 ret = check_inode_item(leaf, key, slot);
1433 case BTRFS_ROOT_ITEM_KEY:
1434 ret = check_root_item(leaf, key, slot);
1436 case BTRFS_EXTENT_ITEM_KEY:
1437 case BTRFS_METADATA_ITEM_KEY:
1438 ret = check_extent_item(leaf, key, slot);
1440 case BTRFS_TREE_BLOCK_REF_KEY:
1441 case BTRFS_SHARED_DATA_REF_KEY:
1442 case BTRFS_SHARED_BLOCK_REF_KEY:
1443 ret = check_simple_keyed_refs(leaf, key, slot);
1445 case BTRFS_EXTENT_DATA_REF_KEY:
1446 ret = check_extent_data_ref(leaf, key, slot);
1452 static int check_leaf(struct extent_buffer *leaf, bool check_item_data)
1454 struct btrfs_fs_info *fs_info = leaf->fs_info;
1455 /* No valid key type is 0, so all key should be larger than this key */
1456 struct btrfs_key prev_key = {0, 0, 0};
1457 struct btrfs_key key;
1458 u32 nritems = btrfs_header_nritems(leaf);
1461 if (btrfs_header_level(leaf) != 0) {
1462 generic_err(leaf, 0,
1463 "invalid level for leaf, have %d expect 0",
1464 btrfs_header_level(leaf));
1469 * Extent buffers from a relocation tree have a owner field that
1470 * corresponds to the subvolume tree they are based on. So just from an
1471 * extent buffer alone we can not find out what is the id of the
1472 * corresponding subvolume tree, so we can not figure out if the extent
1473 * buffer corresponds to the root of the relocation tree or not. So
1474 * skip this check for relocation trees.
1476 if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) {
1477 u64 owner = btrfs_header_owner(leaf);
1479 /* These trees must never be empty */
1480 if (owner == BTRFS_ROOT_TREE_OBJECTID ||
1481 owner == BTRFS_CHUNK_TREE_OBJECTID ||
1482 owner == BTRFS_EXTENT_TREE_OBJECTID ||
1483 owner == BTRFS_DEV_TREE_OBJECTID ||
1484 owner == BTRFS_FS_TREE_OBJECTID ||
1485 owner == BTRFS_DATA_RELOC_TREE_OBJECTID) {
1486 generic_err(leaf, 0,
1487 "invalid root, root %llu must never be empty",
1493 generic_err(leaf, 0,
1494 "invalid owner, root 0 is not defined");
1504 * Check the following things to make sure this is a good leaf, and
1505 * leaf users won't need to bother with similar sanity checks:
1508 * 2) item offset and size
1509 * No overlap, no hole, all inside the leaf.
1511 * If possible, do comprehensive sanity check.
1512 * NOTE: All checks must only rely on the item data itself.
1514 for (slot = 0; slot < nritems; slot++) {
1515 u32 item_end_expected;
1518 btrfs_item_key_to_cpu(leaf, &key, slot);
1520 /* Make sure the keys are in the right order */
1521 if (btrfs_comp_cpu_keys(&prev_key, &key) >= 0) {
1522 generic_err(leaf, slot,
1523 "bad key order, prev (%llu %u %llu) current (%llu %u %llu)",
1524 prev_key.objectid, prev_key.type,
1525 prev_key.offset, key.objectid, key.type,
1531 * Make sure the offset and ends are right, remember that the
1532 * item data starts at the end of the leaf and grows towards the
1536 item_end_expected = BTRFS_LEAF_DATA_SIZE(fs_info);
1538 item_end_expected = btrfs_item_offset_nr(leaf,
1540 if (btrfs_item_end_nr(leaf, slot) != item_end_expected) {
1541 generic_err(leaf, slot,
1542 "unexpected item end, have %u expect %u",
1543 btrfs_item_end_nr(leaf, slot),
1549 * Check to make sure that we don't point outside of the leaf,
1550 * just in case all the items are consistent to each other, but
1551 * all point outside of the leaf.
1553 if (btrfs_item_end_nr(leaf, slot) >
1554 BTRFS_LEAF_DATA_SIZE(fs_info)) {
1555 generic_err(leaf, slot,
1556 "slot end outside of leaf, have %u expect range [0, %u]",
1557 btrfs_item_end_nr(leaf, slot),
1558 BTRFS_LEAF_DATA_SIZE(fs_info));
1562 /* Also check if the item pointer overlaps with btrfs item. */
1563 if (btrfs_item_nr_offset(slot) + sizeof(struct btrfs_item) >
1564 btrfs_item_ptr_offset(leaf, slot)) {
1565 generic_err(leaf, slot,
1566 "slot overlaps with its data, item end %lu data start %lu",
1567 btrfs_item_nr_offset(slot) +
1568 sizeof(struct btrfs_item),
1569 btrfs_item_ptr_offset(leaf, slot));
1573 if (check_item_data) {
1575 * Check if the item size and content meet other
1578 ret = check_leaf_item(leaf, &key, slot, &prev_key);
1583 prev_key.objectid = key.objectid;
1584 prev_key.type = key.type;
1585 prev_key.offset = key.offset;
1591 int btrfs_check_leaf_full(struct extent_buffer *leaf)
1593 return check_leaf(leaf, true);
1595 ALLOW_ERROR_INJECTION(btrfs_check_leaf_full, ERRNO);
1597 int btrfs_check_leaf_relaxed(struct extent_buffer *leaf)
1599 return check_leaf(leaf, false);
1602 int btrfs_check_node(struct extent_buffer *node)
1604 struct btrfs_fs_info *fs_info = node->fs_info;
1605 unsigned long nr = btrfs_header_nritems(node);
1606 struct btrfs_key key, next_key;
1608 int level = btrfs_header_level(node);
1612 if (level <= 0 || level >= BTRFS_MAX_LEVEL) {
1613 generic_err(node, 0,
1614 "invalid level for node, have %d expect [1, %d]",
1615 level, BTRFS_MAX_LEVEL - 1);
1618 if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(fs_info)) {
1620 "corrupt node: root=%llu block=%llu, nritems too %s, have %lu expect range [1,%u]",
1621 btrfs_header_owner(node), node->start,
1622 nr == 0 ? "small" : "large", nr,
1623 BTRFS_NODEPTRS_PER_BLOCK(fs_info));
1627 for (slot = 0; slot < nr - 1; slot++) {
1628 bytenr = btrfs_node_blockptr(node, slot);
1629 btrfs_node_key_to_cpu(node, &key, slot);
1630 btrfs_node_key_to_cpu(node, &next_key, slot + 1);
1633 generic_err(node, slot,
1634 "invalid NULL node pointer");
1638 if (!IS_ALIGNED(bytenr, fs_info->sectorsize)) {
1639 generic_err(node, slot,
1640 "unaligned pointer, have %llu should be aligned to %u",
1641 bytenr, fs_info->sectorsize);
1646 if (btrfs_comp_cpu_keys(&key, &next_key) >= 0) {
1647 generic_err(node, slot,
1648 "bad key order, current (%llu %u %llu) next (%llu %u %llu)",
1649 key.objectid, key.type, key.offset,
1650 next_key.objectid, next_key.type,
1659 ALLOW_ERROR_INJECTION(btrfs_check_node, ERRNO);