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"
28 * Error message should follow the following format:
29 * corrupt <type>: <identifier>, <reason>[, <bad_value>]
32 * @identifier: the necessary info to locate the leaf/node.
33 * It's recommended to decode key.objecitd/offset if it's
35 * @reason: describe the error
36 * @bad_value: optional, it's recommended to output bad value and its
37 * expected value (range).
39 * Since comma is used to separate the components, only space is allowed
40 * inside each component.
44 * Append generic "corrupt leaf/node root=%llu block=%llu slot=%d: " to @fmt.
45 * Allows callers to customize the output.
49 static void generic_err(const struct extent_buffer *eb, int slot,
52 const struct btrfs_fs_info *fs_info = eb->fs_info;
62 "corrupt %s: root=%llu block=%llu slot=%d, %pV",
63 btrfs_header_level(eb) == 0 ? "leaf" : "node",
64 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, &vaf);
69 * Customized reporter for extent data item, since its key objectid and
70 * offset has its own meaning.
74 static void file_extent_err(const struct extent_buffer *eb, int slot,
77 const struct btrfs_fs_info *fs_info = eb->fs_info;
82 btrfs_item_key_to_cpu(eb, &key, slot);
89 "corrupt %s: root=%llu block=%llu slot=%d ino=%llu file_offset=%llu, %pV",
90 btrfs_header_level(eb) == 0 ? "leaf" : "node",
91 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
92 key.objectid, key.offset, &vaf);
97 * Return 0 if the btrfs_file_extent_##name is aligned to @alignment
100 #define CHECK_FE_ALIGNED(leaf, slot, fi, name, alignment) \
102 if (!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment))) \
103 file_extent_err((leaf), (slot), \
104 "invalid %s for file extent, have %llu, should be aligned to %u", \
105 (#name), btrfs_file_extent_##name((leaf), (fi)), \
107 (!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment))); \
110 static u64 file_extent_end(struct extent_buffer *leaf,
111 struct btrfs_key *key,
112 struct btrfs_file_extent_item *extent)
117 if (btrfs_file_extent_type(leaf, extent) == BTRFS_FILE_EXTENT_INLINE) {
118 len = btrfs_file_extent_ram_bytes(leaf, extent);
119 end = ALIGN(key->offset + len, leaf->fs_info->sectorsize);
121 len = btrfs_file_extent_num_bytes(leaf, extent);
122 end = key->offset + len;
127 static int check_extent_data_item(struct extent_buffer *leaf,
128 struct btrfs_key *key, int slot,
129 struct btrfs_key *prev_key)
131 struct btrfs_fs_info *fs_info = leaf->fs_info;
132 struct btrfs_file_extent_item *fi;
133 u32 sectorsize = fs_info->sectorsize;
134 u32 item_size = btrfs_item_size_nr(leaf, slot);
137 if (!IS_ALIGNED(key->offset, sectorsize)) {
138 file_extent_err(leaf, slot,
139 "unaligned file_offset for file extent, have %llu should be aligned to %u",
140 key->offset, sectorsize);
144 fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
146 if (btrfs_file_extent_type(leaf, fi) > BTRFS_FILE_EXTENT_TYPES) {
147 file_extent_err(leaf, slot,
148 "invalid type for file extent, have %u expect range [0, %u]",
149 btrfs_file_extent_type(leaf, fi),
150 BTRFS_FILE_EXTENT_TYPES);
155 * Support for new compression/encryption must introduce incompat flag,
156 * and must be caught in open_ctree().
158 if (btrfs_file_extent_compression(leaf, fi) > BTRFS_COMPRESS_TYPES) {
159 file_extent_err(leaf, slot,
160 "invalid compression for file extent, have %u expect range [0, %u]",
161 btrfs_file_extent_compression(leaf, fi),
162 BTRFS_COMPRESS_TYPES);
165 if (btrfs_file_extent_encryption(leaf, fi)) {
166 file_extent_err(leaf, slot,
167 "invalid encryption for file extent, have %u expect 0",
168 btrfs_file_extent_encryption(leaf, fi));
171 if (btrfs_file_extent_type(leaf, fi) == BTRFS_FILE_EXTENT_INLINE) {
172 /* Inline extent must have 0 as key offset */
174 file_extent_err(leaf, slot,
175 "invalid file_offset for inline file extent, have %llu expect 0",
180 /* Compressed inline extent has no on-disk size, skip it */
181 if (btrfs_file_extent_compression(leaf, fi) !=
185 /* Uncompressed inline extent size must match item size */
186 if (item_size != BTRFS_FILE_EXTENT_INLINE_DATA_START +
187 btrfs_file_extent_ram_bytes(leaf, fi)) {
188 file_extent_err(leaf, slot,
189 "invalid ram_bytes for uncompressed inline extent, have %u expect %llu",
190 item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START +
191 btrfs_file_extent_ram_bytes(leaf, fi));
197 /* Regular or preallocated extent has fixed item size */
198 if (item_size != sizeof(*fi)) {
199 file_extent_err(leaf, slot,
200 "invalid item size for reg/prealloc file extent, have %u expect %zu",
201 item_size, sizeof(*fi));
204 if (CHECK_FE_ALIGNED(leaf, slot, fi, ram_bytes, sectorsize) ||
205 CHECK_FE_ALIGNED(leaf, slot, fi, disk_bytenr, sectorsize) ||
206 CHECK_FE_ALIGNED(leaf, slot, fi, disk_num_bytes, sectorsize) ||
207 CHECK_FE_ALIGNED(leaf, slot, fi, offset, sectorsize) ||
208 CHECK_FE_ALIGNED(leaf, slot, fi, num_bytes, sectorsize))
211 /* Catch extent end overflow */
212 if (check_add_overflow(btrfs_file_extent_num_bytes(leaf, fi),
213 key->offset, &extent_end)) {
214 file_extent_err(leaf, slot,
215 "extent end overflow, have file offset %llu extent num bytes %llu",
217 btrfs_file_extent_num_bytes(leaf, fi));
222 * Check that no two consecutive file extent items, in the same leaf,
223 * present ranges that overlap each other.
226 prev_key->objectid == key->objectid &&
227 prev_key->type == BTRFS_EXTENT_DATA_KEY) {
228 struct btrfs_file_extent_item *prev_fi;
231 prev_fi = btrfs_item_ptr(leaf, slot - 1,
232 struct btrfs_file_extent_item);
233 prev_end = file_extent_end(leaf, prev_key, prev_fi);
234 if (prev_end > key->offset) {
235 file_extent_err(leaf, slot - 1,
236 "file extent end range (%llu) goes beyond start offset (%llu) of the next file extent",
237 prev_end, key->offset);
245 static int check_csum_item(struct extent_buffer *leaf, struct btrfs_key *key,
248 struct btrfs_fs_info *fs_info = leaf->fs_info;
249 u32 sectorsize = fs_info->sectorsize;
250 u32 csumsize = btrfs_super_csum_size(fs_info->super_copy);
252 if (key->objectid != BTRFS_EXTENT_CSUM_OBJECTID) {
253 generic_err(leaf, slot,
254 "invalid key objectid for csum item, have %llu expect %llu",
255 key->objectid, BTRFS_EXTENT_CSUM_OBJECTID);
258 if (!IS_ALIGNED(key->offset, sectorsize)) {
259 generic_err(leaf, slot,
260 "unaligned key offset for csum item, have %llu should be aligned to %u",
261 key->offset, sectorsize);
264 if (!IS_ALIGNED(btrfs_item_size_nr(leaf, slot), csumsize)) {
265 generic_err(leaf, slot,
266 "unaligned item size for csum item, have %u should be aligned to %u",
267 btrfs_item_size_nr(leaf, slot), csumsize);
274 * Customized reported for dir_item, only important new info is key->objectid,
275 * which represents inode number
279 static void dir_item_err(const struct extent_buffer *eb, int slot,
280 const char *fmt, ...)
282 const struct btrfs_fs_info *fs_info = eb->fs_info;
283 struct btrfs_key key;
284 struct va_format vaf;
287 btrfs_item_key_to_cpu(eb, &key, slot);
294 "corrupt %s: root=%llu block=%llu slot=%d ino=%llu, %pV",
295 btrfs_header_level(eb) == 0 ? "leaf" : "node",
296 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
301 static int check_dir_item(struct extent_buffer *leaf,
302 struct btrfs_key *key, int slot)
304 struct btrfs_fs_info *fs_info = leaf->fs_info;
305 struct btrfs_dir_item *di;
306 u32 item_size = btrfs_item_size_nr(leaf, slot);
309 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
310 while (cur < item_size) {
318 /* header itself should not cross item boundary */
319 if (cur + sizeof(*di) > item_size) {
320 dir_item_err(leaf, slot,
321 "dir item header crosses item boundary, have %zu boundary %u",
322 cur + sizeof(*di), item_size);
327 dir_type = btrfs_dir_type(leaf, di);
328 if (dir_type >= BTRFS_FT_MAX) {
329 dir_item_err(leaf, slot,
330 "invalid dir item type, have %u expect [0, %u)",
331 dir_type, BTRFS_FT_MAX);
335 if (key->type == BTRFS_XATTR_ITEM_KEY &&
336 dir_type != BTRFS_FT_XATTR) {
337 dir_item_err(leaf, slot,
338 "invalid dir item type for XATTR key, have %u expect %u",
339 dir_type, BTRFS_FT_XATTR);
342 if (dir_type == BTRFS_FT_XATTR &&
343 key->type != BTRFS_XATTR_ITEM_KEY) {
344 dir_item_err(leaf, slot,
345 "xattr dir type found for non-XATTR key");
348 if (dir_type == BTRFS_FT_XATTR)
349 max_name_len = XATTR_NAME_MAX;
351 max_name_len = BTRFS_NAME_LEN;
353 /* Name/data length check */
354 name_len = btrfs_dir_name_len(leaf, di);
355 data_len = btrfs_dir_data_len(leaf, di);
356 if (name_len > max_name_len) {
357 dir_item_err(leaf, slot,
358 "dir item name len too long, have %u max %u",
359 name_len, max_name_len);
362 if (name_len + data_len > BTRFS_MAX_XATTR_SIZE(fs_info)) {
363 dir_item_err(leaf, slot,
364 "dir item name and data len too long, have %u max %u",
366 BTRFS_MAX_XATTR_SIZE(fs_info));
370 if (data_len && dir_type != BTRFS_FT_XATTR) {
371 dir_item_err(leaf, slot,
372 "dir item with invalid data len, have %u expect 0",
377 total_size = sizeof(*di) + name_len + data_len;
379 /* header and name/data should not cross item boundary */
380 if (cur + total_size > item_size) {
381 dir_item_err(leaf, slot,
382 "dir item data crosses item boundary, have %u boundary %u",
383 cur + total_size, item_size);
388 * Special check for XATTR/DIR_ITEM, as key->offset is name
389 * hash, should match its name
391 if (key->type == BTRFS_DIR_ITEM_KEY ||
392 key->type == BTRFS_XATTR_ITEM_KEY) {
393 char namebuf[max(BTRFS_NAME_LEN, XATTR_NAME_MAX)];
395 read_extent_buffer(leaf, namebuf,
396 (unsigned long)(di + 1), name_len);
397 name_hash = btrfs_name_hash(namebuf, name_len);
398 if (key->offset != name_hash) {
399 dir_item_err(leaf, slot,
400 "name hash mismatch with key, have 0x%016x expect 0x%016llx",
401 name_hash, key->offset);
406 di = (struct btrfs_dir_item *)((void *)di + total_size);
413 static void block_group_err(const struct extent_buffer *eb, int slot,
414 const char *fmt, ...)
416 const struct btrfs_fs_info *fs_info = eb->fs_info;
417 struct btrfs_key key;
418 struct va_format vaf;
421 btrfs_item_key_to_cpu(eb, &key, slot);
428 "corrupt %s: root=%llu block=%llu slot=%d bg_start=%llu bg_len=%llu, %pV",
429 btrfs_header_level(eb) == 0 ? "leaf" : "node",
430 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
431 key.objectid, key.offset, &vaf);
435 static int check_block_group_item(struct extent_buffer *leaf,
436 struct btrfs_key *key, int slot)
438 struct btrfs_block_group_item bgi;
439 u32 item_size = btrfs_item_size_nr(leaf, slot);
444 * Here we don't really care about alignment since extent allocator can
445 * handle it. We care more about the size.
447 if (key->offset == 0) {
448 block_group_err(leaf, slot,
449 "invalid block group size 0");
453 if (item_size != sizeof(bgi)) {
454 block_group_err(leaf, slot,
455 "invalid item size, have %u expect %zu",
456 item_size, sizeof(bgi));
460 read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot),
462 if (btrfs_block_group_chunk_objectid(&bgi) !=
463 BTRFS_FIRST_CHUNK_TREE_OBJECTID) {
464 block_group_err(leaf, slot,
465 "invalid block group chunk objectid, have %llu expect %llu",
466 btrfs_block_group_chunk_objectid(&bgi),
467 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
471 if (btrfs_block_group_used(&bgi) > key->offset) {
472 block_group_err(leaf, slot,
473 "invalid block group used, have %llu expect [0, %llu)",
474 btrfs_block_group_used(&bgi), key->offset);
478 flags = btrfs_block_group_flags(&bgi);
479 if (hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) > 1) {
480 block_group_err(leaf, slot,
481 "invalid profile flags, have 0x%llx (%lu bits set) expect no more than 1 bit set",
482 flags & BTRFS_BLOCK_GROUP_PROFILE_MASK,
483 hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK));
487 type = flags & BTRFS_BLOCK_GROUP_TYPE_MASK;
488 if (type != BTRFS_BLOCK_GROUP_DATA &&
489 type != BTRFS_BLOCK_GROUP_METADATA &&
490 type != BTRFS_BLOCK_GROUP_SYSTEM &&
491 type != (BTRFS_BLOCK_GROUP_METADATA |
492 BTRFS_BLOCK_GROUP_DATA)) {
493 block_group_err(leaf, slot,
494 "invalid type, have 0x%llx (%lu bits set) expect either 0x%llx, 0x%llx, 0x%llx or 0x%llx",
495 type, hweight64(type),
496 BTRFS_BLOCK_GROUP_DATA, BTRFS_BLOCK_GROUP_METADATA,
497 BTRFS_BLOCK_GROUP_SYSTEM,
498 BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA);
506 static void chunk_err(const struct extent_buffer *leaf,
507 const struct btrfs_chunk *chunk, u64 logical,
508 const char *fmt, ...)
510 const struct btrfs_fs_info *fs_info = leaf->fs_info;
512 struct va_format vaf;
517 /* Only superblock eb is able to have such small offset */
518 is_sb = (leaf->start == BTRFS_SUPER_INFO_OFFSET);
522 * Get the slot number by iterating through all slots, this
523 * would provide better readability.
525 for (i = 0; i < btrfs_header_nritems(leaf); i++) {
526 if (btrfs_item_ptr_offset(leaf, i) ==
527 (unsigned long)chunk) {
539 "corrupt superblock syschunk array: chunk_start=%llu, %pV",
543 "corrupt leaf: root=%llu block=%llu slot=%d chunk_start=%llu, %pV",
544 BTRFS_CHUNK_TREE_OBJECTID, leaf->start, slot,
550 * The common chunk check which could also work on super block sys chunk array.
552 * Return -EUCLEAN if anything is corrupted.
553 * Return 0 if everything is OK.
555 int btrfs_check_chunk_valid(struct extent_buffer *leaf,
556 struct btrfs_chunk *chunk, u64 logical)
558 struct btrfs_fs_info *fs_info = leaf->fs_info;
567 length = btrfs_chunk_length(leaf, chunk);
568 stripe_len = btrfs_chunk_stripe_len(leaf, chunk);
569 num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
570 sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk);
571 type = btrfs_chunk_type(leaf, chunk);
574 chunk_err(leaf, chunk, logical,
575 "invalid chunk num_stripes, have %u", num_stripes);
578 if (!IS_ALIGNED(logical, fs_info->sectorsize)) {
579 chunk_err(leaf, chunk, logical,
580 "invalid chunk logical, have %llu should aligned to %u",
581 logical, fs_info->sectorsize);
584 if (btrfs_chunk_sector_size(leaf, chunk) != fs_info->sectorsize) {
585 chunk_err(leaf, chunk, logical,
586 "invalid chunk sectorsize, have %u expect %u",
587 btrfs_chunk_sector_size(leaf, chunk),
588 fs_info->sectorsize);
591 if (!length || !IS_ALIGNED(length, fs_info->sectorsize)) {
592 chunk_err(leaf, chunk, logical,
593 "invalid chunk length, have %llu", length);
596 if (!is_power_of_2(stripe_len) || stripe_len != BTRFS_STRIPE_LEN) {
597 chunk_err(leaf, chunk, logical,
598 "invalid chunk stripe length: %llu",
602 if (~(BTRFS_BLOCK_GROUP_TYPE_MASK | BTRFS_BLOCK_GROUP_PROFILE_MASK) &
604 chunk_err(leaf, chunk, logical,
605 "unrecognized chunk type: 0x%llx",
606 ~(BTRFS_BLOCK_GROUP_TYPE_MASK |
607 BTRFS_BLOCK_GROUP_PROFILE_MASK) &
608 btrfs_chunk_type(leaf, chunk));
612 if (!is_power_of_2(type & BTRFS_BLOCK_GROUP_PROFILE_MASK) &&
613 (type & BTRFS_BLOCK_GROUP_PROFILE_MASK) != 0) {
614 chunk_err(leaf, chunk, logical,
615 "invalid chunk profile flag: 0x%llx, expect 0 or 1 bit set",
616 type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
619 if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == 0) {
620 chunk_err(leaf, chunk, logical,
621 "missing chunk type flag, have 0x%llx one bit must be set in 0x%llx",
622 type, BTRFS_BLOCK_GROUP_TYPE_MASK);
626 if ((type & BTRFS_BLOCK_GROUP_SYSTEM) &&
627 (type & (BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA))) {
628 chunk_err(leaf, chunk, logical,
629 "system chunk with data or metadata type: 0x%llx",
634 features = btrfs_super_incompat_flags(fs_info->super_copy);
635 if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS)
639 if ((type & BTRFS_BLOCK_GROUP_METADATA) &&
640 (type & BTRFS_BLOCK_GROUP_DATA)) {
641 chunk_err(leaf, chunk, logical,
642 "mixed chunk type in non-mixed mode: 0x%llx", type);
647 if ((type & BTRFS_BLOCK_GROUP_RAID10 && sub_stripes != 2) ||
648 (type & BTRFS_BLOCK_GROUP_RAID1 && num_stripes != 2) ||
649 (type & BTRFS_BLOCK_GROUP_RAID5 && num_stripes < 2) ||
650 (type & BTRFS_BLOCK_GROUP_RAID6 && num_stripes < 3) ||
651 (type & BTRFS_BLOCK_GROUP_DUP && num_stripes != 2) ||
652 ((type & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 && num_stripes != 1)) {
653 chunk_err(leaf, chunk, logical,
654 "invalid num_stripes:sub_stripes %u:%u for profile %llu",
655 num_stripes, sub_stripes,
656 type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
665 static void dev_item_err(const struct extent_buffer *eb, int slot,
666 const char *fmt, ...)
668 struct btrfs_key key;
669 struct va_format vaf;
672 btrfs_item_key_to_cpu(eb, &key, slot);
678 btrfs_crit(eb->fs_info,
679 "corrupt %s: root=%llu block=%llu slot=%d devid=%llu %pV",
680 btrfs_header_level(eb) == 0 ? "leaf" : "node",
681 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
686 static int check_dev_item(struct extent_buffer *leaf,
687 struct btrfs_key *key, int slot)
689 struct btrfs_fs_info *fs_info = leaf->fs_info;
690 struct btrfs_dev_item *ditem;
691 u64 max_devid = max(BTRFS_MAX_DEVS(fs_info), BTRFS_MAX_DEVS_SYS_CHUNK);
693 if (key->objectid != BTRFS_DEV_ITEMS_OBJECTID) {
694 dev_item_err(leaf, slot,
695 "invalid objectid: has=%llu expect=%llu",
696 key->objectid, BTRFS_DEV_ITEMS_OBJECTID);
699 if (key->offset > max_devid) {
700 dev_item_err(leaf, slot,
701 "invalid devid: has=%llu expect=[0, %llu]",
702 key->offset, max_devid);
705 ditem = btrfs_item_ptr(leaf, slot, struct btrfs_dev_item);
706 if (btrfs_device_id(leaf, ditem) != key->offset) {
707 dev_item_err(leaf, slot,
708 "devid mismatch: key has=%llu item has=%llu",
709 key->offset, btrfs_device_id(leaf, ditem));
714 * For device total_bytes, we don't have reliable way to check it, as
715 * it can be 0 for device removal. Device size check can only be done
716 * by dev extents check.
718 if (btrfs_device_bytes_used(leaf, ditem) >
719 btrfs_device_total_bytes(leaf, ditem)) {
720 dev_item_err(leaf, slot,
721 "invalid bytes used: have %llu expect [0, %llu]",
722 btrfs_device_bytes_used(leaf, ditem),
723 btrfs_device_total_bytes(leaf, ditem));
727 * Remaining members like io_align/type/gen/dev_group aren't really
728 * utilized. Skip them to make later usage of them easier.
733 /* Inode item error output has the same format as dir_item_err() */
734 #define inode_item_err(fs_info, eb, slot, fmt, ...) \
735 dir_item_err(eb, slot, fmt, __VA_ARGS__)
737 static int check_inode_item(struct extent_buffer *leaf,
738 struct btrfs_key *key, int slot)
740 struct btrfs_fs_info *fs_info = leaf->fs_info;
741 struct btrfs_inode_item *iitem;
742 u64 super_gen = btrfs_super_generation(fs_info->super_copy);
743 u32 valid_mask = (S_IFMT | S_ISUID | S_ISGID | S_ISVTX | 0777);
746 if ((key->objectid < BTRFS_FIRST_FREE_OBJECTID ||
747 key->objectid > BTRFS_LAST_FREE_OBJECTID) &&
748 key->objectid != BTRFS_ROOT_TREE_DIR_OBJECTID &&
749 key->objectid != BTRFS_FREE_INO_OBJECTID) {
750 generic_err(leaf, slot,
751 "invalid key objectid: has %llu expect %llu or [%llu, %llu] or %llu",
752 key->objectid, BTRFS_ROOT_TREE_DIR_OBJECTID,
753 BTRFS_FIRST_FREE_OBJECTID,
754 BTRFS_LAST_FREE_OBJECTID,
755 BTRFS_FREE_INO_OBJECTID);
758 if (key->offset != 0) {
759 inode_item_err(fs_info, leaf, slot,
760 "invalid key offset: has %llu expect 0",
764 iitem = btrfs_item_ptr(leaf, slot, struct btrfs_inode_item);
766 /* Here we use super block generation + 1 to handle log tree */
767 if (btrfs_inode_generation(leaf, iitem) > super_gen + 1) {
768 inode_item_err(fs_info, leaf, slot,
769 "invalid inode generation: has %llu expect (0, %llu]",
770 btrfs_inode_generation(leaf, iitem),
774 /* Note for ROOT_TREE_DIR_ITEM, mkfs could set its transid 0 */
775 if (btrfs_inode_transid(leaf, iitem) > super_gen + 1) {
776 inode_item_err(fs_info, leaf, slot,
777 "invalid inode generation: has %llu expect [0, %llu]",
778 btrfs_inode_transid(leaf, iitem), super_gen + 1);
783 * For size and nbytes it's better not to be too strict, as for dir
784 * item its size/nbytes can easily get wrong, but doesn't affect
785 * anything in the fs. So here we skip the check.
787 mode = btrfs_inode_mode(leaf, iitem);
788 if (mode & ~valid_mask) {
789 inode_item_err(fs_info, leaf, slot,
790 "unknown mode bit detected: 0x%x",
796 * S_IFMT is not bit mapped so we can't completely rely on is_power_of_2,
797 * but is_power_of_2() can save us from checking FIFO/CHR/DIR/REG.
798 * Only needs to check BLK, LNK and SOCKS
800 if (!is_power_of_2(mode & S_IFMT)) {
801 if (!S_ISLNK(mode) && !S_ISBLK(mode) && !S_ISSOCK(mode)) {
802 inode_item_err(fs_info, leaf, slot,
803 "invalid mode: has 0%o expect valid S_IF* bit(s)",
808 if (S_ISDIR(mode) && btrfs_inode_nlink(leaf, iitem) > 1) {
809 inode_item_err(fs_info, leaf, slot,
810 "invalid nlink: has %u expect no more than 1 for dir",
811 btrfs_inode_nlink(leaf, iitem));
814 if (btrfs_inode_flags(leaf, iitem) & ~BTRFS_INODE_FLAG_MASK) {
815 inode_item_err(fs_info, leaf, slot,
816 "unknown flags detected: 0x%llx",
817 btrfs_inode_flags(leaf, iitem) &
818 ~BTRFS_INODE_FLAG_MASK);
824 static int check_root_item(struct extent_buffer *leaf, struct btrfs_key *key,
827 struct btrfs_fs_info *fs_info = leaf->fs_info;
828 struct btrfs_root_item ri;
829 const u64 valid_root_flags = BTRFS_ROOT_SUBVOL_RDONLY |
830 BTRFS_ROOT_SUBVOL_DEAD;
832 /* No such tree id */
833 if (key->objectid == 0) {
834 generic_err(leaf, slot, "invalid root id 0");
839 * Some older kernel may create ROOT_ITEM with non-zero offset, so here
840 * we only check offset for reloc tree whose key->offset must be a
843 if (key->objectid == BTRFS_TREE_RELOC_OBJECTID && key->offset == 0) {
844 generic_err(leaf, slot, "invalid root id 0 for reloc tree");
848 if (btrfs_item_size_nr(leaf, slot) != sizeof(ri)) {
849 generic_err(leaf, slot,
850 "invalid root item size, have %u expect %zu",
851 btrfs_item_size_nr(leaf, slot), sizeof(ri));
854 read_extent_buffer(leaf, &ri, btrfs_item_ptr_offset(leaf, slot),
857 /* Generation related */
858 if (btrfs_root_generation(&ri) >
859 btrfs_super_generation(fs_info->super_copy) + 1) {
860 generic_err(leaf, slot,
861 "invalid root generation, have %llu expect (0, %llu]",
862 btrfs_root_generation(&ri),
863 btrfs_super_generation(fs_info->super_copy) + 1);
866 if (btrfs_root_generation_v2(&ri) >
867 btrfs_super_generation(fs_info->super_copy) + 1) {
868 generic_err(leaf, slot,
869 "invalid root v2 generation, have %llu expect (0, %llu]",
870 btrfs_root_generation_v2(&ri),
871 btrfs_super_generation(fs_info->super_copy) + 1);
874 if (btrfs_root_last_snapshot(&ri) >
875 btrfs_super_generation(fs_info->super_copy) + 1) {
876 generic_err(leaf, slot,
877 "invalid root last_snapshot, have %llu expect (0, %llu]",
878 btrfs_root_last_snapshot(&ri),
879 btrfs_super_generation(fs_info->super_copy) + 1);
883 /* Alignment and level check */
884 if (!IS_ALIGNED(btrfs_root_bytenr(&ri), fs_info->sectorsize)) {
885 generic_err(leaf, slot,
886 "invalid root bytenr, have %llu expect to be aligned to %u",
887 btrfs_root_bytenr(&ri), fs_info->sectorsize);
890 if (btrfs_root_level(&ri) >= BTRFS_MAX_LEVEL) {
891 generic_err(leaf, slot,
892 "invalid root level, have %u expect [0, %u]",
893 btrfs_root_level(&ri), BTRFS_MAX_LEVEL - 1);
896 if (ri.drop_level >= BTRFS_MAX_LEVEL) {
897 generic_err(leaf, slot,
898 "invalid root level, have %u expect [0, %u]",
899 ri.drop_level, BTRFS_MAX_LEVEL - 1);
904 if (btrfs_root_flags(&ri) & ~valid_root_flags) {
905 generic_err(leaf, slot,
906 "invalid root flags, have 0x%llx expect mask 0x%llx",
907 btrfs_root_flags(&ri), valid_root_flags);
915 static void extent_err(const struct extent_buffer *eb, int slot,
916 const char *fmt, ...)
918 struct btrfs_key key;
919 struct va_format vaf;
924 btrfs_item_key_to_cpu(eb, &key, slot);
925 bytenr = key.objectid;
926 if (key.type == BTRFS_METADATA_ITEM_KEY)
927 len = eb->fs_info->nodesize;
935 btrfs_crit(eb->fs_info,
936 "corrupt %s: block=%llu slot=%d extent bytenr=%llu len=%llu %pV",
937 btrfs_header_level(eb) == 0 ? "leaf" : "node",
938 eb->start, slot, bytenr, len, &vaf);
942 static int check_extent_item(struct extent_buffer *leaf,
943 struct btrfs_key *key, int slot)
945 struct btrfs_fs_info *fs_info = leaf->fs_info;
946 struct btrfs_extent_item *ei;
947 bool is_tree_block = false;
948 unsigned long ptr; /* Current pointer inside inline refs */
949 unsigned long end; /* Extent item end */
950 const u32 item_size = btrfs_item_size_nr(leaf, slot);
953 u64 total_refs; /* Total refs in btrfs_extent_item */
954 u64 inline_refs = 0; /* found total inline refs */
956 if (key->type == BTRFS_METADATA_ITEM_KEY &&
957 !btrfs_fs_incompat(fs_info, SKINNY_METADATA)) {
958 generic_err(leaf, slot,
959 "invalid key type, METADATA_ITEM type invalid when SKINNY_METADATA feature disabled");
962 /* key->objectid is the bytenr for both key types */
963 if (!IS_ALIGNED(key->objectid, fs_info->sectorsize)) {
964 generic_err(leaf, slot,
965 "invalid key objectid, have %llu expect to be aligned to %u",
966 key->objectid, fs_info->sectorsize);
970 /* key->offset is tree level for METADATA_ITEM_KEY */
971 if (key->type == BTRFS_METADATA_ITEM_KEY &&
972 key->offset >= BTRFS_MAX_LEVEL) {
973 extent_err(leaf, slot,
974 "invalid tree level, have %llu expect [0, %u]",
975 key->offset, BTRFS_MAX_LEVEL - 1);
980 * EXTENT/METADATA_ITEM consists of:
981 * 1) One btrfs_extent_item
982 * Records the total refs, type and generation of the extent.
984 * 2) One btrfs_tree_block_info (for EXTENT_ITEM and tree backref only)
985 * Records the first key and level of the tree block.
987 * 2) Zero or more btrfs_extent_inline_ref(s)
988 * Each inline ref has one btrfs_extent_inline_ref shows:
989 * 2.1) The ref type, one of the 4
990 * TREE_BLOCK_REF Tree block only
991 * SHARED_BLOCK_REF Tree block only
992 * EXTENT_DATA_REF Data only
993 * SHARED_DATA_REF Data only
994 * 2.2) Ref type specific data
995 * Either using btrfs_extent_inline_ref::offset, or specific
998 if (item_size < sizeof(*ei)) {
999 extent_err(leaf, slot,
1000 "invalid item size, have %u expect [%zu, %u)",
1001 item_size, sizeof(*ei),
1002 BTRFS_LEAF_DATA_SIZE(fs_info));
1005 end = item_size + btrfs_item_ptr_offset(leaf, slot);
1007 /* Checks against extent_item */
1008 ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
1009 flags = btrfs_extent_flags(leaf, ei);
1010 total_refs = btrfs_extent_refs(leaf, ei);
1011 generation = btrfs_extent_generation(leaf, ei);
1012 if (generation > btrfs_super_generation(fs_info->super_copy) + 1) {
1013 extent_err(leaf, slot,
1014 "invalid generation, have %llu expect (0, %llu]",
1016 btrfs_super_generation(fs_info->super_copy) + 1);
1019 if (!is_power_of_2(flags & (BTRFS_EXTENT_FLAG_DATA |
1020 BTRFS_EXTENT_FLAG_TREE_BLOCK))) {
1021 extent_err(leaf, slot,
1022 "invalid extent flag, have 0x%llx expect 1 bit set in 0x%llx",
1023 flags, BTRFS_EXTENT_FLAG_DATA |
1024 BTRFS_EXTENT_FLAG_TREE_BLOCK);
1027 is_tree_block = !!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK);
1028 if (is_tree_block) {
1029 if (key->type == BTRFS_EXTENT_ITEM_KEY &&
1030 key->offset != fs_info->nodesize) {
1031 extent_err(leaf, slot,
1032 "invalid extent length, have %llu expect %u",
1033 key->offset, fs_info->nodesize);
1037 if (key->type != BTRFS_EXTENT_ITEM_KEY) {
1038 extent_err(leaf, slot,
1039 "invalid key type, have %u expect %u for data backref",
1040 key->type, BTRFS_EXTENT_ITEM_KEY);
1043 if (!IS_ALIGNED(key->offset, fs_info->sectorsize)) {
1044 extent_err(leaf, slot,
1045 "invalid extent length, have %llu expect aligned to %u",
1046 key->offset, fs_info->sectorsize);
1050 ptr = (unsigned long)(struct btrfs_extent_item *)(ei + 1);
1052 /* Check the special case of btrfs_tree_block_info */
1053 if (is_tree_block && key->type != BTRFS_METADATA_ITEM_KEY) {
1054 struct btrfs_tree_block_info *info;
1056 info = (struct btrfs_tree_block_info *)ptr;
1057 if (btrfs_tree_block_level(leaf, info) >= BTRFS_MAX_LEVEL) {
1058 extent_err(leaf, slot,
1059 "invalid tree block info level, have %u expect [0, %u]",
1060 btrfs_tree_block_level(leaf, info),
1061 BTRFS_MAX_LEVEL - 1);
1064 ptr = (unsigned long)(struct btrfs_tree_block_info *)(info + 1);
1067 /* Check inline refs */
1069 struct btrfs_extent_inline_ref *iref;
1070 struct btrfs_extent_data_ref *dref;
1071 struct btrfs_shared_data_ref *sref;
1076 if (ptr + sizeof(*iref) > end) {
1077 extent_err(leaf, slot,
1078 "inline ref item overflows extent item, ptr %lu iref size %zu end %lu",
1079 ptr, sizeof(*iref), end);
1082 iref = (struct btrfs_extent_inline_ref *)ptr;
1083 inline_type = btrfs_extent_inline_ref_type(leaf, iref);
1084 inline_offset = btrfs_extent_inline_ref_offset(leaf, iref);
1085 if (ptr + btrfs_extent_inline_ref_size(inline_type) > end) {
1086 extent_err(leaf, slot,
1087 "inline ref item overflows extent item, ptr %lu iref size %u end %lu",
1088 ptr, inline_type, end);
1092 switch (inline_type) {
1093 /* inline_offset is subvolid of the owner, no need to check */
1094 case BTRFS_TREE_BLOCK_REF_KEY:
1097 /* Contains parent bytenr */
1098 case BTRFS_SHARED_BLOCK_REF_KEY:
1099 if (!IS_ALIGNED(inline_offset, fs_info->sectorsize)) {
1100 extent_err(leaf, slot,
1101 "invalid tree parent bytenr, have %llu expect aligned to %u",
1102 inline_offset, fs_info->sectorsize);
1108 * Contains owner subvolid, owner key objectid, adjusted offset.
1109 * The only obvious corruption can happen in that offset.
1111 case BTRFS_EXTENT_DATA_REF_KEY:
1112 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1113 dref_offset = btrfs_extent_data_ref_offset(leaf, dref);
1114 if (!IS_ALIGNED(dref_offset, fs_info->sectorsize)) {
1115 extent_err(leaf, slot,
1116 "invalid data ref offset, have %llu expect aligned to %u",
1117 dref_offset, fs_info->sectorsize);
1120 inline_refs += btrfs_extent_data_ref_count(leaf, dref);
1122 /* Contains parent bytenr and ref count */
1123 case BTRFS_SHARED_DATA_REF_KEY:
1124 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1125 if (!IS_ALIGNED(inline_offset, fs_info->sectorsize)) {
1126 extent_err(leaf, slot,
1127 "invalid data parent bytenr, have %llu expect aligned to %u",
1128 inline_offset, fs_info->sectorsize);
1131 inline_refs += btrfs_shared_data_ref_count(leaf, sref);
1134 extent_err(leaf, slot, "unknown inline ref type: %u",
1138 ptr += btrfs_extent_inline_ref_size(inline_type);
1140 /* No padding is allowed */
1142 extent_err(leaf, slot,
1143 "invalid extent item size, padding bytes found");
1147 /* Finally, check the inline refs against total refs */
1148 if (inline_refs > total_refs) {
1149 extent_err(leaf, slot,
1150 "invalid extent refs, have %llu expect >= inline %llu",
1151 total_refs, inline_refs);
1158 * Common point to switch the item-specific validation.
1160 static int check_leaf_item(struct extent_buffer *leaf,
1161 struct btrfs_key *key, int slot,
1162 struct btrfs_key *prev_key)
1165 struct btrfs_chunk *chunk;
1167 switch (key->type) {
1168 case BTRFS_EXTENT_DATA_KEY:
1169 ret = check_extent_data_item(leaf, key, slot, prev_key);
1171 case BTRFS_EXTENT_CSUM_KEY:
1172 ret = check_csum_item(leaf, key, slot);
1174 case BTRFS_DIR_ITEM_KEY:
1175 case BTRFS_DIR_INDEX_KEY:
1176 case BTRFS_XATTR_ITEM_KEY:
1177 ret = check_dir_item(leaf, key, slot);
1179 case BTRFS_BLOCK_GROUP_ITEM_KEY:
1180 ret = check_block_group_item(leaf, key, slot);
1182 case BTRFS_CHUNK_ITEM_KEY:
1183 chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
1184 ret = btrfs_check_chunk_valid(leaf, chunk, key->offset);
1186 case BTRFS_DEV_ITEM_KEY:
1187 ret = check_dev_item(leaf, key, slot);
1189 case BTRFS_INODE_ITEM_KEY:
1190 ret = check_inode_item(leaf, key, slot);
1192 case BTRFS_ROOT_ITEM_KEY:
1193 ret = check_root_item(leaf, key, slot);
1195 case BTRFS_EXTENT_ITEM_KEY:
1196 case BTRFS_METADATA_ITEM_KEY:
1197 ret = check_extent_item(leaf, key, slot);
1203 static int check_leaf(struct extent_buffer *leaf, bool check_item_data)
1205 struct btrfs_fs_info *fs_info = leaf->fs_info;
1206 /* No valid key type is 0, so all key should be larger than this key */
1207 struct btrfs_key prev_key = {0, 0, 0};
1208 struct btrfs_key key;
1209 u32 nritems = btrfs_header_nritems(leaf);
1212 if (btrfs_header_level(leaf) != 0) {
1213 generic_err(leaf, 0,
1214 "invalid level for leaf, have %d expect 0",
1215 btrfs_header_level(leaf));
1220 * Extent buffers from a relocation tree have a owner field that
1221 * corresponds to the subvolume tree they are based on. So just from an
1222 * extent buffer alone we can not find out what is the id of the
1223 * corresponding subvolume tree, so we can not figure out if the extent
1224 * buffer corresponds to the root of the relocation tree or not. So
1225 * skip this check for relocation trees.
1227 if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) {
1228 u64 owner = btrfs_header_owner(leaf);
1230 /* These trees must never be empty */
1231 if (owner == BTRFS_ROOT_TREE_OBJECTID ||
1232 owner == BTRFS_CHUNK_TREE_OBJECTID ||
1233 owner == BTRFS_EXTENT_TREE_OBJECTID ||
1234 owner == BTRFS_DEV_TREE_OBJECTID ||
1235 owner == BTRFS_FS_TREE_OBJECTID ||
1236 owner == BTRFS_DATA_RELOC_TREE_OBJECTID) {
1237 generic_err(leaf, 0,
1238 "invalid root, root %llu must never be empty",
1249 * Check the following things to make sure this is a good leaf, and
1250 * leaf users won't need to bother with similar sanity checks:
1253 * 2) item offset and size
1254 * No overlap, no hole, all inside the leaf.
1256 * If possible, do comprehensive sanity check.
1257 * NOTE: All checks must only rely on the item data itself.
1259 for (slot = 0; slot < nritems; slot++) {
1260 u32 item_end_expected;
1263 btrfs_item_key_to_cpu(leaf, &key, slot);
1265 /* Make sure the keys are in the right order */
1266 if (btrfs_comp_cpu_keys(&prev_key, &key) >= 0) {
1267 generic_err(leaf, slot,
1268 "bad key order, prev (%llu %u %llu) current (%llu %u %llu)",
1269 prev_key.objectid, prev_key.type,
1270 prev_key.offset, key.objectid, key.type,
1276 * Make sure the offset and ends are right, remember that the
1277 * item data starts at the end of the leaf and grows towards the
1281 item_end_expected = BTRFS_LEAF_DATA_SIZE(fs_info);
1283 item_end_expected = btrfs_item_offset_nr(leaf,
1285 if (btrfs_item_end_nr(leaf, slot) != item_end_expected) {
1286 generic_err(leaf, slot,
1287 "unexpected item end, have %u expect %u",
1288 btrfs_item_end_nr(leaf, slot),
1294 * Check to make sure that we don't point outside of the leaf,
1295 * just in case all the items are consistent to each other, but
1296 * all point outside of the leaf.
1298 if (btrfs_item_end_nr(leaf, slot) >
1299 BTRFS_LEAF_DATA_SIZE(fs_info)) {
1300 generic_err(leaf, slot,
1301 "slot end outside of leaf, have %u expect range [0, %u]",
1302 btrfs_item_end_nr(leaf, slot),
1303 BTRFS_LEAF_DATA_SIZE(fs_info));
1307 /* Also check if the item pointer overlaps with btrfs item. */
1308 if (btrfs_item_nr_offset(slot) + sizeof(struct btrfs_item) >
1309 btrfs_item_ptr_offset(leaf, slot)) {
1310 generic_err(leaf, slot,
1311 "slot overlaps with its data, item end %lu data start %lu",
1312 btrfs_item_nr_offset(slot) +
1313 sizeof(struct btrfs_item),
1314 btrfs_item_ptr_offset(leaf, slot));
1318 if (check_item_data) {
1320 * Check if the item size and content meet other
1323 ret = check_leaf_item(leaf, &key, slot, &prev_key);
1328 prev_key.objectid = key.objectid;
1329 prev_key.type = key.type;
1330 prev_key.offset = key.offset;
1336 int btrfs_check_leaf_full(struct extent_buffer *leaf)
1338 return check_leaf(leaf, true);
1340 ALLOW_ERROR_INJECTION(btrfs_check_leaf_full, ERRNO);
1342 int btrfs_check_leaf_relaxed(struct extent_buffer *leaf)
1344 return check_leaf(leaf, false);
1347 int btrfs_check_node(struct extent_buffer *node)
1349 struct btrfs_fs_info *fs_info = node->fs_info;
1350 unsigned long nr = btrfs_header_nritems(node);
1351 struct btrfs_key key, next_key;
1353 int level = btrfs_header_level(node);
1357 if (level <= 0 || level >= BTRFS_MAX_LEVEL) {
1358 generic_err(node, 0,
1359 "invalid level for node, have %d expect [1, %d]",
1360 level, BTRFS_MAX_LEVEL - 1);
1363 if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(fs_info)) {
1365 "corrupt node: root=%llu block=%llu, nritems too %s, have %lu expect range [1,%u]",
1366 btrfs_header_owner(node), node->start,
1367 nr == 0 ? "small" : "large", nr,
1368 BTRFS_NODEPTRS_PER_BLOCK(fs_info));
1372 for (slot = 0; slot < nr - 1; slot++) {
1373 bytenr = btrfs_node_blockptr(node, slot);
1374 btrfs_node_key_to_cpu(node, &key, slot);
1375 btrfs_node_key_to_cpu(node, &next_key, slot + 1);
1378 generic_err(node, slot,
1379 "invalid NULL node pointer");
1383 if (!IS_ALIGNED(bytenr, fs_info->sectorsize)) {
1384 generic_err(node, slot,
1385 "unaligned pointer, have %llu should be aligned to %u",
1386 bytenr, fs_info->sectorsize);
1391 if (btrfs_comp_cpu_keys(&key, &next_key) >= 0) {
1392 generic_err(node, slot,
1393 "bad key order, current (%llu %u %llu) next (%llu %u %llu)",
1394 key.objectid, key.type, key.offset,
1395 next_key.objectid, next_key.type,
1404 ALLOW_ERROR_INJECTION(btrfs_check_node, ERRNO);