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.
19 #include "tree-checker.h"
21 #include "compression.h"
25 * Error message should follow the following format:
26 * corrupt <type>: <identifier>, <reason>[, <bad_value>]
29 * @identifier: the necessary info to locate the leaf/node.
30 * It's recommened to decode key.objecitd/offset if it's
32 * @reason: describe the error
33 * @bad_value: optional, it's recommened to output bad value and its
34 * expected value (range).
36 * Since comma is used to separate the components, only space is allowed
37 * inside each component.
41 * Append generic "corrupt leaf/node root=%llu block=%llu slot=%d: " to @fmt.
42 * Allows callers to customize the output.
46 static void generic_err(const struct btrfs_fs_info *fs_info,
47 const struct extent_buffer *eb, int slot,
59 "corrupt %s: root=%llu block=%llu slot=%d, %pV",
60 btrfs_header_level(eb) == 0 ? "leaf" : "node",
61 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, &vaf);
66 * Customized reporter for extent data item, since its key objectid and
67 * offset has its own meaning.
71 static void file_extent_err(const struct btrfs_fs_info *fs_info,
72 const struct extent_buffer *eb, int slot,
79 btrfs_item_key_to_cpu(eb, &key, slot);
86 "corrupt %s: root=%llu block=%llu slot=%d ino=%llu file_offset=%llu, %pV",
87 btrfs_header_level(eb) == 0 ? "leaf" : "node",
88 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
89 key.objectid, key.offset, &vaf);
94 * Return 0 if the btrfs_file_extent_##name is aligned to @alignment
97 #define CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, name, alignment) \
99 if (!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment))) \
100 file_extent_err((fs_info), (leaf), (slot), \
101 "invalid %s for file extent, have %llu, should be aligned to %u", \
102 (#name), btrfs_file_extent_##name((leaf), (fi)), \
104 (!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment))); \
107 static int check_extent_data_item(struct btrfs_fs_info *fs_info,
108 struct extent_buffer *leaf,
109 struct btrfs_key *key, int slot)
111 struct btrfs_file_extent_item *fi;
112 u32 sectorsize = fs_info->sectorsize;
113 u32 item_size = btrfs_item_size_nr(leaf, slot);
115 if (!IS_ALIGNED(key->offset, sectorsize)) {
116 file_extent_err(fs_info, leaf, slot,
117 "unaligned file_offset for file extent, have %llu should be aligned to %u",
118 key->offset, sectorsize);
122 fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
124 if (btrfs_file_extent_type(leaf, fi) > BTRFS_FILE_EXTENT_TYPES) {
125 file_extent_err(fs_info, leaf, slot,
126 "invalid type for file extent, have %u expect range [0, %u]",
127 btrfs_file_extent_type(leaf, fi),
128 BTRFS_FILE_EXTENT_TYPES);
133 * Support for new compression/encrption must introduce incompat flag,
134 * and must be caught in open_ctree().
136 if (btrfs_file_extent_compression(leaf, fi) > BTRFS_COMPRESS_TYPES) {
137 file_extent_err(fs_info, leaf, slot,
138 "invalid compression for file extent, have %u expect range [0, %u]",
139 btrfs_file_extent_compression(leaf, fi),
140 BTRFS_COMPRESS_TYPES);
143 if (btrfs_file_extent_encryption(leaf, fi)) {
144 file_extent_err(fs_info, leaf, slot,
145 "invalid encryption for file extent, have %u expect 0",
146 btrfs_file_extent_encryption(leaf, fi));
149 if (btrfs_file_extent_type(leaf, fi) == BTRFS_FILE_EXTENT_INLINE) {
150 /* Inline extent must have 0 as key offset */
152 file_extent_err(fs_info, leaf, slot,
153 "invalid file_offset for inline file extent, have %llu expect 0",
158 /* Compressed inline extent has no on-disk size, skip it */
159 if (btrfs_file_extent_compression(leaf, fi) !=
163 /* Uncompressed inline extent size must match item size */
164 if (item_size != BTRFS_FILE_EXTENT_INLINE_DATA_START +
165 btrfs_file_extent_ram_bytes(leaf, fi)) {
166 file_extent_err(fs_info, leaf, slot,
167 "invalid ram_bytes for uncompressed inline extent, have %u expect %llu",
168 item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START +
169 btrfs_file_extent_ram_bytes(leaf, fi));
175 /* Regular or preallocated extent has fixed item size */
176 if (item_size != sizeof(*fi)) {
177 file_extent_err(fs_info, leaf, slot,
178 "invalid item size for reg/prealloc file extent, have %u expect %zu",
179 item_size, sizeof(*fi));
182 if (CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, ram_bytes, sectorsize) ||
183 CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, disk_bytenr, sectorsize) ||
184 CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, disk_num_bytes, sectorsize) ||
185 CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, offset, sectorsize) ||
186 CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, num_bytes, sectorsize))
191 static int check_csum_item(struct btrfs_fs_info *fs_info,
192 struct extent_buffer *leaf, struct btrfs_key *key,
195 u32 sectorsize = fs_info->sectorsize;
196 u32 csumsize = btrfs_super_csum_size(fs_info->super_copy);
198 if (key->objectid != BTRFS_EXTENT_CSUM_OBJECTID) {
199 generic_err(fs_info, leaf, slot,
200 "invalid key objectid for csum item, have %llu expect %llu",
201 key->objectid, BTRFS_EXTENT_CSUM_OBJECTID);
204 if (!IS_ALIGNED(key->offset, sectorsize)) {
205 generic_err(fs_info, leaf, slot,
206 "unaligned key offset for csum item, have %llu should be aligned to %u",
207 key->offset, sectorsize);
210 if (!IS_ALIGNED(btrfs_item_size_nr(leaf, slot), csumsize)) {
211 generic_err(fs_info, leaf, slot,
212 "unaligned item size for csum item, have %u should be aligned to %u",
213 btrfs_item_size_nr(leaf, slot), csumsize);
220 * Customized reported for dir_item, only important new info is key->objectid,
221 * which represents inode number
225 static void dir_item_err(const struct btrfs_fs_info *fs_info,
226 const struct extent_buffer *eb, int slot,
227 const char *fmt, ...)
229 struct btrfs_key key;
230 struct va_format vaf;
233 btrfs_item_key_to_cpu(eb, &key, slot);
240 "corrupt %s: root=%llu block=%llu slot=%d ino=%llu, %pV",
241 btrfs_header_level(eb) == 0 ? "leaf" : "node",
242 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
247 static int check_dir_item(struct btrfs_fs_info *fs_info,
248 struct extent_buffer *leaf,
249 struct btrfs_key *key, int slot)
251 struct btrfs_dir_item *di;
252 u32 item_size = btrfs_item_size_nr(leaf, slot);
255 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
256 while (cur < item_size) {
264 /* header itself should not cross item boundary */
265 if (cur + sizeof(*di) > item_size) {
266 dir_item_err(fs_info, leaf, slot,
267 "dir item header crosses item boundary, have %zu boundary %u",
268 cur + sizeof(*di), item_size);
273 dir_type = btrfs_dir_type(leaf, di);
274 if (dir_type >= BTRFS_FT_MAX) {
275 dir_item_err(fs_info, leaf, slot,
276 "invalid dir item type, have %u expect [0, %u)",
277 dir_type, BTRFS_FT_MAX);
281 if (key->type == BTRFS_XATTR_ITEM_KEY &&
282 dir_type != BTRFS_FT_XATTR) {
283 dir_item_err(fs_info, leaf, slot,
284 "invalid dir item type for XATTR key, have %u expect %u",
285 dir_type, BTRFS_FT_XATTR);
288 if (dir_type == BTRFS_FT_XATTR &&
289 key->type != BTRFS_XATTR_ITEM_KEY) {
290 dir_item_err(fs_info, leaf, slot,
291 "xattr dir type found for non-XATTR key");
294 if (dir_type == BTRFS_FT_XATTR)
295 max_name_len = XATTR_NAME_MAX;
297 max_name_len = BTRFS_NAME_LEN;
299 /* Name/data length check */
300 name_len = btrfs_dir_name_len(leaf, di);
301 data_len = btrfs_dir_data_len(leaf, di);
302 if (name_len > max_name_len) {
303 dir_item_err(fs_info, leaf, slot,
304 "dir item name len too long, have %u max %u",
305 name_len, max_name_len);
308 if (name_len + data_len > BTRFS_MAX_XATTR_SIZE(fs_info)) {
309 dir_item_err(fs_info, leaf, slot,
310 "dir item name and data len too long, have %u max %u",
312 BTRFS_MAX_XATTR_SIZE(fs_info));
316 if (data_len && dir_type != BTRFS_FT_XATTR) {
317 dir_item_err(fs_info, leaf, slot,
318 "dir item with invalid data len, have %u expect 0",
323 total_size = sizeof(*di) + name_len + data_len;
325 /* header and name/data should not cross item boundary */
326 if (cur + total_size > item_size) {
327 dir_item_err(fs_info, leaf, slot,
328 "dir item data crosses item boundary, have %u boundary %u",
329 cur + total_size, item_size);
334 * Special check for XATTR/DIR_ITEM, as key->offset is name
335 * hash, should match its name
337 if (key->type == BTRFS_DIR_ITEM_KEY ||
338 key->type == BTRFS_XATTR_ITEM_KEY) {
339 char namebuf[max(BTRFS_NAME_LEN, XATTR_NAME_MAX)];
341 read_extent_buffer(leaf, namebuf,
342 (unsigned long)(di + 1), name_len);
343 name_hash = btrfs_name_hash(namebuf, name_len);
344 if (key->offset != name_hash) {
345 dir_item_err(fs_info, leaf, slot,
346 "name hash mismatch with key, have 0x%016x expect 0x%016llx",
347 name_hash, key->offset);
352 di = (struct btrfs_dir_item *)((void *)di + total_size);
359 static void block_group_err(const struct btrfs_fs_info *fs_info,
360 const struct extent_buffer *eb, int slot,
361 const char *fmt, ...)
363 struct btrfs_key key;
364 struct va_format vaf;
367 btrfs_item_key_to_cpu(eb, &key, slot);
374 "corrupt %s: root=%llu block=%llu slot=%d bg_start=%llu bg_len=%llu, %pV",
375 btrfs_header_level(eb) == 0 ? "leaf" : "node",
376 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
377 key.objectid, key.offset, &vaf);
381 static int check_block_group_item(struct btrfs_fs_info *fs_info,
382 struct extent_buffer *leaf,
383 struct btrfs_key *key, int slot)
385 struct btrfs_block_group_item bgi;
386 u32 item_size = btrfs_item_size_nr(leaf, slot);
391 * Here we don't really care about alignment since extent allocator can
392 * handle it. We care more about the size, as if one block group is
393 * larger than maximum size, it's must be some obvious corruption.
395 if (key->offset > BTRFS_MAX_DATA_CHUNK_SIZE || key->offset == 0) {
396 block_group_err(fs_info, leaf, slot,
397 "invalid block group size, have %llu expect (0, %llu]",
398 key->offset, BTRFS_MAX_DATA_CHUNK_SIZE);
402 if (item_size != sizeof(bgi)) {
403 block_group_err(fs_info, leaf, slot,
404 "invalid item size, have %u expect %zu",
405 item_size, sizeof(bgi));
409 read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot),
411 if (btrfs_block_group_chunk_objectid(&bgi) !=
412 BTRFS_FIRST_CHUNK_TREE_OBJECTID) {
413 block_group_err(fs_info, leaf, slot,
414 "invalid block group chunk objectid, have %llu expect %llu",
415 btrfs_block_group_chunk_objectid(&bgi),
416 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
420 if (btrfs_block_group_used(&bgi) > key->offset) {
421 block_group_err(fs_info, leaf, slot,
422 "invalid block group used, have %llu expect [0, %llu)",
423 btrfs_block_group_used(&bgi), key->offset);
427 flags = btrfs_block_group_flags(&bgi);
428 if (hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) > 1) {
429 block_group_err(fs_info, leaf, slot,
430 "invalid profile flags, have 0x%llx (%lu bits set) expect no more than 1 bit set",
431 flags & BTRFS_BLOCK_GROUP_PROFILE_MASK,
432 hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK));
436 type = flags & BTRFS_BLOCK_GROUP_TYPE_MASK;
437 if (type != BTRFS_BLOCK_GROUP_DATA &&
438 type != BTRFS_BLOCK_GROUP_METADATA &&
439 type != BTRFS_BLOCK_GROUP_SYSTEM &&
440 type != (BTRFS_BLOCK_GROUP_METADATA |
441 BTRFS_BLOCK_GROUP_DATA)) {
442 block_group_err(fs_info, leaf, slot,
443 "invalid type, have 0x%llx (%lu bits set) expect either 0x%llx, 0x%llx, 0x%llu or 0x%llx",
444 type, hweight64(type),
445 BTRFS_BLOCK_GROUP_DATA, BTRFS_BLOCK_GROUP_METADATA,
446 BTRFS_BLOCK_GROUP_SYSTEM,
447 BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA);
454 * Common point to switch the item-specific validation.
456 static int check_leaf_item(struct btrfs_fs_info *fs_info,
457 struct extent_buffer *leaf,
458 struct btrfs_key *key, int slot)
463 case BTRFS_EXTENT_DATA_KEY:
464 ret = check_extent_data_item(fs_info, leaf, key, slot);
466 case BTRFS_EXTENT_CSUM_KEY:
467 ret = check_csum_item(fs_info, leaf, key, slot);
469 case BTRFS_DIR_ITEM_KEY:
470 case BTRFS_DIR_INDEX_KEY:
471 case BTRFS_XATTR_ITEM_KEY:
472 ret = check_dir_item(fs_info, leaf, key, slot);
474 case BTRFS_BLOCK_GROUP_ITEM_KEY:
475 ret = check_block_group_item(fs_info, leaf, key, slot);
481 static int check_leaf(struct btrfs_fs_info *fs_info, struct extent_buffer *leaf,
482 bool check_item_data)
484 /* No valid key type is 0, so all key should be larger than this key */
485 struct btrfs_key prev_key = {0, 0, 0};
486 struct btrfs_key key;
487 u32 nritems = btrfs_header_nritems(leaf);
491 * Extent buffers from a relocation tree have a owner field that
492 * corresponds to the subvolume tree they are based on. So just from an
493 * extent buffer alone we can not find out what is the id of the
494 * corresponding subvolume tree, so we can not figure out if the extent
495 * buffer corresponds to the root of the relocation tree or not. So
496 * skip this check for relocation trees.
498 if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) {
499 u64 owner = btrfs_header_owner(leaf);
500 struct btrfs_root *check_root;
502 /* These trees must never be empty */
503 if (owner == BTRFS_ROOT_TREE_OBJECTID ||
504 owner == BTRFS_CHUNK_TREE_OBJECTID ||
505 owner == BTRFS_EXTENT_TREE_OBJECTID ||
506 owner == BTRFS_DEV_TREE_OBJECTID ||
507 owner == BTRFS_FS_TREE_OBJECTID ||
508 owner == BTRFS_DATA_RELOC_TREE_OBJECTID) {
509 generic_err(fs_info, leaf, 0,
510 "invalid root, root %llu must never be empty",
514 key.objectid = owner;
515 key.type = BTRFS_ROOT_ITEM_KEY;
516 key.offset = (u64)-1;
518 check_root = btrfs_get_fs_root(fs_info, &key, false);
520 * The only reason we also check NULL here is that during
521 * open_ctree() some roots has not yet been set up.
523 if (!IS_ERR_OR_NULL(check_root)) {
524 struct extent_buffer *eb;
526 eb = btrfs_root_node(check_root);
527 /* if leaf is the root, then it's fine */
529 generic_err(fs_info, leaf, 0,
530 "invalid nritems, have %u should not be 0 for non-root leaf",
532 free_extent_buffer(eb);
535 free_extent_buffer(eb);
544 * Check the following things to make sure this is a good leaf, and
545 * leaf users won't need to bother with similar sanity checks:
548 * 2) item offset and size
549 * No overlap, no hole, all inside the leaf.
551 * If possible, do comprehensive sanity check.
552 * NOTE: All checks must only rely on the item data itself.
554 for (slot = 0; slot < nritems; slot++) {
555 u32 item_end_expected;
558 btrfs_item_key_to_cpu(leaf, &key, slot);
560 /* Make sure the keys are in the right order */
561 if (btrfs_comp_cpu_keys(&prev_key, &key) >= 0) {
562 generic_err(fs_info, leaf, slot,
563 "bad key order, prev (%llu %u %llu) current (%llu %u %llu)",
564 prev_key.objectid, prev_key.type,
565 prev_key.offset, key.objectid, key.type,
571 * Make sure the offset and ends are right, remember that the
572 * item data starts at the end of the leaf and grows towards the
576 item_end_expected = BTRFS_LEAF_DATA_SIZE(fs_info);
578 item_end_expected = btrfs_item_offset_nr(leaf,
580 if (btrfs_item_end_nr(leaf, slot) != item_end_expected) {
581 generic_err(fs_info, leaf, slot,
582 "unexpected item end, have %u expect %u",
583 btrfs_item_end_nr(leaf, slot),
589 * Check to make sure that we don't point outside of the leaf,
590 * just in case all the items are consistent to each other, but
591 * all point outside of the leaf.
593 if (btrfs_item_end_nr(leaf, slot) >
594 BTRFS_LEAF_DATA_SIZE(fs_info)) {
595 generic_err(fs_info, leaf, slot,
596 "slot end outside of leaf, have %u expect range [0, %u]",
597 btrfs_item_end_nr(leaf, slot),
598 BTRFS_LEAF_DATA_SIZE(fs_info));
602 /* Also check if the item pointer overlaps with btrfs item. */
603 if (btrfs_item_nr_offset(slot) + sizeof(struct btrfs_item) >
604 btrfs_item_ptr_offset(leaf, slot)) {
605 generic_err(fs_info, leaf, slot,
606 "slot overlaps with its data, item end %lu data start %lu",
607 btrfs_item_nr_offset(slot) +
608 sizeof(struct btrfs_item),
609 btrfs_item_ptr_offset(leaf, slot));
613 if (check_item_data) {
615 * Check if the item size and content meet other
618 ret = check_leaf_item(fs_info, leaf, &key, slot);
623 prev_key.objectid = key.objectid;
624 prev_key.type = key.type;
625 prev_key.offset = key.offset;
631 int btrfs_check_leaf_full(struct btrfs_fs_info *fs_info,
632 struct extent_buffer *leaf)
634 return check_leaf(fs_info, leaf, true);
637 int btrfs_check_leaf_relaxed(struct btrfs_fs_info *fs_info,
638 struct extent_buffer *leaf)
640 return check_leaf(fs_info, leaf, false);
643 int btrfs_check_node(struct btrfs_fs_info *fs_info, struct extent_buffer *node)
645 unsigned long nr = btrfs_header_nritems(node);
646 struct btrfs_key key, next_key;
651 if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(fs_info)) {
653 "corrupt node: root=%llu block=%llu, nritems too %s, have %lu expect range [1,%u]",
654 btrfs_header_owner(node), node->start,
655 nr == 0 ? "small" : "large", nr,
656 BTRFS_NODEPTRS_PER_BLOCK(fs_info));
660 for (slot = 0; slot < nr - 1; slot++) {
661 bytenr = btrfs_node_blockptr(node, slot);
662 btrfs_node_key_to_cpu(node, &key, slot);
663 btrfs_node_key_to_cpu(node, &next_key, slot + 1);
666 generic_err(fs_info, node, slot,
667 "invalid NULL node pointer");
671 if (!IS_ALIGNED(bytenr, fs_info->sectorsize)) {
672 generic_err(fs_info, node, slot,
673 "unaligned pointer, have %llu should be aligned to %u",
674 bytenr, fs_info->sectorsize);
679 if (btrfs_comp_cpu_keys(&key, &next_key) >= 0) {
680 generic_err(fs_info, node, slot,
681 "bad key order, current (%llu %u %llu) next (%llu %u %llu)",
682 key.objectid, key.type, key.offset,
683 next_key.objectid, next_key.type,