2 * Copyright (C) Qu Wenruo 2017. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program.
18 * The module is used to catch unexpected/corrupted tree block data.
19 * Such behavior can be caused either by a fuzzed image or bugs.
21 * The objective is to do leaf/node validation checks when tree block is read
22 * from disk, and check *every* possible member, so other code won't
23 * need to checking them again.
25 * Due to the potential and unwanted damage, every checker needs to be
26 * carefully reviewed otherwise so it does not prevent mount of valid images.
30 #include "tree-checker.h"
32 #include "compression.h"
36 #define CORRUPT(reason, eb, root, slot) \
37 btrfs_crit(root->fs_info, \
38 "corrupt %s, %s: block=%llu, root=%llu, slot=%d", \
39 btrfs_header_level(eb) == 0 ? "leaf" : "node", \
40 reason, btrfs_header_bytenr(eb), root->objectid, slot)
43 * Error message should follow the following format:
44 * corrupt <type>: <identifier>, <reason>[, <bad_value>]
47 * @identifier: the necessary info to locate the leaf/node.
48 * It's recommened to decode key.objecitd/offset if it's
50 * @reason: describe the error
51 * @bad_value: optional, it's recommened to output bad value and its
52 * expected value (range).
54 * Since comma is used to separate the components, only space is allowed
55 * inside each component.
59 * Append generic "corrupt leaf/node root=%llu block=%llu slot=%d: " to @fmt.
60 * Allows callers to customize the output.
63 static void generic_err(const struct btrfs_root *root,
64 const struct extent_buffer *eb, int slot,
75 btrfs_crit(root->fs_info,
76 "corrupt %s: root=%llu block=%llu slot=%d, %pV",
77 btrfs_header_level(eb) == 0 ? "leaf" : "node",
78 root->objectid, btrfs_header_bytenr(eb), slot, &vaf);
82 static int check_extent_data_item(struct btrfs_root *root,
83 struct extent_buffer *leaf,
84 struct btrfs_key *key, int slot)
86 struct btrfs_file_extent_item *fi;
87 u32 sectorsize = root->sectorsize;
88 u32 item_size = btrfs_item_size_nr(leaf, slot);
90 if (!IS_ALIGNED(key->offset, sectorsize)) {
91 CORRUPT("unaligned key offset for file extent",
96 fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
98 if (btrfs_file_extent_type(leaf, fi) > BTRFS_FILE_EXTENT_TYPES) {
99 CORRUPT("invalid file extent type", leaf, root, slot);
104 * Support for new compression/encrption must introduce incompat flag,
105 * and must be caught in open_ctree().
107 if (btrfs_file_extent_compression(leaf, fi) > BTRFS_COMPRESS_TYPES) {
108 CORRUPT("invalid file extent compression", leaf, root, slot);
111 if (btrfs_file_extent_encryption(leaf, fi)) {
112 CORRUPT("invalid file extent encryption", leaf, root, slot);
115 if (btrfs_file_extent_type(leaf, fi) == BTRFS_FILE_EXTENT_INLINE) {
116 /* Inline extent must have 0 as key offset */
118 CORRUPT("inline extent has non-zero key offset",
123 /* Compressed inline extent has no on-disk size, skip it */
124 if (btrfs_file_extent_compression(leaf, fi) !=
128 /* Uncompressed inline extent size must match item size */
129 if (item_size != BTRFS_FILE_EXTENT_INLINE_DATA_START +
130 btrfs_file_extent_ram_bytes(leaf, fi)) {
131 CORRUPT("plaintext inline extent has invalid size",
138 /* Regular or preallocated extent has fixed item size */
139 if (item_size != sizeof(*fi)) {
141 "regluar or preallocated extent data item size is invalid",
145 if (!IS_ALIGNED(btrfs_file_extent_ram_bytes(leaf, fi), sectorsize) ||
146 !IS_ALIGNED(btrfs_file_extent_disk_bytenr(leaf, fi), sectorsize) ||
147 !IS_ALIGNED(btrfs_file_extent_disk_num_bytes(leaf, fi), sectorsize) ||
148 !IS_ALIGNED(btrfs_file_extent_offset(leaf, fi), sectorsize) ||
149 !IS_ALIGNED(btrfs_file_extent_num_bytes(leaf, fi), sectorsize)) {
151 "regular or preallocated extent data item has unaligned value",
159 static int check_csum_item(struct btrfs_root *root, struct extent_buffer *leaf,
160 struct btrfs_key *key, int slot)
162 u32 sectorsize = root->sectorsize;
163 u32 csumsize = btrfs_super_csum_size(root->fs_info->super_copy);
165 if (key->objectid != BTRFS_EXTENT_CSUM_OBJECTID) {
166 CORRUPT("invalid objectid for csum item", leaf, root, slot);
169 if (!IS_ALIGNED(key->offset, sectorsize)) {
170 CORRUPT("unaligned key offset for csum item", leaf, root, slot);
173 if (!IS_ALIGNED(btrfs_item_size_nr(leaf, slot), csumsize)) {
174 CORRUPT("unaligned csum item size", leaf, root, slot);
181 * Customized reported for dir_item, only important new info is key->objectid,
182 * which represents inode number
185 static void dir_item_err(const struct btrfs_root *root,
186 const struct extent_buffer *eb, int slot,
187 const char *fmt, ...)
189 struct btrfs_key key;
190 struct va_format vaf;
193 btrfs_item_key_to_cpu(eb, &key, slot);
199 btrfs_crit(root->fs_info,
200 "corrupt %s: root=%llu block=%llu slot=%d ino=%llu, %pV",
201 btrfs_header_level(eb) == 0 ? "leaf" : "node", root->objectid,
202 btrfs_header_bytenr(eb), slot, key.objectid, &vaf);
206 static int check_dir_item(struct btrfs_root *root,
207 struct extent_buffer *leaf,
208 struct btrfs_key *key, int slot)
210 struct btrfs_dir_item *di;
211 u32 item_size = btrfs_item_size_nr(leaf, slot);
214 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
215 while (cur < item_size) {
223 /* header itself should not cross item boundary */
224 if (cur + sizeof(*di) > item_size) {
225 dir_item_err(root, leaf, slot,
226 "dir item header crosses item boundary, have %zu boundary %u",
227 cur + sizeof(*di), item_size);
232 dir_type = btrfs_dir_type(leaf, di);
233 if (dir_type >= BTRFS_FT_MAX) {
234 dir_item_err(root, leaf, slot,
235 "invalid dir item type, have %u expect [0, %u)",
236 dir_type, BTRFS_FT_MAX);
240 if (key->type == BTRFS_XATTR_ITEM_KEY &&
241 dir_type != BTRFS_FT_XATTR) {
242 dir_item_err(root, leaf, slot,
243 "invalid dir item type for XATTR key, have %u expect %u",
244 dir_type, BTRFS_FT_XATTR);
247 if (dir_type == BTRFS_FT_XATTR &&
248 key->type != BTRFS_XATTR_ITEM_KEY) {
249 dir_item_err(root, leaf, slot,
250 "xattr dir type found for non-XATTR key");
253 if (dir_type == BTRFS_FT_XATTR)
254 max_name_len = XATTR_NAME_MAX;
256 max_name_len = BTRFS_NAME_LEN;
258 /* Name/data length check */
259 name_len = btrfs_dir_name_len(leaf, di);
260 data_len = btrfs_dir_data_len(leaf, di);
261 if (name_len > max_name_len) {
262 dir_item_err(root, leaf, slot,
263 "dir item name len too long, have %u max %u",
264 name_len, max_name_len);
267 if (name_len + data_len > BTRFS_MAX_XATTR_SIZE(root)) {
268 dir_item_err(root, leaf, slot,
269 "dir item name and data len too long, have %u max %u",
271 BTRFS_MAX_XATTR_SIZE(root));
275 if (data_len && dir_type != BTRFS_FT_XATTR) {
276 dir_item_err(root, leaf, slot,
277 "dir item with invalid data len, have %u expect 0",
282 total_size = sizeof(*di) + name_len + data_len;
284 /* header and name/data should not cross item boundary */
285 if (cur + total_size > item_size) {
286 dir_item_err(root, leaf, slot,
287 "dir item data crosses item boundary, have %u boundary %u",
288 cur + total_size, item_size);
293 * Special check for XATTR/DIR_ITEM, as key->offset is name
294 * hash, should match its name
296 if (key->type == BTRFS_DIR_ITEM_KEY ||
297 key->type == BTRFS_XATTR_ITEM_KEY) {
298 char namebuf[max(BTRFS_NAME_LEN, XATTR_NAME_MAX)];
300 read_extent_buffer(leaf, namebuf,
301 (unsigned long)(di + 1), name_len);
302 name_hash = btrfs_name_hash(namebuf, name_len);
303 if (key->offset != name_hash) {
304 dir_item_err(root, leaf, slot,
305 "name hash mismatch with key, have 0x%016x expect 0x%016llx",
306 name_hash, key->offset);
311 di = (struct btrfs_dir_item *)((void *)di + total_size);
318 static void block_group_err(const struct btrfs_fs_info *fs_info,
319 const struct extent_buffer *eb, int slot,
320 const char *fmt, ...)
322 struct btrfs_key key;
323 struct va_format vaf;
326 btrfs_item_key_to_cpu(eb, &key, slot);
333 "corrupt %s: root=%llu block=%llu slot=%d bg_start=%llu bg_len=%llu, %pV",
334 btrfs_header_level(eb) == 0 ? "leaf" : "node",
335 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
336 key.objectid, key.offset, &vaf);
340 static int check_block_group_item(struct btrfs_fs_info *fs_info,
341 struct extent_buffer *leaf,
342 struct btrfs_key *key, int slot)
344 struct btrfs_block_group_item bgi;
345 u32 item_size = btrfs_item_size_nr(leaf, slot);
350 * Here we don't really care about alignment since extent allocator can
351 * handle it. We care more about the size, as if one block group is
352 * larger than maximum size, it's must be some obvious corruption.
354 if (key->offset > BTRFS_MAX_DATA_CHUNK_SIZE || key->offset == 0) {
355 block_group_err(fs_info, leaf, slot,
356 "invalid block group size, have %llu expect (0, %llu]",
357 key->offset, BTRFS_MAX_DATA_CHUNK_SIZE);
361 if (item_size != sizeof(bgi)) {
362 block_group_err(fs_info, leaf, slot,
363 "invalid item size, have %u expect %zu",
364 item_size, sizeof(bgi));
368 read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot),
370 if (btrfs_block_group_chunk_objectid(&bgi) !=
371 BTRFS_FIRST_CHUNK_TREE_OBJECTID) {
372 block_group_err(fs_info, leaf, slot,
373 "invalid block group chunk objectid, have %llu expect %llu",
374 btrfs_block_group_chunk_objectid(&bgi),
375 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
379 if (btrfs_block_group_used(&bgi) > key->offset) {
380 block_group_err(fs_info, leaf, slot,
381 "invalid block group used, have %llu expect [0, %llu)",
382 btrfs_block_group_used(&bgi), key->offset);
386 flags = btrfs_block_group_flags(&bgi);
387 if (hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) > 1) {
388 block_group_err(fs_info, leaf, slot,
389 "invalid profile flags, have 0x%llx (%lu bits set) expect no more than 1 bit set",
390 flags & BTRFS_BLOCK_GROUP_PROFILE_MASK,
391 hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK));
395 type = flags & BTRFS_BLOCK_GROUP_TYPE_MASK;
396 if (type != BTRFS_BLOCK_GROUP_DATA &&
397 type != BTRFS_BLOCK_GROUP_METADATA &&
398 type != BTRFS_BLOCK_GROUP_SYSTEM &&
399 type != (BTRFS_BLOCK_GROUP_METADATA |
400 BTRFS_BLOCK_GROUP_DATA)) {
401 block_group_err(fs_info, leaf, slot,
402 "invalid type, have 0x%llx (%lu bits set) expect either 0x%llx, 0x%llx, 0x%llx or 0x%llx",
403 type, hweight64(type),
404 BTRFS_BLOCK_GROUP_DATA, BTRFS_BLOCK_GROUP_METADATA,
405 BTRFS_BLOCK_GROUP_SYSTEM,
406 BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA);
413 * Common point to switch the item-specific validation.
415 static int check_leaf_item(struct btrfs_root *root,
416 struct extent_buffer *leaf,
417 struct btrfs_key *key, int slot)
422 case BTRFS_EXTENT_DATA_KEY:
423 ret = check_extent_data_item(root, leaf, key, slot);
425 case BTRFS_EXTENT_CSUM_KEY:
426 ret = check_csum_item(root, leaf, key, slot);
428 case BTRFS_DIR_ITEM_KEY:
429 case BTRFS_DIR_INDEX_KEY:
430 case BTRFS_XATTR_ITEM_KEY:
431 ret = check_dir_item(root, leaf, key, slot);
433 case BTRFS_BLOCK_GROUP_ITEM_KEY:
434 ret = check_block_group_item(root->fs_info, leaf, key, slot);
440 static int check_leaf(struct btrfs_root *root, struct extent_buffer *leaf,
441 bool check_item_data)
443 struct btrfs_fs_info *fs_info = root->fs_info;
444 /* No valid key type is 0, so all key should be larger than this key */
445 struct btrfs_key prev_key = {0, 0, 0};
446 struct btrfs_key key;
447 u32 nritems = btrfs_header_nritems(leaf);
450 if (btrfs_header_level(leaf) != 0) {
451 generic_err(root, leaf, 0,
452 "invalid level for leaf, have %d expect 0",
453 btrfs_header_level(leaf));
458 * Extent buffers from a relocation tree have a owner field that
459 * corresponds to the subvolume tree they are based on. So just from an
460 * extent buffer alone we can not find out what is the id of the
461 * corresponding subvolume tree, so we can not figure out if the extent
462 * buffer corresponds to the root of the relocation tree or not. So
463 * skip this check for relocation trees.
465 if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) {
466 u64 owner = btrfs_header_owner(leaf);
467 struct btrfs_root *check_root;
469 /* These trees must never be empty */
470 if (owner == BTRFS_ROOT_TREE_OBJECTID ||
471 owner == BTRFS_CHUNK_TREE_OBJECTID ||
472 owner == BTRFS_EXTENT_TREE_OBJECTID ||
473 owner == BTRFS_DEV_TREE_OBJECTID ||
474 owner == BTRFS_FS_TREE_OBJECTID ||
475 owner == BTRFS_DATA_RELOC_TREE_OBJECTID) {
476 generic_err(root, leaf, 0,
477 "invalid root, root %llu must never be empty",
481 key.objectid = owner;
482 key.type = BTRFS_ROOT_ITEM_KEY;
483 key.offset = (u64)-1;
485 check_root = btrfs_get_fs_root(fs_info, &key, false);
487 * The only reason we also check NULL here is that during
488 * open_ctree() some roots has not yet been set up.
490 if (!IS_ERR_OR_NULL(check_root)) {
491 struct extent_buffer *eb;
493 eb = btrfs_root_node(check_root);
494 /* if leaf is the root, then it's fine */
496 CORRUPT("non-root leaf's nritems is 0",
497 leaf, check_root, 0);
498 free_extent_buffer(eb);
501 free_extent_buffer(eb);
510 * Check the following things to make sure this is a good leaf, and
511 * leaf users won't need to bother with similar sanity checks:
514 * 2) item offset and size
515 * No overlap, no hole, all inside the leaf.
517 * If possible, do comprehensive sanity check.
518 * NOTE: All checks must only rely on the item data itself.
520 for (slot = 0; slot < nritems; slot++) {
521 u32 item_end_expected;
524 btrfs_item_key_to_cpu(leaf, &key, slot);
526 /* Make sure the keys are in the right order */
527 if (btrfs_comp_cpu_keys(&prev_key, &key) >= 0) {
528 CORRUPT("bad key order", leaf, root, slot);
533 * Make sure the offset and ends are right, remember that the
534 * item data starts at the end of the leaf and grows towards the
538 item_end_expected = BTRFS_LEAF_DATA_SIZE(root);
540 item_end_expected = btrfs_item_offset_nr(leaf,
542 if (btrfs_item_end_nr(leaf, slot) != item_end_expected) {
543 CORRUPT("slot offset bad", leaf, root, slot);
548 * Check to make sure that we don't point outside of the leaf,
549 * just in case all the items are consistent to each other, but
550 * all point outside of the leaf.
552 if (btrfs_item_end_nr(leaf, slot) >
553 BTRFS_LEAF_DATA_SIZE(root)) {
554 CORRUPT("slot end outside of leaf", leaf, root, slot);
558 /* Also check if the item pointer overlaps with btrfs item. */
559 if (btrfs_item_nr_offset(slot) + sizeof(struct btrfs_item) >
560 btrfs_item_ptr_offset(leaf, slot)) {
561 CORRUPT("slot overlap with its data", leaf, root, slot);
565 if (check_item_data) {
567 * Check if the item size and content meet other
570 ret = check_leaf_item(root, leaf, &key, slot);
575 prev_key.objectid = key.objectid;
576 prev_key.type = key.type;
577 prev_key.offset = key.offset;
583 int btrfs_check_leaf_full(struct btrfs_root *root, struct extent_buffer *leaf)
585 return check_leaf(root, leaf, true);
588 int btrfs_check_leaf_relaxed(struct btrfs_root *root,
589 struct extent_buffer *leaf)
591 return check_leaf(root, leaf, false);
594 int btrfs_check_node(struct btrfs_root *root, struct extent_buffer *node)
596 unsigned long nr = btrfs_header_nritems(node);
597 struct btrfs_key key, next_key;
599 int level = btrfs_header_level(node);
603 if (level <= 0 || level >= BTRFS_MAX_LEVEL) {
604 generic_err(root, node, 0,
605 "invalid level for node, have %d expect [1, %d]",
606 level, BTRFS_MAX_LEVEL - 1);
609 if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(root)) {
610 btrfs_crit(root->fs_info,
611 "corrupt node: root=%llu block=%llu, nritems too %s, have %lu expect range [1,%u]",
612 root->objectid, node->start,
613 nr == 0 ? "small" : "large", nr,
614 BTRFS_NODEPTRS_PER_BLOCK(root));
618 for (slot = 0; slot < nr - 1; slot++) {
619 bytenr = btrfs_node_blockptr(node, slot);
620 btrfs_node_key_to_cpu(node, &key, slot);
621 btrfs_node_key_to_cpu(node, &next_key, slot + 1);
624 generic_err(root, node, slot,
625 "invalid NULL node pointer");
629 if (!IS_ALIGNED(bytenr, root->sectorsize)) {
630 generic_err(root, node, slot,
631 "unaligned pointer, have %llu should be aligned to %u",
632 bytenr, root->sectorsize);
637 if (btrfs_comp_cpu_keys(&key, &next_key) >= 0) {
638 generic_err(root, node, slot,
639 "bad key order, current (%llu %u %llu) next (%llu %u %llu)",
640 key.objectid, key.type, key.offset,
641 next_key.objectid, next_key.type,