1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Copyright (C) Qu Wenruo 2017. All rights reserved.
4 */
5
6 /*
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.
9 *
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.
13 *
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.
16 */
17
18 #include <linux/types.h>
19 #include <linux/stddef.h>
20 #include <linux/error-injection.h>
21 #include "ctree.h"
22 #include "tree-checker.h"
23 #include "disk-io.h"
24 #include "compression.h"
25 #include "volumes.h"
26 #include "misc.h"
27 #include "btrfs_inode.h"
28
29 /*
30 * Error message should follow the following format:
31 * corrupt <type>: <identifier>, <reason>[, <bad_value>]
32 *
33 * @type: leaf or node
34 * @identifier: the necessary info to locate the leaf/node.
35 * It's recommended to decode key.objecitd/offset if it's
36 * meaningful.
37 * @reason: describe the error
38 * @bad_value: optional, it's recommended to output bad value and its
39 * expected value (range).
40 *
41 * Since comma is used to separate the components, only space is allowed
42 * inside each component.
43 */
44
45 /*
46 * Append generic "corrupt leaf/node root=%llu block=%llu slot=%d: " to @fmt.
47 * Allows callers to customize the output.
48 */
49 __printf(3, 4)
50 __cold
generic_err(const struct extent_buffer * eb,int slot,const char * fmt,...)51 static void generic_err(const struct extent_buffer *eb, int slot,
52 const char *fmt, ...)
53 {
54 const struct btrfs_fs_info *fs_info = eb->fs_info;
55 struct va_format vaf;
56 va_list args;
57
58 va_start(args, fmt);
59
60 vaf.fmt = fmt;
61 vaf.va = &args;
62
63 btrfs_crit(fs_info,
64 "corrupt %s: root=%llu block=%llu slot=%d, %pV",
65 btrfs_header_level(eb) == 0 ? "leaf" : "node",
66 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, &vaf);
67 va_end(args);
68 }
69
70 /*
71 * Customized reporter for extent data item, since its key objectid and
72 * offset has its own meaning.
73 */
74 __printf(3, 4)
75 __cold
file_extent_err(const struct extent_buffer * eb,int slot,const char * fmt,...)76 static void file_extent_err(const struct extent_buffer *eb, int slot,
77 const char *fmt, ...)
78 {
79 const struct btrfs_fs_info *fs_info = eb->fs_info;
80 struct btrfs_key key;
81 struct va_format vaf;
82 va_list args;
83
84 btrfs_item_key_to_cpu(eb, &key, slot);
85 va_start(args, fmt);
86
87 vaf.fmt = fmt;
88 vaf.va = &args;
89
90 btrfs_crit(fs_info,
91 "corrupt %s: root=%llu block=%llu slot=%d ino=%llu file_offset=%llu, %pV",
92 btrfs_header_level(eb) == 0 ? "leaf" : "node",
93 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
94 key.objectid, key.offset, &vaf);
95 va_end(args);
96 }
97
98 /*
99 * Return 0 if the btrfs_file_extent_##name is aligned to @alignment
100 * Else return 1
101 */
102 #define CHECK_FE_ALIGNED(leaf, slot, fi, name, alignment) \
103 ({ \
104 if (unlikely(!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), \
105 (alignment)))) \
106 file_extent_err((leaf), (slot), \
107 "invalid %s for file extent, have %llu, should be aligned to %u", \
108 (#name), btrfs_file_extent_##name((leaf), (fi)), \
109 (alignment)); \
110 (!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment))); \
111 })
112
file_extent_end(struct extent_buffer * leaf,struct btrfs_key * key,struct btrfs_file_extent_item * extent)113 static u64 file_extent_end(struct extent_buffer *leaf,
114 struct btrfs_key *key,
115 struct btrfs_file_extent_item *extent)
116 {
117 u64 end;
118 u64 len;
119
120 if (btrfs_file_extent_type(leaf, extent) == BTRFS_FILE_EXTENT_INLINE) {
121 len = btrfs_file_extent_ram_bytes(leaf, extent);
122 end = ALIGN(key->offset + len, leaf->fs_info->sectorsize);
123 } else {
124 len = btrfs_file_extent_num_bytes(leaf, extent);
125 end = key->offset + len;
126 }
127 return end;
128 }
129
130 /*
131 * Customized report for dir_item, the only new important information is
132 * key->objectid, which represents inode number
133 */
134 __printf(3, 4)
135 __cold
dir_item_err(const struct extent_buffer * eb,int slot,const char * fmt,...)136 static void dir_item_err(const struct extent_buffer *eb, int slot,
137 const char *fmt, ...)
138 {
139 const struct btrfs_fs_info *fs_info = eb->fs_info;
140 struct btrfs_key key;
141 struct va_format vaf;
142 va_list args;
143
144 btrfs_item_key_to_cpu(eb, &key, slot);
145 va_start(args, fmt);
146
147 vaf.fmt = fmt;
148 vaf.va = &args;
149
150 btrfs_crit(fs_info,
151 "corrupt %s: root=%llu block=%llu slot=%d ino=%llu, %pV",
152 btrfs_header_level(eb) == 0 ? "leaf" : "node",
153 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
154 key.objectid, &vaf);
155 va_end(args);
156 }
157
158 /*
159 * This functions checks prev_key->objectid, to ensure current key and prev_key
160 * share the same objectid as inode number.
161 *
162 * This is to detect missing INODE_ITEM in subvolume trees.
163 *
164 * Return true if everything is OK or we don't need to check.
165 * Return false if anything is wrong.
166 */
check_prev_ino(struct extent_buffer * leaf,struct btrfs_key * key,int slot,struct btrfs_key * prev_key)167 static bool check_prev_ino(struct extent_buffer *leaf,
168 struct btrfs_key *key, int slot,
169 struct btrfs_key *prev_key)
170 {
171 /* No prev key, skip check */
172 if (slot == 0)
173 return true;
174
175 /* Only these key->types needs to be checked */
176 ASSERT(key->type == BTRFS_XATTR_ITEM_KEY ||
177 key->type == BTRFS_INODE_REF_KEY ||
178 key->type == BTRFS_DIR_INDEX_KEY ||
179 key->type == BTRFS_DIR_ITEM_KEY ||
180 key->type == BTRFS_EXTENT_DATA_KEY);
181
182 /*
183 * Only subvolume trees along with their reloc trees need this check.
184 * Things like log tree doesn't follow this ino requirement.
185 */
186 if (!is_fstree(btrfs_header_owner(leaf)))
187 return true;
188
189 if (key->objectid == prev_key->objectid)
190 return true;
191
192 /* Error found */
193 dir_item_err(leaf, slot,
194 "invalid previous key objectid, have %llu expect %llu",
195 prev_key->objectid, key->objectid);
196 return false;
197 }
check_extent_data_item(struct extent_buffer * leaf,struct btrfs_key * key,int slot,struct btrfs_key * prev_key)198 static int check_extent_data_item(struct extent_buffer *leaf,
199 struct btrfs_key *key, int slot,
200 struct btrfs_key *prev_key)
201 {
202 struct btrfs_fs_info *fs_info = leaf->fs_info;
203 struct btrfs_file_extent_item *fi;
204 u32 sectorsize = fs_info->sectorsize;
205 u32 item_size = btrfs_item_size_nr(leaf, slot);
206 u64 extent_end;
207
208 if (unlikely(!IS_ALIGNED(key->offset, sectorsize))) {
209 file_extent_err(leaf, slot,
210 "unaligned file_offset for file extent, have %llu should be aligned to %u",
211 key->offset, sectorsize);
212 return -EUCLEAN;
213 }
214
215 /*
216 * Previous key must have the same key->objectid (ino).
217 * It can be XATTR_ITEM, INODE_ITEM or just another EXTENT_DATA.
218 * But if objectids mismatch, it means we have a missing
219 * INODE_ITEM.
220 */
221 if (unlikely(!check_prev_ino(leaf, key, slot, prev_key)))
222 return -EUCLEAN;
223
224 fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
225
226 /*
227 * Make sure the item contains at least inline header, so the file
228 * extent type is not some garbage.
229 */
230 if (unlikely(item_size < BTRFS_FILE_EXTENT_INLINE_DATA_START)) {
231 file_extent_err(leaf, slot,
232 "invalid item size, have %u expect [%zu, %u)",
233 item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START,
234 SZ_4K);
235 return -EUCLEAN;
236 }
237 if (unlikely(btrfs_file_extent_type(leaf, fi) >=
238 BTRFS_NR_FILE_EXTENT_TYPES)) {
239 file_extent_err(leaf, slot,
240 "invalid type for file extent, have %u expect range [0, %u]",
241 btrfs_file_extent_type(leaf, fi),
242 BTRFS_NR_FILE_EXTENT_TYPES - 1);
243 return -EUCLEAN;
244 }
245
246 /*
247 * Support for new compression/encryption must introduce incompat flag,
248 * and must be caught in open_ctree().
249 */
250 if (unlikely(btrfs_file_extent_compression(leaf, fi) >=
251 BTRFS_NR_COMPRESS_TYPES)) {
252 file_extent_err(leaf, slot,
253 "invalid compression for file extent, have %u expect range [0, %u]",
254 btrfs_file_extent_compression(leaf, fi),
255 BTRFS_NR_COMPRESS_TYPES - 1);
256 return -EUCLEAN;
257 }
258 if (unlikely(btrfs_file_extent_encryption(leaf, fi))) {
259 file_extent_err(leaf, slot,
260 "invalid encryption for file extent, have %u expect 0",
261 btrfs_file_extent_encryption(leaf, fi));
262 return -EUCLEAN;
263 }
264 if (btrfs_file_extent_type(leaf, fi) == BTRFS_FILE_EXTENT_INLINE) {
265 /* Inline extent must have 0 as key offset */
266 if (unlikely(key->offset)) {
267 file_extent_err(leaf, slot,
268 "invalid file_offset for inline file extent, have %llu expect 0",
269 key->offset);
270 return -EUCLEAN;
271 }
272
273 /* Compressed inline extent has no on-disk size, skip it */
274 if (btrfs_file_extent_compression(leaf, fi) !=
275 BTRFS_COMPRESS_NONE)
276 return 0;
277
278 /* Uncompressed inline extent size must match item size */
279 if (unlikely(item_size != BTRFS_FILE_EXTENT_INLINE_DATA_START +
280 btrfs_file_extent_ram_bytes(leaf, fi))) {
281 file_extent_err(leaf, slot,
282 "invalid ram_bytes for uncompressed inline extent, have %u expect %llu",
283 item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START +
284 btrfs_file_extent_ram_bytes(leaf, fi));
285 return -EUCLEAN;
286 }
287 return 0;
288 }
289
290 /* Regular or preallocated extent has fixed item size */
291 if (unlikely(item_size != sizeof(*fi))) {
292 file_extent_err(leaf, slot,
293 "invalid item size for reg/prealloc file extent, have %u expect %zu",
294 item_size, sizeof(*fi));
295 return -EUCLEAN;
296 }
297 if (unlikely(CHECK_FE_ALIGNED(leaf, slot, fi, ram_bytes, sectorsize) ||
298 CHECK_FE_ALIGNED(leaf, slot, fi, disk_bytenr, sectorsize) ||
299 CHECK_FE_ALIGNED(leaf, slot, fi, disk_num_bytes, sectorsize) ||
300 CHECK_FE_ALIGNED(leaf, slot, fi, offset, sectorsize) ||
301 CHECK_FE_ALIGNED(leaf, slot, fi, num_bytes, sectorsize)))
302 return -EUCLEAN;
303
304 /* Catch extent end overflow */
305 if (unlikely(check_add_overflow(btrfs_file_extent_num_bytes(leaf, fi),
306 key->offset, &extent_end))) {
307 file_extent_err(leaf, slot,
308 "extent end overflow, have file offset %llu extent num bytes %llu",
309 key->offset,
310 btrfs_file_extent_num_bytes(leaf, fi));
311 return -EUCLEAN;
312 }
313
314 /*
315 * Check that no two consecutive file extent items, in the same leaf,
316 * present ranges that overlap each other.
317 */
318 if (slot > 0 &&
319 prev_key->objectid == key->objectid &&
320 prev_key->type == BTRFS_EXTENT_DATA_KEY) {
321 struct btrfs_file_extent_item *prev_fi;
322 u64 prev_end;
323
324 prev_fi = btrfs_item_ptr(leaf, slot - 1,
325 struct btrfs_file_extent_item);
326 prev_end = file_extent_end(leaf, prev_key, prev_fi);
327 if (unlikely(prev_end > key->offset)) {
328 file_extent_err(leaf, slot - 1,
329 "file extent end range (%llu) goes beyond start offset (%llu) of the next file extent",
330 prev_end, key->offset);
331 return -EUCLEAN;
332 }
333 }
334
335 return 0;
336 }
337
check_csum_item(struct extent_buffer * leaf,struct btrfs_key * key,int slot,struct btrfs_key * prev_key)338 static int check_csum_item(struct extent_buffer *leaf, struct btrfs_key *key,
339 int slot, struct btrfs_key *prev_key)
340 {
341 struct btrfs_fs_info *fs_info = leaf->fs_info;
342 u32 sectorsize = fs_info->sectorsize;
343 const u32 csumsize = fs_info->csum_size;
344
345 if (unlikely(key->objectid != BTRFS_EXTENT_CSUM_OBJECTID)) {
346 generic_err(leaf, slot,
347 "invalid key objectid for csum item, have %llu expect %llu",
348 key->objectid, BTRFS_EXTENT_CSUM_OBJECTID);
349 return -EUCLEAN;
350 }
351 if (unlikely(!IS_ALIGNED(key->offset, sectorsize))) {
352 generic_err(leaf, slot,
353 "unaligned key offset for csum item, have %llu should be aligned to %u",
354 key->offset, sectorsize);
355 return -EUCLEAN;
356 }
357 if (unlikely(!IS_ALIGNED(btrfs_item_size_nr(leaf, slot), csumsize))) {
358 generic_err(leaf, slot,
359 "unaligned item size for csum item, have %u should be aligned to %u",
360 btrfs_item_size_nr(leaf, slot), csumsize);
361 return -EUCLEAN;
362 }
363 if (slot > 0 && prev_key->type == BTRFS_EXTENT_CSUM_KEY) {
364 u64 prev_csum_end;
365 u32 prev_item_size;
366
367 prev_item_size = btrfs_item_size_nr(leaf, slot - 1);
368 prev_csum_end = (prev_item_size / csumsize) * sectorsize;
369 prev_csum_end += prev_key->offset;
370 if (unlikely(prev_csum_end > key->offset)) {
371 generic_err(leaf, slot - 1,
372 "csum end range (%llu) goes beyond the start range (%llu) of the next csum item",
373 prev_csum_end, key->offset);
374 return -EUCLEAN;
375 }
376 }
377 return 0;
378 }
379
380 /* Inode item error output has the same format as dir_item_err() */
381 #define inode_item_err(eb, slot, fmt, ...) \
382 dir_item_err(eb, slot, fmt, __VA_ARGS__)
383
check_inode_key(struct extent_buffer * leaf,struct btrfs_key * key,int slot)384 static int check_inode_key(struct extent_buffer *leaf, struct btrfs_key *key,
385 int slot)
386 {
387 struct btrfs_key item_key;
388 bool is_inode_item;
389
390 btrfs_item_key_to_cpu(leaf, &item_key, slot);
391 is_inode_item = (item_key.type == BTRFS_INODE_ITEM_KEY);
392
393 /* For XATTR_ITEM, location key should be all 0 */
394 if (item_key.type == BTRFS_XATTR_ITEM_KEY) {
395 if (unlikely(key->objectid != 0 || key->type != 0 ||
396 key->offset != 0))
397 return -EUCLEAN;
398 return 0;
399 }
400
401 if (unlikely((key->objectid < BTRFS_FIRST_FREE_OBJECTID ||
402 key->objectid > BTRFS_LAST_FREE_OBJECTID) &&
403 key->objectid != BTRFS_ROOT_TREE_DIR_OBJECTID &&
404 key->objectid != BTRFS_FREE_INO_OBJECTID)) {
405 if (is_inode_item) {
406 generic_err(leaf, slot,
407 "invalid key objectid: has %llu expect %llu or [%llu, %llu] or %llu",
408 key->objectid, BTRFS_ROOT_TREE_DIR_OBJECTID,
409 BTRFS_FIRST_FREE_OBJECTID,
410 BTRFS_LAST_FREE_OBJECTID,
411 BTRFS_FREE_INO_OBJECTID);
412 } else {
413 dir_item_err(leaf, slot,
414 "invalid location key objectid: has %llu expect %llu or [%llu, %llu] or %llu",
415 key->objectid, BTRFS_ROOT_TREE_DIR_OBJECTID,
416 BTRFS_FIRST_FREE_OBJECTID,
417 BTRFS_LAST_FREE_OBJECTID,
418 BTRFS_FREE_INO_OBJECTID);
419 }
420 return -EUCLEAN;
421 }
422 if (unlikely(key->offset != 0)) {
423 if (is_inode_item)
424 inode_item_err(leaf, slot,
425 "invalid key offset: has %llu expect 0",
426 key->offset);
427 else
428 dir_item_err(leaf, slot,
429 "invalid location key offset:has %llu expect 0",
430 key->offset);
431 return -EUCLEAN;
432 }
433 return 0;
434 }
435
check_root_key(struct extent_buffer * leaf,struct btrfs_key * key,int slot)436 static int check_root_key(struct extent_buffer *leaf, struct btrfs_key *key,
437 int slot)
438 {
439 struct btrfs_key item_key;
440 bool is_root_item;
441
442 btrfs_item_key_to_cpu(leaf, &item_key, slot);
443 is_root_item = (item_key.type == BTRFS_ROOT_ITEM_KEY);
444
445 /*
446 * Bad rootid for reloc trees.
447 *
448 * Reloc trees are only for subvolume trees, other trees only need
449 * to be COWed to be relocated.
450 */
451 if (unlikely(is_root_item && key->objectid == BTRFS_TREE_RELOC_OBJECTID &&
452 !is_fstree(key->offset))) {
453 generic_err(leaf, slot,
454 "invalid reloc tree for root %lld, root id is not a subvolume tree",
455 key->offset);
456 return -EUCLEAN;
457 }
458
459 /* No such tree id */
460 if (unlikely(key->objectid == 0)) {
461 if (is_root_item)
462 generic_err(leaf, slot, "invalid root id 0");
463 else
464 dir_item_err(leaf, slot,
465 "invalid location key root id 0");
466 return -EUCLEAN;
467 }
468
469 /* DIR_ITEM/INDEX/INODE_REF is not allowed to point to non-fs trees */
470 if (unlikely(!is_fstree(key->objectid) && !is_root_item)) {
471 dir_item_err(leaf, slot,
472 "invalid location key objectid, have %llu expect [%llu, %llu]",
473 key->objectid, BTRFS_FIRST_FREE_OBJECTID,
474 BTRFS_LAST_FREE_OBJECTID);
475 return -EUCLEAN;
476 }
477
478 /*
479 * ROOT_ITEM with non-zero offset means this is a snapshot, created at
480 * @offset transid.
481 * Furthermore, for location key in DIR_ITEM, its offset is always -1.
482 *
483 * So here we only check offset for reloc tree whose key->offset must
484 * be a valid tree.
485 */
486 if (unlikely(key->objectid == BTRFS_TREE_RELOC_OBJECTID &&
487 key->offset == 0)) {
488 generic_err(leaf, slot, "invalid root id 0 for reloc tree");
489 return -EUCLEAN;
490 }
491 return 0;
492 }
493
check_dir_item(struct extent_buffer * leaf,struct btrfs_key * key,struct btrfs_key * prev_key,int slot)494 static int check_dir_item(struct extent_buffer *leaf,
495 struct btrfs_key *key, struct btrfs_key *prev_key,
496 int slot)
497 {
498 struct btrfs_fs_info *fs_info = leaf->fs_info;
499 struct btrfs_dir_item *di;
500 u32 item_size = btrfs_item_size_nr(leaf, slot);
501 u32 cur = 0;
502
503 if (unlikely(!check_prev_ino(leaf, key, slot, prev_key)))
504 return -EUCLEAN;
505
506 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
507 while (cur < item_size) {
508 struct btrfs_key location_key;
509 u32 name_len;
510 u32 data_len;
511 u32 max_name_len;
512 u32 total_size;
513 u32 name_hash;
514 u8 dir_type;
515 int ret;
516
517 /* header itself should not cross item boundary */
518 if (unlikely(cur + sizeof(*di) > item_size)) {
519 dir_item_err(leaf, slot,
520 "dir item header crosses item boundary, have %zu boundary %u",
521 cur + sizeof(*di), item_size);
522 return -EUCLEAN;
523 }
524
525 /* Location key check */
526 btrfs_dir_item_key_to_cpu(leaf, di, &location_key);
527 if (location_key.type == BTRFS_ROOT_ITEM_KEY) {
528 ret = check_root_key(leaf, &location_key, slot);
529 if (unlikely(ret < 0))
530 return ret;
531 } else if (location_key.type == BTRFS_INODE_ITEM_KEY ||
532 location_key.type == 0) {
533 ret = check_inode_key(leaf, &location_key, slot);
534 if (unlikely(ret < 0))
535 return ret;
536 } else {
537 dir_item_err(leaf, slot,
538 "invalid location key type, have %u, expect %u or %u",
539 location_key.type, BTRFS_ROOT_ITEM_KEY,
540 BTRFS_INODE_ITEM_KEY);
541 return -EUCLEAN;
542 }
543
544 /* dir type check */
545 dir_type = btrfs_dir_type(leaf, di);
546 if (unlikely(dir_type >= BTRFS_FT_MAX)) {
547 dir_item_err(leaf, slot,
548 "invalid dir item type, have %u expect [0, %u)",
549 dir_type, BTRFS_FT_MAX);
550 return -EUCLEAN;
551 }
552
553 if (unlikely(key->type == BTRFS_XATTR_ITEM_KEY &&
554 dir_type != BTRFS_FT_XATTR)) {
555 dir_item_err(leaf, slot,
556 "invalid dir item type for XATTR key, have %u expect %u",
557 dir_type, BTRFS_FT_XATTR);
558 return -EUCLEAN;
559 }
560 if (unlikely(dir_type == BTRFS_FT_XATTR &&
561 key->type != BTRFS_XATTR_ITEM_KEY)) {
562 dir_item_err(leaf, slot,
563 "xattr dir type found for non-XATTR key");
564 return -EUCLEAN;
565 }
566 if (dir_type == BTRFS_FT_XATTR)
567 max_name_len = XATTR_NAME_MAX;
568 else
569 max_name_len = BTRFS_NAME_LEN;
570
571 /* Name/data length check */
572 name_len = btrfs_dir_name_len(leaf, di);
573 data_len = btrfs_dir_data_len(leaf, di);
574 if (unlikely(name_len > max_name_len)) {
575 dir_item_err(leaf, slot,
576 "dir item name len too long, have %u max %u",
577 name_len, max_name_len);
578 return -EUCLEAN;
579 }
580 if (unlikely(name_len + data_len > BTRFS_MAX_XATTR_SIZE(fs_info))) {
581 dir_item_err(leaf, slot,
582 "dir item name and data len too long, have %u max %u",
583 name_len + data_len,
584 BTRFS_MAX_XATTR_SIZE(fs_info));
585 return -EUCLEAN;
586 }
587
588 if (unlikely(data_len && dir_type != BTRFS_FT_XATTR)) {
589 dir_item_err(leaf, slot,
590 "dir item with invalid data len, have %u expect 0",
591 data_len);
592 return -EUCLEAN;
593 }
594
595 total_size = sizeof(*di) + name_len + data_len;
596
597 /* header and name/data should not cross item boundary */
598 if (unlikely(cur + total_size > item_size)) {
599 dir_item_err(leaf, slot,
600 "dir item data crosses item boundary, have %u boundary %u",
601 cur + total_size, item_size);
602 return -EUCLEAN;
603 }
604
605 /*
606 * Special check for XATTR/DIR_ITEM, as key->offset is name
607 * hash, should match its name
608 */
609 if (key->type == BTRFS_DIR_ITEM_KEY ||
610 key->type == BTRFS_XATTR_ITEM_KEY) {
611 char namebuf[max(BTRFS_NAME_LEN, XATTR_NAME_MAX)];
612
613 read_extent_buffer(leaf, namebuf,
614 (unsigned long)(di + 1), name_len);
615 name_hash = btrfs_name_hash(namebuf, name_len);
616 if (unlikely(key->offset != name_hash)) {
617 dir_item_err(leaf, slot,
618 "name hash mismatch with key, have 0x%016x expect 0x%016llx",
619 name_hash, key->offset);
620 return -EUCLEAN;
621 }
622 }
623 cur += total_size;
624 di = (struct btrfs_dir_item *)((void *)di + total_size);
625 }
626 return 0;
627 }
628
629 __printf(3, 4)
630 __cold
block_group_err(const struct extent_buffer * eb,int slot,const char * fmt,...)631 static void block_group_err(const struct extent_buffer *eb, int slot,
632 const char *fmt, ...)
633 {
634 const struct btrfs_fs_info *fs_info = eb->fs_info;
635 struct btrfs_key key;
636 struct va_format vaf;
637 va_list args;
638
639 btrfs_item_key_to_cpu(eb, &key, slot);
640 va_start(args, fmt);
641
642 vaf.fmt = fmt;
643 vaf.va = &args;
644
645 btrfs_crit(fs_info,
646 "corrupt %s: root=%llu block=%llu slot=%d bg_start=%llu bg_len=%llu, %pV",
647 btrfs_header_level(eb) == 0 ? "leaf" : "node",
648 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
649 key.objectid, key.offset, &vaf);
650 va_end(args);
651 }
652
check_block_group_item(struct extent_buffer * leaf,struct btrfs_key * key,int slot)653 static int check_block_group_item(struct extent_buffer *leaf,
654 struct btrfs_key *key, int slot)
655 {
656 struct btrfs_block_group_item bgi;
657 u32 item_size = btrfs_item_size_nr(leaf, slot);
658 u64 flags;
659 u64 type;
660
661 /*
662 * Here we don't really care about alignment since extent allocator can
663 * handle it. We care more about the size.
664 */
665 if (unlikely(key->offset == 0)) {
666 block_group_err(leaf, slot,
667 "invalid block group size 0");
668 return -EUCLEAN;
669 }
670
671 if (unlikely(item_size != sizeof(bgi))) {
672 block_group_err(leaf, slot,
673 "invalid item size, have %u expect %zu",
674 item_size, sizeof(bgi));
675 return -EUCLEAN;
676 }
677
678 read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot),
679 sizeof(bgi));
680 if (unlikely(btrfs_stack_block_group_chunk_objectid(&bgi) !=
681 BTRFS_FIRST_CHUNK_TREE_OBJECTID)) {
682 block_group_err(leaf, slot,
683 "invalid block group chunk objectid, have %llu expect %llu",
684 btrfs_stack_block_group_chunk_objectid(&bgi),
685 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
686 return -EUCLEAN;
687 }
688
689 if (unlikely(btrfs_stack_block_group_used(&bgi) > key->offset)) {
690 block_group_err(leaf, slot,
691 "invalid block group used, have %llu expect [0, %llu)",
692 btrfs_stack_block_group_used(&bgi), key->offset);
693 return -EUCLEAN;
694 }
695
696 flags = btrfs_stack_block_group_flags(&bgi);
697 if (unlikely(hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) > 1)) {
698 block_group_err(leaf, slot,
699 "invalid profile flags, have 0x%llx (%lu bits set) expect no more than 1 bit set",
700 flags & BTRFS_BLOCK_GROUP_PROFILE_MASK,
701 hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK));
702 return -EUCLEAN;
703 }
704
705 type = flags & BTRFS_BLOCK_GROUP_TYPE_MASK;
706 if (unlikely(type != BTRFS_BLOCK_GROUP_DATA &&
707 type != BTRFS_BLOCK_GROUP_METADATA &&
708 type != BTRFS_BLOCK_GROUP_SYSTEM &&
709 type != (BTRFS_BLOCK_GROUP_METADATA |
710 BTRFS_BLOCK_GROUP_DATA))) {
711 block_group_err(leaf, slot,
712 "invalid type, have 0x%llx (%lu bits set) expect either 0x%llx, 0x%llx, 0x%llx or 0x%llx",
713 type, hweight64(type),
714 BTRFS_BLOCK_GROUP_DATA, BTRFS_BLOCK_GROUP_METADATA,
715 BTRFS_BLOCK_GROUP_SYSTEM,
716 BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA);
717 return -EUCLEAN;
718 }
719 return 0;
720 }
721
722 __printf(4, 5)
723 __cold
chunk_err(const struct extent_buffer * leaf,const struct btrfs_chunk * chunk,u64 logical,const char * fmt,...)724 static void chunk_err(const struct extent_buffer *leaf,
725 const struct btrfs_chunk *chunk, u64 logical,
726 const char *fmt, ...)
727 {
728 const struct btrfs_fs_info *fs_info = leaf->fs_info;
729 bool is_sb;
730 struct va_format vaf;
731 va_list args;
732 int i;
733 int slot = -1;
734
735 /* Only superblock eb is able to have such small offset */
736 is_sb = (leaf->start == BTRFS_SUPER_INFO_OFFSET);
737
738 if (!is_sb) {
739 /*
740 * Get the slot number by iterating through all slots, this
741 * would provide better readability.
742 */
743 for (i = 0; i < btrfs_header_nritems(leaf); i++) {
744 if (btrfs_item_ptr_offset(leaf, i) ==
745 (unsigned long)chunk) {
746 slot = i;
747 break;
748 }
749 }
750 }
751 va_start(args, fmt);
752 vaf.fmt = fmt;
753 vaf.va = &args;
754
755 if (is_sb)
756 btrfs_crit(fs_info,
757 "corrupt superblock syschunk array: chunk_start=%llu, %pV",
758 logical, &vaf);
759 else
760 btrfs_crit(fs_info,
761 "corrupt leaf: root=%llu block=%llu slot=%d chunk_start=%llu, %pV",
762 BTRFS_CHUNK_TREE_OBJECTID, leaf->start, slot,
763 logical, &vaf);
764 va_end(args);
765 }
766
767 /*
768 * The common chunk check which could also work on super block sys chunk array.
769 *
770 * Return -EUCLEAN if anything is corrupted.
771 * Return 0 if everything is OK.
772 */
btrfs_check_chunk_valid(struct extent_buffer * leaf,struct btrfs_chunk * chunk,u64 logical)773 int btrfs_check_chunk_valid(struct extent_buffer *leaf,
774 struct btrfs_chunk *chunk, u64 logical)
775 {
776 struct btrfs_fs_info *fs_info = leaf->fs_info;
777 u64 length;
778 u64 chunk_end;
779 u64 stripe_len;
780 u16 num_stripes;
781 u16 sub_stripes;
782 u64 type;
783 u64 features;
784 bool mixed = false;
785 int raid_index;
786 int nparity;
787 int ncopies;
788
789 length = btrfs_chunk_length(leaf, chunk);
790 stripe_len = btrfs_chunk_stripe_len(leaf, chunk);
791 num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
792 sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk);
793 type = btrfs_chunk_type(leaf, chunk);
794 raid_index = btrfs_bg_flags_to_raid_index(type);
795 ncopies = btrfs_raid_array[raid_index].ncopies;
796 nparity = btrfs_raid_array[raid_index].nparity;
797
798 if (unlikely(!num_stripes)) {
799 chunk_err(leaf, chunk, logical,
800 "invalid chunk num_stripes, have %u", num_stripes);
801 return -EUCLEAN;
802 }
803 if (unlikely(num_stripes < ncopies)) {
804 chunk_err(leaf, chunk, logical,
805 "invalid chunk num_stripes < ncopies, have %u < %d",
806 num_stripes, ncopies);
807 return -EUCLEAN;
808 }
809 if (unlikely(nparity && num_stripes == nparity)) {
810 chunk_err(leaf, chunk, logical,
811 "invalid chunk num_stripes == nparity, have %u == %d",
812 num_stripes, nparity);
813 return -EUCLEAN;
814 }
815 if (unlikely(!IS_ALIGNED(logical, fs_info->sectorsize))) {
816 chunk_err(leaf, chunk, logical,
817 "invalid chunk logical, have %llu should aligned to %u",
818 logical, fs_info->sectorsize);
819 return -EUCLEAN;
820 }
821 if (unlikely(btrfs_chunk_sector_size(leaf, chunk) != fs_info->sectorsize)) {
822 chunk_err(leaf, chunk, logical,
823 "invalid chunk sectorsize, have %u expect %u",
824 btrfs_chunk_sector_size(leaf, chunk),
825 fs_info->sectorsize);
826 return -EUCLEAN;
827 }
828 if (unlikely(!length || !IS_ALIGNED(length, fs_info->sectorsize))) {
829 chunk_err(leaf, chunk, logical,
830 "invalid chunk length, have %llu", length);
831 return -EUCLEAN;
832 }
833 if (unlikely(check_add_overflow(logical, length, &chunk_end))) {
834 chunk_err(leaf, chunk, logical,
835 "invalid chunk logical start and length, have logical start %llu length %llu",
836 logical, length);
837 return -EUCLEAN;
838 }
839 if (unlikely(!is_power_of_2(stripe_len) || stripe_len != BTRFS_STRIPE_LEN)) {
840 chunk_err(leaf, chunk, logical,
841 "invalid chunk stripe length: %llu",
842 stripe_len);
843 return -EUCLEAN;
844 }
845 if (unlikely(type & ~(BTRFS_BLOCK_GROUP_TYPE_MASK |
846 BTRFS_BLOCK_GROUP_PROFILE_MASK))) {
847 chunk_err(leaf, chunk, logical,
848 "unrecognized chunk type: 0x%llx",
849 ~(BTRFS_BLOCK_GROUP_TYPE_MASK |
850 BTRFS_BLOCK_GROUP_PROFILE_MASK) &
851 btrfs_chunk_type(leaf, chunk));
852 return -EUCLEAN;
853 }
854
855 if (unlikely(!has_single_bit_set(type & BTRFS_BLOCK_GROUP_PROFILE_MASK) &&
856 (type & BTRFS_BLOCK_GROUP_PROFILE_MASK) != 0)) {
857 chunk_err(leaf, chunk, logical,
858 "invalid chunk profile flag: 0x%llx, expect 0 or 1 bit set",
859 type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
860 return -EUCLEAN;
861 }
862 if (unlikely((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == 0)) {
863 chunk_err(leaf, chunk, logical,
864 "missing chunk type flag, have 0x%llx one bit must be set in 0x%llx",
865 type, BTRFS_BLOCK_GROUP_TYPE_MASK);
866 return -EUCLEAN;
867 }
868
869 if (unlikely((type & BTRFS_BLOCK_GROUP_SYSTEM) &&
870 (type & (BTRFS_BLOCK_GROUP_METADATA |
871 BTRFS_BLOCK_GROUP_DATA)))) {
872 chunk_err(leaf, chunk, logical,
873 "system chunk with data or metadata type: 0x%llx",
874 type);
875 return -EUCLEAN;
876 }
877
878 features = btrfs_super_incompat_flags(fs_info->super_copy);
879 if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS)
880 mixed = true;
881
882 if (!mixed) {
883 if (unlikely((type & BTRFS_BLOCK_GROUP_METADATA) &&
884 (type & BTRFS_BLOCK_GROUP_DATA))) {
885 chunk_err(leaf, chunk, logical,
886 "mixed chunk type in non-mixed mode: 0x%llx", type);
887 return -EUCLEAN;
888 }
889 }
890
891 if (unlikely((type & BTRFS_BLOCK_GROUP_RAID10 &&
892 sub_stripes != btrfs_raid_array[BTRFS_RAID_RAID10].sub_stripes) ||
893 (type & BTRFS_BLOCK_GROUP_RAID1 &&
894 num_stripes != btrfs_raid_array[BTRFS_RAID_RAID1].devs_min) ||
895 (type & BTRFS_BLOCK_GROUP_RAID1C3 &&
896 num_stripes != btrfs_raid_array[BTRFS_RAID_RAID1C3].devs_min) ||
897 (type & BTRFS_BLOCK_GROUP_RAID1C4 &&
898 num_stripes != btrfs_raid_array[BTRFS_RAID_RAID1C4].devs_min) ||
899 (type & BTRFS_BLOCK_GROUP_RAID5 &&
900 num_stripes < btrfs_raid_array[BTRFS_RAID_RAID5].devs_min) ||
901 (type & BTRFS_BLOCK_GROUP_RAID6 &&
902 num_stripes < btrfs_raid_array[BTRFS_RAID_RAID6].devs_min) ||
903 (type & BTRFS_BLOCK_GROUP_DUP &&
904 num_stripes != btrfs_raid_array[BTRFS_RAID_DUP].dev_stripes) ||
905 ((type & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 &&
906 num_stripes != btrfs_raid_array[BTRFS_RAID_SINGLE].dev_stripes))) {
907 chunk_err(leaf, chunk, logical,
908 "invalid num_stripes:sub_stripes %u:%u for profile %llu",
909 num_stripes, sub_stripes,
910 type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
911 return -EUCLEAN;
912 }
913
914 return 0;
915 }
916
917 /*
918 * Enhanced version of chunk item checker.
919 *
920 * The common btrfs_check_chunk_valid() doesn't check item size since it needs
921 * to work on super block sys_chunk_array which doesn't have full item ptr.
922 */
check_leaf_chunk_item(struct extent_buffer * leaf,struct btrfs_chunk * chunk,struct btrfs_key * key,int slot)923 static int check_leaf_chunk_item(struct extent_buffer *leaf,
924 struct btrfs_chunk *chunk,
925 struct btrfs_key *key, int slot)
926 {
927 int num_stripes;
928
929 if (unlikely(btrfs_item_size_nr(leaf, slot) < sizeof(struct btrfs_chunk))) {
930 chunk_err(leaf, chunk, key->offset,
931 "invalid chunk item size: have %u expect [%zu, %u)",
932 btrfs_item_size_nr(leaf, slot),
933 sizeof(struct btrfs_chunk),
934 BTRFS_LEAF_DATA_SIZE(leaf->fs_info));
935 return -EUCLEAN;
936 }
937
938 num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
939 /* Let btrfs_check_chunk_valid() handle this error type */
940 if (num_stripes == 0)
941 goto out;
942
943 if (unlikely(btrfs_chunk_item_size(num_stripes) !=
944 btrfs_item_size_nr(leaf, slot))) {
945 chunk_err(leaf, chunk, key->offset,
946 "invalid chunk item size: have %u expect %lu",
947 btrfs_item_size_nr(leaf, slot),
948 btrfs_chunk_item_size(num_stripes));
949 return -EUCLEAN;
950 }
951 out:
952 return btrfs_check_chunk_valid(leaf, chunk, key->offset);
953 }
954
955 __printf(3, 4)
956 __cold
dev_item_err(const struct extent_buffer * eb,int slot,const char * fmt,...)957 static void dev_item_err(const struct extent_buffer *eb, int slot,
958 const char *fmt, ...)
959 {
960 struct btrfs_key key;
961 struct va_format vaf;
962 va_list args;
963
964 btrfs_item_key_to_cpu(eb, &key, slot);
965 va_start(args, fmt);
966
967 vaf.fmt = fmt;
968 vaf.va = &args;
969
970 btrfs_crit(eb->fs_info,
971 "corrupt %s: root=%llu block=%llu slot=%d devid=%llu %pV",
972 btrfs_header_level(eb) == 0 ? "leaf" : "node",
973 btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot,
974 key.objectid, &vaf);
975 va_end(args);
976 }
977
check_dev_item(struct extent_buffer * leaf,struct btrfs_key * key,int slot)978 static int check_dev_item(struct extent_buffer *leaf,
979 struct btrfs_key *key, int slot)
980 {
981 struct btrfs_dev_item *ditem;
982 const u32 item_size = btrfs_item_size_nr(leaf, slot);
983
984 if (unlikely(key->objectid != BTRFS_DEV_ITEMS_OBJECTID)) {
985 dev_item_err(leaf, slot,
986 "invalid objectid: has=%llu expect=%llu",
987 key->objectid, BTRFS_DEV_ITEMS_OBJECTID);
988 return -EUCLEAN;
989 }
990
991 if (unlikely(item_size != sizeof(*ditem))) {
992 dev_item_err(leaf, slot, "invalid item size: has %u expect %zu",
993 item_size, sizeof(*ditem));
994 return -EUCLEAN;
995 }
996
997 ditem = btrfs_item_ptr(leaf, slot, struct btrfs_dev_item);
998 if (unlikely(btrfs_device_id(leaf, ditem) != key->offset)) {
999 dev_item_err(leaf, slot,
1000 "devid mismatch: key has=%llu item has=%llu",
1001 key->offset, btrfs_device_id(leaf, ditem));
1002 return -EUCLEAN;
1003 }
1004
1005 /*
1006 * For device total_bytes, we don't have reliable way to check it, as
1007 * it can be 0 for device removal. Device size check can only be done
1008 * by dev extents check.
1009 */
1010 if (unlikely(btrfs_device_bytes_used(leaf, ditem) >
1011 btrfs_device_total_bytes(leaf, ditem))) {
1012 dev_item_err(leaf, slot,
1013 "invalid bytes used: have %llu expect [0, %llu]",
1014 btrfs_device_bytes_used(leaf, ditem),
1015 btrfs_device_total_bytes(leaf, ditem));
1016 return -EUCLEAN;
1017 }
1018 /*
1019 * Remaining members like io_align/type/gen/dev_group aren't really
1020 * utilized. Skip them to make later usage of them easier.
1021 */
1022 return 0;
1023 }
1024
check_inode_item(struct extent_buffer * leaf,struct btrfs_key * key,int slot)1025 static int check_inode_item(struct extent_buffer *leaf,
1026 struct btrfs_key *key, int slot)
1027 {
1028 struct btrfs_fs_info *fs_info = leaf->fs_info;
1029 struct btrfs_inode_item *iitem;
1030 u64 super_gen = btrfs_super_generation(fs_info->super_copy);
1031 u32 valid_mask = (S_IFMT | S_ISUID | S_ISGID | S_ISVTX | 0777);
1032 const u32 item_size = btrfs_item_size_nr(leaf, slot);
1033 u32 mode;
1034 int ret;
1035 u32 flags;
1036 u32 ro_flags;
1037
1038 ret = check_inode_key(leaf, key, slot);
1039 if (unlikely(ret < 0))
1040 return ret;
1041
1042 if (unlikely(item_size != sizeof(*iitem))) {
1043 generic_err(leaf, slot, "invalid item size: has %u expect %zu",
1044 item_size, sizeof(*iitem));
1045 return -EUCLEAN;
1046 }
1047
1048 iitem = btrfs_item_ptr(leaf, slot, struct btrfs_inode_item);
1049
1050 /* Here we use super block generation + 1 to handle log tree */
1051 if (unlikely(btrfs_inode_generation(leaf, iitem) > super_gen + 1)) {
1052 inode_item_err(leaf, slot,
1053 "invalid inode generation: has %llu expect (0, %llu]",
1054 btrfs_inode_generation(leaf, iitem),
1055 super_gen + 1);
1056 return -EUCLEAN;
1057 }
1058 /* Note for ROOT_TREE_DIR_ITEM, mkfs could set its transid 0 */
1059 if (unlikely(btrfs_inode_transid(leaf, iitem) > super_gen + 1)) {
1060 inode_item_err(leaf, slot,
1061 "invalid inode transid: has %llu expect [0, %llu]",
1062 btrfs_inode_transid(leaf, iitem), super_gen + 1);
1063 return -EUCLEAN;
1064 }
1065
1066 /*
1067 * For size and nbytes it's better not to be too strict, as for dir
1068 * item its size/nbytes can easily get wrong, but doesn't affect
1069 * anything in the fs. So here we skip the check.
1070 */
1071 mode = btrfs_inode_mode(leaf, iitem);
1072 if (unlikely(mode & ~valid_mask)) {
1073 inode_item_err(leaf, slot,
1074 "unknown mode bit detected: 0x%x",
1075 mode & ~valid_mask);
1076 return -EUCLEAN;
1077 }
1078
1079 /*
1080 * S_IFMT is not bit mapped so we can't completely rely on
1081 * is_power_of_2/has_single_bit_set, but it can save us from checking
1082 * FIFO/CHR/DIR/REG. Only needs to check BLK, LNK and SOCKS
1083 */
1084 if (!has_single_bit_set(mode & S_IFMT)) {
1085 if (unlikely(!S_ISLNK(mode) && !S_ISBLK(mode) && !S_ISSOCK(mode))) {
1086 inode_item_err(leaf, slot,
1087 "invalid mode: has 0%o expect valid S_IF* bit(s)",
1088 mode & S_IFMT);
1089 return -EUCLEAN;
1090 }
1091 }
1092 if (unlikely(S_ISDIR(mode) && btrfs_inode_nlink(leaf, iitem) > 1)) {
1093 inode_item_err(leaf, slot,
1094 "invalid nlink: has %u expect no more than 1 for dir",
1095 btrfs_inode_nlink(leaf, iitem));
1096 return -EUCLEAN;
1097 }
1098 btrfs_inode_split_flags(btrfs_inode_flags(leaf, iitem), &flags, &ro_flags);
1099 if (unlikely(flags & ~BTRFS_INODE_FLAG_MASK)) {
1100 inode_item_err(leaf, slot,
1101 "unknown incompat flags detected: 0x%x", flags);
1102 return -EUCLEAN;
1103 }
1104 if (unlikely(!sb_rdonly(fs_info->sb) &&
1105 (ro_flags & ~BTRFS_INODE_RO_FLAG_MASK))) {
1106 inode_item_err(leaf, slot,
1107 "unknown ro-compat flags detected on writeable mount: 0x%x",
1108 ro_flags);
1109 return -EUCLEAN;
1110 }
1111 return 0;
1112 }
1113
check_root_item(struct extent_buffer * leaf,struct btrfs_key * key,int slot)1114 static int check_root_item(struct extent_buffer *leaf, struct btrfs_key *key,
1115 int slot)
1116 {
1117 struct btrfs_fs_info *fs_info = leaf->fs_info;
1118 struct btrfs_root_item ri = { 0 };
1119 const u64 valid_root_flags = BTRFS_ROOT_SUBVOL_RDONLY |
1120 BTRFS_ROOT_SUBVOL_DEAD;
1121 int ret;
1122
1123 ret = check_root_key(leaf, key, slot);
1124 if (unlikely(ret < 0))
1125 return ret;
1126
1127 if (unlikely(btrfs_item_size_nr(leaf, slot) != sizeof(ri) &&
1128 btrfs_item_size_nr(leaf, slot) !=
1129 btrfs_legacy_root_item_size())) {
1130 generic_err(leaf, slot,
1131 "invalid root item size, have %u expect %zu or %u",
1132 btrfs_item_size_nr(leaf, slot), sizeof(ri),
1133 btrfs_legacy_root_item_size());
1134 return -EUCLEAN;
1135 }
1136
1137 /*
1138 * For legacy root item, the members starting at generation_v2 will be
1139 * all filled with 0.
1140 * And since we allow geneartion_v2 as 0, it will still pass the check.
1141 */
1142 read_extent_buffer(leaf, &ri, btrfs_item_ptr_offset(leaf, slot),
1143 btrfs_item_size_nr(leaf, slot));
1144
1145 /* Generation related */
1146 if (unlikely(btrfs_root_generation(&ri) >
1147 btrfs_super_generation(fs_info->super_copy) + 1)) {
1148 generic_err(leaf, slot,
1149 "invalid root generation, have %llu expect (0, %llu]",
1150 btrfs_root_generation(&ri),
1151 btrfs_super_generation(fs_info->super_copy) + 1);
1152 return -EUCLEAN;
1153 }
1154 if (unlikely(btrfs_root_generation_v2(&ri) >
1155 btrfs_super_generation(fs_info->super_copy) + 1)) {
1156 generic_err(leaf, slot,
1157 "invalid root v2 generation, have %llu expect (0, %llu]",
1158 btrfs_root_generation_v2(&ri),
1159 btrfs_super_generation(fs_info->super_copy) + 1);
1160 return -EUCLEAN;
1161 }
1162 if (unlikely(btrfs_root_last_snapshot(&ri) >
1163 btrfs_super_generation(fs_info->super_copy) + 1)) {
1164 generic_err(leaf, slot,
1165 "invalid root last_snapshot, have %llu expect (0, %llu]",
1166 btrfs_root_last_snapshot(&ri),
1167 btrfs_super_generation(fs_info->super_copy) + 1);
1168 return -EUCLEAN;
1169 }
1170
1171 /* Alignment and level check */
1172 if (unlikely(!IS_ALIGNED(btrfs_root_bytenr(&ri), fs_info->sectorsize))) {
1173 generic_err(leaf, slot,
1174 "invalid root bytenr, have %llu expect to be aligned to %u",
1175 btrfs_root_bytenr(&ri), fs_info->sectorsize);
1176 return -EUCLEAN;
1177 }
1178 if (unlikely(btrfs_root_level(&ri) >= BTRFS_MAX_LEVEL)) {
1179 generic_err(leaf, slot,
1180 "invalid root level, have %u expect [0, %u]",
1181 btrfs_root_level(&ri), BTRFS_MAX_LEVEL - 1);
1182 return -EUCLEAN;
1183 }
1184 if (unlikely(btrfs_root_drop_level(&ri) >= BTRFS_MAX_LEVEL)) {
1185 generic_err(leaf, slot,
1186 "invalid root level, have %u expect [0, %u]",
1187 btrfs_root_drop_level(&ri), BTRFS_MAX_LEVEL - 1);
1188 return -EUCLEAN;
1189 }
1190
1191 /* Flags check */
1192 if (unlikely(btrfs_root_flags(&ri) & ~valid_root_flags)) {
1193 generic_err(leaf, slot,
1194 "invalid root flags, have 0x%llx expect mask 0x%llx",
1195 btrfs_root_flags(&ri), valid_root_flags);
1196 return -EUCLEAN;
1197 }
1198 return 0;
1199 }
1200
1201 __printf(3,4)
1202 __cold
extent_err(const struct extent_buffer * eb,int slot,const char * fmt,...)1203 static void extent_err(const struct extent_buffer *eb, int slot,
1204 const char *fmt, ...)
1205 {
1206 struct btrfs_key key;
1207 struct va_format vaf;
1208 va_list args;
1209 u64 bytenr;
1210 u64 len;
1211
1212 btrfs_item_key_to_cpu(eb, &key, slot);
1213 bytenr = key.objectid;
1214 if (key.type == BTRFS_METADATA_ITEM_KEY ||
1215 key.type == BTRFS_TREE_BLOCK_REF_KEY ||
1216 key.type == BTRFS_SHARED_BLOCK_REF_KEY)
1217 len = eb->fs_info->nodesize;
1218 else
1219 len = key.offset;
1220 va_start(args, fmt);
1221
1222 vaf.fmt = fmt;
1223 vaf.va = &args;
1224
1225 btrfs_crit(eb->fs_info,
1226 "corrupt %s: block=%llu slot=%d extent bytenr=%llu len=%llu %pV",
1227 btrfs_header_level(eb) == 0 ? "leaf" : "node",
1228 eb->start, slot, bytenr, len, &vaf);
1229 va_end(args);
1230 }
1231
check_extent_item(struct extent_buffer * leaf,struct btrfs_key * key,int slot,struct btrfs_key * prev_key)1232 static int check_extent_item(struct extent_buffer *leaf,
1233 struct btrfs_key *key, int slot,
1234 struct btrfs_key *prev_key)
1235 {
1236 struct btrfs_fs_info *fs_info = leaf->fs_info;
1237 struct btrfs_extent_item *ei;
1238 bool is_tree_block = false;
1239 unsigned long ptr; /* Current pointer inside inline refs */
1240 unsigned long end; /* Extent item end */
1241 const u32 item_size = btrfs_item_size_nr(leaf, slot);
1242 u64 flags;
1243 u64 generation;
1244 u64 total_refs; /* Total refs in btrfs_extent_item */
1245 u64 inline_refs = 0; /* found total inline refs */
1246
1247 if (unlikely(key->type == BTRFS_METADATA_ITEM_KEY &&
1248 !btrfs_fs_incompat(fs_info, SKINNY_METADATA))) {
1249 generic_err(leaf, slot,
1250 "invalid key type, METADATA_ITEM type invalid when SKINNY_METADATA feature disabled");
1251 return -EUCLEAN;
1252 }
1253 /* key->objectid is the bytenr for both key types */
1254 if (unlikely(!IS_ALIGNED(key->objectid, fs_info->sectorsize))) {
1255 generic_err(leaf, slot,
1256 "invalid key objectid, have %llu expect to be aligned to %u",
1257 key->objectid, fs_info->sectorsize);
1258 return -EUCLEAN;
1259 }
1260
1261 /* key->offset is tree level for METADATA_ITEM_KEY */
1262 if (unlikely(key->type == BTRFS_METADATA_ITEM_KEY &&
1263 key->offset >= BTRFS_MAX_LEVEL)) {
1264 extent_err(leaf, slot,
1265 "invalid tree level, have %llu expect [0, %u]",
1266 key->offset, BTRFS_MAX_LEVEL - 1);
1267 return -EUCLEAN;
1268 }
1269
1270 /*
1271 * EXTENT/METADATA_ITEM consists of:
1272 * 1) One btrfs_extent_item
1273 * Records the total refs, type and generation of the extent.
1274 *
1275 * 2) One btrfs_tree_block_info (for EXTENT_ITEM and tree backref only)
1276 * Records the first key and level of the tree block.
1277 *
1278 * 2) Zero or more btrfs_extent_inline_ref(s)
1279 * Each inline ref has one btrfs_extent_inline_ref shows:
1280 * 2.1) The ref type, one of the 4
1281 * TREE_BLOCK_REF Tree block only
1282 * SHARED_BLOCK_REF Tree block only
1283 * EXTENT_DATA_REF Data only
1284 * SHARED_DATA_REF Data only
1285 * 2.2) Ref type specific data
1286 * Either using btrfs_extent_inline_ref::offset, or specific
1287 * data structure.
1288 */
1289 if (unlikely(item_size < sizeof(*ei))) {
1290 extent_err(leaf, slot,
1291 "invalid item size, have %u expect [%zu, %u)",
1292 item_size, sizeof(*ei),
1293 BTRFS_LEAF_DATA_SIZE(fs_info));
1294 return -EUCLEAN;
1295 }
1296 end = item_size + btrfs_item_ptr_offset(leaf, slot);
1297
1298 /* Checks against extent_item */
1299 ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
1300 flags = btrfs_extent_flags(leaf, ei);
1301 total_refs = btrfs_extent_refs(leaf, ei);
1302 generation = btrfs_extent_generation(leaf, ei);
1303 if (unlikely(generation >
1304 btrfs_super_generation(fs_info->super_copy) + 1)) {
1305 extent_err(leaf, slot,
1306 "invalid generation, have %llu expect (0, %llu]",
1307 generation,
1308 btrfs_super_generation(fs_info->super_copy) + 1);
1309 return -EUCLEAN;
1310 }
1311 if (unlikely(!has_single_bit_set(flags & (BTRFS_EXTENT_FLAG_DATA |
1312 BTRFS_EXTENT_FLAG_TREE_BLOCK)))) {
1313 extent_err(leaf, slot,
1314 "invalid extent flag, have 0x%llx expect 1 bit set in 0x%llx",
1315 flags, BTRFS_EXTENT_FLAG_DATA |
1316 BTRFS_EXTENT_FLAG_TREE_BLOCK);
1317 return -EUCLEAN;
1318 }
1319 is_tree_block = !!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK);
1320 if (is_tree_block) {
1321 if (unlikely(key->type == BTRFS_EXTENT_ITEM_KEY &&
1322 key->offset != fs_info->nodesize)) {
1323 extent_err(leaf, slot,
1324 "invalid extent length, have %llu expect %u",
1325 key->offset, fs_info->nodesize);
1326 return -EUCLEAN;
1327 }
1328 } else {
1329 if (unlikely(key->type != BTRFS_EXTENT_ITEM_KEY)) {
1330 extent_err(leaf, slot,
1331 "invalid key type, have %u expect %u for data backref",
1332 key->type, BTRFS_EXTENT_ITEM_KEY);
1333 return -EUCLEAN;
1334 }
1335 if (unlikely(!IS_ALIGNED(key->offset, fs_info->sectorsize))) {
1336 extent_err(leaf, slot,
1337 "invalid extent length, have %llu expect aligned to %u",
1338 key->offset, fs_info->sectorsize);
1339 return -EUCLEAN;
1340 }
1341 if (unlikely(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)) {
1342 extent_err(leaf, slot,
1343 "invalid extent flag, data has full backref set");
1344 return -EUCLEAN;
1345 }
1346 }
1347 ptr = (unsigned long)(struct btrfs_extent_item *)(ei + 1);
1348
1349 /* Check the special case of btrfs_tree_block_info */
1350 if (is_tree_block && key->type != BTRFS_METADATA_ITEM_KEY) {
1351 struct btrfs_tree_block_info *info;
1352
1353 info = (struct btrfs_tree_block_info *)ptr;
1354 if (unlikely(btrfs_tree_block_level(leaf, info) >= BTRFS_MAX_LEVEL)) {
1355 extent_err(leaf, slot,
1356 "invalid tree block info level, have %u expect [0, %u]",
1357 btrfs_tree_block_level(leaf, info),
1358 BTRFS_MAX_LEVEL - 1);
1359 return -EUCLEAN;
1360 }
1361 ptr = (unsigned long)(struct btrfs_tree_block_info *)(info + 1);
1362 }
1363
1364 /* Check inline refs */
1365 while (ptr < end) {
1366 struct btrfs_extent_inline_ref *iref;
1367 struct btrfs_extent_data_ref *dref;
1368 struct btrfs_shared_data_ref *sref;
1369 u64 dref_offset;
1370 u64 inline_offset;
1371 u8 inline_type;
1372
1373 if (unlikely(ptr + sizeof(*iref) > end)) {
1374 extent_err(leaf, slot,
1375 "inline ref item overflows extent item, ptr %lu iref size %zu end %lu",
1376 ptr, sizeof(*iref), end);
1377 return -EUCLEAN;
1378 }
1379 iref = (struct btrfs_extent_inline_ref *)ptr;
1380 inline_type = btrfs_extent_inline_ref_type(leaf, iref);
1381 inline_offset = btrfs_extent_inline_ref_offset(leaf, iref);
1382 if (unlikely(ptr + btrfs_extent_inline_ref_size(inline_type) > end)) {
1383 extent_err(leaf, slot,
1384 "inline ref item overflows extent item, ptr %lu iref size %u end %lu",
1385 ptr, btrfs_extent_inline_ref_size(inline_type), end);
1386 return -EUCLEAN;
1387 }
1388
1389 switch (inline_type) {
1390 /* inline_offset is subvolid of the owner, no need to check */
1391 case BTRFS_TREE_BLOCK_REF_KEY:
1392 inline_refs++;
1393 break;
1394 /* Contains parent bytenr */
1395 case BTRFS_SHARED_BLOCK_REF_KEY:
1396 if (unlikely(!IS_ALIGNED(inline_offset,
1397 fs_info->sectorsize))) {
1398 extent_err(leaf, slot,
1399 "invalid tree parent bytenr, have %llu expect aligned to %u",
1400 inline_offset, fs_info->sectorsize);
1401 return -EUCLEAN;
1402 }
1403 inline_refs++;
1404 break;
1405 /*
1406 * Contains owner subvolid, owner key objectid, adjusted offset.
1407 * The only obvious corruption can happen in that offset.
1408 */
1409 case BTRFS_EXTENT_DATA_REF_KEY:
1410 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1411 dref_offset = btrfs_extent_data_ref_offset(leaf, dref);
1412 if (unlikely(!IS_ALIGNED(dref_offset,
1413 fs_info->sectorsize))) {
1414 extent_err(leaf, slot,
1415 "invalid data ref offset, have %llu expect aligned to %u",
1416 dref_offset, fs_info->sectorsize);
1417 return -EUCLEAN;
1418 }
1419 inline_refs += btrfs_extent_data_ref_count(leaf, dref);
1420 break;
1421 /* Contains parent bytenr and ref count */
1422 case BTRFS_SHARED_DATA_REF_KEY:
1423 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1424 if (unlikely(!IS_ALIGNED(inline_offset,
1425 fs_info->sectorsize))) {
1426 extent_err(leaf, slot,
1427 "invalid data parent bytenr, have %llu expect aligned to %u",
1428 inline_offset, fs_info->sectorsize);
1429 return -EUCLEAN;
1430 }
1431 inline_refs += btrfs_shared_data_ref_count(leaf, sref);
1432 break;
1433 default:
1434 extent_err(leaf, slot, "unknown inline ref type: %u",
1435 inline_type);
1436 return -EUCLEAN;
1437 }
1438 ptr += btrfs_extent_inline_ref_size(inline_type);
1439 }
1440 /* No padding is allowed */
1441 if (unlikely(ptr != end)) {
1442 extent_err(leaf, slot,
1443 "invalid extent item size, padding bytes found");
1444 return -EUCLEAN;
1445 }
1446
1447 /* Finally, check the inline refs against total refs */
1448 if (unlikely(inline_refs > total_refs)) {
1449 extent_err(leaf, slot,
1450 "invalid extent refs, have %llu expect >= inline %llu",
1451 total_refs, inline_refs);
1452 return -EUCLEAN;
1453 }
1454
1455 if ((prev_key->type == BTRFS_EXTENT_ITEM_KEY) ||
1456 (prev_key->type == BTRFS_METADATA_ITEM_KEY)) {
1457 u64 prev_end = prev_key->objectid;
1458
1459 if (prev_key->type == BTRFS_METADATA_ITEM_KEY)
1460 prev_end += fs_info->nodesize;
1461 else
1462 prev_end += prev_key->offset;
1463
1464 if (unlikely(prev_end > key->objectid)) {
1465 extent_err(leaf, slot,
1466 "previous extent [%llu %u %llu] overlaps current extent [%llu %u %llu]",
1467 prev_key->objectid, prev_key->type,
1468 prev_key->offset, key->objectid, key->type,
1469 key->offset);
1470 return -EUCLEAN;
1471 }
1472 }
1473
1474 return 0;
1475 }
1476
check_simple_keyed_refs(struct extent_buffer * leaf,struct btrfs_key * key,int slot)1477 static int check_simple_keyed_refs(struct extent_buffer *leaf,
1478 struct btrfs_key *key, int slot)
1479 {
1480 u32 expect_item_size = 0;
1481
1482 if (key->type == BTRFS_SHARED_DATA_REF_KEY)
1483 expect_item_size = sizeof(struct btrfs_shared_data_ref);
1484
1485 if (unlikely(btrfs_item_size_nr(leaf, slot) != expect_item_size)) {
1486 generic_err(leaf, slot,
1487 "invalid item size, have %u expect %u for key type %u",
1488 btrfs_item_size_nr(leaf, slot),
1489 expect_item_size, key->type);
1490 return -EUCLEAN;
1491 }
1492 if (unlikely(!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize))) {
1493 generic_err(leaf, slot,
1494 "invalid key objectid for shared block ref, have %llu expect aligned to %u",
1495 key->objectid, leaf->fs_info->sectorsize);
1496 return -EUCLEAN;
1497 }
1498 if (unlikely(key->type != BTRFS_TREE_BLOCK_REF_KEY &&
1499 !IS_ALIGNED(key->offset, leaf->fs_info->sectorsize))) {
1500 extent_err(leaf, slot,
1501 "invalid tree parent bytenr, have %llu expect aligned to %u",
1502 key->offset, leaf->fs_info->sectorsize);
1503 return -EUCLEAN;
1504 }
1505 return 0;
1506 }
1507
check_extent_data_ref(struct extent_buffer * leaf,struct btrfs_key * key,int slot)1508 static int check_extent_data_ref(struct extent_buffer *leaf,
1509 struct btrfs_key *key, int slot)
1510 {
1511 struct btrfs_extent_data_ref *dref;
1512 unsigned long ptr = btrfs_item_ptr_offset(leaf, slot);
1513 const unsigned long end = ptr + btrfs_item_size_nr(leaf, slot);
1514
1515 if (unlikely(btrfs_item_size_nr(leaf, slot) % sizeof(*dref) != 0)) {
1516 generic_err(leaf, slot,
1517 "invalid item size, have %u expect aligned to %zu for key type %u",
1518 btrfs_item_size_nr(leaf, slot),
1519 sizeof(*dref), key->type);
1520 return -EUCLEAN;
1521 }
1522 if (unlikely(!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize))) {
1523 generic_err(leaf, slot,
1524 "invalid key objectid for shared block ref, have %llu expect aligned to %u",
1525 key->objectid, leaf->fs_info->sectorsize);
1526 return -EUCLEAN;
1527 }
1528 for (; ptr < end; ptr += sizeof(*dref)) {
1529 u64 offset;
1530
1531 /*
1532 * We cannot check the extent_data_ref hash due to possible
1533 * overflow from the leaf due to hash collisions.
1534 */
1535 dref = (struct btrfs_extent_data_ref *)ptr;
1536 offset = btrfs_extent_data_ref_offset(leaf, dref);
1537 if (unlikely(!IS_ALIGNED(offset, leaf->fs_info->sectorsize))) {
1538 extent_err(leaf, slot,
1539 "invalid extent data backref offset, have %llu expect aligned to %u",
1540 offset, leaf->fs_info->sectorsize);
1541 return -EUCLEAN;
1542 }
1543 }
1544 return 0;
1545 }
1546
1547 #define inode_ref_err(eb, slot, fmt, args...) \
1548 inode_item_err(eb, slot, fmt, ##args)
check_inode_ref(struct extent_buffer * leaf,struct btrfs_key * key,struct btrfs_key * prev_key,int slot)1549 static int check_inode_ref(struct extent_buffer *leaf,
1550 struct btrfs_key *key, struct btrfs_key *prev_key,
1551 int slot)
1552 {
1553 struct btrfs_inode_ref *iref;
1554 unsigned long ptr;
1555 unsigned long end;
1556
1557 if (unlikely(!check_prev_ino(leaf, key, slot, prev_key)))
1558 return -EUCLEAN;
1559 /* namelen can't be 0, so item_size == sizeof() is also invalid */
1560 if (unlikely(btrfs_item_size_nr(leaf, slot) <= sizeof(*iref))) {
1561 inode_ref_err(leaf, slot,
1562 "invalid item size, have %u expect (%zu, %u)",
1563 btrfs_item_size_nr(leaf, slot),
1564 sizeof(*iref), BTRFS_LEAF_DATA_SIZE(leaf->fs_info));
1565 return -EUCLEAN;
1566 }
1567
1568 ptr = btrfs_item_ptr_offset(leaf, slot);
1569 end = ptr + btrfs_item_size_nr(leaf, slot);
1570 while (ptr < end) {
1571 u16 namelen;
1572
1573 if (unlikely(ptr + sizeof(iref) > end)) {
1574 inode_ref_err(leaf, slot,
1575 "inode ref overflow, ptr %lu end %lu inode_ref_size %zu",
1576 ptr, end, sizeof(iref));
1577 return -EUCLEAN;
1578 }
1579
1580 iref = (struct btrfs_inode_ref *)ptr;
1581 namelen = btrfs_inode_ref_name_len(leaf, iref);
1582 if (unlikely(ptr + sizeof(*iref) + namelen > end)) {
1583 inode_ref_err(leaf, slot,
1584 "inode ref overflow, ptr %lu end %lu namelen %u",
1585 ptr, end, namelen);
1586 return -EUCLEAN;
1587 }
1588
1589 /*
1590 * NOTE: In theory we should record all found index numbers
1591 * to find any duplicated indexes, but that will be too time
1592 * consuming for inodes with too many hard links.
1593 */
1594 ptr += sizeof(*iref) + namelen;
1595 }
1596 return 0;
1597 }
1598
1599 /*
1600 * Common point to switch the item-specific validation.
1601 */
check_leaf_item(struct extent_buffer * leaf,struct btrfs_key * key,int slot,struct btrfs_key * prev_key)1602 static int check_leaf_item(struct extent_buffer *leaf,
1603 struct btrfs_key *key, int slot,
1604 struct btrfs_key *prev_key)
1605 {
1606 int ret = 0;
1607 struct btrfs_chunk *chunk;
1608
1609 switch (key->type) {
1610 case BTRFS_EXTENT_DATA_KEY:
1611 ret = check_extent_data_item(leaf, key, slot, prev_key);
1612 break;
1613 case BTRFS_EXTENT_CSUM_KEY:
1614 ret = check_csum_item(leaf, key, slot, prev_key);
1615 break;
1616 case BTRFS_DIR_ITEM_KEY:
1617 case BTRFS_DIR_INDEX_KEY:
1618 case BTRFS_XATTR_ITEM_KEY:
1619 ret = check_dir_item(leaf, key, prev_key, slot);
1620 break;
1621 case BTRFS_INODE_REF_KEY:
1622 ret = check_inode_ref(leaf, key, prev_key, slot);
1623 break;
1624 case BTRFS_BLOCK_GROUP_ITEM_KEY:
1625 ret = check_block_group_item(leaf, key, slot);
1626 break;
1627 case BTRFS_CHUNK_ITEM_KEY:
1628 chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
1629 ret = check_leaf_chunk_item(leaf, chunk, key, slot);
1630 break;
1631 case BTRFS_DEV_ITEM_KEY:
1632 ret = check_dev_item(leaf, key, slot);
1633 break;
1634 case BTRFS_INODE_ITEM_KEY:
1635 ret = check_inode_item(leaf, key, slot);
1636 break;
1637 case BTRFS_ROOT_ITEM_KEY:
1638 ret = check_root_item(leaf, key, slot);
1639 break;
1640 case BTRFS_EXTENT_ITEM_KEY:
1641 case BTRFS_METADATA_ITEM_KEY:
1642 ret = check_extent_item(leaf, key, slot, prev_key);
1643 break;
1644 case BTRFS_TREE_BLOCK_REF_KEY:
1645 case BTRFS_SHARED_DATA_REF_KEY:
1646 case BTRFS_SHARED_BLOCK_REF_KEY:
1647 ret = check_simple_keyed_refs(leaf, key, slot);
1648 break;
1649 case BTRFS_EXTENT_DATA_REF_KEY:
1650 ret = check_extent_data_ref(leaf, key, slot);
1651 break;
1652 }
1653 return ret;
1654 }
1655
check_leaf(struct extent_buffer * leaf,bool check_item_data)1656 static int check_leaf(struct extent_buffer *leaf, bool check_item_data)
1657 {
1658 struct btrfs_fs_info *fs_info = leaf->fs_info;
1659 /* No valid key type is 0, so all key should be larger than this key */
1660 struct btrfs_key prev_key = {0, 0, 0};
1661 struct btrfs_key key;
1662 u32 nritems = btrfs_header_nritems(leaf);
1663 int slot;
1664
1665 if (unlikely(btrfs_header_level(leaf) != 0)) {
1666 generic_err(leaf, 0,
1667 "invalid level for leaf, have %d expect 0",
1668 btrfs_header_level(leaf));
1669 return -EUCLEAN;
1670 }
1671
1672 /*
1673 * Extent buffers from a relocation tree have a owner field that
1674 * corresponds to the subvolume tree they are based on. So just from an
1675 * extent buffer alone we can not find out what is the id of the
1676 * corresponding subvolume tree, so we can not figure out if the extent
1677 * buffer corresponds to the root of the relocation tree or not. So
1678 * skip this check for relocation trees.
1679 */
1680 if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) {
1681 u64 owner = btrfs_header_owner(leaf);
1682
1683 /* These trees must never be empty */
1684 if (unlikely(owner == BTRFS_ROOT_TREE_OBJECTID ||
1685 owner == BTRFS_CHUNK_TREE_OBJECTID ||
1686 owner == BTRFS_EXTENT_TREE_OBJECTID ||
1687 owner == BTRFS_DEV_TREE_OBJECTID ||
1688 owner == BTRFS_FS_TREE_OBJECTID ||
1689 owner == BTRFS_DATA_RELOC_TREE_OBJECTID)) {
1690 generic_err(leaf, 0,
1691 "invalid root, root %llu must never be empty",
1692 owner);
1693 return -EUCLEAN;
1694 }
1695 /* Unknown tree */
1696 if (unlikely(owner == 0)) {
1697 generic_err(leaf, 0,
1698 "invalid owner, root 0 is not defined");
1699 return -EUCLEAN;
1700 }
1701 return 0;
1702 }
1703
1704 if (unlikely(nritems == 0))
1705 return 0;
1706
1707 /*
1708 * Check the following things to make sure this is a good leaf, and
1709 * leaf users won't need to bother with similar sanity checks:
1710 *
1711 * 1) key ordering
1712 * 2) item offset and size
1713 * No overlap, no hole, all inside the leaf.
1714 * 3) item content
1715 * If possible, do comprehensive sanity check.
1716 * NOTE: All checks must only rely on the item data itself.
1717 */
1718 for (slot = 0; slot < nritems; slot++) {
1719 u32 item_end_expected;
1720 int ret;
1721
1722 btrfs_item_key_to_cpu(leaf, &key, slot);
1723
1724 /* Make sure the keys are in the right order */
1725 if (unlikely(btrfs_comp_cpu_keys(&prev_key, &key) >= 0)) {
1726 generic_err(leaf, slot,
1727 "bad key order, prev (%llu %u %llu) current (%llu %u %llu)",
1728 prev_key.objectid, prev_key.type,
1729 prev_key.offset, key.objectid, key.type,
1730 key.offset);
1731 return -EUCLEAN;
1732 }
1733
1734 /*
1735 * Make sure the offset and ends are right, remember that the
1736 * item data starts at the end of the leaf and grows towards the
1737 * front.
1738 */
1739 if (slot == 0)
1740 item_end_expected = BTRFS_LEAF_DATA_SIZE(fs_info);
1741 else
1742 item_end_expected = btrfs_item_offset_nr(leaf,
1743 slot - 1);
1744 if (unlikely(btrfs_item_end_nr(leaf, slot) != item_end_expected)) {
1745 generic_err(leaf, slot,
1746 "unexpected item end, have %u expect %u",
1747 btrfs_item_end_nr(leaf, slot),
1748 item_end_expected);
1749 return -EUCLEAN;
1750 }
1751
1752 /*
1753 * Check to make sure that we don't point outside of the leaf,
1754 * just in case all the items are consistent to each other, but
1755 * all point outside of the leaf.
1756 */
1757 if (unlikely(btrfs_item_end_nr(leaf, slot) >
1758 BTRFS_LEAF_DATA_SIZE(fs_info))) {
1759 generic_err(leaf, slot,
1760 "slot end outside of leaf, have %u expect range [0, %u]",
1761 btrfs_item_end_nr(leaf, slot),
1762 BTRFS_LEAF_DATA_SIZE(fs_info));
1763 return -EUCLEAN;
1764 }
1765
1766 /* Also check if the item pointer overlaps with btrfs item. */
1767 if (unlikely(btrfs_item_ptr_offset(leaf, slot) <
1768 btrfs_item_nr_offset(slot) + sizeof(struct btrfs_item))) {
1769 generic_err(leaf, slot,
1770 "slot overlaps with its data, item end %lu data start %lu",
1771 btrfs_item_nr_offset(slot) +
1772 sizeof(struct btrfs_item),
1773 btrfs_item_ptr_offset(leaf, slot));
1774 return -EUCLEAN;
1775 }
1776
1777 if (check_item_data) {
1778 /*
1779 * Check if the item size and content meet other
1780 * criteria
1781 */
1782 ret = check_leaf_item(leaf, &key, slot, &prev_key);
1783 if (unlikely(ret < 0))
1784 return ret;
1785 }
1786
1787 prev_key.objectid = key.objectid;
1788 prev_key.type = key.type;
1789 prev_key.offset = key.offset;
1790 }
1791
1792 return 0;
1793 }
1794
btrfs_check_leaf_full(struct extent_buffer * leaf)1795 int btrfs_check_leaf_full(struct extent_buffer *leaf)
1796 {
1797 return check_leaf(leaf, true);
1798 }
1799 ALLOW_ERROR_INJECTION(btrfs_check_leaf_full, ERRNO);
1800
btrfs_check_leaf_relaxed(struct extent_buffer * leaf)1801 int btrfs_check_leaf_relaxed(struct extent_buffer *leaf)
1802 {
1803 return check_leaf(leaf, false);
1804 }
1805
btrfs_check_node(struct extent_buffer * node)1806 int btrfs_check_node(struct extent_buffer *node)
1807 {
1808 struct btrfs_fs_info *fs_info = node->fs_info;
1809 unsigned long nr = btrfs_header_nritems(node);
1810 struct btrfs_key key, next_key;
1811 int slot;
1812 int level = btrfs_header_level(node);
1813 u64 bytenr;
1814 int ret = 0;
1815
1816 if (unlikely(level <= 0 || level >= BTRFS_MAX_LEVEL)) {
1817 generic_err(node, 0,
1818 "invalid level for node, have %d expect [1, %d]",
1819 level, BTRFS_MAX_LEVEL - 1);
1820 return -EUCLEAN;
1821 }
1822 if (unlikely(nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(fs_info))) {
1823 btrfs_crit(fs_info,
1824 "corrupt node: root=%llu block=%llu, nritems too %s, have %lu expect range [1,%u]",
1825 btrfs_header_owner(node), node->start,
1826 nr == 0 ? "small" : "large", nr,
1827 BTRFS_NODEPTRS_PER_BLOCK(fs_info));
1828 return -EUCLEAN;
1829 }
1830
1831 for (slot = 0; slot < nr - 1; slot++) {
1832 bytenr = btrfs_node_blockptr(node, slot);
1833 btrfs_node_key_to_cpu(node, &key, slot);
1834 btrfs_node_key_to_cpu(node, &next_key, slot + 1);
1835
1836 if (unlikely(!bytenr)) {
1837 generic_err(node, slot,
1838 "invalid NULL node pointer");
1839 ret = -EUCLEAN;
1840 goto out;
1841 }
1842 if (unlikely(!IS_ALIGNED(bytenr, fs_info->sectorsize))) {
1843 generic_err(node, slot,
1844 "unaligned pointer, have %llu should be aligned to %u",
1845 bytenr, fs_info->sectorsize);
1846 ret = -EUCLEAN;
1847 goto out;
1848 }
1849
1850 if (unlikely(btrfs_comp_cpu_keys(&key, &next_key) >= 0)) {
1851 generic_err(node, slot,
1852 "bad key order, current (%llu %u %llu) next (%llu %u %llu)",
1853 key.objectid, key.type, key.offset,
1854 next_key.objectid, next_key.type,
1855 next_key.offset);
1856 ret = -EUCLEAN;
1857 goto out;
1858 }
1859 }
1860 out:
1861 return ret;
1862 }
1863 ALLOW_ERROR_INJECTION(btrfs_check_node, ERRNO);
1864