Commit | Line | Data |
---|---|---|
c1d7c514 | 1 | // SPDX-License-Identifier: GPL-2.0 |
557ea5dd QW |
2 | /* |
3 | * Copyright (C) Qu Wenruo 2017. All rights reserved. | |
557ea5dd QW |
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 | ||
02529d7a QW |
18 | #include <linux/types.h> |
19 | #include <linux/stddef.h> | |
20 | #include <linux/error-injection.h> | |
557ea5dd QW |
21 | #include "ctree.h" |
22 | #include "tree-checker.h" | |
23 | #include "disk-io.h" | |
24 | #include "compression.h" | |
fce466ea | 25 | #include "volumes.h" |
c1499166 | 26 | #include "misc.h" |
557ea5dd | 27 | |
bba4f298 QW |
28 | /* |
29 | * Error message should follow the following format: | |
30 | * corrupt <type>: <identifier>, <reason>[, <bad_value>] | |
31 | * | |
32 | * @type: leaf or node | |
33 | * @identifier: the necessary info to locate the leaf/node. | |
52042d8e | 34 | * It's recommended to decode key.objecitd/offset if it's |
bba4f298 QW |
35 | * meaningful. |
36 | * @reason: describe the error | |
52042d8e | 37 | * @bad_value: optional, it's recommended to output bad value and its |
bba4f298 QW |
38 | * expected value (range). |
39 | * | |
40 | * Since comma is used to separate the components, only space is allowed | |
41 | * inside each component. | |
42 | */ | |
43 | ||
44 | /* | |
45 | * Append generic "corrupt leaf/node root=%llu block=%llu slot=%d: " to @fmt. | |
46 | * Allows callers to customize the output. | |
47 | */ | |
86a6be3a | 48 | __printf(3, 4) |
e67c718b | 49 | __cold |
86a6be3a | 50 | static void generic_err(const struct extent_buffer *eb, int slot, |
bba4f298 QW |
51 | const char *fmt, ...) |
52 | { | |
86a6be3a | 53 | const struct btrfs_fs_info *fs_info = eb->fs_info; |
bba4f298 QW |
54 | struct va_format vaf; |
55 | va_list args; | |
56 | ||
57 | va_start(args, fmt); | |
58 | ||
59 | vaf.fmt = fmt; | |
60 | vaf.va = &args; | |
61 | ||
2f659546 | 62 | btrfs_crit(fs_info, |
bba4f298 QW |
63 | "corrupt %s: root=%llu block=%llu slot=%d, %pV", |
64 | btrfs_header_level(eb) == 0 ? "leaf" : "node", | |
2f659546 | 65 | btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, &vaf); |
bba4f298 QW |
66 | va_end(args); |
67 | } | |
68 | ||
8806d718 QW |
69 | /* |
70 | * Customized reporter for extent data item, since its key objectid and | |
71 | * offset has its own meaning. | |
72 | */ | |
1fd715ff | 73 | __printf(3, 4) |
e67c718b | 74 | __cold |
1fd715ff | 75 | static void file_extent_err(const struct extent_buffer *eb, int slot, |
8806d718 QW |
76 | const char *fmt, ...) |
77 | { | |
1fd715ff | 78 | const struct btrfs_fs_info *fs_info = eb->fs_info; |
8806d718 QW |
79 | struct btrfs_key key; |
80 | struct va_format vaf; | |
81 | va_list args; | |
82 | ||
83 | btrfs_item_key_to_cpu(eb, &key, slot); | |
84 | va_start(args, fmt); | |
85 | ||
86 | vaf.fmt = fmt; | |
87 | vaf.va = &args; | |
88 | ||
2f659546 | 89 | btrfs_crit(fs_info, |
8806d718 | 90 | "corrupt %s: root=%llu block=%llu slot=%d ino=%llu file_offset=%llu, %pV", |
2f659546 QW |
91 | btrfs_header_level(eb) == 0 ? "leaf" : "node", |
92 | btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, | |
93 | key.objectid, key.offset, &vaf); | |
8806d718 QW |
94 | va_end(args); |
95 | } | |
96 | ||
97 | /* | |
98 | * Return 0 if the btrfs_file_extent_##name is aligned to @alignment | |
99 | * Else return 1 | |
100 | */ | |
033774dc | 101 | #define CHECK_FE_ALIGNED(leaf, slot, fi, name, alignment) \ |
8806d718 QW |
102 | ({ \ |
103 | if (!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment))) \ | |
1fd715ff | 104 | file_extent_err((leaf), (slot), \ |
8806d718 QW |
105 | "invalid %s for file extent, have %llu, should be aligned to %u", \ |
106 | (#name), btrfs_file_extent_##name((leaf), (fi)), \ | |
107 | (alignment)); \ | |
108 | (!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment))); \ | |
109 | }) | |
110 | ||
4e9845ef FM |
111 | static u64 file_extent_end(struct extent_buffer *leaf, |
112 | struct btrfs_key *key, | |
113 | struct btrfs_file_extent_item *extent) | |
114 | { | |
115 | u64 end; | |
116 | u64 len; | |
117 | ||
118 | if (btrfs_file_extent_type(leaf, extent) == BTRFS_FILE_EXTENT_INLINE) { | |
119 | len = btrfs_file_extent_ram_bytes(leaf, extent); | |
120 | end = ALIGN(key->offset + len, leaf->fs_info->sectorsize); | |
121 | } else { | |
122 | len = btrfs_file_extent_num_bytes(leaf, extent); | |
123 | end = key->offset + len; | |
124 | } | |
125 | return end; | |
126 | } | |
127 | ||
80d7fd1e QW |
128 | /* |
129 | * Customized report for dir_item, the only new important information is | |
130 | * key->objectid, which represents inode number | |
131 | */ | |
132 | __printf(3, 4) | |
133 | __cold | |
134 | static void dir_item_err(const struct extent_buffer *eb, int slot, | |
135 | const char *fmt, ...) | |
136 | { | |
137 | const struct btrfs_fs_info *fs_info = eb->fs_info; | |
138 | struct btrfs_key key; | |
139 | struct va_format vaf; | |
140 | va_list args; | |
141 | ||
142 | btrfs_item_key_to_cpu(eb, &key, slot); | |
143 | va_start(args, fmt); | |
144 | ||
145 | vaf.fmt = fmt; | |
146 | vaf.va = &args; | |
147 | ||
148 | btrfs_crit(fs_info, | |
149 | "corrupt %s: root=%llu block=%llu slot=%d ino=%llu, %pV", | |
150 | btrfs_header_level(eb) == 0 ? "leaf" : "node", | |
151 | btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, | |
152 | key.objectid, &vaf); | |
153 | va_end(args); | |
154 | } | |
155 | ||
156 | /* | |
157 | * This functions checks prev_key->objectid, to ensure current key and prev_key | |
158 | * share the same objectid as inode number. | |
159 | * | |
160 | * This is to detect missing INODE_ITEM in subvolume trees. | |
161 | * | |
162 | * Return true if everything is OK or we don't need to check. | |
163 | * Return false if anything is wrong. | |
164 | */ | |
165 | static bool check_prev_ino(struct extent_buffer *leaf, | |
166 | struct btrfs_key *key, int slot, | |
167 | struct btrfs_key *prev_key) | |
168 | { | |
169 | /* No prev key, skip check */ | |
170 | if (slot == 0) | |
171 | return true; | |
172 | ||
173 | /* Only these key->types needs to be checked */ | |
174 | ASSERT(key->type == BTRFS_XATTR_ITEM_KEY || | |
175 | key->type == BTRFS_INODE_REF_KEY || | |
176 | key->type == BTRFS_DIR_INDEX_KEY || | |
177 | key->type == BTRFS_DIR_ITEM_KEY || | |
178 | key->type == BTRFS_EXTENT_DATA_KEY); | |
179 | ||
180 | /* | |
181 | * Only subvolume trees along with their reloc trees need this check. | |
182 | * Things like log tree doesn't follow this ino requirement. | |
183 | */ | |
184 | if (!is_fstree(btrfs_header_owner(leaf))) | |
185 | return true; | |
186 | ||
187 | if (key->objectid == prev_key->objectid) | |
188 | return true; | |
189 | ||
190 | /* Error found */ | |
191 | dir_item_err(leaf, slot, | |
192 | "invalid previous key objectid, have %llu expect %llu", | |
193 | prev_key->objectid, key->objectid); | |
194 | return false; | |
195 | } | |
ae2a19d8 | 196 | static int check_extent_data_item(struct extent_buffer *leaf, |
4e9845ef FM |
197 | struct btrfs_key *key, int slot, |
198 | struct btrfs_key *prev_key) | |
557ea5dd | 199 | { |
ae2a19d8 | 200 | struct btrfs_fs_info *fs_info = leaf->fs_info; |
557ea5dd | 201 | struct btrfs_file_extent_item *fi; |
2f659546 | 202 | u32 sectorsize = fs_info->sectorsize; |
557ea5dd | 203 | u32 item_size = btrfs_item_size_nr(leaf, slot); |
4c094c33 | 204 | u64 extent_end; |
557ea5dd QW |
205 | |
206 | if (!IS_ALIGNED(key->offset, sectorsize)) { | |
1fd715ff | 207 | file_extent_err(leaf, slot, |
8806d718 QW |
208 | "unaligned file_offset for file extent, have %llu should be aligned to %u", |
209 | key->offset, sectorsize); | |
557ea5dd QW |
210 | return -EUCLEAN; |
211 | } | |
212 | ||
c18679eb QW |
213 | /* |
214 | * Previous key must have the same key->objectid (ino). | |
215 | * It can be XATTR_ITEM, INODE_ITEM or just another EXTENT_DATA. | |
216 | * But if objectids mismatch, it means we have a missing | |
217 | * INODE_ITEM. | |
218 | */ | |
80d7fd1e | 219 | if (!check_prev_ino(leaf, key, slot, prev_key)) |
c18679eb | 220 | return -EUCLEAN; |
c18679eb | 221 | |
557ea5dd QW |
222 | fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item); |
223 | ||
153a6d29 QW |
224 | /* |
225 | * Make sure the item contains at least inline header, so the file | |
226 | * extent type is not some garbage. | |
227 | */ | |
228 | if (item_size < BTRFS_FILE_EXTENT_INLINE_DATA_START) { | |
229 | file_extent_err(leaf, slot, | |
994bf9cd | 230 | "invalid item size, have %u expect [%zu, %u)", |
153a6d29 QW |
231 | item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START, |
232 | SZ_4K); | |
233 | return -EUCLEAN; | |
234 | } | |
b9b1a53e | 235 | if (btrfs_file_extent_type(leaf, fi) >= BTRFS_NR_FILE_EXTENT_TYPES) { |
1fd715ff | 236 | file_extent_err(leaf, slot, |
8806d718 QW |
237 | "invalid type for file extent, have %u expect range [0, %u]", |
238 | btrfs_file_extent_type(leaf, fi), | |
b9b1a53e | 239 | BTRFS_NR_FILE_EXTENT_TYPES - 1); |
557ea5dd QW |
240 | return -EUCLEAN; |
241 | } | |
242 | ||
243 | /* | |
52042d8e | 244 | * Support for new compression/encryption must introduce incompat flag, |
557ea5dd QW |
245 | * and must be caught in open_ctree(). |
246 | */ | |
ce96b7ff | 247 | if (btrfs_file_extent_compression(leaf, fi) >= BTRFS_NR_COMPRESS_TYPES) { |
1fd715ff | 248 | file_extent_err(leaf, slot, |
8806d718 QW |
249 | "invalid compression for file extent, have %u expect range [0, %u]", |
250 | btrfs_file_extent_compression(leaf, fi), | |
ce96b7ff | 251 | BTRFS_NR_COMPRESS_TYPES - 1); |
557ea5dd QW |
252 | return -EUCLEAN; |
253 | } | |
254 | if (btrfs_file_extent_encryption(leaf, fi)) { | |
1fd715ff | 255 | file_extent_err(leaf, slot, |
8806d718 QW |
256 | "invalid encryption for file extent, have %u expect 0", |
257 | btrfs_file_extent_encryption(leaf, fi)); | |
557ea5dd QW |
258 | return -EUCLEAN; |
259 | } | |
260 | if (btrfs_file_extent_type(leaf, fi) == BTRFS_FILE_EXTENT_INLINE) { | |
261 | /* Inline extent must have 0 as key offset */ | |
262 | if (key->offset) { | |
1fd715ff | 263 | file_extent_err(leaf, slot, |
8806d718 QW |
264 | "invalid file_offset for inline file extent, have %llu expect 0", |
265 | key->offset); | |
557ea5dd QW |
266 | return -EUCLEAN; |
267 | } | |
268 | ||
269 | /* Compressed inline extent has no on-disk size, skip it */ | |
270 | if (btrfs_file_extent_compression(leaf, fi) != | |
271 | BTRFS_COMPRESS_NONE) | |
272 | return 0; | |
273 | ||
274 | /* Uncompressed inline extent size must match item size */ | |
275 | if (item_size != BTRFS_FILE_EXTENT_INLINE_DATA_START + | |
276 | btrfs_file_extent_ram_bytes(leaf, fi)) { | |
1fd715ff | 277 | file_extent_err(leaf, slot, |
8806d718 QW |
278 | "invalid ram_bytes for uncompressed inline extent, have %u expect %llu", |
279 | item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START + | |
280 | btrfs_file_extent_ram_bytes(leaf, fi)); | |
557ea5dd QW |
281 | return -EUCLEAN; |
282 | } | |
283 | return 0; | |
284 | } | |
285 | ||
286 | /* Regular or preallocated extent has fixed item size */ | |
287 | if (item_size != sizeof(*fi)) { | |
1fd715ff | 288 | file_extent_err(leaf, slot, |
709a95c3 | 289 | "invalid item size for reg/prealloc file extent, have %u expect %zu", |
8806d718 | 290 | item_size, sizeof(*fi)); |
557ea5dd QW |
291 | return -EUCLEAN; |
292 | } | |
033774dc DS |
293 | if (CHECK_FE_ALIGNED(leaf, slot, fi, ram_bytes, sectorsize) || |
294 | CHECK_FE_ALIGNED(leaf, slot, fi, disk_bytenr, sectorsize) || | |
295 | CHECK_FE_ALIGNED(leaf, slot, fi, disk_num_bytes, sectorsize) || | |
296 | CHECK_FE_ALIGNED(leaf, slot, fi, offset, sectorsize) || | |
297 | CHECK_FE_ALIGNED(leaf, slot, fi, num_bytes, sectorsize)) | |
557ea5dd | 298 | return -EUCLEAN; |
4e9845ef | 299 | |
4c094c33 QW |
300 | /* Catch extent end overflow */ |
301 | if (check_add_overflow(btrfs_file_extent_num_bytes(leaf, fi), | |
302 | key->offset, &extent_end)) { | |
303 | file_extent_err(leaf, slot, | |
304 | "extent end overflow, have file offset %llu extent num bytes %llu", | |
305 | key->offset, | |
306 | btrfs_file_extent_num_bytes(leaf, fi)); | |
307 | return -EUCLEAN; | |
308 | } | |
309 | ||
4e9845ef FM |
310 | /* |
311 | * Check that no two consecutive file extent items, in the same leaf, | |
312 | * present ranges that overlap each other. | |
313 | */ | |
314 | if (slot > 0 && | |
315 | prev_key->objectid == key->objectid && | |
316 | prev_key->type == BTRFS_EXTENT_DATA_KEY) { | |
317 | struct btrfs_file_extent_item *prev_fi; | |
318 | u64 prev_end; | |
319 | ||
320 | prev_fi = btrfs_item_ptr(leaf, slot - 1, | |
321 | struct btrfs_file_extent_item); | |
322 | prev_end = file_extent_end(leaf, prev_key, prev_fi); | |
323 | if (prev_end > key->offset) { | |
324 | file_extent_err(leaf, slot - 1, | |
325 | "file extent end range (%llu) goes beyond start offset (%llu) of the next file extent", | |
326 | prev_end, key->offset); | |
327 | return -EUCLEAN; | |
328 | } | |
329 | } | |
330 | ||
557ea5dd QW |
331 | return 0; |
332 | } | |
333 | ||
68128ce7 | 334 | static int check_csum_item(struct extent_buffer *leaf, struct btrfs_key *key, |
ad1d8c43 | 335 | int slot, struct btrfs_key *prev_key) |
557ea5dd | 336 | { |
68128ce7 | 337 | struct btrfs_fs_info *fs_info = leaf->fs_info; |
2f659546 QW |
338 | u32 sectorsize = fs_info->sectorsize; |
339 | u32 csumsize = btrfs_super_csum_size(fs_info->super_copy); | |
557ea5dd QW |
340 | |
341 | if (key->objectid != BTRFS_EXTENT_CSUM_OBJECTID) { | |
86a6be3a | 342 | generic_err(leaf, slot, |
d508c5f0 QW |
343 | "invalid key objectid for csum item, have %llu expect %llu", |
344 | key->objectid, BTRFS_EXTENT_CSUM_OBJECTID); | |
557ea5dd QW |
345 | return -EUCLEAN; |
346 | } | |
347 | if (!IS_ALIGNED(key->offset, sectorsize)) { | |
86a6be3a | 348 | generic_err(leaf, slot, |
d508c5f0 QW |
349 | "unaligned key offset for csum item, have %llu should be aligned to %u", |
350 | key->offset, sectorsize); | |
557ea5dd QW |
351 | return -EUCLEAN; |
352 | } | |
353 | if (!IS_ALIGNED(btrfs_item_size_nr(leaf, slot), csumsize)) { | |
86a6be3a | 354 | generic_err(leaf, slot, |
d508c5f0 QW |
355 | "unaligned item size for csum item, have %u should be aligned to %u", |
356 | btrfs_item_size_nr(leaf, slot), csumsize); | |
557ea5dd QW |
357 | return -EUCLEAN; |
358 | } | |
ad1d8c43 FM |
359 | if (slot > 0 && prev_key->type == BTRFS_EXTENT_CSUM_KEY) { |
360 | u64 prev_csum_end; | |
361 | u32 prev_item_size; | |
362 | ||
363 | prev_item_size = btrfs_item_size_nr(leaf, slot - 1); | |
364 | prev_csum_end = (prev_item_size / csumsize) * sectorsize; | |
365 | prev_csum_end += prev_key->offset; | |
366 | if (prev_csum_end > key->offset) { | |
367 | generic_err(leaf, slot - 1, | |
368 | "csum end range (%llu) goes beyond the start range (%llu) of the next csum item", | |
369 | prev_csum_end, key->offset); | |
370 | return -EUCLEAN; | |
371 | } | |
372 | } | |
557ea5dd QW |
373 | return 0; |
374 | } | |
375 | ||
c23c77b0 QW |
376 | /* Inode item error output has the same format as dir_item_err() */ |
377 | #define inode_item_err(eb, slot, fmt, ...) \ | |
378 | dir_item_err(eb, slot, fmt, __VA_ARGS__) | |
379 | ||
380 | static int check_inode_key(struct extent_buffer *leaf, struct btrfs_key *key, | |
381 | int slot) | |
382 | { | |
383 | struct btrfs_key item_key; | |
384 | bool is_inode_item; | |
385 | ||
386 | btrfs_item_key_to_cpu(leaf, &item_key, slot); | |
387 | is_inode_item = (item_key.type == BTRFS_INODE_ITEM_KEY); | |
388 | ||
389 | /* For XATTR_ITEM, location key should be all 0 */ | |
390 | if (item_key.type == BTRFS_XATTR_ITEM_KEY) { | |
391 | if (key->type != 0 || key->objectid != 0 || key->offset != 0) | |
392 | return -EUCLEAN; | |
393 | return 0; | |
394 | } | |
395 | ||
396 | if ((key->objectid < BTRFS_FIRST_FREE_OBJECTID || | |
397 | key->objectid > BTRFS_LAST_FREE_OBJECTID) && | |
398 | key->objectid != BTRFS_ROOT_TREE_DIR_OBJECTID && | |
399 | key->objectid != BTRFS_FREE_INO_OBJECTID) { | |
400 | if (is_inode_item) { | |
401 | generic_err(leaf, slot, | |
402 | "invalid key objectid: has %llu expect %llu or [%llu, %llu] or %llu", | |
403 | key->objectid, BTRFS_ROOT_TREE_DIR_OBJECTID, | |
404 | BTRFS_FIRST_FREE_OBJECTID, | |
405 | BTRFS_LAST_FREE_OBJECTID, | |
406 | BTRFS_FREE_INO_OBJECTID); | |
407 | } else { | |
408 | dir_item_err(leaf, slot, | |
409 | "invalid location key objectid: has %llu expect %llu or [%llu, %llu] or %llu", | |
410 | key->objectid, BTRFS_ROOT_TREE_DIR_OBJECTID, | |
411 | BTRFS_FIRST_FREE_OBJECTID, | |
412 | BTRFS_LAST_FREE_OBJECTID, | |
413 | BTRFS_FREE_INO_OBJECTID); | |
414 | } | |
415 | return -EUCLEAN; | |
416 | } | |
417 | if (key->offset != 0) { | |
418 | if (is_inode_item) | |
419 | inode_item_err(leaf, slot, | |
420 | "invalid key offset: has %llu expect 0", | |
421 | key->offset); | |
422 | else | |
423 | dir_item_err(leaf, slot, | |
424 | "invalid location key offset:has %llu expect 0", | |
425 | key->offset); | |
426 | return -EUCLEAN; | |
427 | } | |
428 | return 0; | |
429 | } | |
430 | ||
57a0e674 QW |
431 | static int check_root_key(struct extent_buffer *leaf, struct btrfs_key *key, |
432 | int slot) | |
433 | { | |
434 | struct btrfs_key item_key; | |
435 | bool is_root_item; | |
436 | ||
437 | btrfs_item_key_to_cpu(leaf, &item_key, slot); | |
438 | is_root_item = (item_key.type == BTRFS_ROOT_ITEM_KEY); | |
439 | ||
440 | /* No such tree id */ | |
441 | if (key->objectid == 0) { | |
442 | if (is_root_item) | |
443 | generic_err(leaf, slot, "invalid root id 0"); | |
444 | else | |
445 | dir_item_err(leaf, slot, | |
446 | "invalid location key root id 0"); | |
447 | return -EUCLEAN; | |
448 | } | |
449 | ||
450 | /* DIR_ITEM/INDEX/INODE_REF is not allowed to point to non-fs trees */ | |
451 | if (!is_fstree(key->objectid) && !is_root_item) { | |
452 | dir_item_err(leaf, slot, | |
453 | "invalid location key objectid, have %llu expect [%llu, %llu]", | |
454 | key->objectid, BTRFS_FIRST_FREE_OBJECTID, | |
455 | BTRFS_LAST_FREE_OBJECTID); | |
456 | return -EUCLEAN; | |
457 | } | |
458 | ||
459 | /* | |
460 | * ROOT_ITEM with non-zero offset means this is a snapshot, created at | |
461 | * @offset transid. | |
462 | * Furthermore, for location key in DIR_ITEM, its offset is always -1. | |
463 | * | |
464 | * So here we only check offset for reloc tree whose key->offset must | |
465 | * be a valid tree. | |
466 | */ | |
467 | if (key->objectid == BTRFS_TREE_RELOC_OBJECTID && key->offset == 0) { | |
468 | generic_err(leaf, slot, "invalid root id 0 for reloc tree"); | |
469 | return -EUCLEAN; | |
470 | } | |
471 | return 0; | |
472 | } | |
473 | ||
ce4252c0 | 474 | static int check_dir_item(struct extent_buffer *leaf, |
c18679eb QW |
475 | struct btrfs_key *key, struct btrfs_key *prev_key, |
476 | int slot) | |
ad7b0368 | 477 | { |
ce4252c0 | 478 | struct btrfs_fs_info *fs_info = leaf->fs_info; |
ad7b0368 QW |
479 | struct btrfs_dir_item *di; |
480 | u32 item_size = btrfs_item_size_nr(leaf, slot); | |
481 | u32 cur = 0; | |
482 | ||
80d7fd1e | 483 | if (!check_prev_ino(leaf, key, slot, prev_key)) |
c18679eb | 484 | return -EUCLEAN; |
ad7b0368 QW |
485 | di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item); |
486 | while (cur < item_size) { | |
147a097c | 487 | struct btrfs_key location_key; |
ad7b0368 QW |
488 | u32 name_len; |
489 | u32 data_len; | |
490 | u32 max_name_len; | |
491 | u32 total_size; | |
492 | u32 name_hash; | |
493 | u8 dir_type; | |
147a097c | 494 | int ret; |
ad7b0368 QW |
495 | |
496 | /* header itself should not cross item boundary */ | |
497 | if (cur + sizeof(*di) > item_size) { | |
d98ced68 | 498 | dir_item_err(leaf, slot, |
7cfad652 | 499 | "dir item header crosses item boundary, have %zu boundary %u", |
ad7b0368 QW |
500 | cur + sizeof(*di), item_size); |
501 | return -EUCLEAN; | |
502 | } | |
503 | ||
147a097c QW |
504 | /* Location key check */ |
505 | btrfs_dir_item_key_to_cpu(leaf, di, &location_key); | |
506 | if (location_key.type == BTRFS_ROOT_ITEM_KEY) { | |
507 | ret = check_root_key(leaf, &location_key, slot); | |
508 | if (ret < 0) | |
509 | return ret; | |
510 | } else if (location_key.type == BTRFS_INODE_ITEM_KEY || | |
511 | location_key.type == 0) { | |
512 | ret = check_inode_key(leaf, &location_key, slot); | |
513 | if (ret < 0) | |
514 | return ret; | |
515 | } else { | |
516 | dir_item_err(leaf, slot, | |
517 | "invalid location key type, have %u, expect %u or %u", | |
518 | location_key.type, BTRFS_ROOT_ITEM_KEY, | |
519 | BTRFS_INODE_ITEM_KEY); | |
520 | return -EUCLEAN; | |
521 | } | |
522 | ||
ad7b0368 QW |
523 | /* dir type check */ |
524 | dir_type = btrfs_dir_type(leaf, di); | |
525 | if (dir_type >= BTRFS_FT_MAX) { | |
d98ced68 | 526 | dir_item_err(leaf, slot, |
ad7b0368 QW |
527 | "invalid dir item type, have %u expect [0, %u)", |
528 | dir_type, BTRFS_FT_MAX); | |
529 | return -EUCLEAN; | |
530 | } | |
531 | ||
532 | if (key->type == BTRFS_XATTR_ITEM_KEY && | |
533 | dir_type != BTRFS_FT_XATTR) { | |
d98ced68 | 534 | dir_item_err(leaf, slot, |
ad7b0368 QW |
535 | "invalid dir item type for XATTR key, have %u expect %u", |
536 | dir_type, BTRFS_FT_XATTR); | |
537 | return -EUCLEAN; | |
538 | } | |
539 | if (dir_type == BTRFS_FT_XATTR && | |
540 | key->type != BTRFS_XATTR_ITEM_KEY) { | |
d98ced68 | 541 | dir_item_err(leaf, slot, |
ad7b0368 QW |
542 | "xattr dir type found for non-XATTR key"); |
543 | return -EUCLEAN; | |
544 | } | |
545 | if (dir_type == BTRFS_FT_XATTR) | |
546 | max_name_len = XATTR_NAME_MAX; | |
547 | else | |
548 | max_name_len = BTRFS_NAME_LEN; | |
549 | ||
550 | /* Name/data length check */ | |
551 | name_len = btrfs_dir_name_len(leaf, di); | |
552 | data_len = btrfs_dir_data_len(leaf, di); | |
553 | if (name_len > max_name_len) { | |
d98ced68 | 554 | dir_item_err(leaf, slot, |
ad7b0368 QW |
555 | "dir item name len too long, have %u max %u", |
556 | name_len, max_name_len); | |
557 | return -EUCLEAN; | |
558 | } | |
2f659546 | 559 | if (name_len + data_len > BTRFS_MAX_XATTR_SIZE(fs_info)) { |
d98ced68 | 560 | dir_item_err(leaf, slot, |
ad7b0368 QW |
561 | "dir item name and data len too long, have %u max %u", |
562 | name_len + data_len, | |
2f659546 | 563 | BTRFS_MAX_XATTR_SIZE(fs_info)); |
ad7b0368 QW |
564 | return -EUCLEAN; |
565 | } | |
566 | ||
567 | if (data_len && dir_type != BTRFS_FT_XATTR) { | |
d98ced68 | 568 | dir_item_err(leaf, slot, |
ad7b0368 QW |
569 | "dir item with invalid data len, have %u expect 0", |
570 | data_len); | |
571 | return -EUCLEAN; | |
572 | } | |
573 | ||
574 | total_size = sizeof(*di) + name_len + data_len; | |
575 | ||
576 | /* header and name/data should not cross item boundary */ | |
577 | if (cur + total_size > item_size) { | |
d98ced68 | 578 | dir_item_err(leaf, slot, |
ad7b0368 QW |
579 | "dir item data crosses item boundary, have %u boundary %u", |
580 | cur + total_size, item_size); | |
581 | return -EUCLEAN; | |
582 | } | |
583 | ||
584 | /* | |
585 | * Special check for XATTR/DIR_ITEM, as key->offset is name | |
586 | * hash, should match its name | |
587 | */ | |
588 | if (key->type == BTRFS_DIR_ITEM_KEY || | |
589 | key->type == BTRFS_XATTR_ITEM_KEY) { | |
e2683fc9 DS |
590 | char namebuf[max(BTRFS_NAME_LEN, XATTR_NAME_MAX)]; |
591 | ||
ad7b0368 QW |
592 | read_extent_buffer(leaf, namebuf, |
593 | (unsigned long)(di + 1), name_len); | |
594 | name_hash = btrfs_name_hash(namebuf, name_len); | |
595 | if (key->offset != name_hash) { | |
d98ced68 | 596 | dir_item_err(leaf, slot, |
ad7b0368 QW |
597 | "name hash mismatch with key, have 0x%016x expect 0x%016llx", |
598 | name_hash, key->offset); | |
599 | return -EUCLEAN; | |
600 | } | |
601 | } | |
602 | cur += total_size; | |
603 | di = (struct btrfs_dir_item *)((void *)di + total_size); | |
604 | } | |
605 | return 0; | |
606 | } | |
607 | ||
4806bd88 | 608 | __printf(3, 4) |
fce466ea | 609 | __cold |
4806bd88 | 610 | static void block_group_err(const struct extent_buffer *eb, int slot, |
fce466ea QW |
611 | const char *fmt, ...) |
612 | { | |
4806bd88 | 613 | const struct btrfs_fs_info *fs_info = eb->fs_info; |
fce466ea QW |
614 | struct btrfs_key key; |
615 | struct va_format vaf; | |
616 | va_list args; | |
617 | ||
618 | btrfs_item_key_to_cpu(eb, &key, slot); | |
619 | va_start(args, fmt); | |
620 | ||
621 | vaf.fmt = fmt; | |
622 | vaf.va = &args; | |
623 | ||
624 | btrfs_crit(fs_info, | |
625 | "corrupt %s: root=%llu block=%llu slot=%d bg_start=%llu bg_len=%llu, %pV", | |
626 | btrfs_header_level(eb) == 0 ? "leaf" : "node", | |
627 | btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, | |
628 | key.objectid, key.offset, &vaf); | |
629 | va_end(args); | |
630 | } | |
631 | ||
af60ce2b | 632 | static int check_block_group_item(struct extent_buffer *leaf, |
fce466ea QW |
633 | struct btrfs_key *key, int slot) |
634 | { | |
635 | struct btrfs_block_group_item bgi; | |
636 | u32 item_size = btrfs_item_size_nr(leaf, slot); | |
637 | u64 flags; | |
638 | u64 type; | |
639 | ||
640 | /* | |
641 | * Here we don't really care about alignment since extent allocator can | |
10950929 | 642 | * handle it. We care more about the size. |
fce466ea | 643 | */ |
10950929 | 644 | if (key->offset == 0) { |
4806bd88 | 645 | block_group_err(leaf, slot, |
10950929 | 646 | "invalid block group size 0"); |
fce466ea QW |
647 | return -EUCLEAN; |
648 | } | |
649 | ||
650 | if (item_size != sizeof(bgi)) { | |
4806bd88 | 651 | block_group_err(leaf, slot, |
fce466ea QW |
652 | "invalid item size, have %u expect %zu", |
653 | item_size, sizeof(bgi)); | |
654 | return -EUCLEAN; | |
655 | } | |
656 | ||
657 | read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot), | |
658 | sizeof(bgi)); | |
de0dc456 | 659 | if (btrfs_stack_block_group_chunk_objectid(&bgi) != |
fce466ea | 660 | BTRFS_FIRST_CHUNK_TREE_OBJECTID) { |
4806bd88 | 661 | block_group_err(leaf, slot, |
fce466ea | 662 | "invalid block group chunk objectid, have %llu expect %llu", |
de0dc456 | 663 | btrfs_stack_block_group_chunk_objectid(&bgi), |
fce466ea QW |
664 | BTRFS_FIRST_CHUNK_TREE_OBJECTID); |
665 | return -EUCLEAN; | |
666 | } | |
667 | ||
de0dc456 | 668 | if (btrfs_stack_block_group_used(&bgi) > key->offset) { |
4806bd88 | 669 | block_group_err(leaf, slot, |
fce466ea | 670 | "invalid block group used, have %llu expect [0, %llu)", |
de0dc456 | 671 | btrfs_stack_block_group_used(&bgi), key->offset); |
fce466ea QW |
672 | return -EUCLEAN; |
673 | } | |
674 | ||
de0dc456 | 675 | flags = btrfs_stack_block_group_flags(&bgi); |
fce466ea | 676 | if (hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) > 1) { |
4806bd88 | 677 | block_group_err(leaf, slot, |
fce466ea QW |
678 | "invalid profile flags, have 0x%llx (%lu bits set) expect no more than 1 bit set", |
679 | flags & BTRFS_BLOCK_GROUP_PROFILE_MASK, | |
680 | hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)); | |
681 | return -EUCLEAN; | |
682 | } | |
683 | ||
684 | type = flags & BTRFS_BLOCK_GROUP_TYPE_MASK; | |
685 | if (type != BTRFS_BLOCK_GROUP_DATA && | |
686 | type != BTRFS_BLOCK_GROUP_METADATA && | |
687 | type != BTRFS_BLOCK_GROUP_SYSTEM && | |
688 | type != (BTRFS_BLOCK_GROUP_METADATA | | |
689 | BTRFS_BLOCK_GROUP_DATA)) { | |
4806bd88 | 690 | block_group_err(leaf, slot, |
761333f2 | 691 | "invalid type, have 0x%llx (%lu bits set) expect either 0x%llx, 0x%llx, 0x%llx or 0x%llx", |
fce466ea QW |
692 | type, hweight64(type), |
693 | BTRFS_BLOCK_GROUP_DATA, BTRFS_BLOCK_GROUP_METADATA, | |
694 | BTRFS_BLOCK_GROUP_SYSTEM, | |
695 | BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA); | |
696 | return -EUCLEAN; | |
697 | } | |
698 | return 0; | |
82fc28fb QW |
699 | } |
700 | ||
d001e4a3 | 701 | __printf(4, 5) |
f1140243 | 702 | __cold |
d001e4a3 | 703 | static void chunk_err(const struct extent_buffer *leaf, |
f1140243 QW |
704 | const struct btrfs_chunk *chunk, u64 logical, |
705 | const char *fmt, ...) | |
706 | { | |
d001e4a3 | 707 | const struct btrfs_fs_info *fs_info = leaf->fs_info; |
f1140243 QW |
708 | bool is_sb; |
709 | struct va_format vaf; | |
710 | va_list args; | |
711 | int i; | |
712 | int slot = -1; | |
713 | ||
714 | /* Only superblock eb is able to have such small offset */ | |
715 | is_sb = (leaf->start == BTRFS_SUPER_INFO_OFFSET); | |
716 | ||
717 | if (!is_sb) { | |
718 | /* | |
719 | * Get the slot number by iterating through all slots, this | |
720 | * would provide better readability. | |
721 | */ | |
722 | for (i = 0; i < btrfs_header_nritems(leaf); i++) { | |
723 | if (btrfs_item_ptr_offset(leaf, i) == | |
724 | (unsigned long)chunk) { | |
725 | slot = i; | |
726 | break; | |
727 | } | |
728 | } | |
729 | } | |
730 | va_start(args, fmt); | |
731 | vaf.fmt = fmt; | |
732 | vaf.va = &args; | |
733 | ||
734 | if (is_sb) | |
735 | btrfs_crit(fs_info, | |
736 | "corrupt superblock syschunk array: chunk_start=%llu, %pV", | |
737 | logical, &vaf); | |
738 | else | |
739 | btrfs_crit(fs_info, | |
740 | "corrupt leaf: root=%llu block=%llu slot=%d chunk_start=%llu, %pV", | |
741 | BTRFS_CHUNK_TREE_OBJECTID, leaf->start, slot, | |
742 | logical, &vaf); | |
743 | va_end(args); | |
744 | } | |
745 | ||
82fc28fb QW |
746 | /* |
747 | * The common chunk check which could also work on super block sys chunk array. | |
748 | * | |
bf871c3b | 749 | * Return -EUCLEAN if anything is corrupted. |
82fc28fb QW |
750 | * Return 0 if everything is OK. |
751 | */ | |
ddaf1d5a | 752 | int btrfs_check_chunk_valid(struct extent_buffer *leaf, |
82fc28fb QW |
753 | struct btrfs_chunk *chunk, u64 logical) |
754 | { | |
ddaf1d5a | 755 | struct btrfs_fs_info *fs_info = leaf->fs_info; |
82fc28fb QW |
756 | u64 length; |
757 | u64 stripe_len; | |
758 | u16 num_stripes; | |
759 | u16 sub_stripes; | |
760 | u64 type; | |
761 | u64 features; | |
762 | bool mixed = false; | |
763 | ||
764 | length = btrfs_chunk_length(leaf, chunk); | |
765 | stripe_len = btrfs_chunk_stripe_len(leaf, chunk); | |
766 | num_stripes = btrfs_chunk_num_stripes(leaf, chunk); | |
767 | sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk); | |
768 | type = btrfs_chunk_type(leaf, chunk); | |
769 | ||
770 | if (!num_stripes) { | |
d001e4a3 | 771 | chunk_err(leaf, chunk, logical, |
f1140243 | 772 | "invalid chunk num_stripes, have %u", num_stripes); |
bf871c3b | 773 | return -EUCLEAN; |
82fc28fb QW |
774 | } |
775 | if (!IS_ALIGNED(logical, fs_info->sectorsize)) { | |
d001e4a3 | 776 | chunk_err(leaf, chunk, logical, |
f1140243 QW |
777 | "invalid chunk logical, have %llu should aligned to %u", |
778 | logical, fs_info->sectorsize); | |
bf871c3b | 779 | return -EUCLEAN; |
82fc28fb QW |
780 | } |
781 | if (btrfs_chunk_sector_size(leaf, chunk) != fs_info->sectorsize) { | |
d001e4a3 | 782 | chunk_err(leaf, chunk, logical, |
f1140243 QW |
783 | "invalid chunk sectorsize, have %u expect %u", |
784 | btrfs_chunk_sector_size(leaf, chunk), | |
785 | fs_info->sectorsize); | |
bf871c3b | 786 | return -EUCLEAN; |
82fc28fb QW |
787 | } |
788 | if (!length || !IS_ALIGNED(length, fs_info->sectorsize)) { | |
d001e4a3 | 789 | chunk_err(leaf, chunk, logical, |
f1140243 | 790 | "invalid chunk length, have %llu", length); |
bf871c3b | 791 | return -EUCLEAN; |
82fc28fb QW |
792 | } |
793 | if (!is_power_of_2(stripe_len) || stripe_len != BTRFS_STRIPE_LEN) { | |
d001e4a3 | 794 | chunk_err(leaf, chunk, logical, |
f1140243 | 795 | "invalid chunk stripe length: %llu", |
82fc28fb | 796 | stripe_len); |
bf871c3b | 797 | return -EUCLEAN; |
82fc28fb QW |
798 | } |
799 | if (~(BTRFS_BLOCK_GROUP_TYPE_MASK | BTRFS_BLOCK_GROUP_PROFILE_MASK) & | |
800 | type) { | |
d001e4a3 | 801 | chunk_err(leaf, chunk, logical, |
f1140243 | 802 | "unrecognized chunk type: 0x%llx", |
82fc28fb QW |
803 | ~(BTRFS_BLOCK_GROUP_TYPE_MASK | |
804 | BTRFS_BLOCK_GROUP_PROFILE_MASK) & | |
805 | btrfs_chunk_type(leaf, chunk)); | |
bf871c3b | 806 | return -EUCLEAN; |
82fc28fb QW |
807 | } |
808 | ||
c1499166 | 809 | if (!has_single_bit_set(type & BTRFS_BLOCK_GROUP_PROFILE_MASK) && |
80e46cf2 | 810 | (type & BTRFS_BLOCK_GROUP_PROFILE_MASK) != 0) { |
d001e4a3 | 811 | chunk_err(leaf, chunk, logical, |
80e46cf2 QW |
812 | "invalid chunk profile flag: 0x%llx, expect 0 or 1 bit set", |
813 | type & BTRFS_BLOCK_GROUP_PROFILE_MASK); | |
814 | return -EUCLEAN; | |
815 | } | |
82fc28fb | 816 | if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == 0) { |
d001e4a3 | 817 | chunk_err(leaf, chunk, logical, |
f1140243 QW |
818 | "missing chunk type flag, have 0x%llx one bit must be set in 0x%llx", |
819 | type, BTRFS_BLOCK_GROUP_TYPE_MASK); | |
bf871c3b | 820 | return -EUCLEAN; |
82fc28fb QW |
821 | } |
822 | ||
823 | if ((type & BTRFS_BLOCK_GROUP_SYSTEM) && | |
824 | (type & (BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA))) { | |
d001e4a3 | 825 | chunk_err(leaf, chunk, logical, |
f1140243 QW |
826 | "system chunk with data or metadata type: 0x%llx", |
827 | type); | |
bf871c3b | 828 | return -EUCLEAN; |
82fc28fb QW |
829 | } |
830 | ||
831 | features = btrfs_super_incompat_flags(fs_info->super_copy); | |
832 | if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) | |
833 | mixed = true; | |
834 | ||
835 | if (!mixed) { | |
836 | if ((type & BTRFS_BLOCK_GROUP_METADATA) && | |
837 | (type & BTRFS_BLOCK_GROUP_DATA)) { | |
d001e4a3 | 838 | chunk_err(leaf, chunk, logical, |
82fc28fb | 839 | "mixed chunk type in non-mixed mode: 0x%llx", type); |
bf871c3b | 840 | return -EUCLEAN; |
82fc28fb QW |
841 | } |
842 | } | |
843 | ||
844 | if ((type & BTRFS_BLOCK_GROUP_RAID10 && sub_stripes != 2) || | |
845 | (type & BTRFS_BLOCK_GROUP_RAID1 && num_stripes != 2) || | |
846 | (type & BTRFS_BLOCK_GROUP_RAID5 && num_stripes < 2) || | |
847 | (type & BTRFS_BLOCK_GROUP_RAID6 && num_stripes < 3) || | |
848 | (type & BTRFS_BLOCK_GROUP_DUP && num_stripes != 2) || | |
849 | ((type & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 && num_stripes != 1)) { | |
d001e4a3 | 850 | chunk_err(leaf, chunk, logical, |
82fc28fb QW |
851 | "invalid num_stripes:sub_stripes %u:%u for profile %llu", |
852 | num_stripes, sub_stripes, | |
853 | type & BTRFS_BLOCK_GROUP_PROFILE_MASK); | |
bf871c3b | 854 | return -EUCLEAN; |
82fc28fb QW |
855 | } |
856 | ||
857 | return 0; | |
fce466ea QW |
858 | } |
859 | ||
f6d2a5c2 QW |
860 | /* |
861 | * Enhanced version of chunk item checker. | |
862 | * | |
863 | * The common btrfs_check_chunk_valid() doesn't check item size since it needs | |
864 | * to work on super block sys_chunk_array which doesn't have full item ptr. | |
865 | */ | |
866 | static int check_leaf_chunk_item(struct extent_buffer *leaf, | |
867 | struct btrfs_chunk *chunk, | |
868 | struct btrfs_key *key, int slot) | |
869 | { | |
870 | int num_stripes; | |
871 | ||
872 | if (btrfs_item_size_nr(leaf, slot) < sizeof(struct btrfs_chunk)) { | |
873 | chunk_err(leaf, chunk, key->offset, | |
874 | "invalid chunk item size: have %u expect [%zu, %u)", | |
875 | btrfs_item_size_nr(leaf, slot), | |
876 | sizeof(struct btrfs_chunk), | |
877 | BTRFS_LEAF_DATA_SIZE(leaf->fs_info)); | |
878 | return -EUCLEAN; | |
879 | } | |
880 | ||
881 | num_stripes = btrfs_chunk_num_stripes(leaf, chunk); | |
882 | /* Let btrfs_check_chunk_valid() handle this error type */ | |
883 | if (num_stripes == 0) | |
884 | goto out; | |
885 | ||
886 | if (btrfs_chunk_item_size(num_stripes) != | |
887 | btrfs_item_size_nr(leaf, slot)) { | |
888 | chunk_err(leaf, chunk, key->offset, | |
889 | "invalid chunk item size: have %u expect %lu", | |
890 | btrfs_item_size_nr(leaf, slot), | |
891 | btrfs_chunk_item_size(num_stripes)); | |
892 | return -EUCLEAN; | |
893 | } | |
894 | out: | |
895 | return btrfs_check_chunk_valid(leaf, chunk, key->offset); | |
896 | } | |
897 | ||
5617ed80 | 898 | __printf(3, 4) |
ab4ba2e1 | 899 | __cold |
5617ed80 | 900 | static void dev_item_err(const struct extent_buffer *eb, int slot, |
ab4ba2e1 QW |
901 | const char *fmt, ...) |
902 | { | |
903 | struct btrfs_key key; | |
904 | struct va_format vaf; | |
905 | va_list args; | |
906 | ||
907 | btrfs_item_key_to_cpu(eb, &key, slot); | |
908 | va_start(args, fmt); | |
909 | ||
910 | vaf.fmt = fmt; | |
911 | vaf.va = &args; | |
912 | ||
5617ed80 | 913 | btrfs_crit(eb->fs_info, |
ab4ba2e1 QW |
914 | "corrupt %s: root=%llu block=%llu slot=%d devid=%llu %pV", |
915 | btrfs_header_level(eb) == 0 ? "leaf" : "node", | |
916 | btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, | |
917 | key.objectid, &vaf); | |
918 | va_end(args); | |
919 | } | |
920 | ||
412a2312 | 921 | static int check_dev_item(struct extent_buffer *leaf, |
ab4ba2e1 QW |
922 | struct btrfs_key *key, int slot) |
923 | { | |
924 | struct btrfs_dev_item *ditem; | |
ab4ba2e1 QW |
925 | |
926 | if (key->objectid != BTRFS_DEV_ITEMS_OBJECTID) { | |
5617ed80 | 927 | dev_item_err(leaf, slot, |
ab4ba2e1 QW |
928 | "invalid objectid: has=%llu expect=%llu", |
929 | key->objectid, BTRFS_DEV_ITEMS_OBJECTID); | |
930 | return -EUCLEAN; | |
931 | } | |
ab4ba2e1 QW |
932 | ditem = btrfs_item_ptr(leaf, slot, struct btrfs_dev_item); |
933 | if (btrfs_device_id(leaf, ditem) != key->offset) { | |
5617ed80 | 934 | dev_item_err(leaf, slot, |
ab4ba2e1 QW |
935 | "devid mismatch: key has=%llu item has=%llu", |
936 | key->offset, btrfs_device_id(leaf, ditem)); | |
937 | return -EUCLEAN; | |
938 | } | |
939 | ||
940 | /* | |
941 | * For device total_bytes, we don't have reliable way to check it, as | |
942 | * it can be 0 for device removal. Device size check can only be done | |
943 | * by dev extents check. | |
944 | */ | |
945 | if (btrfs_device_bytes_used(leaf, ditem) > | |
946 | btrfs_device_total_bytes(leaf, ditem)) { | |
5617ed80 | 947 | dev_item_err(leaf, slot, |
ab4ba2e1 QW |
948 | "invalid bytes used: have %llu expect [0, %llu]", |
949 | btrfs_device_bytes_used(leaf, ditem), | |
950 | btrfs_device_total_bytes(leaf, ditem)); | |
951 | return -EUCLEAN; | |
952 | } | |
953 | /* | |
954 | * Remaining members like io_align/type/gen/dev_group aren't really | |
955 | * utilized. Skip them to make later usage of them easier. | |
956 | */ | |
957 | return 0; | |
958 | } | |
959 | ||
39e57f49 | 960 | static int check_inode_item(struct extent_buffer *leaf, |
496245ca QW |
961 | struct btrfs_key *key, int slot) |
962 | { | |
39e57f49 | 963 | struct btrfs_fs_info *fs_info = leaf->fs_info; |
496245ca QW |
964 | struct btrfs_inode_item *iitem; |
965 | u64 super_gen = btrfs_super_generation(fs_info->super_copy); | |
966 | u32 valid_mask = (S_IFMT | S_ISUID | S_ISGID | S_ISVTX | 0777); | |
967 | u32 mode; | |
c23c77b0 QW |
968 | int ret; |
969 | ||
970 | ret = check_inode_key(leaf, key, slot); | |
971 | if (ret < 0) | |
972 | return ret; | |
496245ca | 973 | |
496245ca QW |
974 | iitem = btrfs_item_ptr(leaf, slot, struct btrfs_inode_item); |
975 | ||
976 | /* Here we use super block generation + 1 to handle log tree */ | |
977 | if (btrfs_inode_generation(leaf, iitem) > super_gen + 1) { | |
c3053ebb | 978 | inode_item_err(leaf, slot, |
496245ca QW |
979 | "invalid inode generation: has %llu expect (0, %llu]", |
980 | btrfs_inode_generation(leaf, iitem), | |
981 | super_gen + 1); | |
982 | return -EUCLEAN; | |
983 | } | |
984 | /* Note for ROOT_TREE_DIR_ITEM, mkfs could set its transid 0 */ | |
985 | if (btrfs_inode_transid(leaf, iitem) > super_gen + 1) { | |
c3053ebb | 986 | inode_item_err(leaf, slot, |
f96d6960 | 987 | "invalid inode transid: has %llu expect [0, %llu]", |
496245ca QW |
988 | btrfs_inode_transid(leaf, iitem), super_gen + 1); |
989 | return -EUCLEAN; | |
990 | } | |
991 | ||
992 | /* | |
993 | * For size and nbytes it's better not to be too strict, as for dir | |
994 | * item its size/nbytes can easily get wrong, but doesn't affect | |
995 | * anything in the fs. So here we skip the check. | |
996 | */ | |
997 | mode = btrfs_inode_mode(leaf, iitem); | |
998 | if (mode & ~valid_mask) { | |
c3053ebb | 999 | inode_item_err(leaf, slot, |
496245ca QW |
1000 | "unknown mode bit detected: 0x%x", |
1001 | mode & ~valid_mask); | |
1002 | return -EUCLEAN; | |
1003 | } | |
1004 | ||
1005 | /* | |
c1499166 DS |
1006 | * S_IFMT is not bit mapped so we can't completely rely on |
1007 | * is_power_of_2/has_single_bit_set, but it can save us from checking | |
1008 | * FIFO/CHR/DIR/REG. Only needs to check BLK, LNK and SOCKS | |
496245ca | 1009 | */ |
c1499166 | 1010 | if (!has_single_bit_set(mode & S_IFMT)) { |
496245ca | 1011 | if (!S_ISLNK(mode) && !S_ISBLK(mode) && !S_ISSOCK(mode)) { |
c3053ebb | 1012 | inode_item_err(leaf, slot, |
496245ca QW |
1013 | "invalid mode: has 0%o expect valid S_IF* bit(s)", |
1014 | mode & S_IFMT); | |
1015 | return -EUCLEAN; | |
1016 | } | |
1017 | } | |
1018 | if (S_ISDIR(mode) && btrfs_inode_nlink(leaf, iitem) > 1) { | |
c3053ebb | 1019 | inode_item_err(leaf, slot, |
496245ca QW |
1020 | "invalid nlink: has %u expect no more than 1 for dir", |
1021 | btrfs_inode_nlink(leaf, iitem)); | |
1022 | return -EUCLEAN; | |
1023 | } | |
1024 | if (btrfs_inode_flags(leaf, iitem) & ~BTRFS_INODE_FLAG_MASK) { | |
c3053ebb | 1025 | inode_item_err(leaf, slot, |
496245ca QW |
1026 | "unknown flags detected: 0x%llx", |
1027 | btrfs_inode_flags(leaf, iitem) & | |
1028 | ~BTRFS_INODE_FLAG_MASK); | |
1029 | return -EUCLEAN; | |
1030 | } | |
1031 | return 0; | |
1032 | } | |
1033 | ||
259ee775 QW |
1034 | static int check_root_item(struct extent_buffer *leaf, struct btrfs_key *key, |
1035 | int slot) | |
1036 | { | |
1037 | struct btrfs_fs_info *fs_info = leaf->fs_info; | |
1038 | struct btrfs_root_item ri; | |
1039 | const u64 valid_root_flags = BTRFS_ROOT_SUBVOL_RDONLY | | |
1040 | BTRFS_ROOT_SUBVOL_DEAD; | |
57a0e674 | 1041 | int ret; |
259ee775 | 1042 | |
57a0e674 QW |
1043 | ret = check_root_key(leaf, key, slot); |
1044 | if (ret < 0) | |
1045 | return ret; | |
259ee775 QW |
1046 | |
1047 | if (btrfs_item_size_nr(leaf, slot) != sizeof(ri)) { | |
1048 | generic_err(leaf, slot, | |
1049 | "invalid root item size, have %u expect %zu", | |
1050 | btrfs_item_size_nr(leaf, slot), sizeof(ri)); | |
1051 | } | |
1052 | ||
1053 | read_extent_buffer(leaf, &ri, btrfs_item_ptr_offset(leaf, slot), | |
1054 | sizeof(ri)); | |
1055 | ||
1056 | /* Generation related */ | |
1057 | if (btrfs_root_generation(&ri) > | |
1058 | btrfs_super_generation(fs_info->super_copy) + 1) { | |
1059 | generic_err(leaf, slot, | |
1060 | "invalid root generation, have %llu expect (0, %llu]", | |
1061 | btrfs_root_generation(&ri), | |
1062 | btrfs_super_generation(fs_info->super_copy) + 1); | |
1063 | return -EUCLEAN; | |
1064 | } | |
1065 | if (btrfs_root_generation_v2(&ri) > | |
1066 | btrfs_super_generation(fs_info->super_copy) + 1) { | |
1067 | generic_err(leaf, slot, | |
1068 | "invalid root v2 generation, have %llu expect (0, %llu]", | |
1069 | btrfs_root_generation_v2(&ri), | |
1070 | btrfs_super_generation(fs_info->super_copy) + 1); | |
1071 | return -EUCLEAN; | |
1072 | } | |
1073 | if (btrfs_root_last_snapshot(&ri) > | |
1074 | btrfs_super_generation(fs_info->super_copy) + 1) { | |
1075 | generic_err(leaf, slot, | |
1076 | "invalid root last_snapshot, have %llu expect (0, %llu]", | |
1077 | btrfs_root_last_snapshot(&ri), | |
1078 | btrfs_super_generation(fs_info->super_copy) + 1); | |
1079 | return -EUCLEAN; | |
1080 | } | |
1081 | ||
1082 | /* Alignment and level check */ | |
1083 | if (!IS_ALIGNED(btrfs_root_bytenr(&ri), fs_info->sectorsize)) { | |
1084 | generic_err(leaf, slot, | |
1085 | "invalid root bytenr, have %llu expect to be aligned to %u", | |
1086 | btrfs_root_bytenr(&ri), fs_info->sectorsize); | |
1087 | return -EUCLEAN; | |
1088 | } | |
1089 | if (btrfs_root_level(&ri) >= BTRFS_MAX_LEVEL) { | |
1090 | generic_err(leaf, slot, | |
1091 | "invalid root level, have %u expect [0, %u]", | |
1092 | btrfs_root_level(&ri), BTRFS_MAX_LEVEL - 1); | |
1093 | return -EUCLEAN; | |
1094 | } | |
1095 | if (ri.drop_level >= BTRFS_MAX_LEVEL) { | |
1096 | generic_err(leaf, slot, | |
1097 | "invalid root level, have %u expect [0, %u]", | |
1098 | ri.drop_level, BTRFS_MAX_LEVEL - 1); | |
1099 | return -EUCLEAN; | |
1100 | } | |
1101 | ||
1102 | /* Flags check */ | |
1103 | if (btrfs_root_flags(&ri) & ~valid_root_flags) { | |
1104 | generic_err(leaf, slot, | |
1105 | "invalid root flags, have 0x%llx expect mask 0x%llx", | |
1106 | btrfs_root_flags(&ri), valid_root_flags); | |
1107 | return -EUCLEAN; | |
1108 | } | |
1109 | return 0; | |
1110 | } | |
1111 | ||
f82d1c7c QW |
1112 | __printf(3,4) |
1113 | __cold | |
1114 | static void extent_err(const struct extent_buffer *eb, int slot, | |
1115 | const char *fmt, ...) | |
1116 | { | |
1117 | struct btrfs_key key; | |
1118 | struct va_format vaf; | |
1119 | va_list args; | |
1120 | u64 bytenr; | |
1121 | u64 len; | |
1122 | ||
1123 | btrfs_item_key_to_cpu(eb, &key, slot); | |
1124 | bytenr = key.objectid; | |
e2406a6f QW |
1125 | if (key.type == BTRFS_METADATA_ITEM_KEY || |
1126 | key.type == BTRFS_TREE_BLOCK_REF_KEY || | |
1127 | key.type == BTRFS_SHARED_BLOCK_REF_KEY) | |
f82d1c7c QW |
1128 | len = eb->fs_info->nodesize; |
1129 | else | |
1130 | len = key.offset; | |
1131 | va_start(args, fmt); | |
1132 | ||
1133 | vaf.fmt = fmt; | |
1134 | vaf.va = &args; | |
1135 | ||
1136 | btrfs_crit(eb->fs_info, | |
1137 | "corrupt %s: block=%llu slot=%d extent bytenr=%llu len=%llu %pV", | |
1138 | btrfs_header_level(eb) == 0 ? "leaf" : "node", | |
1139 | eb->start, slot, bytenr, len, &vaf); | |
1140 | va_end(args); | |
1141 | } | |
1142 | ||
1143 | static int check_extent_item(struct extent_buffer *leaf, | |
1144 | struct btrfs_key *key, int slot) | |
1145 | { | |
1146 | struct btrfs_fs_info *fs_info = leaf->fs_info; | |
1147 | struct btrfs_extent_item *ei; | |
1148 | bool is_tree_block = false; | |
1149 | unsigned long ptr; /* Current pointer inside inline refs */ | |
1150 | unsigned long end; /* Extent item end */ | |
1151 | const u32 item_size = btrfs_item_size_nr(leaf, slot); | |
1152 | u64 flags; | |
1153 | u64 generation; | |
1154 | u64 total_refs; /* Total refs in btrfs_extent_item */ | |
1155 | u64 inline_refs = 0; /* found total inline refs */ | |
1156 | ||
1157 | if (key->type == BTRFS_METADATA_ITEM_KEY && | |
1158 | !btrfs_fs_incompat(fs_info, SKINNY_METADATA)) { | |
1159 | generic_err(leaf, slot, | |
1160 | "invalid key type, METADATA_ITEM type invalid when SKINNY_METADATA feature disabled"); | |
1161 | return -EUCLEAN; | |
1162 | } | |
1163 | /* key->objectid is the bytenr for both key types */ | |
1164 | if (!IS_ALIGNED(key->objectid, fs_info->sectorsize)) { | |
1165 | generic_err(leaf, slot, | |
1166 | "invalid key objectid, have %llu expect to be aligned to %u", | |
1167 | key->objectid, fs_info->sectorsize); | |
1168 | return -EUCLEAN; | |
1169 | } | |
1170 | ||
1171 | /* key->offset is tree level for METADATA_ITEM_KEY */ | |
1172 | if (key->type == BTRFS_METADATA_ITEM_KEY && | |
1173 | key->offset >= BTRFS_MAX_LEVEL) { | |
1174 | extent_err(leaf, slot, | |
1175 | "invalid tree level, have %llu expect [0, %u]", | |
1176 | key->offset, BTRFS_MAX_LEVEL - 1); | |
1177 | return -EUCLEAN; | |
1178 | } | |
1179 | ||
1180 | /* | |
1181 | * EXTENT/METADATA_ITEM consists of: | |
1182 | * 1) One btrfs_extent_item | |
1183 | * Records the total refs, type and generation of the extent. | |
1184 | * | |
1185 | * 2) One btrfs_tree_block_info (for EXTENT_ITEM and tree backref only) | |
1186 | * Records the first key and level of the tree block. | |
1187 | * | |
1188 | * 2) Zero or more btrfs_extent_inline_ref(s) | |
1189 | * Each inline ref has one btrfs_extent_inline_ref shows: | |
1190 | * 2.1) The ref type, one of the 4 | |
1191 | * TREE_BLOCK_REF Tree block only | |
1192 | * SHARED_BLOCK_REF Tree block only | |
1193 | * EXTENT_DATA_REF Data only | |
1194 | * SHARED_DATA_REF Data only | |
1195 | * 2.2) Ref type specific data | |
1196 | * Either using btrfs_extent_inline_ref::offset, or specific | |
1197 | * data structure. | |
1198 | */ | |
1199 | if (item_size < sizeof(*ei)) { | |
1200 | extent_err(leaf, slot, | |
1201 | "invalid item size, have %u expect [%zu, %u)", | |
1202 | item_size, sizeof(*ei), | |
1203 | BTRFS_LEAF_DATA_SIZE(fs_info)); | |
1204 | return -EUCLEAN; | |
1205 | } | |
1206 | end = item_size + btrfs_item_ptr_offset(leaf, slot); | |
1207 | ||
1208 | /* Checks against extent_item */ | |
1209 | ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item); | |
1210 | flags = btrfs_extent_flags(leaf, ei); | |
1211 | total_refs = btrfs_extent_refs(leaf, ei); | |
1212 | generation = btrfs_extent_generation(leaf, ei); | |
1213 | if (generation > btrfs_super_generation(fs_info->super_copy) + 1) { | |
1214 | extent_err(leaf, slot, | |
1215 | "invalid generation, have %llu expect (0, %llu]", | |
1216 | generation, | |
1217 | btrfs_super_generation(fs_info->super_copy) + 1); | |
1218 | return -EUCLEAN; | |
1219 | } | |
c1499166 DS |
1220 | if (!has_single_bit_set(flags & (BTRFS_EXTENT_FLAG_DATA | |
1221 | BTRFS_EXTENT_FLAG_TREE_BLOCK))) { | |
f82d1c7c QW |
1222 | extent_err(leaf, slot, |
1223 | "invalid extent flag, have 0x%llx expect 1 bit set in 0x%llx", | |
1224 | flags, BTRFS_EXTENT_FLAG_DATA | | |
1225 | BTRFS_EXTENT_FLAG_TREE_BLOCK); | |
1226 | return -EUCLEAN; | |
1227 | } | |
1228 | is_tree_block = !!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK); | |
1229 | if (is_tree_block) { | |
1230 | if (key->type == BTRFS_EXTENT_ITEM_KEY && | |
1231 | key->offset != fs_info->nodesize) { | |
1232 | extent_err(leaf, slot, | |
1233 | "invalid extent length, have %llu expect %u", | |
1234 | key->offset, fs_info->nodesize); | |
1235 | return -EUCLEAN; | |
1236 | } | |
1237 | } else { | |
1238 | if (key->type != BTRFS_EXTENT_ITEM_KEY) { | |
1239 | extent_err(leaf, slot, | |
1240 | "invalid key type, have %u expect %u for data backref", | |
1241 | key->type, BTRFS_EXTENT_ITEM_KEY); | |
1242 | return -EUCLEAN; | |
1243 | } | |
1244 | if (!IS_ALIGNED(key->offset, fs_info->sectorsize)) { | |
1245 | extent_err(leaf, slot, | |
1246 | "invalid extent length, have %llu expect aligned to %u", | |
1247 | key->offset, fs_info->sectorsize); | |
1248 | return -EUCLEAN; | |
1249 | } | |
1250 | } | |
1251 | ptr = (unsigned long)(struct btrfs_extent_item *)(ei + 1); | |
1252 | ||
1253 | /* Check the special case of btrfs_tree_block_info */ | |
1254 | if (is_tree_block && key->type != BTRFS_METADATA_ITEM_KEY) { | |
1255 | struct btrfs_tree_block_info *info; | |
1256 | ||
1257 | info = (struct btrfs_tree_block_info *)ptr; | |
1258 | if (btrfs_tree_block_level(leaf, info) >= BTRFS_MAX_LEVEL) { | |
1259 | extent_err(leaf, slot, | |
1260 | "invalid tree block info level, have %u expect [0, %u]", | |
1261 | btrfs_tree_block_level(leaf, info), | |
1262 | BTRFS_MAX_LEVEL - 1); | |
1263 | return -EUCLEAN; | |
1264 | } | |
1265 | ptr = (unsigned long)(struct btrfs_tree_block_info *)(info + 1); | |
1266 | } | |
1267 | ||
1268 | /* Check inline refs */ | |
1269 | while (ptr < end) { | |
1270 | struct btrfs_extent_inline_ref *iref; | |
1271 | struct btrfs_extent_data_ref *dref; | |
1272 | struct btrfs_shared_data_ref *sref; | |
1273 | u64 dref_offset; | |
1274 | u64 inline_offset; | |
1275 | u8 inline_type; | |
1276 | ||
1277 | if (ptr + sizeof(*iref) > end) { | |
1278 | extent_err(leaf, slot, | |
1279 | "inline ref item overflows extent item, ptr %lu iref size %zu end %lu", | |
1280 | ptr, sizeof(*iref), end); | |
1281 | return -EUCLEAN; | |
1282 | } | |
1283 | iref = (struct btrfs_extent_inline_ref *)ptr; | |
1284 | inline_type = btrfs_extent_inline_ref_type(leaf, iref); | |
1285 | inline_offset = btrfs_extent_inline_ref_offset(leaf, iref); | |
1286 | if (ptr + btrfs_extent_inline_ref_size(inline_type) > end) { | |
1287 | extent_err(leaf, slot, | |
1288 | "inline ref item overflows extent item, ptr %lu iref size %u end %lu", | |
1289 | ptr, inline_type, end); | |
1290 | return -EUCLEAN; | |
1291 | } | |
1292 | ||
1293 | switch (inline_type) { | |
1294 | /* inline_offset is subvolid of the owner, no need to check */ | |
1295 | case BTRFS_TREE_BLOCK_REF_KEY: | |
1296 | inline_refs++; | |
1297 | break; | |
1298 | /* Contains parent bytenr */ | |
1299 | case BTRFS_SHARED_BLOCK_REF_KEY: | |
1300 | if (!IS_ALIGNED(inline_offset, fs_info->sectorsize)) { | |
1301 | extent_err(leaf, slot, | |
1302 | "invalid tree parent bytenr, have %llu expect aligned to %u", | |
1303 | inline_offset, fs_info->sectorsize); | |
1304 | return -EUCLEAN; | |
1305 | } | |
1306 | inline_refs++; | |
1307 | break; | |
1308 | /* | |
1309 | * Contains owner subvolid, owner key objectid, adjusted offset. | |
1310 | * The only obvious corruption can happen in that offset. | |
1311 | */ | |
1312 | case BTRFS_EXTENT_DATA_REF_KEY: | |
1313 | dref = (struct btrfs_extent_data_ref *)(&iref->offset); | |
1314 | dref_offset = btrfs_extent_data_ref_offset(leaf, dref); | |
1315 | if (!IS_ALIGNED(dref_offset, fs_info->sectorsize)) { | |
1316 | extent_err(leaf, slot, | |
1317 | "invalid data ref offset, have %llu expect aligned to %u", | |
1318 | dref_offset, fs_info->sectorsize); | |
1319 | return -EUCLEAN; | |
1320 | } | |
1321 | inline_refs += btrfs_extent_data_ref_count(leaf, dref); | |
1322 | break; | |
1323 | /* Contains parent bytenr and ref count */ | |
1324 | case BTRFS_SHARED_DATA_REF_KEY: | |
1325 | sref = (struct btrfs_shared_data_ref *)(iref + 1); | |
1326 | if (!IS_ALIGNED(inline_offset, fs_info->sectorsize)) { | |
1327 | extent_err(leaf, slot, | |
1328 | "invalid data parent bytenr, have %llu expect aligned to %u", | |
1329 | inline_offset, fs_info->sectorsize); | |
1330 | return -EUCLEAN; | |
1331 | } | |
1332 | inline_refs += btrfs_shared_data_ref_count(leaf, sref); | |
1333 | break; | |
1334 | default: | |
1335 | extent_err(leaf, slot, "unknown inline ref type: %u", | |
1336 | inline_type); | |
1337 | return -EUCLEAN; | |
1338 | } | |
1339 | ptr += btrfs_extent_inline_ref_size(inline_type); | |
1340 | } | |
1341 | /* No padding is allowed */ | |
1342 | if (ptr != end) { | |
1343 | extent_err(leaf, slot, | |
1344 | "invalid extent item size, padding bytes found"); | |
1345 | return -EUCLEAN; | |
1346 | } | |
1347 | ||
1348 | /* Finally, check the inline refs against total refs */ | |
1349 | if (inline_refs > total_refs) { | |
1350 | extent_err(leaf, slot, | |
1351 | "invalid extent refs, have %llu expect >= inline %llu", | |
1352 | total_refs, inline_refs); | |
1353 | return -EUCLEAN; | |
1354 | } | |
1355 | return 0; | |
1356 | } | |
1357 | ||
e2406a6f QW |
1358 | static int check_simple_keyed_refs(struct extent_buffer *leaf, |
1359 | struct btrfs_key *key, int slot) | |
1360 | { | |
1361 | u32 expect_item_size = 0; | |
1362 | ||
1363 | if (key->type == BTRFS_SHARED_DATA_REF_KEY) | |
1364 | expect_item_size = sizeof(struct btrfs_shared_data_ref); | |
1365 | ||
1366 | if (btrfs_item_size_nr(leaf, slot) != expect_item_size) { | |
1367 | generic_err(leaf, slot, | |
1368 | "invalid item size, have %u expect %u for key type %u", | |
1369 | btrfs_item_size_nr(leaf, slot), | |
1370 | expect_item_size, key->type); | |
1371 | return -EUCLEAN; | |
1372 | } | |
1373 | if (!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize)) { | |
1374 | generic_err(leaf, slot, | |
1375 | "invalid key objectid for shared block ref, have %llu expect aligned to %u", | |
1376 | key->objectid, leaf->fs_info->sectorsize); | |
1377 | return -EUCLEAN; | |
1378 | } | |
1379 | if (key->type != BTRFS_TREE_BLOCK_REF_KEY && | |
1380 | !IS_ALIGNED(key->offset, leaf->fs_info->sectorsize)) { | |
1381 | extent_err(leaf, slot, | |
1382 | "invalid tree parent bytenr, have %llu expect aligned to %u", | |
1383 | key->offset, leaf->fs_info->sectorsize); | |
1384 | return -EUCLEAN; | |
1385 | } | |
1386 | return 0; | |
1387 | } | |
1388 | ||
0785a9aa QW |
1389 | static int check_extent_data_ref(struct extent_buffer *leaf, |
1390 | struct btrfs_key *key, int slot) | |
1391 | { | |
1392 | struct btrfs_extent_data_ref *dref; | |
1393 | unsigned long ptr = btrfs_item_ptr_offset(leaf, slot); | |
1394 | const unsigned long end = ptr + btrfs_item_size_nr(leaf, slot); | |
1395 | ||
1396 | if (btrfs_item_size_nr(leaf, slot) % sizeof(*dref) != 0) { | |
1397 | generic_err(leaf, slot, | |
1398 | "invalid item size, have %u expect aligned to %zu for key type %u", | |
1399 | btrfs_item_size_nr(leaf, slot), | |
1400 | sizeof(*dref), key->type); | |
1401 | } | |
1402 | if (!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize)) { | |
1403 | generic_err(leaf, slot, | |
1404 | "invalid key objectid for shared block ref, have %llu expect aligned to %u", | |
1405 | key->objectid, leaf->fs_info->sectorsize); | |
1406 | return -EUCLEAN; | |
1407 | } | |
1408 | for (; ptr < end; ptr += sizeof(*dref)) { | |
1409 | u64 root_objectid; | |
1410 | u64 owner; | |
1411 | u64 offset; | |
1412 | u64 hash; | |
1413 | ||
1414 | dref = (struct btrfs_extent_data_ref *)ptr; | |
1415 | root_objectid = btrfs_extent_data_ref_root(leaf, dref); | |
1416 | owner = btrfs_extent_data_ref_objectid(leaf, dref); | |
1417 | offset = btrfs_extent_data_ref_offset(leaf, dref); | |
1418 | hash = hash_extent_data_ref(root_objectid, owner, offset); | |
1419 | if (hash != key->offset) { | |
1420 | extent_err(leaf, slot, | |
1421 | "invalid extent data ref hash, item has 0x%016llx key has 0x%016llx", | |
1422 | hash, key->offset); | |
1423 | return -EUCLEAN; | |
1424 | } | |
1425 | if (!IS_ALIGNED(offset, leaf->fs_info->sectorsize)) { | |
1426 | extent_err(leaf, slot, | |
1427 | "invalid extent data backref offset, have %llu expect aligned to %u", | |
1428 | offset, leaf->fs_info->sectorsize); | |
1429 | } | |
1430 | } | |
1431 | return 0; | |
1432 | } | |
1433 | ||
c3053ebb QW |
1434 | #define inode_ref_err(eb, slot, fmt, args...) \ |
1435 | inode_item_err(eb, slot, fmt, ##args) | |
71bf92a9 QW |
1436 | static int check_inode_ref(struct extent_buffer *leaf, |
1437 | struct btrfs_key *key, struct btrfs_key *prev_key, | |
1438 | int slot) | |
1439 | { | |
1440 | struct btrfs_inode_ref *iref; | |
1441 | unsigned long ptr; | |
1442 | unsigned long end; | |
1443 | ||
80d7fd1e QW |
1444 | if (!check_prev_ino(leaf, key, slot, prev_key)) |
1445 | return -EUCLEAN; | |
71bf92a9 QW |
1446 | /* namelen can't be 0, so item_size == sizeof() is also invalid */ |
1447 | if (btrfs_item_size_nr(leaf, slot) <= sizeof(*iref)) { | |
c3053ebb | 1448 | inode_ref_err(leaf, slot, |
71bf92a9 QW |
1449 | "invalid item size, have %u expect (%zu, %u)", |
1450 | btrfs_item_size_nr(leaf, slot), | |
1451 | sizeof(*iref), BTRFS_LEAF_DATA_SIZE(leaf->fs_info)); | |
1452 | return -EUCLEAN; | |
1453 | } | |
1454 | ||
1455 | ptr = btrfs_item_ptr_offset(leaf, slot); | |
1456 | end = ptr + btrfs_item_size_nr(leaf, slot); | |
1457 | while (ptr < end) { | |
1458 | u16 namelen; | |
1459 | ||
1460 | if (ptr + sizeof(iref) > end) { | |
c3053ebb | 1461 | inode_ref_err(leaf, slot, |
71bf92a9 QW |
1462 | "inode ref overflow, ptr %lu end %lu inode_ref_size %zu", |
1463 | ptr, end, sizeof(iref)); | |
1464 | return -EUCLEAN; | |
1465 | } | |
1466 | ||
1467 | iref = (struct btrfs_inode_ref *)ptr; | |
1468 | namelen = btrfs_inode_ref_name_len(leaf, iref); | |
1469 | if (ptr + sizeof(*iref) + namelen > end) { | |
c3053ebb | 1470 | inode_ref_err(leaf, slot, |
71bf92a9 QW |
1471 | "inode ref overflow, ptr %lu end %lu namelen %u", |
1472 | ptr, end, namelen); | |
1473 | return -EUCLEAN; | |
1474 | } | |
1475 | ||
1476 | /* | |
1477 | * NOTE: In theory we should record all found index numbers | |
1478 | * to find any duplicated indexes, but that will be too time | |
1479 | * consuming for inodes with too many hard links. | |
1480 | */ | |
1481 | ptr += sizeof(*iref) + namelen; | |
1482 | } | |
1483 | return 0; | |
1484 | } | |
1485 | ||
557ea5dd QW |
1486 | /* |
1487 | * Common point to switch the item-specific validation. | |
1488 | */ | |
0076bc89 | 1489 | static int check_leaf_item(struct extent_buffer *leaf, |
4e9845ef FM |
1490 | struct btrfs_key *key, int slot, |
1491 | struct btrfs_key *prev_key) | |
557ea5dd QW |
1492 | { |
1493 | int ret = 0; | |
075cb3c7 | 1494 | struct btrfs_chunk *chunk; |
557ea5dd QW |
1495 | |
1496 | switch (key->type) { | |
1497 | case BTRFS_EXTENT_DATA_KEY: | |
4e9845ef | 1498 | ret = check_extent_data_item(leaf, key, slot, prev_key); |
557ea5dd QW |
1499 | break; |
1500 | case BTRFS_EXTENT_CSUM_KEY: | |
ad1d8c43 | 1501 | ret = check_csum_item(leaf, key, slot, prev_key); |
557ea5dd | 1502 | break; |
ad7b0368 QW |
1503 | case BTRFS_DIR_ITEM_KEY: |
1504 | case BTRFS_DIR_INDEX_KEY: | |
1505 | case BTRFS_XATTR_ITEM_KEY: | |
c18679eb | 1506 | ret = check_dir_item(leaf, key, prev_key, slot); |
ad7b0368 | 1507 | break; |
71bf92a9 QW |
1508 | case BTRFS_INODE_REF_KEY: |
1509 | ret = check_inode_ref(leaf, key, prev_key, slot); | |
1510 | break; | |
fce466ea | 1511 | case BTRFS_BLOCK_GROUP_ITEM_KEY: |
af60ce2b | 1512 | ret = check_block_group_item(leaf, key, slot); |
fce466ea | 1513 | break; |
075cb3c7 QW |
1514 | case BTRFS_CHUNK_ITEM_KEY: |
1515 | chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk); | |
f6d2a5c2 | 1516 | ret = check_leaf_chunk_item(leaf, chunk, key, slot); |
075cb3c7 | 1517 | break; |
ab4ba2e1 | 1518 | case BTRFS_DEV_ITEM_KEY: |
412a2312 | 1519 | ret = check_dev_item(leaf, key, slot); |
ab4ba2e1 | 1520 | break; |
496245ca | 1521 | case BTRFS_INODE_ITEM_KEY: |
39e57f49 | 1522 | ret = check_inode_item(leaf, key, slot); |
496245ca | 1523 | break; |
259ee775 QW |
1524 | case BTRFS_ROOT_ITEM_KEY: |
1525 | ret = check_root_item(leaf, key, slot); | |
1526 | break; | |
f82d1c7c QW |
1527 | case BTRFS_EXTENT_ITEM_KEY: |
1528 | case BTRFS_METADATA_ITEM_KEY: | |
1529 | ret = check_extent_item(leaf, key, slot); | |
1530 | break; | |
e2406a6f QW |
1531 | case BTRFS_TREE_BLOCK_REF_KEY: |
1532 | case BTRFS_SHARED_DATA_REF_KEY: | |
1533 | case BTRFS_SHARED_BLOCK_REF_KEY: | |
1534 | ret = check_simple_keyed_refs(leaf, key, slot); | |
1535 | break; | |
0785a9aa QW |
1536 | case BTRFS_EXTENT_DATA_REF_KEY: |
1537 | ret = check_extent_data_ref(leaf, key, slot); | |
1538 | break; | |
557ea5dd QW |
1539 | } |
1540 | return ret; | |
1541 | } | |
1542 | ||
e2ccd361 | 1543 | static int check_leaf(struct extent_buffer *leaf, bool check_item_data) |
557ea5dd | 1544 | { |
e2ccd361 | 1545 | struct btrfs_fs_info *fs_info = leaf->fs_info; |
557ea5dd QW |
1546 | /* No valid key type is 0, so all key should be larger than this key */ |
1547 | struct btrfs_key prev_key = {0, 0, 0}; | |
1548 | struct btrfs_key key; | |
1549 | u32 nritems = btrfs_header_nritems(leaf); | |
1550 | int slot; | |
1551 | ||
f556faa4 | 1552 | if (btrfs_header_level(leaf) != 0) { |
86a6be3a | 1553 | generic_err(leaf, 0, |
f556faa4 QW |
1554 | "invalid level for leaf, have %d expect 0", |
1555 | btrfs_header_level(leaf)); | |
1556 | return -EUCLEAN; | |
1557 | } | |
1558 | ||
557ea5dd QW |
1559 | /* |
1560 | * Extent buffers from a relocation tree have a owner field that | |
1561 | * corresponds to the subvolume tree they are based on. So just from an | |
1562 | * extent buffer alone we can not find out what is the id of the | |
1563 | * corresponding subvolume tree, so we can not figure out if the extent | |
1564 | * buffer corresponds to the root of the relocation tree or not. So | |
1565 | * skip this check for relocation trees. | |
1566 | */ | |
1567 | if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) { | |
ba480dd4 | 1568 | u64 owner = btrfs_header_owner(leaf); |
557ea5dd | 1569 | |
ba480dd4 QW |
1570 | /* These trees must never be empty */ |
1571 | if (owner == BTRFS_ROOT_TREE_OBJECTID || | |
1572 | owner == BTRFS_CHUNK_TREE_OBJECTID || | |
1573 | owner == BTRFS_EXTENT_TREE_OBJECTID || | |
1574 | owner == BTRFS_DEV_TREE_OBJECTID || | |
1575 | owner == BTRFS_FS_TREE_OBJECTID || | |
1576 | owner == BTRFS_DATA_RELOC_TREE_OBJECTID) { | |
86a6be3a | 1577 | generic_err(leaf, 0, |
ba480dd4 QW |
1578 | "invalid root, root %llu must never be empty", |
1579 | owner); | |
1580 | return -EUCLEAN; | |
1581 | } | |
62fdaa52 QW |
1582 | /* Unknown tree */ |
1583 | if (owner == 0) { | |
1584 | generic_err(leaf, 0, | |
1585 | "invalid owner, root 0 is not defined"); | |
1586 | return -EUCLEAN; | |
1587 | } | |
557ea5dd QW |
1588 | return 0; |
1589 | } | |
1590 | ||
1591 | if (nritems == 0) | |
1592 | return 0; | |
1593 | ||
1594 | /* | |
1595 | * Check the following things to make sure this is a good leaf, and | |
1596 | * leaf users won't need to bother with similar sanity checks: | |
1597 | * | |
1598 | * 1) key ordering | |
1599 | * 2) item offset and size | |
1600 | * No overlap, no hole, all inside the leaf. | |
1601 | * 3) item content | |
1602 | * If possible, do comprehensive sanity check. | |
1603 | * NOTE: All checks must only rely on the item data itself. | |
1604 | */ | |
1605 | for (slot = 0; slot < nritems; slot++) { | |
1606 | u32 item_end_expected; | |
1607 | int ret; | |
1608 | ||
1609 | btrfs_item_key_to_cpu(leaf, &key, slot); | |
1610 | ||
1611 | /* Make sure the keys are in the right order */ | |
1612 | if (btrfs_comp_cpu_keys(&prev_key, &key) >= 0) { | |
86a6be3a | 1613 | generic_err(leaf, slot, |
478d01b3 QW |
1614 | "bad key order, prev (%llu %u %llu) current (%llu %u %llu)", |
1615 | prev_key.objectid, prev_key.type, | |
1616 | prev_key.offset, key.objectid, key.type, | |
1617 | key.offset); | |
557ea5dd QW |
1618 | return -EUCLEAN; |
1619 | } | |
1620 | ||
1621 | /* | |
1622 | * Make sure the offset and ends are right, remember that the | |
1623 | * item data starts at the end of the leaf and grows towards the | |
1624 | * front. | |
1625 | */ | |
1626 | if (slot == 0) | |
1627 | item_end_expected = BTRFS_LEAF_DATA_SIZE(fs_info); | |
1628 | else | |
1629 | item_end_expected = btrfs_item_offset_nr(leaf, | |
1630 | slot - 1); | |
1631 | if (btrfs_item_end_nr(leaf, slot) != item_end_expected) { | |
86a6be3a | 1632 | generic_err(leaf, slot, |
478d01b3 QW |
1633 | "unexpected item end, have %u expect %u", |
1634 | btrfs_item_end_nr(leaf, slot), | |
1635 | item_end_expected); | |
557ea5dd QW |
1636 | return -EUCLEAN; |
1637 | } | |
1638 | ||
1639 | /* | |
1640 | * Check to make sure that we don't point outside of the leaf, | |
1641 | * just in case all the items are consistent to each other, but | |
1642 | * all point outside of the leaf. | |
1643 | */ | |
1644 | if (btrfs_item_end_nr(leaf, slot) > | |
1645 | BTRFS_LEAF_DATA_SIZE(fs_info)) { | |
86a6be3a | 1646 | generic_err(leaf, slot, |
478d01b3 QW |
1647 | "slot end outside of leaf, have %u expect range [0, %u]", |
1648 | btrfs_item_end_nr(leaf, slot), | |
1649 | BTRFS_LEAF_DATA_SIZE(fs_info)); | |
557ea5dd QW |
1650 | return -EUCLEAN; |
1651 | } | |
1652 | ||
1653 | /* Also check if the item pointer overlaps with btrfs item. */ | |
1654 | if (btrfs_item_nr_offset(slot) + sizeof(struct btrfs_item) > | |
1655 | btrfs_item_ptr_offset(leaf, slot)) { | |
86a6be3a | 1656 | generic_err(leaf, slot, |
478d01b3 QW |
1657 | "slot overlaps with its data, item end %lu data start %lu", |
1658 | btrfs_item_nr_offset(slot) + | |
1659 | sizeof(struct btrfs_item), | |
1660 | btrfs_item_ptr_offset(leaf, slot)); | |
557ea5dd QW |
1661 | return -EUCLEAN; |
1662 | } | |
1663 | ||
69fc6cbb QW |
1664 | if (check_item_data) { |
1665 | /* | |
1666 | * Check if the item size and content meet other | |
1667 | * criteria | |
1668 | */ | |
4e9845ef | 1669 | ret = check_leaf_item(leaf, &key, slot, &prev_key); |
69fc6cbb QW |
1670 | if (ret < 0) |
1671 | return ret; | |
1672 | } | |
557ea5dd QW |
1673 | |
1674 | prev_key.objectid = key.objectid; | |
1675 | prev_key.type = key.type; | |
1676 | prev_key.offset = key.offset; | |
1677 | } | |
1678 | ||
1679 | return 0; | |
1680 | } | |
1681 | ||
1c4360ee | 1682 | int btrfs_check_leaf_full(struct extent_buffer *leaf) |
69fc6cbb | 1683 | { |
e2ccd361 | 1684 | return check_leaf(leaf, true); |
69fc6cbb | 1685 | } |
02529d7a | 1686 | ALLOW_ERROR_INJECTION(btrfs_check_leaf_full, ERRNO); |
69fc6cbb | 1687 | |
cfdaad5e | 1688 | int btrfs_check_leaf_relaxed(struct extent_buffer *leaf) |
69fc6cbb | 1689 | { |
e2ccd361 | 1690 | return check_leaf(leaf, false); |
69fc6cbb QW |
1691 | } |
1692 | ||
813fd1dc | 1693 | int btrfs_check_node(struct extent_buffer *node) |
557ea5dd | 1694 | { |
813fd1dc | 1695 | struct btrfs_fs_info *fs_info = node->fs_info; |
557ea5dd QW |
1696 | unsigned long nr = btrfs_header_nritems(node); |
1697 | struct btrfs_key key, next_key; | |
1698 | int slot; | |
f556faa4 | 1699 | int level = btrfs_header_level(node); |
557ea5dd QW |
1700 | u64 bytenr; |
1701 | int ret = 0; | |
1702 | ||
f556faa4 | 1703 | if (level <= 0 || level >= BTRFS_MAX_LEVEL) { |
86a6be3a | 1704 | generic_err(node, 0, |
f556faa4 QW |
1705 | "invalid level for node, have %d expect [1, %d]", |
1706 | level, BTRFS_MAX_LEVEL - 1); | |
1707 | return -EUCLEAN; | |
1708 | } | |
2f659546 QW |
1709 | if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(fs_info)) { |
1710 | btrfs_crit(fs_info, | |
bba4f298 | 1711 | "corrupt node: root=%llu block=%llu, nritems too %s, have %lu expect range [1,%u]", |
2f659546 | 1712 | btrfs_header_owner(node), node->start, |
bba4f298 | 1713 | nr == 0 ? "small" : "large", nr, |
2f659546 | 1714 | BTRFS_NODEPTRS_PER_BLOCK(fs_info)); |
bba4f298 | 1715 | return -EUCLEAN; |
557ea5dd QW |
1716 | } |
1717 | ||
1718 | for (slot = 0; slot < nr - 1; slot++) { | |
1719 | bytenr = btrfs_node_blockptr(node, slot); | |
1720 | btrfs_node_key_to_cpu(node, &key, slot); | |
1721 | btrfs_node_key_to_cpu(node, &next_key, slot + 1); | |
1722 | ||
1723 | if (!bytenr) { | |
86a6be3a | 1724 | generic_err(node, slot, |
bba4f298 QW |
1725 | "invalid NULL node pointer"); |
1726 | ret = -EUCLEAN; | |
1727 | goto out; | |
1728 | } | |
2f659546 | 1729 | if (!IS_ALIGNED(bytenr, fs_info->sectorsize)) { |
86a6be3a | 1730 | generic_err(node, slot, |
bba4f298 | 1731 | "unaligned pointer, have %llu should be aligned to %u", |
2f659546 | 1732 | bytenr, fs_info->sectorsize); |
bba4f298 | 1733 | ret = -EUCLEAN; |
557ea5dd QW |
1734 | goto out; |
1735 | } | |
1736 | ||
1737 | if (btrfs_comp_cpu_keys(&key, &next_key) >= 0) { | |
86a6be3a | 1738 | generic_err(node, slot, |
bba4f298 QW |
1739 | "bad key order, current (%llu %u %llu) next (%llu %u %llu)", |
1740 | key.objectid, key.type, key.offset, | |
1741 | next_key.objectid, next_key.type, | |
1742 | next_key.offset); | |
1743 | ret = -EUCLEAN; | |
557ea5dd QW |
1744 | goto out; |
1745 | } | |
1746 | } | |
1747 | out: | |
1748 | return ret; | |
1749 | } | |
02529d7a | 1750 | ALLOW_ERROR_INJECTION(btrfs_check_node, ERRNO); |