]>
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" |
77eea05e | 27 | #include "btrfs_inode.h" |
557ea5dd | 28 | |
bba4f298 QW |
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. | |
52042d8e | 35 | * It's recommended to decode key.objecitd/offset if it's |
bba4f298 QW |
36 | * meaningful. |
37 | * @reason: describe the error | |
52042d8e | 38 | * @bad_value: optional, it's recommended to output bad value and its |
bba4f298 QW |
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 | */ | |
86a6be3a | 49 | __printf(3, 4) |
e67c718b | 50 | __cold |
86a6be3a | 51 | static void generic_err(const struct extent_buffer *eb, int slot, |
bba4f298 QW |
52 | const char *fmt, ...) |
53 | { | |
86a6be3a | 54 | const struct btrfs_fs_info *fs_info = eb->fs_info; |
bba4f298 QW |
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 | ||
2f659546 | 63 | btrfs_crit(fs_info, |
bba4f298 QW |
64 | "corrupt %s: root=%llu block=%llu slot=%d, %pV", |
65 | btrfs_header_level(eb) == 0 ? "leaf" : "node", | |
2f659546 | 66 | btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, &vaf); |
bba4f298 QW |
67 | va_end(args); |
68 | } | |
69 | ||
8806d718 QW |
70 | /* |
71 | * Customized reporter for extent data item, since its key objectid and | |
72 | * offset has its own meaning. | |
73 | */ | |
1fd715ff | 74 | __printf(3, 4) |
e67c718b | 75 | __cold |
1fd715ff | 76 | static void file_extent_err(const struct extent_buffer *eb, int slot, |
8806d718 QW |
77 | const char *fmt, ...) |
78 | { | |
1fd715ff | 79 | const struct btrfs_fs_info *fs_info = eb->fs_info; |
8806d718 QW |
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 | ||
2f659546 | 90 | btrfs_crit(fs_info, |
8806d718 | 91 | "corrupt %s: root=%llu block=%llu slot=%d ino=%llu file_offset=%llu, %pV", |
2f659546 QW |
92 | btrfs_header_level(eb) == 0 ? "leaf" : "node", |
93 | btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, | |
94 | key.objectid, key.offset, &vaf); | |
8806d718 QW |
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 | */ | |
033774dc | 102 | #define CHECK_FE_ALIGNED(leaf, slot, fi, name, alignment) \ |
8806d718 | 103 | ({ \ |
c7c01a4a DS |
104 | if (unlikely(!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), \ |
105 | (alignment)))) \ | |
1fd715ff | 106 | file_extent_err((leaf), (slot), \ |
8806d718 QW |
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 | ||
4e9845ef FM |
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 | ||
80d7fd1e QW |
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 | |
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 | */ | |
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 | } | |
ae2a19d8 | 198 | static int check_extent_data_item(struct extent_buffer *leaf, |
4e9845ef FM |
199 | struct btrfs_key *key, int slot, |
200 | struct btrfs_key *prev_key) | |
557ea5dd | 201 | { |
ae2a19d8 | 202 | struct btrfs_fs_info *fs_info = leaf->fs_info; |
557ea5dd | 203 | struct btrfs_file_extent_item *fi; |
2f659546 | 204 | u32 sectorsize = fs_info->sectorsize; |
3212fa14 | 205 | u32 item_size = btrfs_item_size(leaf, slot); |
4c094c33 | 206 | u64 extent_end; |
557ea5dd | 207 | |
c7c01a4a | 208 | if (unlikely(!IS_ALIGNED(key->offset, sectorsize))) { |
1fd715ff | 209 | file_extent_err(leaf, slot, |
8806d718 QW |
210 | "unaligned file_offset for file extent, have %llu should be aligned to %u", |
211 | key->offset, sectorsize); | |
557ea5dd QW |
212 | return -EUCLEAN; |
213 | } | |
214 | ||
c18679eb QW |
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 | */ | |
c7c01a4a | 221 | if (unlikely(!check_prev_ino(leaf, key, slot, prev_key))) |
c18679eb | 222 | return -EUCLEAN; |
c18679eb | 223 | |
557ea5dd QW |
224 | fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item); |
225 | ||
153a6d29 QW |
226 | /* |
227 | * Make sure the item contains at least inline header, so the file | |
228 | * extent type is not some garbage. | |
229 | */ | |
c7c01a4a | 230 | if (unlikely(item_size < BTRFS_FILE_EXTENT_INLINE_DATA_START)) { |
153a6d29 | 231 | file_extent_err(leaf, slot, |
994bf9cd | 232 | "invalid item size, have %u expect [%zu, %u)", |
153a6d29 QW |
233 | item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START, |
234 | SZ_4K); | |
235 | return -EUCLEAN; | |
236 | } | |
c7c01a4a DS |
237 | if (unlikely(btrfs_file_extent_type(leaf, fi) >= |
238 | BTRFS_NR_FILE_EXTENT_TYPES)) { | |
1fd715ff | 239 | file_extent_err(leaf, slot, |
8806d718 QW |
240 | "invalid type for file extent, have %u expect range [0, %u]", |
241 | btrfs_file_extent_type(leaf, fi), | |
b9b1a53e | 242 | BTRFS_NR_FILE_EXTENT_TYPES - 1); |
557ea5dd QW |
243 | return -EUCLEAN; |
244 | } | |
245 | ||
246 | /* | |
52042d8e | 247 | * Support for new compression/encryption must introduce incompat flag, |
557ea5dd QW |
248 | * and must be caught in open_ctree(). |
249 | */ | |
c7c01a4a DS |
250 | if (unlikely(btrfs_file_extent_compression(leaf, fi) >= |
251 | BTRFS_NR_COMPRESS_TYPES)) { | |
1fd715ff | 252 | file_extent_err(leaf, slot, |
8806d718 QW |
253 | "invalid compression for file extent, have %u expect range [0, %u]", |
254 | btrfs_file_extent_compression(leaf, fi), | |
ce96b7ff | 255 | BTRFS_NR_COMPRESS_TYPES - 1); |
557ea5dd QW |
256 | return -EUCLEAN; |
257 | } | |
c7c01a4a | 258 | if (unlikely(btrfs_file_extent_encryption(leaf, fi))) { |
1fd715ff | 259 | file_extent_err(leaf, slot, |
8806d718 QW |
260 | "invalid encryption for file extent, have %u expect 0", |
261 | btrfs_file_extent_encryption(leaf, fi)); | |
557ea5dd QW |
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 */ | |
c7c01a4a | 266 | if (unlikely(key->offset)) { |
1fd715ff | 267 | file_extent_err(leaf, slot, |
8806d718 QW |
268 | "invalid file_offset for inline file extent, have %llu expect 0", |
269 | key->offset); | |
557ea5dd QW |
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 */ | |
c7c01a4a DS |
279 | if (unlikely(item_size != BTRFS_FILE_EXTENT_INLINE_DATA_START + |
280 | btrfs_file_extent_ram_bytes(leaf, fi))) { | |
1fd715ff | 281 | file_extent_err(leaf, slot, |
8806d718 QW |
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)); | |
557ea5dd QW |
285 | return -EUCLEAN; |
286 | } | |
287 | return 0; | |
288 | } | |
289 | ||
290 | /* Regular or preallocated extent has fixed item size */ | |
c7c01a4a | 291 | if (unlikely(item_size != sizeof(*fi))) { |
1fd715ff | 292 | file_extent_err(leaf, slot, |
709a95c3 | 293 | "invalid item size for reg/prealloc file extent, have %u expect %zu", |
8806d718 | 294 | item_size, sizeof(*fi)); |
557ea5dd QW |
295 | return -EUCLEAN; |
296 | } | |
c7c01a4a DS |
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))) | |
557ea5dd | 302 | return -EUCLEAN; |
4e9845ef | 303 | |
4c094c33 | 304 | /* Catch extent end overflow */ |
c7c01a4a DS |
305 | if (unlikely(check_add_overflow(btrfs_file_extent_num_bytes(leaf, fi), |
306 | key->offset, &extent_end))) { | |
4c094c33 QW |
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 | ||
4e9845ef FM |
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); | |
c7c01a4a | 327 | if (unlikely(prev_end > key->offset)) { |
4e9845ef FM |
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 | ||
557ea5dd QW |
335 | return 0; |
336 | } | |
337 | ||
68128ce7 | 338 | static int check_csum_item(struct extent_buffer *leaf, struct btrfs_key *key, |
ad1d8c43 | 339 | int slot, struct btrfs_key *prev_key) |
557ea5dd | 340 | { |
68128ce7 | 341 | struct btrfs_fs_info *fs_info = leaf->fs_info; |
2f659546 | 342 | u32 sectorsize = fs_info->sectorsize; |
223486c2 | 343 | const u32 csumsize = fs_info->csum_size; |
557ea5dd | 344 | |
c7c01a4a | 345 | if (unlikely(key->objectid != BTRFS_EXTENT_CSUM_OBJECTID)) { |
86a6be3a | 346 | generic_err(leaf, slot, |
d508c5f0 QW |
347 | "invalid key objectid for csum item, have %llu expect %llu", |
348 | key->objectid, BTRFS_EXTENT_CSUM_OBJECTID); | |
557ea5dd QW |
349 | return -EUCLEAN; |
350 | } | |
c7c01a4a | 351 | if (unlikely(!IS_ALIGNED(key->offset, sectorsize))) { |
86a6be3a | 352 | generic_err(leaf, slot, |
d508c5f0 QW |
353 | "unaligned key offset for csum item, have %llu should be aligned to %u", |
354 | key->offset, sectorsize); | |
557ea5dd QW |
355 | return -EUCLEAN; |
356 | } | |
3212fa14 | 357 | if (unlikely(!IS_ALIGNED(btrfs_item_size(leaf, slot), csumsize))) { |
86a6be3a | 358 | generic_err(leaf, slot, |
d508c5f0 | 359 | "unaligned item size for csum item, have %u should be aligned to %u", |
3212fa14 | 360 | btrfs_item_size(leaf, slot), csumsize); |
557ea5dd QW |
361 | return -EUCLEAN; |
362 | } | |
ad1d8c43 FM |
363 | if (slot > 0 && prev_key->type == BTRFS_EXTENT_CSUM_KEY) { |
364 | u64 prev_csum_end; | |
365 | u32 prev_item_size; | |
366 | ||
3212fa14 | 367 | prev_item_size = btrfs_item_size(leaf, slot - 1); |
ad1d8c43 FM |
368 | prev_csum_end = (prev_item_size / csumsize) * sectorsize; |
369 | prev_csum_end += prev_key->offset; | |
c7c01a4a | 370 | if (unlikely(prev_csum_end > key->offset)) { |
ad1d8c43 FM |
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 | } | |
557ea5dd QW |
377 | return 0; |
378 | } | |
379 | ||
c23c77b0 QW |
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 | ||
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) { | |
c7c01a4a DS |
395 | if (unlikely(key->objectid != 0 || key->type != 0 || |
396 | key->offset != 0)) | |
c23c77b0 QW |
397 | return -EUCLEAN; |
398 | return 0; | |
399 | } | |
400 | ||
c7c01a4a DS |
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)) { | |
c23c77b0 QW |
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 | } | |
c7c01a4a | 422 | if (unlikely(key->offset != 0)) { |
c23c77b0 QW |
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 | ||
57a0e674 QW |
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 | /* No such tree id */ | |
c7c01a4a | 446 | if (unlikely(key->objectid == 0)) { |
57a0e674 QW |
447 | if (is_root_item) |
448 | generic_err(leaf, slot, "invalid root id 0"); | |
449 | else | |
450 | dir_item_err(leaf, slot, | |
451 | "invalid location key root id 0"); | |
452 | return -EUCLEAN; | |
453 | } | |
454 | ||
455 | /* DIR_ITEM/INDEX/INODE_REF is not allowed to point to non-fs trees */ | |
c7c01a4a | 456 | if (unlikely(!is_fstree(key->objectid) && !is_root_item)) { |
57a0e674 QW |
457 | dir_item_err(leaf, slot, |
458 | "invalid location key objectid, have %llu expect [%llu, %llu]", | |
459 | key->objectid, BTRFS_FIRST_FREE_OBJECTID, | |
460 | BTRFS_LAST_FREE_OBJECTID); | |
461 | return -EUCLEAN; | |
462 | } | |
463 | ||
464 | /* | |
465 | * ROOT_ITEM with non-zero offset means this is a snapshot, created at | |
466 | * @offset transid. | |
467 | * Furthermore, for location key in DIR_ITEM, its offset is always -1. | |
468 | * | |
469 | * So here we only check offset for reloc tree whose key->offset must | |
470 | * be a valid tree. | |
471 | */ | |
c7c01a4a DS |
472 | if (unlikely(key->objectid == BTRFS_TREE_RELOC_OBJECTID && |
473 | key->offset == 0)) { | |
57a0e674 QW |
474 | generic_err(leaf, slot, "invalid root id 0 for reloc tree"); |
475 | return -EUCLEAN; | |
476 | } | |
477 | return 0; | |
478 | } | |
479 | ||
ce4252c0 | 480 | static int check_dir_item(struct extent_buffer *leaf, |
c18679eb QW |
481 | struct btrfs_key *key, struct btrfs_key *prev_key, |
482 | int slot) | |
ad7b0368 | 483 | { |
ce4252c0 | 484 | struct btrfs_fs_info *fs_info = leaf->fs_info; |
ad7b0368 | 485 | struct btrfs_dir_item *di; |
3212fa14 | 486 | u32 item_size = btrfs_item_size(leaf, slot); |
ad7b0368 QW |
487 | u32 cur = 0; |
488 | ||
c7c01a4a | 489 | if (unlikely(!check_prev_ino(leaf, key, slot, prev_key))) |
c18679eb | 490 | return -EUCLEAN; |
c7c01a4a | 491 | |
ad7b0368 QW |
492 | di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item); |
493 | while (cur < item_size) { | |
147a097c | 494 | struct btrfs_key location_key; |
ad7b0368 QW |
495 | u32 name_len; |
496 | u32 data_len; | |
497 | u32 max_name_len; | |
498 | u32 total_size; | |
499 | u32 name_hash; | |
500 | u8 dir_type; | |
147a097c | 501 | int ret; |
ad7b0368 QW |
502 | |
503 | /* header itself should not cross item boundary */ | |
c7c01a4a | 504 | if (unlikely(cur + sizeof(*di) > item_size)) { |
d98ced68 | 505 | dir_item_err(leaf, slot, |
7cfad652 | 506 | "dir item header crosses item boundary, have %zu boundary %u", |
ad7b0368 QW |
507 | cur + sizeof(*di), item_size); |
508 | return -EUCLEAN; | |
509 | } | |
510 | ||
147a097c QW |
511 | /* Location key check */ |
512 | btrfs_dir_item_key_to_cpu(leaf, di, &location_key); | |
513 | if (location_key.type == BTRFS_ROOT_ITEM_KEY) { | |
514 | ret = check_root_key(leaf, &location_key, slot); | |
c7c01a4a | 515 | if (unlikely(ret < 0)) |
147a097c QW |
516 | return ret; |
517 | } else if (location_key.type == BTRFS_INODE_ITEM_KEY || | |
518 | location_key.type == 0) { | |
519 | ret = check_inode_key(leaf, &location_key, slot); | |
c7c01a4a | 520 | if (unlikely(ret < 0)) |
147a097c QW |
521 | return ret; |
522 | } else { | |
523 | dir_item_err(leaf, slot, | |
524 | "invalid location key type, have %u, expect %u or %u", | |
525 | location_key.type, BTRFS_ROOT_ITEM_KEY, | |
526 | BTRFS_INODE_ITEM_KEY); | |
527 | return -EUCLEAN; | |
528 | } | |
529 | ||
ad7b0368 QW |
530 | /* dir type check */ |
531 | dir_type = btrfs_dir_type(leaf, di); | |
c7c01a4a | 532 | if (unlikely(dir_type >= BTRFS_FT_MAX)) { |
d98ced68 | 533 | dir_item_err(leaf, slot, |
ad7b0368 QW |
534 | "invalid dir item type, have %u expect [0, %u)", |
535 | dir_type, BTRFS_FT_MAX); | |
536 | return -EUCLEAN; | |
537 | } | |
538 | ||
c7c01a4a DS |
539 | if (unlikely(key->type == BTRFS_XATTR_ITEM_KEY && |
540 | dir_type != BTRFS_FT_XATTR)) { | |
d98ced68 | 541 | dir_item_err(leaf, slot, |
ad7b0368 QW |
542 | "invalid dir item type for XATTR key, have %u expect %u", |
543 | dir_type, BTRFS_FT_XATTR); | |
544 | return -EUCLEAN; | |
545 | } | |
c7c01a4a DS |
546 | if (unlikely(dir_type == BTRFS_FT_XATTR && |
547 | key->type != BTRFS_XATTR_ITEM_KEY)) { | |
d98ced68 | 548 | dir_item_err(leaf, slot, |
ad7b0368 QW |
549 | "xattr dir type found for non-XATTR key"); |
550 | return -EUCLEAN; | |
551 | } | |
552 | if (dir_type == BTRFS_FT_XATTR) | |
553 | max_name_len = XATTR_NAME_MAX; | |
554 | else | |
555 | max_name_len = BTRFS_NAME_LEN; | |
556 | ||
557 | /* Name/data length check */ | |
558 | name_len = btrfs_dir_name_len(leaf, di); | |
559 | data_len = btrfs_dir_data_len(leaf, di); | |
c7c01a4a | 560 | if (unlikely(name_len > max_name_len)) { |
d98ced68 | 561 | dir_item_err(leaf, slot, |
ad7b0368 QW |
562 | "dir item name len too long, have %u max %u", |
563 | name_len, max_name_len); | |
564 | return -EUCLEAN; | |
565 | } | |
c7c01a4a | 566 | if (unlikely(name_len + data_len > BTRFS_MAX_XATTR_SIZE(fs_info))) { |
d98ced68 | 567 | dir_item_err(leaf, slot, |
ad7b0368 QW |
568 | "dir item name and data len too long, have %u max %u", |
569 | name_len + data_len, | |
2f659546 | 570 | BTRFS_MAX_XATTR_SIZE(fs_info)); |
ad7b0368 QW |
571 | return -EUCLEAN; |
572 | } | |
573 | ||
c7c01a4a | 574 | if (unlikely(data_len && dir_type != BTRFS_FT_XATTR)) { |
d98ced68 | 575 | dir_item_err(leaf, slot, |
ad7b0368 QW |
576 | "dir item with invalid data len, have %u expect 0", |
577 | data_len); | |
578 | return -EUCLEAN; | |
579 | } | |
580 | ||
581 | total_size = sizeof(*di) + name_len + data_len; | |
582 | ||
583 | /* header and name/data should not cross item boundary */ | |
c7c01a4a | 584 | if (unlikely(cur + total_size > item_size)) { |
d98ced68 | 585 | dir_item_err(leaf, slot, |
ad7b0368 QW |
586 | "dir item data crosses item boundary, have %u boundary %u", |
587 | cur + total_size, item_size); | |
588 | return -EUCLEAN; | |
589 | } | |
590 | ||
591 | /* | |
592 | * Special check for XATTR/DIR_ITEM, as key->offset is name | |
593 | * hash, should match its name | |
594 | */ | |
595 | if (key->type == BTRFS_DIR_ITEM_KEY || | |
596 | key->type == BTRFS_XATTR_ITEM_KEY) { | |
e2683fc9 DS |
597 | char namebuf[max(BTRFS_NAME_LEN, XATTR_NAME_MAX)]; |
598 | ||
ad7b0368 QW |
599 | read_extent_buffer(leaf, namebuf, |
600 | (unsigned long)(di + 1), name_len); | |
601 | name_hash = btrfs_name_hash(namebuf, name_len); | |
c7c01a4a | 602 | if (unlikely(key->offset != name_hash)) { |
d98ced68 | 603 | dir_item_err(leaf, slot, |
ad7b0368 QW |
604 | "name hash mismatch with key, have 0x%016x expect 0x%016llx", |
605 | name_hash, key->offset); | |
606 | return -EUCLEAN; | |
607 | } | |
608 | } | |
609 | cur += total_size; | |
610 | di = (struct btrfs_dir_item *)((void *)di + total_size); | |
611 | } | |
612 | return 0; | |
613 | } | |
614 | ||
4806bd88 | 615 | __printf(3, 4) |
fce466ea | 616 | __cold |
4806bd88 | 617 | static void block_group_err(const struct extent_buffer *eb, int slot, |
fce466ea QW |
618 | const char *fmt, ...) |
619 | { | |
4806bd88 | 620 | const struct btrfs_fs_info *fs_info = eb->fs_info; |
fce466ea QW |
621 | struct btrfs_key key; |
622 | struct va_format vaf; | |
623 | va_list args; | |
624 | ||
625 | btrfs_item_key_to_cpu(eb, &key, slot); | |
626 | va_start(args, fmt); | |
627 | ||
628 | vaf.fmt = fmt; | |
629 | vaf.va = &args; | |
630 | ||
631 | btrfs_crit(fs_info, | |
632 | "corrupt %s: root=%llu block=%llu slot=%d bg_start=%llu bg_len=%llu, %pV", | |
633 | btrfs_header_level(eb) == 0 ? "leaf" : "node", | |
634 | btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, | |
635 | key.objectid, key.offset, &vaf); | |
636 | va_end(args); | |
637 | } | |
638 | ||
af60ce2b | 639 | static int check_block_group_item(struct extent_buffer *leaf, |
fce466ea QW |
640 | struct btrfs_key *key, int slot) |
641 | { | |
f7238e50 | 642 | struct btrfs_fs_info *fs_info = leaf->fs_info; |
fce466ea | 643 | struct btrfs_block_group_item bgi; |
3212fa14 | 644 | u32 item_size = btrfs_item_size(leaf, slot); |
f7238e50 | 645 | u64 chunk_objectid; |
fce466ea QW |
646 | u64 flags; |
647 | u64 type; | |
648 | ||
649 | /* | |
650 | * Here we don't really care about alignment since extent allocator can | |
10950929 | 651 | * handle it. We care more about the size. |
fce466ea | 652 | */ |
c7c01a4a | 653 | if (unlikely(key->offset == 0)) { |
4806bd88 | 654 | block_group_err(leaf, slot, |
10950929 | 655 | "invalid block group size 0"); |
fce466ea QW |
656 | return -EUCLEAN; |
657 | } | |
658 | ||
c7c01a4a | 659 | if (unlikely(item_size != sizeof(bgi))) { |
4806bd88 | 660 | block_group_err(leaf, slot, |
fce466ea QW |
661 | "invalid item size, have %u expect %zu", |
662 | item_size, sizeof(bgi)); | |
663 | return -EUCLEAN; | |
664 | } | |
665 | ||
666 | read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot), | |
667 | sizeof(bgi)); | |
f7238e50 JB |
668 | chunk_objectid = btrfs_stack_block_group_chunk_objectid(&bgi); |
669 | if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { | |
670 | /* | |
671 | * We don't init the nr_global_roots until we load the global | |
672 | * roots, so this could be 0 at mount time. If it's 0 we'll | |
673 | * just assume we're fine, and later we'll check against our | |
674 | * actual value. | |
675 | */ | |
676 | if (unlikely(fs_info->nr_global_roots && | |
677 | chunk_objectid >= fs_info->nr_global_roots)) { | |
678 | block_group_err(leaf, slot, | |
679 | "invalid block group global root id, have %llu, needs to be <= %llu", | |
680 | chunk_objectid, | |
681 | fs_info->nr_global_roots); | |
682 | return -EUCLEAN; | |
683 | } | |
684 | } else if (unlikely(chunk_objectid != BTRFS_FIRST_CHUNK_TREE_OBJECTID)) { | |
4806bd88 | 685 | block_group_err(leaf, slot, |
fce466ea | 686 | "invalid block group chunk objectid, have %llu expect %llu", |
de0dc456 | 687 | btrfs_stack_block_group_chunk_objectid(&bgi), |
fce466ea QW |
688 | BTRFS_FIRST_CHUNK_TREE_OBJECTID); |
689 | return -EUCLEAN; | |
690 | } | |
691 | ||
c7c01a4a | 692 | if (unlikely(btrfs_stack_block_group_used(&bgi) > key->offset)) { |
4806bd88 | 693 | block_group_err(leaf, slot, |
fce466ea | 694 | "invalid block group used, have %llu expect [0, %llu)", |
de0dc456 | 695 | btrfs_stack_block_group_used(&bgi), key->offset); |
fce466ea QW |
696 | return -EUCLEAN; |
697 | } | |
698 | ||
de0dc456 | 699 | flags = btrfs_stack_block_group_flags(&bgi); |
c7c01a4a | 700 | if (unlikely(hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) > 1)) { |
4806bd88 | 701 | block_group_err(leaf, slot, |
fce466ea QW |
702 | "invalid profile flags, have 0x%llx (%lu bits set) expect no more than 1 bit set", |
703 | flags & BTRFS_BLOCK_GROUP_PROFILE_MASK, | |
704 | hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)); | |
705 | return -EUCLEAN; | |
706 | } | |
707 | ||
708 | type = flags & BTRFS_BLOCK_GROUP_TYPE_MASK; | |
c7c01a4a DS |
709 | if (unlikely(type != BTRFS_BLOCK_GROUP_DATA && |
710 | type != BTRFS_BLOCK_GROUP_METADATA && | |
711 | type != BTRFS_BLOCK_GROUP_SYSTEM && | |
712 | type != (BTRFS_BLOCK_GROUP_METADATA | | |
713 | BTRFS_BLOCK_GROUP_DATA))) { | |
4806bd88 | 714 | block_group_err(leaf, slot, |
761333f2 | 715 | "invalid type, have 0x%llx (%lu bits set) expect either 0x%llx, 0x%llx, 0x%llx or 0x%llx", |
fce466ea QW |
716 | type, hweight64(type), |
717 | BTRFS_BLOCK_GROUP_DATA, BTRFS_BLOCK_GROUP_METADATA, | |
718 | BTRFS_BLOCK_GROUP_SYSTEM, | |
719 | BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA); | |
720 | return -EUCLEAN; | |
721 | } | |
722 | return 0; | |
82fc28fb QW |
723 | } |
724 | ||
d001e4a3 | 725 | __printf(4, 5) |
f1140243 | 726 | __cold |
d001e4a3 | 727 | static void chunk_err(const struct extent_buffer *leaf, |
f1140243 QW |
728 | const struct btrfs_chunk *chunk, u64 logical, |
729 | const char *fmt, ...) | |
730 | { | |
d001e4a3 | 731 | const struct btrfs_fs_info *fs_info = leaf->fs_info; |
f1140243 QW |
732 | bool is_sb; |
733 | struct va_format vaf; | |
734 | va_list args; | |
735 | int i; | |
736 | int slot = -1; | |
737 | ||
738 | /* Only superblock eb is able to have such small offset */ | |
739 | is_sb = (leaf->start == BTRFS_SUPER_INFO_OFFSET); | |
740 | ||
741 | if (!is_sb) { | |
742 | /* | |
743 | * Get the slot number by iterating through all slots, this | |
744 | * would provide better readability. | |
745 | */ | |
746 | for (i = 0; i < btrfs_header_nritems(leaf); i++) { | |
747 | if (btrfs_item_ptr_offset(leaf, i) == | |
748 | (unsigned long)chunk) { | |
749 | slot = i; | |
750 | break; | |
751 | } | |
752 | } | |
753 | } | |
754 | va_start(args, fmt); | |
755 | vaf.fmt = fmt; | |
756 | vaf.va = &args; | |
757 | ||
758 | if (is_sb) | |
759 | btrfs_crit(fs_info, | |
760 | "corrupt superblock syschunk array: chunk_start=%llu, %pV", | |
761 | logical, &vaf); | |
762 | else | |
763 | btrfs_crit(fs_info, | |
764 | "corrupt leaf: root=%llu block=%llu slot=%d chunk_start=%llu, %pV", | |
765 | BTRFS_CHUNK_TREE_OBJECTID, leaf->start, slot, | |
766 | logical, &vaf); | |
767 | va_end(args); | |
768 | } | |
769 | ||
82fc28fb QW |
770 | /* |
771 | * The common chunk check which could also work on super block sys chunk array. | |
772 | * | |
bf871c3b | 773 | * Return -EUCLEAN if anything is corrupted. |
82fc28fb QW |
774 | * Return 0 if everything is OK. |
775 | */ | |
ddaf1d5a | 776 | int btrfs_check_chunk_valid(struct extent_buffer *leaf, |
82fc28fb QW |
777 | struct btrfs_chunk *chunk, u64 logical) |
778 | { | |
ddaf1d5a | 779 | struct btrfs_fs_info *fs_info = leaf->fs_info; |
82fc28fb | 780 | u64 length; |
347fb0cf | 781 | u64 chunk_end; |
82fc28fb QW |
782 | u64 stripe_len; |
783 | u16 num_stripes; | |
784 | u16 sub_stripes; | |
785 | u64 type; | |
786 | u64 features; | |
787 | bool mixed = false; | |
85d07fbe DX |
788 | int raid_index; |
789 | int nparity; | |
790 | int ncopies; | |
82fc28fb QW |
791 | |
792 | length = btrfs_chunk_length(leaf, chunk); | |
793 | stripe_len = btrfs_chunk_stripe_len(leaf, chunk); | |
794 | num_stripes = btrfs_chunk_num_stripes(leaf, chunk); | |
795 | sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk); | |
796 | type = btrfs_chunk_type(leaf, chunk); | |
85d07fbe DX |
797 | raid_index = btrfs_bg_flags_to_raid_index(type); |
798 | ncopies = btrfs_raid_array[raid_index].ncopies; | |
799 | nparity = btrfs_raid_array[raid_index].nparity; | |
82fc28fb | 800 | |
c7c01a4a | 801 | if (unlikely(!num_stripes)) { |
d001e4a3 | 802 | chunk_err(leaf, chunk, logical, |
f1140243 | 803 | "invalid chunk num_stripes, have %u", num_stripes); |
bf871c3b | 804 | return -EUCLEAN; |
82fc28fb | 805 | } |
c7c01a4a | 806 | if (unlikely(num_stripes < ncopies)) { |
85d07fbe DX |
807 | chunk_err(leaf, chunk, logical, |
808 | "invalid chunk num_stripes < ncopies, have %u < %d", | |
809 | num_stripes, ncopies); | |
810 | return -EUCLEAN; | |
811 | } | |
c7c01a4a | 812 | if (unlikely(nparity && num_stripes == nparity)) { |
85d07fbe DX |
813 | chunk_err(leaf, chunk, logical, |
814 | "invalid chunk num_stripes == nparity, have %u == %d", | |
815 | num_stripes, nparity); | |
816 | return -EUCLEAN; | |
817 | } | |
c7c01a4a | 818 | if (unlikely(!IS_ALIGNED(logical, fs_info->sectorsize))) { |
d001e4a3 | 819 | chunk_err(leaf, chunk, logical, |
f1140243 QW |
820 | "invalid chunk logical, have %llu should aligned to %u", |
821 | logical, fs_info->sectorsize); | |
bf871c3b | 822 | return -EUCLEAN; |
82fc28fb | 823 | } |
c7c01a4a | 824 | if (unlikely(btrfs_chunk_sector_size(leaf, chunk) != fs_info->sectorsize)) { |
d001e4a3 | 825 | chunk_err(leaf, chunk, logical, |
f1140243 QW |
826 | "invalid chunk sectorsize, have %u expect %u", |
827 | btrfs_chunk_sector_size(leaf, chunk), | |
828 | fs_info->sectorsize); | |
bf871c3b | 829 | return -EUCLEAN; |
82fc28fb | 830 | } |
c7c01a4a | 831 | if (unlikely(!length || !IS_ALIGNED(length, fs_info->sectorsize))) { |
d001e4a3 | 832 | chunk_err(leaf, chunk, logical, |
f1140243 | 833 | "invalid chunk length, have %llu", length); |
bf871c3b | 834 | return -EUCLEAN; |
82fc28fb | 835 | } |
347fb0cf SY |
836 | if (unlikely(check_add_overflow(logical, length, &chunk_end))) { |
837 | chunk_err(leaf, chunk, logical, | |
838 | "invalid chunk logical start and length, have logical start %llu length %llu", | |
839 | logical, length); | |
840 | return -EUCLEAN; | |
841 | } | |
c7c01a4a | 842 | if (unlikely(!is_power_of_2(stripe_len) || stripe_len != BTRFS_STRIPE_LEN)) { |
d001e4a3 | 843 | chunk_err(leaf, chunk, logical, |
f1140243 | 844 | "invalid chunk stripe length: %llu", |
82fc28fb | 845 | stripe_len); |
bf871c3b | 846 | return -EUCLEAN; |
82fc28fb | 847 | } |
c7c01a4a DS |
848 | if (unlikely(type & ~(BTRFS_BLOCK_GROUP_TYPE_MASK | |
849 | BTRFS_BLOCK_GROUP_PROFILE_MASK))) { | |
d001e4a3 | 850 | chunk_err(leaf, chunk, logical, |
f1140243 | 851 | "unrecognized chunk type: 0x%llx", |
82fc28fb QW |
852 | ~(BTRFS_BLOCK_GROUP_TYPE_MASK | |
853 | BTRFS_BLOCK_GROUP_PROFILE_MASK) & | |
854 | btrfs_chunk_type(leaf, chunk)); | |
bf871c3b | 855 | return -EUCLEAN; |
82fc28fb QW |
856 | } |
857 | ||
c7c01a4a DS |
858 | if (unlikely(!has_single_bit_set(type & BTRFS_BLOCK_GROUP_PROFILE_MASK) && |
859 | (type & BTRFS_BLOCK_GROUP_PROFILE_MASK) != 0)) { | |
d001e4a3 | 860 | chunk_err(leaf, chunk, logical, |
80e46cf2 QW |
861 | "invalid chunk profile flag: 0x%llx, expect 0 or 1 bit set", |
862 | type & BTRFS_BLOCK_GROUP_PROFILE_MASK); | |
863 | return -EUCLEAN; | |
864 | } | |
c7c01a4a | 865 | if (unlikely((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == 0)) { |
d001e4a3 | 866 | chunk_err(leaf, chunk, logical, |
f1140243 QW |
867 | "missing chunk type flag, have 0x%llx one bit must be set in 0x%llx", |
868 | type, BTRFS_BLOCK_GROUP_TYPE_MASK); | |
bf871c3b | 869 | return -EUCLEAN; |
82fc28fb QW |
870 | } |
871 | ||
c7c01a4a DS |
872 | if (unlikely((type & BTRFS_BLOCK_GROUP_SYSTEM) && |
873 | (type & (BTRFS_BLOCK_GROUP_METADATA | | |
874 | BTRFS_BLOCK_GROUP_DATA)))) { | |
d001e4a3 | 875 | chunk_err(leaf, chunk, logical, |
f1140243 QW |
876 | "system chunk with data or metadata type: 0x%llx", |
877 | type); | |
bf871c3b | 878 | return -EUCLEAN; |
82fc28fb QW |
879 | } |
880 | ||
881 | features = btrfs_super_incompat_flags(fs_info->super_copy); | |
882 | if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) | |
883 | mixed = true; | |
884 | ||
885 | if (!mixed) { | |
c7c01a4a DS |
886 | if (unlikely((type & BTRFS_BLOCK_GROUP_METADATA) && |
887 | (type & BTRFS_BLOCK_GROUP_DATA))) { | |
d001e4a3 | 888 | chunk_err(leaf, chunk, logical, |
82fc28fb | 889 | "mixed chunk type in non-mixed mode: 0x%llx", type); |
bf871c3b | 890 | return -EUCLEAN; |
82fc28fb QW |
891 | } |
892 | } | |
893 | ||
0ac6e06b DS |
894 | if (unlikely((type & BTRFS_BLOCK_GROUP_RAID10 && |
895 | sub_stripes != btrfs_raid_array[BTRFS_RAID_RAID10].sub_stripes) || | |
896 | (type & BTRFS_BLOCK_GROUP_RAID1 && | |
897 | num_stripes != btrfs_raid_array[BTRFS_RAID_RAID1].devs_min) || | |
6c154ba4 DS |
898 | (type & BTRFS_BLOCK_GROUP_RAID1C3 && |
899 | num_stripes != btrfs_raid_array[BTRFS_RAID_RAID1C3].devs_min) || | |
900 | (type & BTRFS_BLOCK_GROUP_RAID1C4 && | |
901 | num_stripes != btrfs_raid_array[BTRFS_RAID_RAID1C4].devs_min) || | |
0ac6e06b DS |
902 | (type & BTRFS_BLOCK_GROUP_RAID5 && |
903 | num_stripes < btrfs_raid_array[BTRFS_RAID_RAID5].devs_min) || | |
904 | (type & BTRFS_BLOCK_GROUP_RAID6 && | |
905 | num_stripes < btrfs_raid_array[BTRFS_RAID_RAID6].devs_min) || | |
906 | (type & BTRFS_BLOCK_GROUP_DUP && | |
907 | num_stripes != btrfs_raid_array[BTRFS_RAID_DUP].dev_stripes) || | |
c7c01a4a | 908 | ((type & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 && |
0ac6e06b | 909 | num_stripes != btrfs_raid_array[BTRFS_RAID_SINGLE].dev_stripes))) { |
d001e4a3 | 910 | chunk_err(leaf, chunk, logical, |
82fc28fb QW |
911 | "invalid num_stripes:sub_stripes %u:%u for profile %llu", |
912 | num_stripes, sub_stripes, | |
913 | type & BTRFS_BLOCK_GROUP_PROFILE_MASK); | |
bf871c3b | 914 | return -EUCLEAN; |
82fc28fb QW |
915 | } |
916 | ||
917 | return 0; | |
fce466ea QW |
918 | } |
919 | ||
f6d2a5c2 QW |
920 | /* |
921 | * Enhanced version of chunk item checker. | |
922 | * | |
923 | * The common btrfs_check_chunk_valid() doesn't check item size since it needs | |
924 | * to work on super block sys_chunk_array which doesn't have full item ptr. | |
925 | */ | |
926 | static int check_leaf_chunk_item(struct extent_buffer *leaf, | |
927 | struct btrfs_chunk *chunk, | |
928 | struct btrfs_key *key, int slot) | |
929 | { | |
930 | int num_stripes; | |
931 | ||
3212fa14 | 932 | if (unlikely(btrfs_item_size(leaf, slot) < sizeof(struct btrfs_chunk))) { |
f6d2a5c2 QW |
933 | chunk_err(leaf, chunk, key->offset, |
934 | "invalid chunk item size: have %u expect [%zu, %u)", | |
3212fa14 | 935 | btrfs_item_size(leaf, slot), |
f6d2a5c2 QW |
936 | sizeof(struct btrfs_chunk), |
937 | BTRFS_LEAF_DATA_SIZE(leaf->fs_info)); | |
938 | return -EUCLEAN; | |
939 | } | |
940 | ||
941 | num_stripes = btrfs_chunk_num_stripes(leaf, chunk); | |
942 | /* Let btrfs_check_chunk_valid() handle this error type */ | |
943 | if (num_stripes == 0) | |
944 | goto out; | |
945 | ||
c7c01a4a | 946 | if (unlikely(btrfs_chunk_item_size(num_stripes) != |
3212fa14 | 947 | btrfs_item_size(leaf, slot))) { |
f6d2a5c2 QW |
948 | chunk_err(leaf, chunk, key->offset, |
949 | "invalid chunk item size: have %u expect %lu", | |
3212fa14 | 950 | btrfs_item_size(leaf, slot), |
f6d2a5c2 QW |
951 | btrfs_chunk_item_size(num_stripes)); |
952 | return -EUCLEAN; | |
953 | } | |
954 | out: | |
955 | return btrfs_check_chunk_valid(leaf, chunk, key->offset); | |
956 | } | |
957 | ||
5617ed80 | 958 | __printf(3, 4) |
ab4ba2e1 | 959 | __cold |
5617ed80 | 960 | static void dev_item_err(const struct extent_buffer *eb, int slot, |
ab4ba2e1 QW |
961 | const char *fmt, ...) |
962 | { | |
963 | struct btrfs_key key; | |
964 | struct va_format vaf; | |
965 | va_list args; | |
966 | ||
967 | btrfs_item_key_to_cpu(eb, &key, slot); | |
968 | va_start(args, fmt); | |
969 | ||
970 | vaf.fmt = fmt; | |
971 | vaf.va = &args; | |
972 | ||
5617ed80 | 973 | btrfs_crit(eb->fs_info, |
ab4ba2e1 QW |
974 | "corrupt %s: root=%llu block=%llu slot=%d devid=%llu %pV", |
975 | btrfs_header_level(eb) == 0 ? "leaf" : "node", | |
976 | btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, | |
977 | key.objectid, &vaf); | |
978 | va_end(args); | |
979 | } | |
980 | ||
412a2312 | 981 | static int check_dev_item(struct extent_buffer *leaf, |
ab4ba2e1 QW |
982 | struct btrfs_key *key, int slot) |
983 | { | |
984 | struct btrfs_dev_item *ditem; | |
ea1d1ca4 | 985 | const u32 item_size = btrfs_item_size(leaf, slot); |
ab4ba2e1 | 986 | |
c7c01a4a | 987 | if (unlikely(key->objectid != BTRFS_DEV_ITEMS_OBJECTID)) { |
5617ed80 | 988 | dev_item_err(leaf, slot, |
ab4ba2e1 QW |
989 | "invalid objectid: has=%llu expect=%llu", |
990 | key->objectid, BTRFS_DEV_ITEMS_OBJECTID); | |
991 | return -EUCLEAN; | |
992 | } | |
ea1d1ca4 SY |
993 | |
994 | if (unlikely(item_size != sizeof(*ditem))) { | |
995 | dev_item_err(leaf, slot, "invalid item size: has %u expect %zu", | |
996 | item_size, sizeof(*ditem)); | |
997 | return -EUCLEAN; | |
998 | } | |
999 | ||
ab4ba2e1 | 1000 | ditem = btrfs_item_ptr(leaf, slot, struct btrfs_dev_item); |
c7c01a4a | 1001 | if (unlikely(btrfs_device_id(leaf, ditem) != key->offset)) { |
5617ed80 | 1002 | dev_item_err(leaf, slot, |
ab4ba2e1 QW |
1003 | "devid mismatch: key has=%llu item has=%llu", |
1004 | key->offset, btrfs_device_id(leaf, ditem)); | |
1005 | return -EUCLEAN; | |
1006 | } | |
1007 | ||
1008 | /* | |
1009 | * For device total_bytes, we don't have reliable way to check it, as | |
1010 | * it can be 0 for device removal. Device size check can only be done | |
1011 | * by dev extents check. | |
1012 | */ | |
c7c01a4a DS |
1013 | if (unlikely(btrfs_device_bytes_used(leaf, ditem) > |
1014 | btrfs_device_total_bytes(leaf, ditem))) { | |
5617ed80 | 1015 | dev_item_err(leaf, slot, |
ab4ba2e1 QW |
1016 | "invalid bytes used: have %llu expect [0, %llu]", |
1017 | btrfs_device_bytes_used(leaf, ditem), | |
1018 | btrfs_device_total_bytes(leaf, ditem)); | |
1019 | return -EUCLEAN; | |
1020 | } | |
1021 | /* | |
1022 | * Remaining members like io_align/type/gen/dev_group aren't really | |
1023 | * utilized. Skip them to make later usage of them easier. | |
1024 | */ | |
1025 | return 0; | |
1026 | } | |
1027 | ||
39e57f49 | 1028 | static int check_inode_item(struct extent_buffer *leaf, |
496245ca QW |
1029 | struct btrfs_key *key, int slot) |
1030 | { | |
39e57f49 | 1031 | struct btrfs_fs_info *fs_info = leaf->fs_info; |
496245ca QW |
1032 | struct btrfs_inode_item *iitem; |
1033 | u64 super_gen = btrfs_super_generation(fs_info->super_copy); | |
1034 | u32 valid_mask = (S_IFMT | S_ISUID | S_ISGID | S_ISVTX | 0777); | |
0c982944 | 1035 | const u32 item_size = btrfs_item_size(leaf, slot); |
496245ca | 1036 | u32 mode; |
c23c77b0 | 1037 | int ret; |
77eea05e BB |
1038 | u32 flags; |
1039 | u32 ro_flags; | |
c23c77b0 QW |
1040 | |
1041 | ret = check_inode_key(leaf, key, slot); | |
c7c01a4a | 1042 | if (unlikely(ret < 0)) |
c23c77b0 | 1043 | return ret; |
496245ca | 1044 | |
0c982944 SY |
1045 | if (unlikely(item_size != sizeof(*iitem))) { |
1046 | generic_err(leaf, slot, "invalid item size: has %u expect %zu", | |
1047 | item_size, sizeof(*iitem)); | |
1048 | return -EUCLEAN; | |
1049 | } | |
1050 | ||
496245ca QW |
1051 | iitem = btrfs_item_ptr(leaf, slot, struct btrfs_inode_item); |
1052 | ||
1053 | /* Here we use super block generation + 1 to handle log tree */ | |
c7c01a4a | 1054 | if (unlikely(btrfs_inode_generation(leaf, iitem) > super_gen + 1)) { |
c3053ebb | 1055 | inode_item_err(leaf, slot, |
496245ca QW |
1056 | "invalid inode generation: has %llu expect (0, %llu]", |
1057 | btrfs_inode_generation(leaf, iitem), | |
1058 | super_gen + 1); | |
1059 | return -EUCLEAN; | |
1060 | } | |
1061 | /* Note for ROOT_TREE_DIR_ITEM, mkfs could set its transid 0 */ | |
c7c01a4a | 1062 | if (unlikely(btrfs_inode_transid(leaf, iitem) > super_gen + 1)) { |
c3053ebb | 1063 | inode_item_err(leaf, slot, |
f96d6960 | 1064 | "invalid inode transid: has %llu expect [0, %llu]", |
496245ca QW |
1065 | btrfs_inode_transid(leaf, iitem), super_gen + 1); |
1066 | return -EUCLEAN; | |
1067 | } | |
1068 | ||
1069 | /* | |
1070 | * For size and nbytes it's better not to be too strict, as for dir | |
1071 | * item its size/nbytes can easily get wrong, but doesn't affect | |
1072 | * anything in the fs. So here we skip the check. | |
1073 | */ | |
1074 | mode = btrfs_inode_mode(leaf, iitem); | |
c7c01a4a | 1075 | if (unlikely(mode & ~valid_mask)) { |
c3053ebb | 1076 | inode_item_err(leaf, slot, |
496245ca QW |
1077 | "unknown mode bit detected: 0x%x", |
1078 | mode & ~valid_mask); | |
1079 | return -EUCLEAN; | |
1080 | } | |
1081 | ||
1082 | /* | |
c1499166 DS |
1083 | * S_IFMT is not bit mapped so we can't completely rely on |
1084 | * is_power_of_2/has_single_bit_set, but it can save us from checking | |
1085 | * FIFO/CHR/DIR/REG. Only needs to check BLK, LNK and SOCKS | |
496245ca | 1086 | */ |
c1499166 | 1087 | if (!has_single_bit_set(mode & S_IFMT)) { |
c7c01a4a | 1088 | if (unlikely(!S_ISLNK(mode) && !S_ISBLK(mode) && !S_ISSOCK(mode))) { |
c3053ebb | 1089 | inode_item_err(leaf, slot, |
496245ca QW |
1090 | "invalid mode: has 0%o expect valid S_IF* bit(s)", |
1091 | mode & S_IFMT); | |
1092 | return -EUCLEAN; | |
1093 | } | |
1094 | } | |
c7c01a4a | 1095 | if (unlikely(S_ISDIR(mode) && btrfs_inode_nlink(leaf, iitem) > 1)) { |
c3053ebb | 1096 | inode_item_err(leaf, slot, |
496245ca QW |
1097 | "invalid nlink: has %u expect no more than 1 for dir", |
1098 | btrfs_inode_nlink(leaf, iitem)); | |
1099 | return -EUCLEAN; | |
1100 | } | |
77eea05e BB |
1101 | btrfs_inode_split_flags(btrfs_inode_flags(leaf, iitem), &flags, &ro_flags); |
1102 | if (unlikely(flags & ~BTRFS_INODE_FLAG_MASK)) { | |
c3053ebb | 1103 | inode_item_err(leaf, slot, |
77eea05e BB |
1104 | "unknown incompat flags detected: 0x%x", flags); |
1105 | return -EUCLEAN; | |
1106 | } | |
1107 | if (unlikely(!sb_rdonly(fs_info->sb) && | |
1108 | (ro_flags & ~BTRFS_INODE_RO_FLAG_MASK))) { | |
1109 | inode_item_err(leaf, slot, | |
1110 | "unknown ro-compat flags detected on writeable mount: 0x%x", | |
1111 | ro_flags); | |
496245ca QW |
1112 | return -EUCLEAN; |
1113 | } | |
1114 | return 0; | |
1115 | } | |
1116 | ||
259ee775 QW |
1117 | static int check_root_item(struct extent_buffer *leaf, struct btrfs_key *key, |
1118 | int slot) | |
1119 | { | |
1120 | struct btrfs_fs_info *fs_info = leaf->fs_info; | |
1465af12 | 1121 | struct btrfs_root_item ri = { 0 }; |
259ee775 QW |
1122 | const u64 valid_root_flags = BTRFS_ROOT_SUBVOL_RDONLY | |
1123 | BTRFS_ROOT_SUBVOL_DEAD; | |
57a0e674 | 1124 | int ret; |
259ee775 | 1125 | |
57a0e674 | 1126 | ret = check_root_key(leaf, key, slot); |
c7c01a4a | 1127 | if (unlikely(ret < 0)) |
57a0e674 | 1128 | return ret; |
259ee775 | 1129 | |
3212fa14 JB |
1130 | if (unlikely(btrfs_item_size(leaf, slot) != sizeof(ri) && |
1131 | btrfs_item_size(leaf, slot) != | |
c7c01a4a | 1132 | btrfs_legacy_root_item_size())) { |
259ee775 | 1133 | generic_err(leaf, slot, |
1465af12 | 1134 | "invalid root item size, have %u expect %zu or %u", |
3212fa14 | 1135 | btrfs_item_size(leaf, slot), sizeof(ri), |
1465af12 | 1136 | btrfs_legacy_root_item_size()); |
1a49a97d | 1137 | return -EUCLEAN; |
259ee775 QW |
1138 | } |
1139 | ||
1465af12 QW |
1140 | /* |
1141 | * For legacy root item, the members starting at generation_v2 will be | |
1142 | * all filled with 0. | |
1143 | * And since we allow geneartion_v2 as 0, it will still pass the check. | |
1144 | */ | |
259ee775 | 1145 | read_extent_buffer(leaf, &ri, btrfs_item_ptr_offset(leaf, slot), |
3212fa14 | 1146 | btrfs_item_size(leaf, slot)); |
259ee775 QW |
1147 | |
1148 | /* Generation related */ | |
c7c01a4a DS |
1149 | if (unlikely(btrfs_root_generation(&ri) > |
1150 | btrfs_super_generation(fs_info->super_copy) + 1)) { | |
259ee775 QW |
1151 | generic_err(leaf, slot, |
1152 | "invalid root generation, have %llu expect (0, %llu]", | |
1153 | btrfs_root_generation(&ri), | |
1154 | btrfs_super_generation(fs_info->super_copy) + 1); | |
1155 | return -EUCLEAN; | |
1156 | } | |
c7c01a4a DS |
1157 | if (unlikely(btrfs_root_generation_v2(&ri) > |
1158 | btrfs_super_generation(fs_info->super_copy) + 1)) { | |
259ee775 QW |
1159 | generic_err(leaf, slot, |
1160 | "invalid root v2 generation, have %llu expect (0, %llu]", | |
1161 | btrfs_root_generation_v2(&ri), | |
1162 | btrfs_super_generation(fs_info->super_copy) + 1); | |
1163 | return -EUCLEAN; | |
1164 | } | |
c7c01a4a DS |
1165 | if (unlikely(btrfs_root_last_snapshot(&ri) > |
1166 | btrfs_super_generation(fs_info->super_copy) + 1)) { | |
259ee775 QW |
1167 | generic_err(leaf, slot, |
1168 | "invalid root last_snapshot, have %llu expect (0, %llu]", | |
1169 | btrfs_root_last_snapshot(&ri), | |
1170 | btrfs_super_generation(fs_info->super_copy) + 1); | |
1171 | return -EUCLEAN; | |
1172 | } | |
1173 | ||
1174 | /* Alignment and level check */ | |
c7c01a4a | 1175 | if (unlikely(!IS_ALIGNED(btrfs_root_bytenr(&ri), fs_info->sectorsize))) { |
259ee775 QW |
1176 | generic_err(leaf, slot, |
1177 | "invalid root bytenr, have %llu expect to be aligned to %u", | |
1178 | btrfs_root_bytenr(&ri), fs_info->sectorsize); | |
1179 | return -EUCLEAN; | |
1180 | } | |
c7c01a4a | 1181 | if (unlikely(btrfs_root_level(&ri) >= BTRFS_MAX_LEVEL)) { |
259ee775 QW |
1182 | generic_err(leaf, slot, |
1183 | "invalid root level, have %u expect [0, %u]", | |
1184 | btrfs_root_level(&ri), BTRFS_MAX_LEVEL - 1); | |
1185 | return -EUCLEAN; | |
1186 | } | |
c7c01a4a | 1187 | if (unlikely(btrfs_root_drop_level(&ri) >= BTRFS_MAX_LEVEL)) { |
259ee775 QW |
1188 | generic_err(leaf, slot, |
1189 | "invalid root level, have %u expect [0, %u]", | |
c8422684 | 1190 | btrfs_root_drop_level(&ri), BTRFS_MAX_LEVEL - 1); |
259ee775 QW |
1191 | return -EUCLEAN; |
1192 | } | |
1193 | ||
1194 | /* Flags check */ | |
c7c01a4a | 1195 | if (unlikely(btrfs_root_flags(&ri) & ~valid_root_flags)) { |
259ee775 QW |
1196 | generic_err(leaf, slot, |
1197 | "invalid root flags, have 0x%llx expect mask 0x%llx", | |
1198 | btrfs_root_flags(&ri), valid_root_flags); | |
1199 | return -EUCLEAN; | |
1200 | } | |
1201 | return 0; | |
1202 | } | |
1203 | ||
f82d1c7c QW |
1204 | __printf(3,4) |
1205 | __cold | |
1206 | static void extent_err(const struct extent_buffer *eb, int slot, | |
1207 | const char *fmt, ...) | |
1208 | { | |
1209 | struct btrfs_key key; | |
1210 | struct va_format vaf; | |
1211 | va_list args; | |
1212 | u64 bytenr; | |
1213 | u64 len; | |
1214 | ||
1215 | btrfs_item_key_to_cpu(eb, &key, slot); | |
1216 | bytenr = key.objectid; | |
e2406a6f QW |
1217 | if (key.type == BTRFS_METADATA_ITEM_KEY || |
1218 | key.type == BTRFS_TREE_BLOCK_REF_KEY || | |
1219 | key.type == BTRFS_SHARED_BLOCK_REF_KEY) | |
f82d1c7c QW |
1220 | len = eb->fs_info->nodesize; |
1221 | else | |
1222 | len = key.offset; | |
1223 | va_start(args, fmt); | |
1224 | ||
1225 | vaf.fmt = fmt; | |
1226 | vaf.va = &args; | |
1227 | ||
1228 | btrfs_crit(eb->fs_info, | |
1229 | "corrupt %s: block=%llu slot=%d extent bytenr=%llu len=%llu %pV", | |
1230 | btrfs_header_level(eb) == 0 ? "leaf" : "node", | |
1231 | eb->start, slot, bytenr, len, &vaf); | |
1232 | va_end(args); | |
1233 | } | |
1234 | ||
1235 | static int check_extent_item(struct extent_buffer *leaf, | |
899b7f69 JB |
1236 | struct btrfs_key *key, int slot, |
1237 | struct btrfs_key *prev_key) | |
f82d1c7c QW |
1238 | { |
1239 | struct btrfs_fs_info *fs_info = leaf->fs_info; | |
1240 | struct btrfs_extent_item *ei; | |
1241 | bool is_tree_block = false; | |
1242 | unsigned long ptr; /* Current pointer inside inline refs */ | |
1243 | unsigned long end; /* Extent item end */ | |
3212fa14 | 1244 | const u32 item_size = btrfs_item_size(leaf, slot); |
f82d1c7c QW |
1245 | u64 flags; |
1246 | u64 generation; | |
1247 | u64 total_refs; /* Total refs in btrfs_extent_item */ | |
1248 | u64 inline_refs = 0; /* found total inline refs */ | |
1249 | ||
c7c01a4a DS |
1250 | if (unlikely(key->type == BTRFS_METADATA_ITEM_KEY && |
1251 | !btrfs_fs_incompat(fs_info, SKINNY_METADATA))) { | |
f82d1c7c QW |
1252 | generic_err(leaf, slot, |
1253 | "invalid key type, METADATA_ITEM type invalid when SKINNY_METADATA feature disabled"); | |
1254 | return -EUCLEAN; | |
1255 | } | |
1256 | /* key->objectid is the bytenr for both key types */ | |
c7c01a4a | 1257 | if (unlikely(!IS_ALIGNED(key->objectid, fs_info->sectorsize))) { |
f82d1c7c QW |
1258 | generic_err(leaf, slot, |
1259 | "invalid key objectid, have %llu expect to be aligned to %u", | |
1260 | key->objectid, fs_info->sectorsize); | |
1261 | return -EUCLEAN; | |
1262 | } | |
1263 | ||
1264 | /* key->offset is tree level for METADATA_ITEM_KEY */ | |
c7c01a4a DS |
1265 | if (unlikely(key->type == BTRFS_METADATA_ITEM_KEY && |
1266 | key->offset >= BTRFS_MAX_LEVEL)) { | |
f82d1c7c QW |
1267 | extent_err(leaf, slot, |
1268 | "invalid tree level, have %llu expect [0, %u]", | |
1269 | key->offset, BTRFS_MAX_LEVEL - 1); | |
1270 | return -EUCLEAN; | |
1271 | } | |
1272 | ||
1273 | /* | |
1274 | * EXTENT/METADATA_ITEM consists of: | |
1275 | * 1) One btrfs_extent_item | |
1276 | * Records the total refs, type and generation of the extent. | |
1277 | * | |
1278 | * 2) One btrfs_tree_block_info (for EXTENT_ITEM and tree backref only) | |
1279 | * Records the first key and level of the tree block. | |
1280 | * | |
1281 | * 2) Zero or more btrfs_extent_inline_ref(s) | |
1282 | * Each inline ref has one btrfs_extent_inline_ref shows: | |
1283 | * 2.1) The ref type, one of the 4 | |
1284 | * TREE_BLOCK_REF Tree block only | |
1285 | * SHARED_BLOCK_REF Tree block only | |
1286 | * EXTENT_DATA_REF Data only | |
1287 | * SHARED_DATA_REF Data only | |
1288 | * 2.2) Ref type specific data | |
1289 | * Either using btrfs_extent_inline_ref::offset, or specific | |
1290 | * data structure. | |
1291 | */ | |
c7c01a4a | 1292 | if (unlikely(item_size < sizeof(*ei))) { |
f82d1c7c QW |
1293 | extent_err(leaf, slot, |
1294 | "invalid item size, have %u expect [%zu, %u)", | |
1295 | item_size, sizeof(*ei), | |
1296 | BTRFS_LEAF_DATA_SIZE(fs_info)); | |
1297 | return -EUCLEAN; | |
1298 | } | |
1299 | end = item_size + btrfs_item_ptr_offset(leaf, slot); | |
1300 | ||
1301 | /* Checks against extent_item */ | |
1302 | ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item); | |
1303 | flags = btrfs_extent_flags(leaf, ei); | |
1304 | total_refs = btrfs_extent_refs(leaf, ei); | |
1305 | generation = btrfs_extent_generation(leaf, ei); | |
c7c01a4a DS |
1306 | if (unlikely(generation > |
1307 | btrfs_super_generation(fs_info->super_copy) + 1)) { | |
f82d1c7c QW |
1308 | extent_err(leaf, slot, |
1309 | "invalid generation, have %llu expect (0, %llu]", | |
1310 | generation, | |
1311 | btrfs_super_generation(fs_info->super_copy) + 1); | |
1312 | return -EUCLEAN; | |
1313 | } | |
c7c01a4a DS |
1314 | if (unlikely(!has_single_bit_set(flags & (BTRFS_EXTENT_FLAG_DATA | |
1315 | BTRFS_EXTENT_FLAG_TREE_BLOCK)))) { | |
f82d1c7c QW |
1316 | extent_err(leaf, slot, |
1317 | "invalid extent flag, have 0x%llx expect 1 bit set in 0x%llx", | |
1318 | flags, BTRFS_EXTENT_FLAG_DATA | | |
1319 | BTRFS_EXTENT_FLAG_TREE_BLOCK); | |
1320 | return -EUCLEAN; | |
1321 | } | |
1322 | is_tree_block = !!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK); | |
1323 | if (is_tree_block) { | |
c7c01a4a DS |
1324 | if (unlikely(key->type == BTRFS_EXTENT_ITEM_KEY && |
1325 | key->offset != fs_info->nodesize)) { | |
f82d1c7c QW |
1326 | extent_err(leaf, slot, |
1327 | "invalid extent length, have %llu expect %u", | |
1328 | key->offset, fs_info->nodesize); | |
1329 | return -EUCLEAN; | |
1330 | } | |
1331 | } else { | |
c7c01a4a | 1332 | if (unlikely(key->type != BTRFS_EXTENT_ITEM_KEY)) { |
f82d1c7c QW |
1333 | extent_err(leaf, slot, |
1334 | "invalid key type, have %u expect %u for data backref", | |
1335 | key->type, BTRFS_EXTENT_ITEM_KEY); | |
1336 | return -EUCLEAN; | |
1337 | } | |
c7c01a4a | 1338 | if (unlikely(!IS_ALIGNED(key->offset, fs_info->sectorsize))) { |
f82d1c7c QW |
1339 | extent_err(leaf, slot, |
1340 | "invalid extent length, have %llu expect aligned to %u", | |
1341 | key->offset, fs_info->sectorsize); | |
1342 | return -EUCLEAN; | |
1343 | } | |
0ebb6bbb JB |
1344 | if (unlikely(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)) { |
1345 | extent_err(leaf, slot, | |
1346 | "invalid extent flag, data has full backref set"); | |
1347 | return -EUCLEAN; | |
1348 | } | |
f82d1c7c QW |
1349 | } |
1350 | ptr = (unsigned long)(struct btrfs_extent_item *)(ei + 1); | |
1351 | ||
1352 | /* Check the special case of btrfs_tree_block_info */ | |
1353 | if (is_tree_block && key->type != BTRFS_METADATA_ITEM_KEY) { | |
1354 | struct btrfs_tree_block_info *info; | |
1355 | ||
1356 | info = (struct btrfs_tree_block_info *)ptr; | |
c7c01a4a | 1357 | if (unlikely(btrfs_tree_block_level(leaf, info) >= BTRFS_MAX_LEVEL)) { |
f82d1c7c QW |
1358 | extent_err(leaf, slot, |
1359 | "invalid tree block info level, have %u expect [0, %u]", | |
1360 | btrfs_tree_block_level(leaf, info), | |
1361 | BTRFS_MAX_LEVEL - 1); | |
1362 | return -EUCLEAN; | |
1363 | } | |
1364 | ptr = (unsigned long)(struct btrfs_tree_block_info *)(info + 1); | |
1365 | } | |
1366 | ||
1367 | /* Check inline refs */ | |
1368 | while (ptr < end) { | |
1369 | struct btrfs_extent_inline_ref *iref; | |
1370 | struct btrfs_extent_data_ref *dref; | |
1371 | struct btrfs_shared_data_ref *sref; | |
1372 | u64 dref_offset; | |
1373 | u64 inline_offset; | |
1374 | u8 inline_type; | |
1375 | ||
c7c01a4a | 1376 | if (unlikely(ptr + sizeof(*iref) > end)) { |
f82d1c7c QW |
1377 | extent_err(leaf, slot, |
1378 | "inline ref item overflows extent item, ptr %lu iref size %zu end %lu", | |
1379 | ptr, sizeof(*iref), end); | |
1380 | return -EUCLEAN; | |
1381 | } | |
1382 | iref = (struct btrfs_extent_inline_ref *)ptr; | |
1383 | inline_type = btrfs_extent_inline_ref_type(leaf, iref); | |
1384 | inline_offset = btrfs_extent_inline_ref_offset(leaf, iref); | |
c7c01a4a | 1385 | if (unlikely(ptr + btrfs_extent_inline_ref_size(inline_type) > end)) { |
f82d1c7c QW |
1386 | extent_err(leaf, slot, |
1387 | "inline ref item overflows extent item, ptr %lu iref size %u end %lu", | |
1388 | ptr, inline_type, end); | |
1389 | return -EUCLEAN; | |
1390 | } | |
1391 | ||
1392 | switch (inline_type) { | |
1393 | /* inline_offset is subvolid of the owner, no need to check */ | |
1394 | case BTRFS_TREE_BLOCK_REF_KEY: | |
1395 | inline_refs++; | |
1396 | break; | |
1397 | /* Contains parent bytenr */ | |
1398 | case BTRFS_SHARED_BLOCK_REF_KEY: | |
c7c01a4a DS |
1399 | if (unlikely(!IS_ALIGNED(inline_offset, |
1400 | fs_info->sectorsize))) { | |
f82d1c7c QW |
1401 | extent_err(leaf, slot, |
1402 | "invalid tree parent bytenr, have %llu expect aligned to %u", | |
1403 | inline_offset, fs_info->sectorsize); | |
1404 | return -EUCLEAN; | |
1405 | } | |
1406 | inline_refs++; | |
1407 | break; | |
1408 | /* | |
1409 | * Contains owner subvolid, owner key objectid, adjusted offset. | |
1410 | * The only obvious corruption can happen in that offset. | |
1411 | */ | |
1412 | case BTRFS_EXTENT_DATA_REF_KEY: | |
1413 | dref = (struct btrfs_extent_data_ref *)(&iref->offset); | |
1414 | dref_offset = btrfs_extent_data_ref_offset(leaf, dref); | |
c7c01a4a DS |
1415 | if (unlikely(!IS_ALIGNED(dref_offset, |
1416 | fs_info->sectorsize))) { | |
f82d1c7c QW |
1417 | extent_err(leaf, slot, |
1418 | "invalid data ref offset, have %llu expect aligned to %u", | |
1419 | dref_offset, fs_info->sectorsize); | |
1420 | return -EUCLEAN; | |
1421 | } | |
1422 | inline_refs += btrfs_extent_data_ref_count(leaf, dref); | |
1423 | break; | |
1424 | /* Contains parent bytenr and ref count */ | |
1425 | case BTRFS_SHARED_DATA_REF_KEY: | |
1426 | sref = (struct btrfs_shared_data_ref *)(iref + 1); | |
c7c01a4a DS |
1427 | if (unlikely(!IS_ALIGNED(inline_offset, |
1428 | fs_info->sectorsize))) { | |
f82d1c7c QW |
1429 | extent_err(leaf, slot, |
1430 | "invalid data parent bytenr, have %llu expect aligned to %u", | |
1431 | inline_offset, fs_info->sectorsize); | |
1432 | return -EUCLEAN; | |
1433 | } | |
1434 | inline_refs += btrfs_shared_data_ref_count(leaf, sref); | |
1435 | break; | |
1436 | default: | |
1437 | extent_err(leaf, slot, "unknown inline ref type: %u", | |
1438 | inline_type); | |
1439 | return -EUCLEAN; | |
1440 | } | |
1441 | ptr += btrfs_extent_inline_ref_size(inline_type); | |
1442 | } | |
1443 | /* No padding is allowed */ | |
c7c01a4a | 1444 | if (unlikely(ptr != end)) { |
f82d1c7c QW |
1445 | extent_err(leaf, slot, |
1446 | "invalid extent item size, padding bytes found"); | |
1447 | return -EUCLEAN; | |
1448 | } | |
1449 | ||
1450 | /* Finally, check the inline refs against total refs */ | |
c7c01a4a | 1451 | if (unlikely(inline_refs > total_refs)) { |
f82d1c7c QW |
1452 | extent_err(leaf, slot, |
1453 | "invalid extent refs, have %llu expect >= inline %llu", | |
1454 | total_refs, inline_refs); | |
1455 | return -EUCLEAN; | |
1456 | } | |
899b7f69 JB |
1457 | |
1458 | if ((prev_key->type == BTRFS_EXTENT_ITEM_KEY) || | |
1459 | (prev_key->type == BTRFS_METADATA_ITEM_KEY)) { | |
1460 | u64 prev_end = prev_key->objectid; | |
1461 | ||
1462 | if (prev_key->type == BTRFS_METADATA_ITEM_KEY) | |
1463 | prev_end += fs_info->nodesize; | |
1464 | else | |
1465 | prev_end += prev_key->offset; | |
1466 | ||
1467 | if (unlikely(prev_end > key->objectid)) { | |
1468 | extent_err(leaf, slot, | |
1469 | "previous extent [%llu %u %llu] overlaps current extent [%llu %u %llu]", | |
1470 | prev_key->objectid, prev_key->type, | |
1471 | prev_key->offset, key->objectid, key->type, | |
1472 | key->offset); | |
1473 | return -EUCLEAN; | |
1474 | } | |
1475 | } | |
1476 | ||
f82d1c7c QW |
1477 | return 0; |
1478 | } | |
1479 | ||
e2406a6f QW |
1480 | static int check_simple_keyed_refs(struct extent_buffer *leaf, |
1481 | struct btrfs_key *key, int slot) | |
1482 | { | |
1483 | u32 expect_item_size = 0; | |
1484 | ||
1485 | if (key->type == BTRFS_SHARED_DATA_REF_KEY) | |
1486 | expect_item_size = sizeof(struct btrfs_shared_data_ref); | |
1487 | ||
3212fa14 | 1488 | if (unlikely(btrfs_item_size(leaf, slot) != expect_item_size)) { |
e2406a6f QW |
1489 | generic_err(leaf, slot, |
1490 | "invalid item size, have %u expect %u for key type %u", | |
3212fa14 | 1491 | btrfs_item_size(leaf, slot), |
e2406a6f QW |
1492 | expect_item_size, key->type); |
1493 | return -EUCLEAN; | |
1494 | } | |
c7c01a4a | 1495 | if (unlikely(!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize))) { |
e2406a6f QW |
1496 | generic_err(leaf, slot, |
1497 | "invalid key objectid for shared block ref, have %llu expect aligned to %u", | |
1498 | key->objectid, leaf->fs_info->sectorsize); | |
1499 | return -EUCLEAN; | |
1500 | } | |
c7c01a4a DS |
1501 | if (unlikely(key->type != BTRFS_TREE_BLOCK_REF_KEY && |
1502 | !IS_ALIGNED(key->offset, leaf->fs_info->sectorsize))) { | |
e2406a6f QW |
1503 | extent_err(leaf, slot, |
1504 | "invalid tree parent bytenr, have %llu expect aligned to %u", | |
1505 | key->offset, leaf->fs_info->sectorsize); | |
1506 | return -EUCLEAN; | |
1507 | } | |
1508 | return 0; | |
1509 | } | |
1510 | ||
0785a9aa QW |
1511 | static int check_extent_data_ref(struct extent_buffer *leaf, |
1512 | struct btrfs_key *key, int slot) | |
1513 | { | |
1514 | struct btrfs_extent_data_ref *dref; | |
1515 | unsigned long ptr = btrfs_item_ptr_offset(leaf, slot); | |
3212fa14 | 1516 | const unsigned long end = ptr + btrfs_item_size(leaf, slot); |
0785a9aa | 1517 | |
3212fa14 | 1518 | if (unlikely(btrfs_item_size(leaf, slot) % sizeof(*dref) != 0)) { |
0785a9aa QW |
1519 | generic_err(leaf, slot, |
1520 | "invalid item size, have %u expect aligned to %zu for key type %u", | |
3212fa14 | 1521 | btrfs_item_size(leaf, slot), |
0785a9aa | 1522 | sizeof(*dref), key->type); |
6d06b0ad | 1523 | return -EUCLEAN; |
0785a9aa | 1524 | } |
c7c01a4a | 1525 | if (unlikely(!IS_ALIGNED(key->objectid, leaf->fs_info->sectorsize))) { |
0785a9aa QW |
1526 | generic_err(leaf, slot, |
1527 | "invalid key objectid for shared block ref, have %llu expect aligned to %u", | |
1528 | key->objectid, leaf->fs_info->sectorsize); | |
1529 | return -EUCLEAN; | |
1530 | } | |
1531 | for (; ptr < end; ptr += sizeof(*dref)) { | |
0785a9aa | 1532 | u64 offset; |
0785a9aa | 1533 | |
1119a72e JB |
1534 | /* |
1535 | * We cannot check the extent_data_ref hash due to possible | |
1536 | * overflow from the leaf due to hash collisions. | |
1537 | */ | |
0785a9aa | 1538 | dref = (struct btrfs_extent_data_ref *)ptr; |
0785a9aa | 1539 | offset = btrfs_extent_data_ref_offset(leaf, dref); |
c7c01a4a | 1540 | if (unlikely(!IS_ALIGNED(offset, leaf->fs_info->sectorsize))) { |
0785a9aa QW |
1541 | extent_err(leaf, slot, |
1542 | "invalid extent data backref offset, have %llu expect aligned to %u", | |
1543 | offset, leaf->fs_info->sectorsize); | |
6d06b0ad | 1544 | return -EUCLEAN; |
0785a9aa QW |
1545 | } |
1546 | } | |
1547 | return 0; | |
1548 | } | |
1549 | ||
c3053ebb QW |
1550 | #define inode_ref_err(eb, slot, fmt, args...) \ |
1551 | inode_item_err(eb, slot, fmt, ##args) | |
71bf92a9 QW |
1552 | static int check_inode_ref(struct extent_buffer *leaf, |
1553 | struct btrfs_key *key, struct btrfs_key *prev_key, | |
1554 | int slot) | |
1555 | { | |
1556 | struct btrfs_inode_ref *iref; | |
1557 | unsigned long ptr; | |
1558 | unsigned long end; | |
1559 | ||
c7c01a4a | 1560 | if (unlikely(!check_prev_ino(leaf, key, slot, prev_key))) |
80d7fd1e | 1561 | return -EUCLEAN; |
71bf92a9 | 1562 | /* namelen can't be 0, so item_size == sizeof() is also invalid */ |
3212fa14 | 1563 | if (unlikely(btrfs_item_size(leaf, slot) <= sizeof(*iref))) { |
c3053ebb | 1564 | inode_ref_err(leaf, slot, |
71bf92a9 | 1565 | "invalid item size, have %u expect (%zu, %u)", |
3212fa14 | 1566 | btrfs_item_size(leaf, slot), |
71bf92a9 QW |
1567 | sizeof(*iref), BTRFS_LEAF_DATA_SIZE(leaf->fs_info)); |
1568 | return -EUCLEAN; | |
1569 | } | |
1570 | ||
1571 | ptr = btrfs_item_ptr_offset(leaf, slot); | |
3212fa14 | 1572 | end = ptr + btrfs_item_size(leaf, slot); |
71bf92a9 QW |
1573 | while (ptr < end) { |
1574 | u16 namelen; | |
1575 | ||
c7c01a4a | 1576 | if (unlikely(ptr + sizeof(iref) > end)) { |
c3053ebb | 1577 | inode_ref_err(leaf, slot, |
71bf92a9 QW |
1578 | "inode ref overflow, ptr %lu end %lu inode_ref_size %zu", |
1579 | ptr, end, sizeof(iref)); | |
1580 | return -EUCLEAN; | |
1581 | } | |
1582 | ||
1583 | iref = (struct btrfs_inode_ref *)ptr; | |
1584 | namelen = btrfs_inode_ref_name_len(leaf, iref); | |
c7c01a4a | 1585 | if (unlikely(ptr + sizeof(*iref) + namelen > end)) { |
c3053ebb | 1586 | inode_ref_err(leaf, slot, |
71bf92a9 QW |
1587 | "inode ref overflow, ptr %lu end %lu namelen %u", |
1588 | ptr, end, namelen); | |
1589 | return -EUCLEAN; | |
1590 | } | |
1591 | ||
1592 | /* | |
1593 | * NOTE: In theory we should record all found index numbers | |
1594 | * to find any duplicated indexes, but that will be too time | |
1595 | * consuming for inodes with too many hard links. | |
1596 | */ | |
1597 | ptr += sizeof(*iref) + namelen; | |
1598 | } | |
1599 | return 0; | |
1600 | } | |
1601 | ||
557ea5dd QW |
1602 | /* |
1603 | * Common point to switch the item-specific validation. | |
1604 | */ | |
0076bc89 | 1605 | static int check_leaf_item(struct extent_buffer *leaf, |
4e9845ef FM |
1606 | struct btrfs_key *key, int slot, |
1607 | struct btrfs_key *prev_key) | |
557ea5dd QW |
1608 | { |
1609 | int ret = 0; | |
075cb3c7 | 1610 | struct btrfs_chunk *chunk; |
557ea5dd QW |
1611 | |
1612 | switch (key->type) { | |
1613 | case BTRFS_EXTENT_DATA_KEY: | |
4e9845ef | 1614 | ret = check_extent_data_item(leaf, key, slot, prev_key); |
557ea5dd QW |
1615 | break; |
1616 | case BTRFS_EXTENT_CSUM_KEY: | |
ad1d8c43 | 1617 | ret = check_csum_item(leaf, key, slot, prev_key); |
557ea5dd | 1618 | break; |
ad7b0368 QW |
1619 | case BTRFS_DIR_ITEM_KEY: |
1620 | case BTRFS_DIR_INDEX_KEY: | |
1621 | case BTRFS_XATTR_ITEM_KEY: | |
c18679eb | 1622 | ret = check_dir_item(leaf, key, prev_key, slot); |
ad7b0368 | 1623 | break; |
71bf92a9 QW |
1624 | case BTRFS_INODE_REF_KEY: |
1625 | ret = check_inode_ref(leaf, key, prev_key, slot); | |
1626 | break; | |
fce466ea | 1627 | case BTRFS_BLOCK_GROUP_ITEM_KEY: |
af60ce2b | 1628 | ret = check_block_group_item(leaf, key, slot); |
fce466ea | 1629 | break; |
075cb3c7 QW |
1630 | case BTRFS_CHUNK_ITEM_KEY: |
1631 | chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk); | |
f6d2a5c2 | 1632 | ret = check_leaf_chunk_item(leaf, chunk, key, slot); |
075cb3c7 | 1633 | break; |
ab4ba2e1 | 1634 | case BTRFS_DEV_ITEM_KEY: |
412a2312 | 1635 | ret = check_dev_item(leaf, key, slot); |
ab4ba2e1 | 1636 | break; |
496245ca | 1637 | case BTRFS_INODE_ITEM_KEY: |
39e57f49 | 1638 | ret = check_inode_item(leaf, key, slot); |
496245ca | 1639 | break; |
259ee775 QW |
1640 | case BTRFS_ROOT_ITEM_KEY: |
1641 | ret = check_root_item(leaf, key, slot); | |
1642 | break; | |
f82d1c7c QW |
1643 | case BTRFS_EXTENT_ITEM_KEY: |
1644 | case BTRFS_METADATA_ITEM_KEY: | |
899b7f69 | 1645 | ret = check_extent_item(leaf, key, slot, prev_key); |
f82d1c7c | 1646 | break; |
e2406a6f QW |
1647 | case BTRFS_TREE_BLOCK_REF_KEY: |
1648 | case BTRFS_SHARED_DATA_REF_KEY: | |
1649 | case BTRFS_SHARED_BLOCK_REF_KEY: | |
1650 | ret = check_simple_keyed_refs(leaf, key, slot); | |
1651 | break; | |
0785a9aa QW |
1652 | case BTRFS_EXTENT_DATA_REF_KEY: |
1653 | ret = check_extent_data_ref(leaf, key, slot); | |
1654 | break; | |
557ea5dd QW |
1655 | } |
1656 | return ret; | |
1657 | } | |
1658 | ||
e2ccd361 | 1659 | static int check_leaf(struct extent_buffer *leaf, bool check_item_data) |
557ea5dd | 1660 | { |
e2ccd361 | 1661 | struct btrfs_fs_info *fs_info = leaf->fs_info; |
557ea5dd QW |
1662 | /* No valid key type is 0, so all key should be larger than this key */ |
1663 | struct btrfs_key prev_key = {0, 0, 0}; | |
1664 | struct btrfs_key key; | |
1665 | u32 nritems = btrfs_header_nritems(leaf); | |
1666 | int slot; | |
1667 | ||
c7c01a4a | 1668 | if (unlikely(btrfs_header_level(leaf) != 0)) { |
86a6be3a | 1669 | generic_err(leaf, 0, |
f556faa4 QW |
1670 | "invalid level for leaf, have %d expect 0", |
1671 | btrfs_header_level(leaf)); | |
1672 | return -EUCLEAN; | |
1673 | } | |
1674 | ||
557ea5dd QW |
1675 | /* |
1676 | * Extent buffers from a relocation tree have a owner field that | |
1677 | * corresponds to the subvolume tree they are based on. So just from an | |
1678 | * extent buffer alone we can not find out what is the id of the | |
1679 | * corresponding subvolume tree, so we can not figure out if the extent | |
1680 | * buffer corresponds to the root of the relocation tree or not. So | |
1681 | * skip this check for relocation trees. | |
1682 | */ | |
1683 | if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) { | |
ba480dd4 | 1684 | u64 owner = btrfs_header_owner(leaf); |
557ea5dd | 1685 | |
ba480dd4 | 1686 | /* These trees must never be empty */ |
c7c01a4a DS |
1687 | if (unlikely(owner == BTRFS_ROOT_TREE_OBJECTID || |
1688 | owner == BTRFS_CHUNK_TREE_OBJECTID || | |
c7c01a4a DS |
1689 | owner == BTRFS_DEV_TREE_OBJECTID || |
1690 | owner == BTRFS_FS_TREE_OBJECTID || | |
1691 | owner == BTRFS_DATA_RELOC_TREE_OBJECTID)) { | |
86a6be3a | 1692 | generic_err(leaf, 0, |
ba480dd4 QW |
1693 | "invalid root, root %llu must never be empty", |
1694 | owner); | |
1695 | return -EUCLEAN; | |
1696 | } | |
c2fa821c | 1697 | |
62fdaa52 | 1698 | /* Unknown tree */ |
c7c01a4a | 1699 | if (unlikely(owner == 0)) { |
62fdaa52 QW |
1700 | generic_err(leaf, 0, |
1701 | "invalid owner, root 0 is not defined"); | |
1702 | return -EUCLEAN; | |
1703 | } | |
c2fa821c JB |
1704 | |
1705 | /* EXTENT_TREE_V2 can have empty extent trees. */ | |
1706 | if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) | |
1707 | return 0; | |
1708 | ||
1709 | if (unlikely(owner == BTRFS_EXTENT_TREE_OBJECTID)) { | |
1710 | generic_err(leaf, 0, | |
1711 | "invalid root, root %llu must never be empty", | |
1712 | owner); | |
1713 | return -EUCLEAN; | |
1714 | } | |
1715 | ||
557ea5dd QW |
1716 | return 0; |
1717 | } | |
1718 | ||
c7c01a4a | 1719 | if (unlikely(nritems == 0)) |
557ea5dd QW |
1720 | return 0; |
1721 | ||
1722 | /* | |
1723 | * Check the following things to make sure this is a good leaf, and | |
1724 | * leaf users won't need to bother with similar sanity checks: | |
1725 | * | |
1726 | * 1) key ordering | |
1727 | * 2) item offset and size | |
1728 | * No overlap, no hole, all inside the leaf. | |
1729 | * 3) item content | |
1730 | * If possible, do comprehensive sanity check. | |
1731 | * NOTE: All checks must only rely on the item data itself. | |
1732 | */ | |
1733 | for (slot = 0; slot < nritems; slot++) { | |
1734 | u32 item_end_expected; | |
a6ab66eb | 1735 | u64 item_data_end; |
557ea5dd QW |
1736 | int ret; |
1737 | ||
1738 | btrfs_item_key_to_cpu(leaf, &key, slot); | |
1739 | ||
1740 | /* Make sure the keys are in the right order */ | |
c7c01a4a | 1741 | if (unlikely(btrfs_comp_cpu_keys(&prev_key, &key) >= 0)) { |
86a6be3a | 1742 | generic_err(leaf, slot, |
478d01b3 QW |
1743 | "bad key order, prev (%llu %u %llu) current (%llu %u %llu)", |
1744 | prev_key.objectid, prev_key.type, | |
1745 | prev_key.offset, key.objectid, key.type, | |
1746 | key.offset); | |
557ea5dd QW |
1747 | return -EUCLEAN; |
1748 | } | |
1749 | ||
a6ab66eb SY |
1750 | item_data_end = (u64)btrfs_item_offset(leaf, slot) + |
1751 | btrfs_item_size(leaf, slot); | |
557ea5dd QW |
1752 | /* |
1753 | * Make sure the offset and ends are right, remember that the | |
1754 | * item data starts at the end of the leaf and grows towards the | |
1755 | * front. | |
1756 | */ | |
1757 | if (slot == 0) | |
1758 | item_end_expected = BTRFS_LEAF_DATA_SIZE(fs_info); | |
1759 | else | |
3212fa14 | 1760 | item_end_expected = btrfs_item_offset(leaf, |
557ea5dd | 1761 | slot - 1); |
a6ab66eb | 1762 | if (unlikely(item_data_end != item_end_expected)) { |
86a6be3a | 1763 | generic_err(leaf, slot, |
a6ab66eb SY |
1764 | "unexpected item end, have %llu expect %u", |
1765 | item_data_end, item_end_expected); | |
557ea5dd QW |
1766 | return -EUCLEAN; |
1767 | } | |
1768 | ||
1769 | /* | |
1770 | * Check to make sure that we don't point outside of the leaf, | |
1771 | * just in case all the items are consistent to each other, but | |
1772 | * all point outside of the leaf. | |
1773 | */ | |
a6ab66eb | 1774 | if (unlikely(item_data_end > BTRFS_LEAF_DATA_SIZE(fs_info))) { |
86a6be3a | 1775 | generic_err(leaf, slot, |
a6ab66eb SY |
1776 | "slot end outside of leaf, have %llu expect range [0, %u]", |
1777 | item_data_end, BTRFS_LEAF_DATA_SIZE(fs_info)); | |
557ea5dd QW |
1778 | return -EUCLEAN; |
1779 | } | |
1780 | ||
1781 | /* Also check if the item pointer overlaps with btrfs item. */ | |
c7c01a4a DS |
1782 | if (unlikely(btrfs_item_ptr_offset(leaf, slot) < |
1783 | btrfs_item_nr_offset(slot) + sizeof(struct btrfs_item))) { | |
86a6be3a | 1784 | generic_err(leaf, slot, |
478d01b3 QW |
1785 | "slot overlaps with its data, item end %lu data start %lu", |
1786 | btrfs_item_nr_offset(slot) + | |
1787 | sizeof(struct btrfs_item), | |
1788 | btrfs_item_ptr_offset(leaf, slot)); | |
557ea5dd QW |
1789 | return -EUCLEAN; |
1790 | } | |
1791 | ||
69fc6cbb QW |
1792 | if (check_item_data) { |
1793 | /* | |
1794 | * Check if the item size and content meet other | |
1795 | * criteria | |
1796 | */ | |
4e9845ef | 1797 | ret = check_leaf_item(leaf, &key, slot, &prev_key); |
c7c01a4a | 1798 | if (unlikely(ret < 0)) |
69fc6cbb QW |
1799 | return ret; |
1800 | } | |
557ea5dd QW |
1801 | |
1802 | prev_key.objectid = key.objectid; | |
1803 | prev_key.type = key.type; | |
1804 | prev_key.offset = key.offset; | |
1805 | } | |
1806 | ||
1807 | return 0; | |
1808 | } | |
1809 | ||
1c4360ee | 1810 | int btrfs_check_leaf_full(struct extent_buffer *leaf) |
69fc6cbb | 1811 | { |
e2ccd361 | 1812 | return check_leaf(leaf, true); |
69fc6cbb | 1813 | } |
02529d7a | 1814 | ALLOW_ERROR_INJECTION(btrfs_check_leaf_full, ERRNO); |
69fc6cbb | 1815 | |
cfdaad5e | 1816 | int btrfs_check_leaf_relaxed(struct extent_buffer *leaf) |
69fc6cbb | 1817 | { |
e2ccd361 | 1818 | return check_leaf(leaf, false); |
69fc6cbb QW |
1819 | } |
1820 | ||
813fd1dc | 1821 | int btrfs_check_node(struct extent_buffer *node) |
557ea5dd | 1822 | { |
813fd1dc | 1823 | struct btrfs_fs_info *fs_info = node->fs_info; |
557ea5dd QW |
1824 | unsigned long nr = btrfs_header_nritems(node); |
1825 | struct btrfs_key key, next_key; | |
1826 | int slot; | |
f556faa4 | 1827 | int level = btrfs_header_level(node); |
557ea5dd QW |
1828 | u64 bytenr; |
1829 | int ret = 0; | |
1830 | ||
c7c01a4a | 1831 | if (unlikely(level <= 0 || level >= BTRFS_MAX_LEVEL)) { |
86a6be3a | 1832 | generic_err(node, 0, |
f556faa4 QW |
1833 | "invalid level for node, have %d expect [1, %d]", |
1834 | level, BTRFS_MAX_LEVEL - 1); | |
1835 | return -EUCLEAN; | |
1836 | } | |
c7c01a4a | 1837 | if (unlikely(nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(fs_info))) { |
2f659546 | 1838 | btrfs_crit(fs_info, |
bba4f298 | 1839 | "corrupt node: root=%llu block=%llu, nritems too %s, have %lu expect range [1,%u]", |
2f659546 | 1840 | btrfs_header_owner(node), node->start, |
bba4f298 | 1841 | nr == 0 ? "small" : "large", nr, |
2f659546 | 1842 | BTRFS_NODEPTRS_PER_BLOCK(fs_info)); |
bba4f298 | 1843 | return -EUCLEAN; |
557ea5dd QW |
1844 | } |
1845 | ||
1846 | for (slot = 0; slot < nr - 1; slot++) { | |
1847 | bytenr = btrfs_node_blockptr(node, slot); | |
1848 | btrfs_node_key_to_cpu(node, &key, slot); | |
1849 | btrfs_node_key_to_cpu(node, &next_key, slot + 1); | |
1850 | ||
c7c01a4a | 1851 | if (unlikely(!bytenr)) { |
86a6be3a | 1852 | generic_err(node, slot, |
bba4f298 QW |
1853 | "invalid NULL node pointer"); |
1854 | ret = -EUCLEAN; | |
1855 | goto out; | |
1856 | } | |
c7c01a4a | 1857 | if (unlikely(!IS_ALIGNED(bytenr, fs_info->sectorsize))) { |
86a6be3a | 1858 | generic_err(node, slot, |
bba4f298 | 1859 | "unaligned pointer, have %llu should be aligned to %u", |
2f659546 | 1860 | bytenr, fs_info->sectorsize); |
bba4f298 | 1861 | ret = -EUCLEAN; |
557ea5dd QW |
1862 | goto out; |
1863 | } | |
1864 | ||
c7c01a4a | 1865 | if (unlikely(btrfs_comp_cpu_keys(&key, &next_key) >= 0)) { |
86a6be3a | 1866 | generic_err(node, slot, |
bba4f298 QW |
1867 | "bad key order, current (%llu %u %llu) next (%llu %u %llu)", |
1868 | key.objectid, key.type, key.offset, | |
1869 | next_key.objectid, next_key.type, | |
1870 | next_key.offset); | |
1871 | ret = -EUCLEAN; | |
557ea5dd QW |
1872 | goto out; |
1873 | } | |
1874 | } | |
1875 | out: | |
1876 | return ret; | |
1877 | } | |
02529d7a | 1878 | ALLOW_ERROR_INJECTION(btrfs_check_node, ERRNO); |
88c602ab QW |
1879 | |
1880 | int btrfs_check_eb_owner(const struct extent_buffer *eb, u64 root_owner) | |
1881 | { | |
1882 | const bool is_subvol = is_fstree(root_owner); | |
1883 | const u64 eb_owner = btrfs_header_owner(eb); | |
1884 | ||
1885 | /* | |
1886 | * Skip dummy fs, as selftests don't create unique ebs for each dummy | |
1887 | * root. | |
1888 | */ | |
1889 | if (test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &eb->fs_info->fs_state)) | |
1890 | return 0; | |
1891 | /* | |
1892 | * There are several call sites (backref walking, qgroup, and data | |
1893 | * reloc) passing 0 as @root_owner, as they are not holding the | |
1894 | * tree root. In that case, we can not do a reliable ownership check, | |
1895 | * so just exit. | |
1896 | */ | |
1897 | if (root_owner == 0) | |
1898 | return 0; | |
1899 | /* | |
1900 | * These trees use key.offset as their owner, our callers don't have | |
1901 | * the extra capacity to pass key.offset here. So we just skip them. | |
1902 | */ | |
1903 | if (root_owner == BTRFS_TREE_LOG_OBJECTID || | |
1904 | root_owner == BTRFS_TREE_RELOC_OBJECTID) | |
1905 | return 0; | |
1906 | ||
1907 | if (!is_subvol) { | |
1908 | /* For non-subvolume trees, the eb owner should match root owner */ | |
1909 | if (unlikely(root_owner != eb_owner)) { | |
1910 | btrfs_crit(eb->fs_info, | |
1911 | "corrupted %s, root=%llu block=%llu owner mismatch, have %llu expect %llu", | |
1912 | btrfs_header_level(eb) == 0 ? "leaf" : "node", | |
1913 | root_owner, btrfs_header_bytenr(eb), eb_owner, | |
1914 | root_owner); | |
1915 | return -EUCLEAN; | |
1916 | } | |
1917 | return 0; | |
1918 | } | |
1919 | ||
1920 | /* | |
1921 | * For subvolume trees, owners can mismatch, but they should all belong | |
1922 | * to subvolume trees. | |
1923 | */ | |
1924 | if (unlikely(is_subvol != is_fstree(eb_owner))) { | |
1925 | btrfs_crit(eb->fs_info, | |
1926 | "corrupted %s, root=%llu block=%llu owner mismatch, have %llu expect [%llu, %llu]", | |
1927 | btrfs_header_level(eb) == 0 ? "leaf" : "node", | |
1928 | root_owner, btrfs_header_bytenr(eb), eb_owner, | |
1929 | BTRFS_FIRST_FREE_OBJECTID, BTRFS_LAST_FREE_OBJECTID); | |
1930 | return -EUCLEAN; | |
1931 | } | |
1932 | return 0; | |
1933 | } |