]>
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 | ||
18 | #include "ctree.h" | |
19 | #include "tree-checker.h" | |
20 | #include "disk-io.h" | |
21 | #include "compression.h" | |
fce466ea | 22 | #include "volumes.h" |
557ea5dd | 23 | |
bba4f298 QW |
24 | /* |
25 | * Error message should follow the following format: | |
26 | * corrupt <type>: <identifier>, <reason>[, <bad_value>] | |
27 | * | |
28 | * @type: leaf or node | |
29 | * @identifier: the necessary info to locate the leaf/node. | |
52042d8e | 30 | * It's recommended to decode key.objecitd/offset if it's |
bba4f298 QW |
31 | * meaningful. |
32 | * @reason: describe the error | |
52042d8e | 33 | * @bad_value: optional, it's recommended to output bad value and its |
bba4f298 QW |
34 | * expected value (range). |
35 | * | |
36 | * Since comma is used to separate the components, only space is allowed | |
37 | * inside each component. | |
38 | */ | |
39 | ||
40 | /* | |
41 | * Append generic "corrupt leaf/node root=%llu block=%llu slot=%d: " to @fmt. | |
42 | * Allows callers to customize the output. | |
43 | */ | |
44 | __printf(4, 5) | |
e67c718b | 45 | __cold |
2f659546 | 46 | static void generic_err(const struct btrfs_fs_info *fs_info, |
bba4f298 QW |
47 | const struct extent_buffer *eb, int slot, |
48 | const char *fmt, ...) | |
49 | { | |
50 | struct va_format vaf; | |
51 | va_list args; | |
52 | ||
53 | va_start(args, fmt); | |
54 | ||
55 | vaf.fmt = fmt; | |
56 | vaf.va = &args; | |
57 | ||
2f659546 | 58 | btrfs_crit(fs_info, |
bba4f298 QW |
59 | "corrupt %s: root=%llu block=%llu slot=%d, %pV", |
60 | btrfs_header_level(eb) == 0 ? "leaf" : "node", | |
2f659546 | 61 | btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, &vaf); |
bba4f298 QW |
62 | va_end(args); |
63 | } | |
64 | ||
8806d718 QW |
65 | /* |
66 | * Customized reporter for extent data item, since its key objectid and | |
67 | * offset has its own meaning. | |
68 | */ | |
69 | __printf(4, 5) | |
e67c718b | 70 | __cold |
2f659546 | 71 | static void file_extent_err(const struct btrfs_fs_info *fs_info, |
8806d718 QW |
72 | const struct extent_buffer *eb, int slot, |
73 | const char *fmt, ...) | |
74 | { | |
75 | struct btrfs_key key; | |
76 | struct va_format vaf; | |
77 | va_list args; | |
78 | ||
79 | btrfs_item_key_to_cpu(eb, &key, slot); | |
80 | va_start(args, fmt); | |
81 | ||
82 | vaf.fmt = fmt; | |
83 | vaf.va = &args; | |
84 | ||
2f659546 | 85 | btrfs_crit(fs_info, |
8806d718 | 86 | "corrupt %s: root=%llu block=%llu slot=%d ino=%llu file_offset=%llu, %pV", |
2f659546 QW |
87 | btrfs_header_level(eb) == 0 ? "leaf" : "node", |
88 | btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, | |
89 | key.objectid, key.offset, &vaf); | |
8806d718 QW |
90 | va_end(args); |
91 | } | |
92 | ||
93 | /* | |
94 | * Return 0 if the btrfs_file_extent_##name is aligned to @alignment | |
95 | * Else return 1 | |
96 | */ | |
2f659546 | 97 | #define CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, name, alignment) \ |
8806d718 QW |
98 | ({ \ |
99 | if (!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment))) \ | |
2f659546 | 100 | file_extent_err((fs_info), (leaf), (slot), \ |
8806d718 QW |
101 | "invalid %s for file extent, have %llu, should be aligned to %u", \ |
102 | (#name), btrfs_file_extent_##name((leaf), (fi)), \ | |
103 | (alignment)); \ | |
104 | (!IS_ALIGNED(btrfs_file_extent_##name((leaf), (fi)), (alignment))); \ | |
105 | }) | |
106 | ||
2f659546 | 107 | static int check_extent_data_item(struct btrfs_fs_info *fs_info, |
557ea5dd QW |
108 | struct extent_buffer *leaf, |
109 | struct btrfs_key *key, int slot) | |
110 | { | |
111 | struct btrfs_file_extent_item *fi; | |
2f659546 | 112 | u32 sectorsize = fs_info->sectorsize; |
557ea5dd QW |
113 | u32 item_size = btrfs_item_size_nr(leaf, slot); |
114 | ||
115 | if (!IS_ALIGNED(key->offset, sectorsize)) { | |
2f659546 | 116 | file_extent_err(fs_info, leaf, slot, |
8806d718 QW |
117 | "unaligned file_offset for file extent, have %llu should be aligned to %u", |
118 | key->offset, sectorsize); | |
557ea5dd QW |
119 | return -EUCLEAN; |
120 | } | |
121 | ||
122 | fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item); | |
123 | ||
124 | if (btrfs_file_extent_type(leaf, fi) > BTRFS_FILE_EXTENT_TYPES) { | |
2f659546 | 125 | file_extent_err(fs_info, leaf, slot, |
8806d718 QW |
126 | "invalid type for file extent, have %u expect range [0, %u]", |
127 | btrfs_file_extent_type(leaf, fi), | |
128 | BTRFS_FILE_EXTENT_TYPES); | |
557ea5dd QW |
129 | return -EUCLEAN; |
130 | } | |
131 | ||
132 | /* | |
52042d8e | 133 | * Support for new compression/encryption must introduce incompat flag, |
557ea5dd QW |
134 | * and must be caught in open_ctree(). |
135 | */ | |
136 | if (btrfs_file_extent_compression(leaf, fi) > BTRFS_COMPRESS_TYPES) { | |
2f659546 | 137 | file_extent_err(fs_info, leaf, slot, |
8806d718 QW |
138 | "invalid compression for file extent, have %u expect range [0, %u]", |
139 | btrfs_file_extent_compression(leaf, fi), | |
140 | BTRFS_COMPRESS_TYPES); | |
557ea5dd QW |
141 | return -EUCLEAN; |
142 | } | |
143 | if (btrfs_file_extent_encryption(leaf, fi)) { | |
2f659546 | 144 | file_extent_err(fs_info, leaf, slot, |
8806d718 QW |
145 | "invalid encryption for file extent, have %u expect 0", |
146 | btrfs_file_extent_encryption(leaf, fi)); | |
557ea5dd QW |
147 | return -EUCLEAN; |
148 | } | |
149 | if (btrfs_file_extent_type(leaf, fi) == BTRFS_FILE_EXTENT_INLINE) { | |
150 | /* Inline extent must have 0 as key offset */ | |
151 | if (key->offset) { | |
2f659546 | 152 | file_extent_err(fs_info, leaf, slot, |
8806d718 QW |
153 | "invalid file_offset for inline file extent, have %llu expect 0", |
154 | key->offset); | |
557ea5dd QW |
155 | return -EUCLEAN; |
156 | } | |
157 | ||
158 | /* Compressed inline extent has no on-disk size, skip it */ | |
159 | if (btrfs_file_extent_compression(leaf, fi) != | |
160 | BTRFS_COMPRESS_NONE) | |
161 | return 0; | |
162 | ||
163 | /* Uncompressed inline extent size must match item size */ | |
164 | if (item_size != BTRFS_FILE_EXTENT_INLINE_DATA_START + | |
165 | btrfs_file_extent_ram_bytes(leaf, fi)) { | |
2f659546 | 166 | file_extent_err(fs_info, leaf, slot, |
8806d718 QW |
167 | "invalid ram_bytes for uncompressed inline extent, have %u expect %llu", |
168 | item_size, BTRFS_FILE_EXTENT_INLINE_DATA_START + | |
169 | btrfs_file_extent_ram_bytes(leaf, fi)); | |
557ea5dd QW |
170 | return -EUCLEAN; |
171 | } | |
172 | return 0; | |
173 | } | |
174 | ||
175 | /* Regular or preallocated extent has fixed item size */ | |
176 | if (item_size != sizeof(*fi)) { | |
2f659546 | 177 | file_extent_err(fs_info, leaf, slot, |
709a95c3 | 178 | "invalid item size for reg/prealloc file extent, have %u expect %zu", |
8806d718 | 179 | item_size, sizeof(*fi)); |
557ea5dd QW |
180 | return -EUCLEAN; |
181 | } | |
2f659546 QW |
182 | if (CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, ram_bytes, sectorsize) || |
183 | CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, disk_bytenr, sectorsize) || | |
184 | CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, disk_num_bytes, sectorsize) || | |
185 | CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, offset, sectorsize) || | |
186 | CHECK_FE_ALIGNED(fs_info, leaf, slot, fi, num_bytes, sectorsize)) | |
557ea5dd | 187 | return -EUCLEAN; |
557ea5dd QW |
188 | return 0; |
189 | } | |
190 | ||
2f659546 QW |
191 | static int check_csum_item(struct btrfs_fs_info *fs_info, |
192 | struct extent_buffer *leaf, struct btrfs_key *key, | |
193 | int slot) | |
557ea5dd | 194 | { |
2f659546 QW |
195 | u32 sectorsize = fs_info->sectorsize; |
196 | u32 csumsize = btrfs_super_csum_size(fs_info->super_copy); | |
557ea5dd QW |
197 | |
198 | if (key->objectid != BTRFS_EXTENT_CSUM_OBJECTID) { | |
2f659546 | 199 | generic_err(fs_info, leaf, slot, |
d508c5f0 QW |
200 | "invalid key objectid for csum item, have %llu expect %llu", |
201 | key->objectid, BTRFS_EXTENT_CSUM_OBJECTID); | |
557ea5dd QW |
202 | return -EUCLEAN; |
203 | } | |
204 | if (!IS_ALIGNED(key->offset, sectorsize)) { | |
2f659546 | 205 | generic_err(fs_info, leaf, slot, |
d508c5f0 QW |
206 | "unaligned key offset for csum item, have %llu should be aligned to %u", |
207 | key->offset, sectorsize); | |
557ea5dd QW |
208 | return -EUCLEAN; |
209 | } | |
210 | if (!IS_ALIGNED(btrfs_item_size_nr(leaf, slot), csumsize)) { | |
2f659546 | 211 | generic_err(fs_info, leaf, slot, |
d508c5f0 QW |
212 | "unaligned item size for csum item, have %u should be aligned to %u", |
213 | btrfs_item_size_nr(leaf, slot), csumsize); | |
557ea5dd QW |
214 | return -EUCLEAN; |
215 | } | |
216 | return 0; | |
217 | } | |
218 | ||
ad7b0368 QW |
219 | /* |
220 | * Customized reported for dir_item, only important new info is key->objectid, | |
221 | * which represents inode number | |
222 | */ | |
223 | __printf(4, 5) | |
e67c718b | 224 | __cold |
2f659546 | 225 | static void dir_item_err(const struct btrfs_fs_info *fs_info, |
ad7b0368 QW |
226 | const struct extent_buffer *eb, int slot, |
227 | const char *fmt, ...) | |
228 | { | |
229 | struct btrfs_key key; | |
230 | struct va_format vaf; | |
231 | va_list args; | |
232 | ||
233 | btrfs_item_key_to_cpu(eb, &key, slot); | |
234 | va_start(args, fmt); | |
235 | ||
236 | vaf.fmt = fmt; | |
237 | vaf.va = &args; | |
238 | ||
2f659546 | 239 | btrfs_crit(fs_info, |
ad7b0368 | 240 | "corrupt %s: root=%llu block=%llu slot=%d ino=%llu, %pV", |
2f659546 QW |
241 | btrfs_header_level(eb) == 0 ? "leaf" : "node", |
242 | btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, | |
243 | key.objectid, &vaf); | |
ad7b0368 QW |
244 | va_end(args); |
245 | } | |
246 | ||
2f659546 | 247 | static int check_dir_item(struct btrfs_fs_info *fs_info, |
ad7b0368 QW |
248 | struct extent_buffer *leaf, |
249 | struct btrfs_key *key, int slot) | |
250 | { | |
251 | struct btrfs_dir_item *di; | |
252 | u32 item_size = btrfs_item_size_nr(leaf, slot); | |
253 | u32 cur = 0; | |
254 | ||
255 | di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item); | |
256 | while (cur < item_size) { | |
ad7b0368 QW |
257 | u32 name_len; |
258 | u32 data_len; | |
259 | u32 max_name_len; | |
260 | u32 total_size; | |
261 | u32 name_hash; | |
262 | u8 dir_type; | |
263 | ||
264 | /* header itself should not cross item boundary */ | |
265 | if (cur + sizeof(*di) > item_size) { | |
2f659546 | 266 | dir_item_err(fs_info, leaf, slot, |
7cfad652 | 267 | "dir item header crosses item boundary, have %zu boundary %u", |
ad7b0368 QW |
268 | cur + sizeof(*di), item_size); |
269 | return -EUCLEAN; | |
270 | } | |
271 | ||
272 | /* dir type check */ | |
273 | dir_type = btrfs_dir_type(leaf, di); | |
274 | if (dir_type >= BTRFS_FT_MAX) { | |
2f659546 | 275 | dir_item_err(fs_info, leaf, slot, |
ad7b0368 QW |
276 | "invalid dir item type, have %u expect [0, %u)", |
277 | dir_type, BTRFS_FT_MAX); | |
278 | return -EUCLEAN; | |
279 | } | |
280 | ||
281 | if (key->type == BTRFS_XATTR_ITEM_KEY && | |
282 | dir_type != BTRFS_FT_XATTR) { | |
2f659546 | 283 | dir_item_err(fs_info, leaf, slot, |
ad7b0368 QW |
284 | "invalid dir item type for XATTR key, have %u expect %u", |
285 | dir_type, BTRFS_FT_XATTR); | |
286 | return -EUCLEAN; | |
287 | } | |
288 | if (dir_type == BTRFS_FT_XATTR && | |
289 | key->type != BTRFS_XATTR_ITEM_KEY) { | |
2f659546 | 290 | dir_item_err(fs_info, leaf, slot, |
ad7b0368 QW |
291 | "xattr dir type found for non-XATTR key"); |
292 | return -EUCLEAN; | |
293 | } | |
294 | if (dir_type == BTRFS_FT_XATTR) | |
295 | max_name_len = XATTR_NAME_MAX; | |
296 | else | |
297 | max_name_len = BTRFS_NAME_LEN; | |
298 | ||
299 | /* Name/data length check */ | |
300 | name_len = btrfs_dir_name_len(leaf, di); | |
301 | data_len = btrfs_dir_data_len(leaf, di); | |
302 | if (name_len > max_name_len) { | |
2f659546 | 303 | dir_item_err(fs_info, leaf, slot, |
ad7b0368 QW |
304 | "dir item name len too long, have %u max %u", |
305 | name_len, max_name_len); | |
306 | return -EUCLEAN; | |
307 | } | |
2f659546 QW |
308 | if (name_len + data_len > BTRFS_MAX_XATTR_SIZE(fs_info)) { |
309 | dir_item_err(fs_info, leaf, slot, | |
ad7b0368 QW |
310 | "dir item name and data len too long, have %u max %u", |
311 | name_len + data_len, | |
2f659546 | 312 | BTRFS_MAX_XATTR_SIZE(fs_info)); |
ad7b0368 QW |
313 | return -EUCLEAN; |
314 | } | |
315 | ||
316 | if (data_len && dir_type != BTRFS_FT_XATTR) { | |
2f659546 | 317 | dir_item_err(fs_info, leaf, slot, |
ad7b0368 QW |
318 | "dir item with invalid data len, have %u expect 0", |
319 | data_len); | |
320 | return -EUCLEAN; | |
321 | } | |
322 | ||
323 | total_size = sizeof(*di) + name_len + data_len; | |
324 | ||
325 | /* header and name/data should not cross item boundary */ | |
326 | if (cur + total_size > item_size) { | |
2f659546 | 327 | dir_item_err(fs_info, leaf, slot, |
ad7b0368 QW |
328 | "dir item data crosses item boundary, have %u boundary %u", |
329 | cur + total_size, item_size); | |
330 | return -EUCLEAN; | |
331 | } | |
332 | ||
333 | /* | |
334 | * Special check for XATTR/DIR_ITEM, as key->offset is name | |
335 | * hash, should match its name | |
336 | */ | |
337 | if (key->type == BTRFS_DIR_ITEM_KEY || | |
338 | key->type == BTRFS_XATTR_ITEM_KEY) { | |
e2683fc9 DS |
339 | char namebuf[max(BTRFS_NAME_LEN, XATTR_NAME_MAX)]; |
340 | ||
ad7b0368 QW |
341 | read_extent_buffer(leaf, namebuf, |
342 | (unsigned long)(di + 1), name_len); | |
343 | name_hash = btrfs_name_hash(namebuf, name_len); | |
344 | if (key->offset != name_hash) { | |
2f659546 | 345 | dir_item_err(fs_info, leaf, slot, |
ad7b0368 QW |
346 | "name hash mismatch with key, have 0x%016x expect 0x%016llx", |
347 | name_hash, key->offset); | |
348 | return -EUCLEAN; | |
349 | } | |
350 | } | |
351 | cur += total_size; | |
352 | di = (struct btrfs_dir_item *)((void *)di + total_size); | |
353 | } | |
354 | return 0; | |
355 | } | |
356 | ||
fce466ea QW |
357 | __printf(4, 5) |
358 | __cold | |
359 | static void block_group_err(const struct btrfs_fs_info *fs_info, | |
360 | const struct extent_buffer *eb, int slot, | |
361 | const char *fmt, ...) | |
362 | { | |
363 | struct btrfs_key key; | |
364 | struct va_format vaf; | |
365 | va_list args; | |
366 | ||
367 | btrfs_item_key_to_cpu(eb, &key, slot); | |
368 | va_start(args, fmt); | |
369 | ||
370 | vaf.fmt = fmt; | |
371 | vaf.va = &args; | |
372 | ||
373 | btrfs_crit(fs_info, | |
374 | "corrupt %s: root=%llu block=%llu slot=%d bg_start=%llu bg_len=%llu, %pV", | |
375 | btrfs_header_level(eb) == 0 ? "leaf" : "node", | |
376 | btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, | |
377 | key.objectid, key.offset, &vaf); | |
378 | va_end(args); | |
379 | } | |
380 | ||
381 | static int check_block_group_item(struct btrfs_fs_info *fs_info, | |
382 | struct extent_buffer *leaf, | |
383 | struct btrfs_key *key, int slot) | |
384 | { | |
385 | struct btrfs_block_group_item bgi; | |
386 | u32 item_size = btrfs_item_size_nr(leaf, slot); | |
387 | u64 flags; | |
388 | u64 type; | |
389 | ||
390 | /* | |
391 | * Here we don't really care about alignment since extent allocator can | |
10950929 | 392 | * handle it. We care more about the size. |
fce466ea | 393 | */ |
10950929 | 394 | if (key->offset == 0) { |
fce466ea | 395 | block_group_err(fs_info, leaf, slot, |
10950929 | 396 | "invalid block group size 0"); |
fce466ea QW |
397 | return -EUCLEAN; |
398 | } | |
399 | ||
400 | if (item_size != sizeof(bgi)) { | |
401 | block_group_err(fs_info, leaf, slot, | |
402 | "invalid item size, have %u expect %zu", | |
403 | item_size, sizeof(bgi)); | |
404 | return -EUCLEAN; | |
405 | } | |
406 | ||
407 | read_extent_buffer(leaf, &bgi, btrfs_item_ptr_offset(leaf, slot), | |
408 | sizeof(bgi)); | |
409 | if (btrfs_block_group_chunk_objectid(&bgi) != | |
410 | BTRFS_FIRST_CHUNK_TREE_OBJECTID) { | |
411 | block_group_err(fs_info, leaf, slot, | |
412 | "invalid block group chunk objectid, have %llu expect %llu", | |
413 | btrfs_block_group_chunk_objectid(&bgi), | |
414 | BTRFS_FIRST_CHUNK_TREE_OBJECTID); | |
415 | return -EUCLEAN; | |
416 | } | |
417 | ||
418 | if (btrfs_block_group_used(&bgi) > key->offset) { | |
419 | block_group_err(fs_info, leaf, slot, | |
420 | "invalid block group used, have %llu expect [0, %llu)", | |
421 | btrfs_block_group_used(&bgi), key->offset); | |
422 | return -EUCLEAN; | |
423 | } | |
424 | ||
425 | flags = btrfs_block_group_flags(&bgi); | |
426 | if (hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) > 1) { | |
427 | block_group_err(fs_info, leaf, slot, | |
428 | "invalid profile flags, have 0x%llx (%lu bits set) expect no more than 1 bit set", | |
429 | flags & BTRFS_BLOCK_GROUP_PROFILE_MASK, | |
430 | hweight64(flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)); | |
431 | return -EUCLEAN; | |
432 | } | |
433 | ||
434 | type = flags & BTRFS_BLOCK_GROUP_TYPE_MASK; | |
435 | if (type != BTRFS_BLOCK_GROUP_DATA && | |
436 | type != BTRFS_BLOCK_GROUP_METADATA && | |
437 | type != BTRFS_BLOCK_GROUP_SYSTEM && | |
438 | type != (BTRFS_BLOCK_GROUP_METADATA | | |
439 | BTRFS_BLOCK_GROUP_DATA)) { | |
440 | block_group_err(fs_info, leaf, slot, | |
761333f2 | 441 | "invalid type, have 0x%llx (%lu bits set) expect either 0x%llx, 0x%llx, 0x%llx or 0x%llx", |
fce466ea QW |
442 | type, hweight64(type), |
443 | BTRFS_BLOCK_GROUP_DATA, BTRFS_BLOCK_GROUP_METADATA, | |
444 | BTRFS_BLOCK_GROUP_SYSTEM, | |
445 | BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA); | |
446 | return -EUCLEAN; | |
447 | } | |
448 | return 0; | |
82fc28fb QW |
449 | } |
450 | ||
f1140243 QW |
451 | __printf(5, 6) |
452 | __cold | |
453 | static void chunk_err(const struct btrfs_fs_info *fs_info, | |
454 | const struct extent_buffer *leaf, | |
455 | const struct btrfs_chunk *chunk, u64 logical, | |
456 | const char *fmt, ...) | |
457 | { | |
458 | bool is_sb; | |
459 | struct va_format vaf; | |
460 | va_list args; | |
461 | int i; | |
462 | int slot = -1; | |
463 | ||
464 | /* Only superblock eb is able to have such small offset */ | |
465 | is_sb = (leaf->start == BTRFS_SUPER_INFO_OFFSET); | |
466 | ||
467 | if (!is_sb) { | |
468 | /* | |
469 | * Get the slot number by iterating through all slots, this | |
470 | * would provide better readability. | |
471 | */ | |
472 | for (i = 0; i < btrfs_header_nritems(leaf); i++) { | |
473 | if (btrfs_item_ptr_offset(leaf, i) == | |
474 | (unsigned long)chunk) { | |
475 | slot = i; | |
476 | break; | |
477 | } | |
478 | } | |
479 | } | |
480 | va_start(args, fmt); | |
481 | vaf.fmt = fmt; | |
482 | vaf.va = &args; | |
483 | ||
484 | if (is_sb) | |
485 | btrfs_crit(fs_info, | |
486 | "corrupt superblock syschunk array: chunk_start=%llu, %pV", | |
487 | logical, &vaf); | |
488 | else | |
489 | btrfs_crit(fs_info, | |
490 | "corrupt leaf: root=%llu block=%llu slot=%d chunk_start=%llu, %pV", | |
491 | BTRFS_CHUNK_TREE_OBJECTID, leaf->start, slot, | |
492 | logical, &vaf); | |
493 | va_end(args); | |
494 | } | |
495 | ||
82fc28fb QW |
496 | /* |
497 | * The common chunk check which could also work on super block sys chunk array. | |
498 | * | |
bf871c3b | 499 | * Return -EUCLEAN if anything is corrupted. |
82fc28fb QW |
500 | * Return 0 if everything is OK. |
501 | */ | |
502 | int btrfs_check_chunk_valid(struct btrfs_fs_info *fs_info, | |
503 | struct extent_buffer *leaf, | |
504 | struct btrfs_chunk *chunk, u64 logical) | |
505 | { | |
506 | u64 length; | |
507 | u64 stripe_len; | |
508 | u16 num_stripes; | |
509 | u16 sub_stripes; | |
510 | u64 type; | |
511 | u64 features; | |
512 | bool mixed = false; | |
513 | ||
514 | length = btrfs_chunk_length(leaf, chunk); | |
515 | stripe_len = btrfs_chunk_stripe_len(leaf, chunk); | |
516 | num_stripes = btrfs_chunk_num_stripes(leaf, chunk); | |
517 | sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk); | |
518 | type = btrfs_chunk_type(leaf, chunk); | |
519 | ||
520 | if (!num_stripes) { | |
f1140243 QW |
521 | chunk_err(fs_info, leaf, chunk, logical, |
522 | "invalid chunk num_stripes, have %u", num_stripes); | |
bf871c3b | 523 | return -EUCLEAN; |
82fc28fb QW |
524 | } |
525 | if (!IS_ALIGNED(logical, fs_info->sectorsize)) { | |
f1140243 QW |
526 | chunk_err(fs_info, leaf, chunk, logical, |
527 | "invalid chunk logical, have %llu should aligned to %u", | |
528 | logical, fs_info->sectorsize); | |
bf871c3b | 529 | return -EUCLEAN; |
82fc28fb QW |
530 | } |
531 | if (btrfs_chunk_sector_size(leaf, chunk) != fs_info->sectorsize) { | |
f1140243 QW |
532 | chunk_err(fs_info, leaf, chunk, logical, |
533 | "invalid chunk sectorsize, have %u expect %u", | |
534 | btrfs_chunk_sector_size(leaf, chunk), | |
535 | fs_info->sectorsize); | |
bf871c3b | 536 | return -EUCLEAN; |
82fc28fb QW |
537 | } |
538 | if (!length || !IS_ALIGNED(length, fs_info->sectorsize)) { | |
f1140243 QW |
539 | chunk_err(fs_info, leaf, chunk, logical, |
540 | "invalid chunk length, have %llu", length); | |
bf871c3b | 541 | return -EUCLEAN; |
82fc28fb QW |
542 | } |
543 | if (!is_power_of_2(stripe_len) || stripe_len != BTRFS_STRIPE_LEN) { | |
f1140243 QW |
544 | chunk_err(fs_info, leaf, chunk, logical, |
545 | "invalid chunk stripe length: %llu", | |
82fc28fb | 546 | stripe_len); |
bf871c3b | 547 | return -EUCLEAN; |
82fc28fb QW |
548 | } |
549 | if (~(BTRFS_BLOCK_GROUP_TYPE_MASK | BTRFS_BLOCK_GROUP_PROFILE_MASK) & | |
550 | type) { | |
f1140243 QW |
551 | chunk_err(fs_info, leaf, chunk, logical, |
552 | "unrecognized chunk type: 0x%llx", | |
82fc28fb QW |
553 | ~(BTRFS_BLOCK_GROUP_TYPE_MASK | |
554 | BTRFS_BLOCK_GROUP_PROFILE_MASK) & | |
555 | btrfs_chunk_type(leaf, chunk)); | |
bf871c3b | 556 | return -EUCLEAN; |
82fc28fb QW |
557 | } |
558 | ||
559 | if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == 0) { | |
f1140243 QW |
560 | chunk_err(fs_info, leaf, chunk, logical, |
561 | "missing chunk type flag, have 0x%llx one bit must be set in 0x%llx", | |
562 | type, BTRFS_BLOCK_GROUP_TYPE_MASK); | |
bf871c3b | 563 | return -EUCLEAN; |
82fc28fb QW |
564 | } |
565 | ||
566 | if ((type & BTRFS_BLOCK_GROUP_SYSTEM) && | |
567 | (type & (BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA))) { | |
f1140243 QW |
568 | chunk_err(fs_info, leaf, chunk, logical, |
569 | "system chunk with data or metadata type: 0x%llx", | |
570 | type); | |
bf871c3b | 571 | return -EUCLEAN; |
82fc28fb QW |
572 | } |
573 | ||
574 | features = btrfs_super_incompat_flags(fs_info->super_copy); | |
575 | if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) | |
576 | mixed = true; | |
577 | ||
578 | if (!mixed) { | |
579 | if ((type & BTRFS_BLOCK_GROUP_METADATA) && | |
580 | (type & BTRFS_BLOCK_GROUP_DATA)) { | |
f1140243 | 581 | chunk_err(fs_info, leaf, chunk, logical, |
82fc28fb | 582 | "mixed chunk type in non-mixed mode: 0x%llx", type); |
bf871c3b | 583 | return -EUCLEAN; |
82fc28fb QW |
584 | } |
585 | } | |
586 | ||
587 | if ((type & BTRFS_BLOCK_GROUP_RAID10 && sub_stripes != 2) || | |
588 | (type & BTRFS_BLOCK_GROUP_RAID1 && num_stripes != 2) || | |
589 | (type & BTRFS_BLOCK_GROUP_RAID5 && num_stripes < 2) || | |
590 | (type & BTRFS_BLOCK_GROUP_RAID6 && num_stripes < 3) || | |
591 | (type & BTRFS_BLOCK_GROUP_DUP && num_stripes != 2) || | |
592 | ((type & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 && num_stripes != 1)) { | |
f1140243 | 593 | chunk_err(fs_info, leaf, chunk, logical, |
82fc28fb QW |
594 | "invalid num_stripes:sub_stripes %u:%u for profile %llu", |
595 | num_stripes, sub_stripes, | |
596 | type & BTRFS_BLOCK_GROUP_PROFILE_MASK); | |
bf871c3b | 597 | return -EUCLEAN; |
82fc28fb QW |
598 | } |
599 | ||
600 | return 0; | |
fce466ea QW |
601 | } |
602 | ||
ab4ba2e1 QW |
603 | __printf(4, 5) |
604 | __cold | |
605 | static void dev_item_err(const struct btrfs_fs_info *fs_info, | |
606 | const struct extent_buffer *eb, int slot, | |
607 | const char *fmt, ...) | |
608 | { | |
609 | struct btrfs_key key; | |
610 | struct va_format vaf; | |
611 | va_list args; | |
612 | ||
613 | btrfs_item_key_to_cpu(eb, &key, slot); | |
614 | va_start(args, fmt); | |
615 | ||
616 | vaf.fmt = fmt; | |
617 | vaf.va = &args; | |
618 | ||
619 | btrfs_crit(fs_info, | |
620 | "corrupt %s: root=%llu block=%llu slot=%d devid=%llu %pV", | |
621 | btrfs_header_level(eb) == 0 ? "leaf" : "node", | |
622 | btrfs_header_owner(eb), btrfs_header_bytenr(eb), slot, | |
623 | key.objectid, &vaf); | |
624 | va_end(args); | |
625 | } | |
626 | ||
627 | static int check_dev_item(struct btrfs_fs_info *fs_info, | |
628 | struct extent_buffer *leaf, | |
629 | struct btrfs_key *key, int slot) | |
630 | { | |
631 | struct btrfs_dev_item *ditem; | |
632 | u64 max_devid = max(BTRFS_MAX_DEVS(fs_info), BTRFS_MAX_DEVS_SYS_CHUNK); | |
633 | ||
634 | if (key->objectid != BTRFS_DEV_ITEMS_OBJECTID) { | |
635 | dev_item_err(fs_info, leaf, slot, | |
636 | "invalid objectid: has=%llu expect=%llu", | |
637 | key->objectid, BTRFS_DEV_ITEMS_OBJECTID); | |
638 | return -EUCLEAN; | |
639 | } | |
640 | if (key->offset > max_devid) { | |
641 | dev_item_err(fs_info, leaf, slot, | |
642 | "invalid devid: has=%llu expect=[0, %llu]", | |
643 | key->offset, max_devid); | |
644 | return -EUCLEAN; | |
645 | } | |
646 | ditem = btrfs_item_ptr(leaf, slot, struct btrfs_dev_item); | |
647 | if (btrfs_device_id(leaf, ditem) != key->offset) { | |
648 | dev_item_err(fs_info, leaf, slot, | |
649 | "devid mismatch: key has=%llu item has=%llu", | |
650 | key->offset, btrfs_device_id(leaf, ditem)); | |
651 | return -EUCLEAN; | |
652 | } | |
653 | ||
654 | /* | |
655 | * For device total_bytes, we don't have reliable way to check it, as | |
656 | * it can be 0 for device removal. Device size check can only be done | |
657 | * by dev extents check. | |
658 | */ | |
659 | if (btrfs_device_bytes_used(leaf, ditem) > | |
660 | btrfs_device_total_bytes(leaf, ditem)) { | |
661 | dev_item_err(fs_info, leaf, slot, | |
662 | "invalid bytes used: have %llu expect [0, %llu]", | |
663 | btrfs_device_bytes_used(leaf, ditem), | |
664 | btrfs_device_total_bytes(leaf, ditem)); | |
665 | return -EUCLEAN; | |
666 | } | |
667 | /* | |
668 | * Remaining members like io_align/type/gen/dev_group aren't really | |
669 | * utilized. Skip them to make later usage of them easier. | |
670 | */ | |
671 | return 0; | |
672 | } | |
673 | ||
557ea5dd QW |
674 | /* |
675 | * Common point to switch the item-specific validation. | |
676 | */ | |
2f659546 | 677 | static int check_leaf_item(struct btrfs_fs_info *fs_info, |
557ea5dd QW |
678 | struct extent_buffer *leaf, |
679 | struct btrfs_key *key, int slot) | |
680 | { | |
681 | int ret = 0; | |
075cb3c7 | 682 | struct btrfs_chunk *chunk; |
557ea5dd QW |
683 | |
684 | switch (key->type) { | |
685 | case BTRFS_EXTENT_DATA_KEY: | |
2f659546 | 686 | ret = check_extent_data_item(fs_info, leaf, key, slot); |
557ea5dd QW |
687 | break; |
688 | case BTRFS_EXTENT_CSUM_KEY: | |
2f659546 | 689 | ret = check_csum_item(fs_info, leaf, key, slot); |
557ea5dd | 690 | break; |
ad7b0368 QW |
691 | case BTRFS_DIR_ITEM_KEY: |
692 | case BTRFS_DIR_INDEX_KEY: | |
693 | case BTRFS_XATTR_ITEM_KEY: | |
2f659546 | 694 | ret = check_dir_item(fs_info, leaf, key, slot); |
ad7b0368 | 695 | break; |
fce466ea QW |
696 | case BTRFS_BLOCK_GROUP_ITEM_KEY: |
697 | ret = check_block_group_item(fs_info, leaf, key, slot); | |
698 | break; | |
075cb3c7 QW |
699 | case BTRFS_CHUNK_ITEM_KEY: |
700 | chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk); | |
701 | ret = btrfs_check_chunk_valid(fs_info, leaf, chunk, | |
702 | key->offset); | |
703 | break; | |
ab4ba2e1 QW |
704 | case BTRFS_DEV_ITEM_KEY: |
705 | ret = check_dev_item(fs_info, leaf, key, slot); | |
706 | break; | |
557ea5dd QW |
707 | } |
708 | return ret; | |
709 | } | |
710 | ||
2f659546 | 711 | static int check_leaf(struct btrfs_fs_info *fs_info, struct extent_buffer *leaf, |
69fc6cbb | 712 | bool check_item_data) |
557ea5dd | 713 | { |
557ea5dd QW |
714 | /* No valid key type is 0, so all key should be larger than this key */ |
715 | struct btrfs_key prev_key = {0, 0, 0}; | |
716 | struct btrfs_key key; | |
717 | u32 nritems = btrfs_header_nritems(leaf); | |
718 | int slot; | |
719 | ||
f556faa4 QW |
720 | if (btrfs_header_level(leaf) != 0) { |
721 | generic_err(fs_info, leaf, 0, | |
722 | "invalid level for leaf, have %d expect 0", | |
723 | btrfs_header_level(leaf)); | |
724 | return -EUCLEAN; | |
725 | } | |
726 | ||
557ea5dd QW |
727 | /* |
728 | * Extent buffers from a relocation tree have a owner field that | |
729 | * corresponds to the subvolume tree they are based on. So just from an | |
730 | * extent buffer alone we can not find out what is the id of the | |
731 | * corresponding subvolume tree, so we can not figure out if the extent | |
732 | * buffer corresponds to the root of the relocation tree or not. So | |
733 | * skip this check for relocation trees. | |
734 | */ | |
735 | if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) { | |
ba480dd4 | 736 | u64 owner = btrfs_header_owner(leaf); |
557ea5dd QW |
737 | struct btrfs_root *check_root; |
738 | ||
ba480dd4 QW |
739 | /* These trees must never be empty */ |
740 | if (owner == BTRFS_ROOT_TREE_OBJECTID || | |
741 | owner == BTRFS_CHUNK_TREE_OBJECTID || | |
742 | owner == BTRFS_EXTENT_TREE_OBJECTID || | |
743 | owner == BTRFS_DEV_TREE_OBJECTID || | |
744 | owner == BTRFS_FS_TREE_OBJECTID || | |
745 | owner == BTRFS_DATA_RELOC_TREE_OBJECTID) { | |
746 | generic_err(fs_info, leaf, 0, | |
747 | "invalid root, root %llu must never be empty", | |
748 | owner); | |
749 | return -EUCLEAN; | |
750 | } | |
751 | key.objectid = owner; | |
557ea5dd QW |
752 | key.type = BTRFS_ROOT_ITEM_KEY; |
753 | key.offset = (u64)-1; | |
754 | ||
755 | check_root = btrfs_get_fs_root(fs_info, &key, false); | |
756 | /* | |
757 | * The only reason we also check NULL here is that during | |
758 | * open_ctree() some roots has not yet been set up. | |
759 | */ | |
760 | if (!IS_ERR_OR_NULL(check_root)) { | |
761 | struct extent_buffer *eb; | |
762 | ||
763 | eb = btrfs_root_node(check_root); | |
764 | /* if leaf is the root, then it's fine */ | |
765 | if (leaf != eb) { | |
2f659546 | 766 | generic_err(fs_info, leaf, 0, |
478d01b3 QW |
767 | "invalid nritems, have %u should not be 0 for non-root leaf", |
768 | nritems); | |
557ea5dd QW |
769 | free_extent_buffer(eb); |
770 | return -EUCLEAN; | |
771 | } | |
772 | free_extent_buffer(eb); | |
773 | } | |
774 | return 0; | |
775 | } | |
776 | ||
777 | if (nritems == 0) | |
778 | return 0; | |
779 | ||
780 | /* | |
781 | * Check the following things to make sure this is a good leaf, and | |
782 | * leaf users won't need to bother with similar sanity checks: | |
783 | * | |
784 | * 1) key ordering | |
785 | * 2) item offset and size | |
786 | * No overlap, no hole, all inside the leaf. | |
787 | * 3) item content | |
788 | * If possible, do comprehensive sanity check. | |
789 | * NOTE: All checks must only rely on the item data itself. | |
790 | */ | |
791 | for (slot = 0; slot < nritems; slot++) { | |
792 | u32 item_end_expected; | |
793 | int ret; | |
794 | ||
795 | btrfs_item_key_to_cpu(leaf, &key, slot); | |
796 | ||
797 | /* Make sure the keys are in the right order */ | |
798 | if (btrfs_comp_cpu_keys(&prev_key, &key) >= 0) { | |
2f659546 | 799 | generic_err(fs_info, leaf, slot, |
478d01b3 QW |
800 | "bad key order, prev (%llu %u %llu) current (%llu %u %llu)", |
801 | prev_key.objectid, prev_key.type, | |
802 | prev_key.offset, key.objectid, key.type, | |
803 | key.offset); | |
557ea5dd QW |
804 | return -EUCLEAN; |
805 | } | |
806 | ||
807 | /* | |
808 | * Make sure the offset and ends are right, remember that the | |
809 | * item data starts at the end of the leaf and grows towards the | |
810 | * front. | |
811 | */ | |
812 | if (slot == 0) | |
813 | item_end_expected = BTRFS_LEAF_DATA_SIZE(fs_info); | |
814 | else | |
815 | item_end_expected = btrfs_item_offset_nr(leaf, | |
816 | slot - 1); | |
817 | if (btrfs_item_end_nr(leaf, slot) != item_end_expected) { | |
2f659546 | 818 | generic_err(fs_info, leaf, slot, |
478d01b3 QW |
819 | "unexpected item end, have %u expect %u", |
820 | btrfs_item_end_nr(leaf, slot), | |
821 | item_end_expected); | |
557ea5dd QW |
822 | return -EUCLEAN; |
823 | } | |
824 | ||
825 | /* | |
826 | * Check to make sure that we don't point outside of the leaf, | |
827 | * just in case all the items are consistent to each other, but | |
828 | * all point outside of the leaf. | |
829 | */ | |
830 | if (btrfs_item_end_nr(leaf, slot) > | |
831 | BTRFS_LEAF_DATA_SIZE(fs_info)) { | |
2f659546 | 832 | generic_err(fs_info, leaf, slot, |
478d01b3 QW |
833 | "slot end outside of leaf, have %u expect range [0, %u]", |
834 | btrfs_item_end_nr(leaf, slot), | |
835 | BTRFS_LEAF_DATA_SIZE(fs_info)); | |
557ea5dd QW |
836 | return -EUCLEAN; |
837 | } | |
838 | ||
839 | /* Also check if the item pointer overlaps with btrfs item. */ | |
840 | if (btrfs_item_nr_offset(slot) + sizeof(struct btrfs_item) > | |
841 | btrfs_item_ptr_offset(leaf, slot)) { | |
2f659546 | 842 | generic_err(fs_info, leaf, slot, |
478d01b3 QW |
843 | "slot overlaps with its data, item end %lu data start %lu", |
844 | btrfs_item_nr_offset(slot) + | |
845 | sizeof(struct btrfs_item), | |
846 | btrfs_item_ptr_offset(leaf, slot)); | |
557ea5dd QW |
847 | return -EUCLEAN; |
848 | } | |
849 | ||
69fc6cbb QW |
850 | if (check_item_data) { |
851 | /* | |
852 | * Check if the item size and content meet other | |
853 | * criteria | |
854 | */ | |
2f659546 | 855 | ret = check_leaf_item(fs_info, leaf, &key, slot); |
69fc6cbb QW |
856 | if (ret < 0) |
857 | return ret; | |
858 | } | |
557ea5dd QW |
859 | |
860 | prev_key.objectid = key.objectid; | |
861 | prev_key.type = key.type; | |
862 | prev_key.offset = key.offset; | |
863 | } | |
864 | ||
865 | return 0; | |
866 | } | |
867 | ||
2f659546 QW |
868 | int btrfs_check_leaf_full(struct btrfs_fs_info *fs_info, |
869 | struct extent_buffer *leaf) | |
69fc6cbb | 870 | { |
2f659546 | 871 | return check_leaf(fs_info, leaf, true); |
69fc6cbb QW |
872 | } |
873 | ||
2f659546 | 874 | int btrfs_check_leaf_relaxed(struct btrfs_fs_info *fs_info, |
69fc6cbb QW |
875 | struct extent_buffer *leaf) |
876 | { | |
2f659546 | 877 | return check_leaf(fs_info, leaf, false); |
69fc6cbb QW |
878 | } |
879 | ||
2f659546 | 880 | int btrfs_check_node(struct btrfs_fs_info *fs_info, struct extent_buffer *node) |
557ea5dd QW |
881 | { |
882 | unsigned long nr = btrfs_header_nritems(node); | |
883 | struct btrfs_key key, next_key; | |
884 | int slot; | |
f556faa4 | 885 | int level = btrfs_header_level(node); |
557ea5dd QW |
886 | u64 bytenr; |
887 | int ret = 0; | |
888 | ||
f556faa4 QW |
889 | if (level <= 0 || level >= BTRFS_MAX_LEVEL) { |
890 | generic_err(fs_info, node, 0, | |
891 | "invalid level for node, have %d expect [1, %d]", | |
892 | level, BTRFS_MAX_LEVEL - 1); | |
893 | return -EUCLEAN; | |
894 | } | |
2f659546 QW |
895 | if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(fs_info)) { |
896 | btrfs_crit(fs_info, | |
bba4f298 | 897 | "corrupt node: root=%llu block=%llu, nritems too %s, have %lu expect range [1,%u]", |
2f659546 | 898 | btrfs_header_owner(node), node->start, |
bba4f298 | 899 | nr == 0 ? "small" : "large", nr, |
2f659546 | 900 | BTRFS_NODEPTRS_PER_BLOCK(fs_info)); |
bba4f298 | 901 | return -EUCLEAN; |
557ea5dd QW |
902 | } |
903 | ||
904 | for (slot = 0; slot < nr - 1; slot++) { | |
905 | bytenr = btrfs_node_blockptr(node, slot); | |
906 | btrfs_node_key_to_cpu(node, &key, slot); | |
907 | btrfs_node_key_to_cpu(node, &next_key, slot + 1); | |
908 | ||
909 | if (!bytenr) { | |
2f659546 | 910 | generic_err(fs_info, node, slot, |
bba4f298 QW |
911 | "invalid NULL node pointer"); |
912 | ret = -EUCLEAN; | |
913 | goto out; | |
914 | } | |
2f659546 QW |
915 | if (!IS_ALIGNED(bytenr, fs_info->sectorsize)) { |
916 | generic_err(fs_info, node, slot, | |
bba4f298 | 917 | "unaligned pointer, have %llu should be aligned to %u", |
2f659546 | 918 | bytenr, fs_info->sectorsize); |
bba4f298 | 919 | ret = -EUCLEAN; |
557ea5dd QW |
920 | goto out; |
921 | } | |
922 | ||
923 | if (btrfs_comp_cpu_keys(&key, &next_key) >= 0) { | |
2f659546 | 924 | generic_err(fs_info, node, slot, |
bba4f298 QW |
925 | "bad key order, current (%llu %u %llu) next (%llu %u %llu)", |
926 | key.objectid, key.type, key.offset, | |
927 | next_key.objectid, next_key.type, | |
928 | next_key.offset); | |
929 | ret = -EUCLEAN; | |
557ea5dd QW |
930 | goto out; |
931 | } | |
932 | } | |
933 | out: | |
934 | return ret; | |
935 | } |