?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
s3-smbd: longer explanation of smbd_shim code
[samba/gbeck.git] / lib / tdb2 / check.c
blob52fb188764dfa0e2c85250e3658200db2fc67894
1 /*
2 Trivial Database 2: free list/block handling
3 Copyright (C) Rusty Russell 2010
4
5 This library is free software; you can redistribute it and/or
6 modify it under the terms of the GNU Lesser General Public
7 License as published by the Free Software Foundation; either
8 version 3 of the License, or (at your option) any later version.
9
10 This library is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 Lesser General Public License for more details.
14
15 You should have received a copy of the GNU Lesser General Public
16 License along with this library; if not, see <http://www.gnu.org/licenses/>.
17 */
18 #include "private.h"
19 #include <ccan/likely/likely.h>
20 #include <ccan/asearch/asearch.h>
21
22 /* We keep an ordered array of offsets. */
23 static bool append(tdb_off_t **arr, size_t *num, tdb_off_t off)
24 {
25 tdb_off_t *new = realloc(*arr, (*num + 1) * sizeof(tdb_off_t));
26 if (!new)
27 return false;
28 new[(*num)++] = off;
29 *arr = new;
30 return true;
31 }
32
33 static enum TDB_ERROR check_header(struct tdb_context *tdb, tdb_off_t *recovery,
34 uint64_t *features)
35 {
36 uint64_t hash_test;
37 struct tdb_header hdr;
38 enum TDB_ERROR ecode;
39
40 ecode = tdb_read_convert(tdb, 0, &hdr, sizeof(hdr));
41 if (ecode != TDB_SUCCESS) {
42 return ecode;
43 }
44 /* magic food should not be converted, so convert back. */
45 tdb_convert(tdb, hdr.magic_food, sizeof(hdr.magic_food));
46
47 hash_test = TDB_HASH_MAGIC;
48 hash_test = tdb_hash(tdb, &hash_test, sizeof(hash_test));
49 if (hdr.hash_test != hash_test) {
50 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
51 "check: hash test %llu should be %llu",
52 (long long)hdr.hash_test,
53 (long long)hash_test);
54 }
55
56 if (strcmp(hdr.magic_food, TDB_MAGIC_FOOD) != 0) {
57 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
58 "check: bad magic '%.*s'",
59 (unsigned)sizeof(hdr.magic_food),
60 hdr.magic_food);
61 }
62
63 /* Features which are used must be a subset of features offered. */
64 if (hdr.features_used & ~hdr.features_offered) {
65 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
66 "check: features used (0x%llx) which"
67 " are not offered (0x%llx)",
68 (long long)hdr.features_used,
69 (long long)hdr.features_offered);
70 }
71
72 *features = hdr.features_offered;
73 *recovery = hdr.recovery;
74 if (*recovery) {
75 if (*recovery < sizeof(hdr)
76 || *recovery > tdb->file->map_size) {
77 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
78 "tdb_check:"
79 " invalid recovery offset %zu",
80 (size_t)*recovery);
81 }
82 }
83
84 /* Don't check reserved: they *can* be used later. */
85 return TDB_SUCCESS;
86 }
87
88 static enum TDB_ERROR check_hash_tree(struct tdb_context *tdb,
89 tdb_off_t off, unsigned int group_bits,
90 uint64_t hprefix,
91 unsigned hprefix_bits,
92 tdb_off_t used[],
93 size_t num_used,
94 size_t *num_found,
95 enum TDB_ERROR (*check)(TDB_DATA,
96 TDB_DATA, void *),
97 void *data);
98
99 static enum TDB_ERROR check_hash_chain(struct tdb_context *tdb,
100 tdb_off_t off,
101 uint64_t hash,
102 tdb_off_t used[],
103 size_t num_used,
104 size_t *num_found,
105 enum TDB_ERROR (*check)(TDB_DATA,
106 TDB_DATA,
107 void *),
108 void *data)
109 {
110 struct tdb_used_record rec;
111 enum TDB_ERROR ecode;
112
113 ecode = tdb_read_convert(tdb, off, &rec, sizeof(rec));
114 if (ecode != TDB_SUCCESS) {
115 return ecode;
116 }
117
118 if (rec_magic(&rec) != TDB_CHAIN_MAGIC) {
119 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
120 "tdb_check: Bad hash chain magic %llu",
121 (long long)rec_magic(&rec));
122 }
123
124 if (rec_data_length(&rec) != sizeof(struct tdb_chain)) {
125 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
126 "tdb_check:"
127 " Bad hash chain length %llu vs %zu",
128 (long long)rec_data_length(&rec),
129 sizeof(struct tdb_chain));
130 }
131 if (rec_key_length(&rec) != 0) {
132 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
133 "tdb_check: Bad hash chain key length %llu",
134 (long long)rec_key_length(&rec));
135 }
136 if (rec_hash(&rec) != 0) {
137 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
138 "tdb_check: Bad hash chain hash value %llu",
139 (long long)rec_hash(&rec));
140 }
141
142 off += sizeof(rec);
143 ecode = check_hash_tree(tdb, off, 0, hash, 64,
144 used, num_used, num_found, check, data);
145 if (ecode != TDB_SUCCESS) {
146 return ecode;
147 }
148
149 off = tdb_read_off(tdb, off + offsetof(struct tdb_chain, next));
150 if (TDB_OFF_IS_ERR(off)) {
151 return off;
152 }
153 if (off == 0)
154 return TDB_SUCCESS;
155 (*num_found)++;
156 return check_hash_chain(tdb, off, hash, used, num_used, num_found,
157 check, data);
158 }
159
160 static enum TDB_ERROR check_hash_record(struct tdb_context *tdb,
161 tdb_off_t off,
162 uint64_t hprefix,
163 unsigned hprefix_bits,
164 tdb_off_t used[],
165 size_t num_used,
166 size_t *num_found,
167 enum TDB_ERROR (*check)(TDB_DATA,
168 TDB_DATA,
169 void *),
170 void *data)
171 {
172 struct tdb_used_record rec;
173 enum TDB_ERROR ecode;
174
175 if (hprefix_bits >= 64)
176 return check_hash_chain(tdb, off, hprefix, used, num_used,
177 num_found, check, data);
178
179 ecode = tdb_read_convert(tdb, off, &rec, sizeof(rec));
180 if (ecode != TDB_SUCCESS) {
181 return ecode;
182 }
183
184 if (rec_magic(&rec) != TDB_HTABLE_MAGIC) {
185 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
186 "tdb_check: Bad hash table magic %llu",
187 (long long)rec_magic(&rec));
188 }
189 if (rec_data_length(&rec)
190 != sizeof(tdb_off_t) << TDB_SUBLEVEL_HASH_BITS) {
191 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
192 "tdb_check:"
193 " Bad hash table length %llu vs %llu",
194 (long long)rec_data_length(&rec),
195 (long long)sizeof(tdb_off_t)
196 << TDB_SUBLEVEL_HASH_BITS);
197 }
198 if (rec_key_length(&rec) != 0) {
199 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
200 "tdb_check: Bad hash table key length %llu",
201 (long long)rec_key_length(&rec));
202 }
203 if (rec_hash(&rec) != 0) {
204 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
205 "tdb_check: Bad hash table hash value %llu",
206 (long long)rec_hash(&rec));
207 }
208
209 off += sizeof(rec);
210 return check_hash_tree(tdb, off,
211 TDB_SUBLEVEL_HASH_BITS-TDB_HASH_GROUP_BITS,
212 hprefix, hprefix_bits,
213 used, num_used, num_found, check, data);
214 }
215
216 static int off_cmp(const tdb_off_t *a, const tdb_off_t *b)
217 {
218 /* Can overflow an int. */
219 return *a > *b ? 1
220 : *a < *b ? -1
221 : 0;
222 }
223
224 static uint64_t get_bits(uint64_t h, unsigned num, unsigned *used)
225 {
226 *used += num;
227
228 return (h >> (64 - *used)) & ((1U << num) - 1);
229 }
230
231 static enum TDB_ERROR check_hash_tree(struct tdb_context *tdb,
232 tdb_off_t off, unsigned int group_bits,
233 uint64_t hprefix,
234 unsigned hprefix_bits,
235 tdb_off_t used[],
236 size_t num_used,
237 size_t *num_found,
238 enum TDB_ERROR (*check)(TDB_DATA,
239 TDB_DATA, void *),
240 void *data)
241 {
242 unsigned int g, b;
243 const tdb_off_t *hash;
244 struct tdb_used_record rec;
245 enum TDB_ERROR ecode;
246
247 hash = tdb_access_read(tdb, off,
248 sizeof(tdb_off_t)
249 << (group_bits + TDB_HASH_GROUP_BITS),
250 true);
251 if (TDB_PTR_IS_ERR(hash)) {
252 return TDB_PTR_ERR(hash);
253 }
254
255 for (g = 0; g < (1 << group_bits); g++) {
256 const tdb_off_t *group = hash + (g << TDB_HASH_GROUP_BITS);
257 for (b = 0; b < (1 << TDB_HASH_GROUP_BITS); b++) {
258 unsigned int bucket, i, used_bits;
259 uint64_t h;
260 tdb_off_t *p;
261 if (group[b] == 0)
262 continue;
263
264 off = group[b] & TDB_OFF_MASK;
265 p = asearch(&off, used, num_used, off_cmp);
266 if (!p) {
267 ecode = tdb_logerr(tdb, TDB_ERR_CORRUPT,
268 TDB_LOG_ERROR,
269 "tdb_check: Invalid offset"
270 " %llu in hash",
271 (long long)off);
272 goto fail;
273 }
274 /* Mark it invalid. */
275 *p ^= 1;
276 (*num_found)++;
277
278 if (hprefix_bits == 64) {
279 /* Chained entries are unordered. */
280 if (is_subhash(group[b])) {
281 ecode = TDB_ERR_CORRUPT;
282 tdb_logerr(tdb, ecode,
283 TDB_LOG_ERROR,
284 "tdb_check: Invalid chain"
285 " entry subhash");
286 goto fail;
287 }
288 h = hash_record(tdb, off);
289 if (h != hprefix) {
290 ecode = TDB_ERR_CORRUPT;
291 tdb_logerr(tdb, ecode,
292 TDB_LOG_ERROR,
293 "check: bad hash chain"
294 " placement"
295 " 0x%llx vs 0x%llx",
296 (long long)h,
297 (long long)hprefix);
298 goto fail;
299 }
300 ecode = tdb_read_convert(tdb, off, &rec,
301 sizeof(rec));
302 if (ecode != TDB_SUCCESS) {
303 goto fail;
304 }
305 goto check;
306 }
307
308 if (is_subhash(group[b])) {
309 uint64_t subprefix;
310 subprefix = (hprefix
311 << (group_bits + TDB_HASH_GROUP_BITS))
312 + g * (1 << TDB_HASH_GROUP_BITS) + b;
313
314 ecode = check_hash_record(tdb,
315 group[b] & TDB_OFF_MASK,
316 subprefix,
317 hprefix_bits
318 + group_bits
319 + TDB_HASH_GROUP_BITS,
320 used, num_used, num_found,
321 check, data);
322 if (ecode != TDB_SUCCESS) {
323 goto fail;
324 }
325 continue;
326 }
327 /* A normal entry */
328
329 /* Does it belong here at all? */
330 h = hash_record(tdb, off);
331 used_bits = 0;
332 if (get_bits(h, hprefix_bits, &used_bits) != hprefix
333 && hprefix_bits) {
334 ecode = tdb_logerr(tdb, TDB_ERR_CORRUPT,
335 TDB_LOG_ERROR,
336 "check: bad hash placement"
337 " 0x%llx vs 0x%llx",
338 (long long)h,
339 (long long)hprefix);
340 goto fail;
341 }
342
343 /* Does it belong in this group? */
344 if (get_bits(h, group_bits, &used_bits) != g) {
345 ecode = tdb_logerr(tdb, TDB_ERR_CORRUPT,
346 TDB_LOG_ERROR,
347 "check: bad group %llu"
348 " vs %u",
349 (long long)h, g);
350 goto fail;
351 }
352
353 /* Are bucket bits correct? */
354 bucket = group[b] & TDB_OFF_HASH_GROUP_MASK;
355 if (get_bits(h, TDB_HASH_GROUP_BITS, &used_bits)
356 != bucket) {
357 used_bits -= TDB_HASH_GROUP_BITS;
358 ecode = tdb_logerr(tdb, TDB_ERR_CORRUPT,
359 TDB_LOG_ERROR,
360 "check: bad bucket %u vs %u",
361 (unsigned)get_bits(h,
362 TDB_HASH_GROUP_BITS,
363 &used_bits),
364 bucket);
365 goto fail;
366 }
367
368 /* There must not be any zero entries between
369 * the bucket it belongs in and this one! */
370 for (i = bucket;
371 i != b;
372 i = (i + 1) % (1 << TDB_HASH_GROUP_BITS)) {
373 if (group[i] == 0) {
374 ecode = TDB_ERR_CORRUPT;
375 tdb_logerr(tdb, ecode,
376 TDB_LOG_ERROR,
377 "check: bad group placement"
378 " %u vs %u",
379 b, bucket);
380 goto fail;
381 }
382 }
383
384 ecode = tdb_read_convert(tdb, off, &rec, sizeof(rec));
385 if (ecode != TDB_SUCCESS) {
386 goto fail;
387 }
388
389 /* Bottom bits must match header. */
390 if ((h & ((1 << 11)-1)) != rec_hash(&rec)) {
391 ecode = tdb_logerr(tdb, TDB_ERR_CORRUPT,
392 TDB_LOG_ERROR,
393 "tdb_check: Bad hash magic"
394 " at offset %llu"
395 " (0x%llx vs 0x%llx)",
396 (long long)off,
397 (long long)h,
398 (long long)rec_hash(&rec));
399 goto fail;
400 }
401
402 check:
403 if (check) {
404 TDB_DATA k, d;
405 const unsigned char *kptr;
406
407 kptr = tdb_access_read(tdb,
408 off + sizeof(rec),
409 rec_key_length(&rec)
410 + rec_data_length(&rec),
411 false);
412 if (TDB_PTR_IS_ERR(kptr)) {
413 ecode = TDB_PTR_ERR(kptr);
414 goto fail;
415 }
416
417 k = tdb_mkdata(kptr, rec_key_length(&rec));
418 d = tdb_mkdata(kptr + k.dsize,
419 rec_data_length(&rec));
420 ecode = check(k, d, data);
421 tdb_access_release(tdb, kptr);
422 if (ecode != TDB_SUCCESS) {
423 goto fail;
424 }
425 }
426 }
427 }
428 tdb_access_release(tdb, hash);
429 return TDB_SUCCESS;
430
431 fail:
432 tdb_access_release(tdb, hash);
433 return ecode;
434 }
435
436 static enum TDB_ERROR check_hash(struct tdb_context *tdb,
437 tdb_off_t used[],
438 size_t num_used, size_t num_ftables,
439 int (*check)(TDB_DATA, TDB_DATA, void *),
440 void *data)
441 {
442 /* Free tables also show up as used. */
443 size_t num_found = num_ftables;
444 enum TDB_ERROR ecode;
445
446 ecode = check_hash_tree(tdb, offsetof(struct tdb_header, hashtable),
447 TDB_TOPLEVEL_HASH_BITS-TDB_HASH_GROUP_BITS,
448 0, 0, used, num_used, &num_found,
449 check, data);
450 if (ecode == TDB_SUCCESS) {
451 if (num_found != num_used) {
452 ecode = tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
453 "tdb_check: Not all entries"
454 " are in hash");
455 }
456 }
457 return ecode;
458 }
459
460 static enum TDB_ERROR check_free(struct tdb_context *tdb,
461 tdb_off_t off,
462 const struct tdb_free_record *frec,
463 tdb_off_t prev, unsigned int ftable,
464 unsigned int bucket)
465 {
466 enum TDB_ERROR ecode;
467
468 if (frec_magic(frec) != TDB_FREE_MAGIC) {
469 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
470 "tdb_check: offset %llu bad magic 0x%llx",
471 (long long)off,
472 (long long)frec->magic_and_prev);
473 }
474 if (frec_ftable(frec) != ftable) {
475 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
476 "tdb_check: offset %llu bad freetable %u",
477 (long long)off, frec_ftable(frec));
478
479 }
480
481 ecode = tdb->methods->oob(tdb, off
482 + frec_len(frec)
483 + sizeof(struct tdb_used_record),
484 false);
485 if (ecode != TDB_SUCCESS) {
486 return ecode;
487 }
488 if (size_to_bucket(frec_len(frec)) != bucket) {
489 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
490 "tdb_check: offset %llu in wrong bucket"
491 " (%u vs %u)",
492 (long long)off,
493 bucket, size_to_bucket(frec_len(frec)));
494 }
495 if (prev && prev != frec_prev(frec)) {
496 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
497 "tdb_check: offset %llu bad prev"
498 " (%llu vs %llu)",
499 (long long)off,
500 (long long)prev, (long long)frec_len(frec));
501 }
502 return TDB_SUCCESS;
503 }
504
505 static enum TDB_ERROR check_free_table(struct tdb_context *tdb,
506 tdb_off_t ftable_off,
507 unsigned ftable_num,
508 tdb_off_t fr[],
509 size_t num_free,
510 size_t *num_found)
511 {
512 struct tdb_freetable ft;
513 tdb_off_t h;
514 unsigned int i;
515 enum TDB_ERROR ecode;
516
517 ecode = tdb_read_convert(tdb, ftable_off, &ft, sizeof(ft));
518 if (ecode != TDB_SUCCESS) {
519 return ecode;
520 }
521
522 if (rec_magic(&ft.hdr) != TDB_FTABLE_MAGIC
523 || rec_key_length(&ft.hdr) != 0
524 || rec_data_length(&ft.hdr) != sizeof(ft) - sizeof(ft.hdr)
525 || rec_hash(&ft.hdr) != 0) {
526 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
527 "tdb_check: Invalid header on free table");
528 }
529
530 for (i = 0; i < TDB_FREE_BUCKETS; i++) {
531 tdb_off_t off, prev = 0, *p, first = 0;
532 struct tdb_free_record f;
533
534 h = bucket_off(ftable_off, i);
535 for (off = tdb_read_off(tdb, h); off; off = f.next) {
536 if (TDB_OFF_IS_ERR(off)) {
537 return off;
538 }
539 if (!first) {
540 off &= TDB_OFF_MASK;
541 first = off;
542 }
543 ecode = tdb_read_convert(tdb, off, &f, sizeof(f));
544 if (ecode != TDB_SUCCESS) {
545 return ecode;
546 }
547 ecode = check_free(tdb, off, &f, prev, ftable_num, i);
548 if (ecode != TDB_SUCCESS) {
549 return ecode;
550 }
551
552 /* FIXME: Check hash bits */
553 p = asearch(&off, fr, num_free, off_cmp);
554 if (!p) {
555 return tdb_logerr(tdb, TDB_ERR_CORRUPT,
556 TDB_LOG_ERROR,
557 "tdb_check: Invalid offset"
558 " %llu in free table",
559 (long long)off);
560 }
561 /* Mark it invalid. */
562 *p ^= 1;
563 (*num_found)++;
564 prev = off;
565 }
566
567 if (first) {
568 /* Now we can check first back pointer. */
569 ecode = tdb_read_convert(tdb, first, &f, sizeof(f));
570 if (ecode != TDB_SUCCESS) {
571 return ecode;
572 }
573 ecode = check_free(tdb, first, &f, prev, ftable_num, i);
574 if (ecode != TDB_SUCCESS) {
575 return ecode;
576 }
577 }
578 }
579 return TDB_SUCCESS;
580 }
581
582 /* Slow, but should be very rare. */
583 tdb_off_t dead_space(struct tdb_context *tdb, tdb_off_t off)
584 {
585 size_t len;
586 enum TDB_ERROR ecode;
587
588 for (len = 0; off + len < tdb->file->map_size; len++) {
589 char c;
590 ecode = tdb->methods->tread(tdb, off, &c, 1);
591 if (ecode != TDB_SUCCESS) {
592 return ecode;
593 }
594 if (c != 0 && c != 0x43)
595 break;
596 }
597 return len;
598 }
599
600 static enum TDB_ERROR check_linear(struct tdb_context *tdb,
601 tdb_off_t **used, size_t *num_used,
602 tdb_off_t **fr, size_t *num_free,
603 uint64_t features, tdb_off_t recovery)
604 {
605 tdb_off_t off;
606 tdb_len_t len;
607 enum TDB_ERROR ecode;
608 bool found_recovery = false;
609
610 for (off = sizeof(struct tdb_header);
611 off < tdb->file->map_size;
612 off += len) {
613 union {
614 struct tdb_used_record u;
615 struct tdb_free_record f;
616 struct tdb_recovery_record r;
617 } rec;
618 /* r is larger: only get that if we need to. */
619 ecode = tdb_read_convert(tdb, off, &rec, sizeof(rec.f));
620 if (ecode != TDB_SUCCESS) {
621 return ecode;
622 }
623
624 /* If we crash after ftruncate, we can get zeroes or fill. */
625 if (rec.r.magic == TDB_RECOVERY_INVALID_MAGIC
626 || rec.r.magic == 0x4343434343434343ULL) {
627 ecode = tdb_read_convert(tdb, off, &rec, sizeof(rec.r));
628 if (ecode != TDB_SUCCESS) {
629 return ecode;
630 }
631 if (recovery == off) {
632 found_recovery = true;
633 len = sizeof(rec.r) + rec.r.max_len;
634 } else {
635 len = dead_space(tdb, off);
636 if (TDB_OFF_IS_ERR(len)) {
637 return len;
638 }
639 if (len < sizeof(rec.r)) {
640 return tdb_logerr(tdb, TDB_ERR_CORRUPT,
641 TDB_LOG_ERROR,
642 "tdb_check: invalid"
643 " dead space at %zu",
644 (size_t)off);
645 }
646
647 tdb_logerr(tdb, TDB_SUCCESS, TDB_LOG_WARNING,
648 "Dead space at %zu-%zu (of %zu)",
649 (size_t)off, (size_t)(off + len),
650 (size_t)tdb->file->map_size);
651 }
652 } else if (rec.r.magic == TDB_RECOVERY_MAGIC) {
653 ecode = tdb_read_convert(tdb, off, &rec, sizeof(rec.r));
654 if (ecode != TDB_SUCCESS) {
655 return ecode;
656 }
657 if (recovery != off) {
658 return tdb_logerr(tdb, TDB_ERR_CORRUPT,
659 TDB_LOG_ERROR,
660 "tdb_check: unexpected"
661 " recovery record at offset"
662 " %zu",
663 (size_t)off);
664 }
665 if (rec.r.len > rec.r.max_len) {
666 return tdb_logerr(tdb, TDB_ERR_CORRUPT,
667 TDB_LOG_ERROR,
668 "tdb_check: invalid recovery"
669 " length %zu",
670 (size_t)rec.r.len);
671 }
672 if (rec.r.eof > tdb->file->map_size) {
673 return tdb_logerr(tdb, TDB_ERR_CORRUPT,
674 TDB_LOG_ERROR,
675 "tdb_check: invalid old EOF"
676 " %zu", (size_t)rec.r.eof);
677 }
678 found_recovery = true;
679 len = sizeof(rec.r) + rec.r.max_len;
680 } else if (frec_magic(&rec.f) == TDB_FREE_MAGIC) {
681 len = sizeof(rec.u) + frec_len(&rec.f);
682 if (off + len > tdb->file->map_size) {
683 return tdb_logerr(tdb, TDB_ERR_CORRUPT,
684 TDB_LOG_ERROR,
685 "tdb_check: free overlength"
686 " %llu at offset %llu",
687 (long long)len,
688 (long long)off);
689 }
690 /* This record should be in free lists. */
691 if (frec_ftable(&rec.f) != TDB_FTABLE_NONE
692 && !append(fr, num_free, off)) {
693 return tdb_logerr(tdb, TDB_ERR_OOM,
694 TDB_LOG_ERROR,
695 "tdb_check: tracking %zu'th"
696 " free record.", *num_free);
697 }
698 } else if (rec_magic(&rec.u) == TDB_USED_MAGIC
699 || rec_magic(&rec.u) == TDB_CHAIN_MAGIC
700 || rec_magic(&rec.u) == TDB_HTABLE_MAGIC
701 || rec_magic(&rec.u) == TDB_FTABLE_MAGIC) {
702 uint64_t klen, dlen, extra;
703
704 /* This record is used! */
705 if (!append(used, num_used, off)) {
706 return tdb_logerr(tdb, TDB_ERR_OOM,
707 TDB_LOG_ERROR,
708 "tdb_check: tracking %zu'th"
709 " used record.", *num_used);
710 }
711
712 klen = rec_key_length(&rec.u);
713 dlen = rec_data_length(&rec.u);
714 extra = rec_extra_padding(&rec.u);
715
716 len = sizeof(rec.u) + klen + dlen + extra;
717 if (off + len > tdb->file->map_size) {
718 return tdb_logerr(tdb, TDB_ERR_CORRUPT,
719 TDB_LOG_ERROR,
720 "tdb_check: used overlength"
721 " %llu at offset %llu",
722 (long long)len,
723 (long long)off);
724 }
725
726 if (len < sizeof(rec.f)) {
727 return tdb_logerr(tdb, TDB_ERR_CORRUPT,
728 TDB_LOG_ERROR,
729 "tdb_check: too short record"
730 " %llu at %llu",
731 (long long)len,
732 (long long)off);
733 }
734
735 /* Check that records have correct 0 at end (but may
736 * not in future). */
737 if (extra && !features) {
738 const char *p;
739 char c;
740 p = tdb_access_read(tdb, off + sizeof(rec.u)
741 + klen + dlen, 1, false);
742 if (TDB_PTR_IS_ERR(p))
743 return TDB_PTR_ERR(p);
744 c = *p;
745 tdb_access_release(tdb, p);
746
747 if (c != '\0') {
748 return tdb_logerr(tdb, TDB_ERR_CORRUPT,
749 TDB_LOG_ERROR,
750 "tdb_check:"
751 " non-zero extra"
752 " at %llu",
753 (long long)off);
754 }
755 }
756 } else {
757 return tdb_logerr(tdb, TDB_ERR_CORRUPT,
758 TDB_LOG_ERROR,
759 "tdb_check: Bad magic 0x%llx"
760 " at offset %zu",
761 (long long)rec_magic(&rec.u),
762 (size_t)off);
763 }
764 }
765
766 /* We must have found recovery area if there was one. */
767 if (recovery != 0 && !found_recovery) {
768 return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
769 "tdb_check: expected a recovery area at %zu",
770 (size_t)recovery);
771 }
772
773 return TDB_SUCCESS;
774 }
775
776 enum TDB_ERROR tdb_check_(struct tdb_context *tdb,
777 enum TDB_ERROR (*check)(TDB_DATA, TDB_DATA, void *),
778 void *data)
779 {
780 tdb_off_t *fr = NULL, *used = NULL, ft, recovery;
781 size_t num_free = 0, num_used = 0, num_found = 0, num_ftables = 0;
782 uint64_t features;
783 enum TDB_ERROR ecode;
784
785 ecode = tdb_allrecord_lock(tdb, F_RDLCK, TDB_LOCK_WAIT, false);
786 if (ecode != TDB_SUCCESS) {
787 return tdb->last_error = ecode;
788 }
789
790 ecode = tdb_lock_expand(tdb, F_RDLCK);
791 if (ecode != TDB_SUCCESS) {
792 tdb_allrecord_unlock(tdb, F_RDLCK);
793 return tdb->last_error = ecode;
794 }
795
796 ecode = check_header(tdb, &recovery, &features);
797 if (ecode != TDB_SUCCESS)
798 goto out;
799
800 /* First we do a linear scan, checking all records. */
801 ecode = check_linear(tdb, &used, &num_used, &fr, &num_free, features,
802 recovery);
803 if (ecode != TDB_SUCCESS)
804 goto out;
805
806 for (ft = first_ftable(tdb); ft; ft = next_ftable(tdb, ft)) {
807 if (TDB_OFF_IS_ERR(ft)) {
808 ecode = ft;
809 goto out;
810 }
811 ecode = check_free_table(tdb, ft, num_ftables, fr, num_free,
812 &num_found);
813 if (ecode != TDB_SUCCESS)
814 goto out;
815 num_ftables++;
816 }
817
818 /* FIXME: Check key uniqueness? */
819 ecode = check_hash(tdb, used, num_used, num_ftables, check, data);
820 if (ecode != TDB_SUCCESS)
821 goto out;
822
823 if (num_found != num_free) {
824 ecode = tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
825 "tdb_check: Not all entries are in"
826 " free table");
827 }
828
829 out:
830 tdb_allrecord_unlock(tdb, F_RDLCK);
831 tdb_unlock_expand(tdb, F_RDLCK);
832 free(fr);
833 free(used);
834 return tdb->last_error = ecode;
835 }
WIP