2 Trivial Database 2: free list/block handling
3 Copyright (C) Rusty Russell 2010
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.
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.
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/>.
19 #include <ccan/likely/likely.h>
20 #include <ccan/asearch/asearch.h>
22 /* We keep an ordered array of offsets. */
23 static bool append(tdb_off_t **arr, size_t *num, tdb_off_t off)
25 tdb_off_t *new = realloc(*arr, (*num + 1) * sizeof(tdb_off_t));
33 static bool check_header(struct tdb_context *tdb)
36 struct tdb_header hdr;
38 if (tdb_read_convert(tdb, 0, &hdr, sizeof(hdr)) == -1)
40 /* magic food should not be converted, so convert back. */
41 tdb_convert(tdb, hdr.magic_food, sizeof(hdr.magic_food));
43 hash_test = TDB_HASH_MAGIC;
44 hash_test = tdb_hash(tdb, &hash_test, sizeof(hash_test));
45 if (hdr.hash_test != hash_test) {
46 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
47 "check: hash test %llu should be %llu\n",
48 (long long)hdr.hash_test,
49 (long long)hash_test);
53 if (strcmp(hdr.magic_food, TDB_MAGIC_FOOD) != 0) {
54 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
55 "check: bad magic '%.*s'\n",
56 (unsigned)sizeof(hdr.magic_food), hdr.magic_food);
60 /* Don't check reserved: they *can* be used later. */
64 static bool check_hash_tree(struct tdb_context *tdb,
65 tdb_off_t off, unsigned int group_bits,
67 unsigned hprefix_bits,
72 static bool check_hash_record(struct tdb_context *tdb,
75 unsigned hprefix_bits,
80 struct tdb_used_record rec;
82 if (tdb_read_convert(tdb, off, &rec, sizeof(rec)) == -1)
85 if (rec_data_length(&rec)
86 != sizeof(tdb_off_t) << TDB_SUBLEVEL_HASH_BITS) {
87 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
88 "tdb_check: Bad hash table length %llu vs %llu\n",
89 (long long)rec_data_length(&rec),
90 (long long)sizeof(tdb_off_t)<<TDB_SUBLEVEL_HASH_BITS);
93 if (rec_key_length(&rec) != 0) {
94 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
95 "tdb_check: Bad hash table key length %llu\n",
96 (long long)rec_key_length(&rec));
99 if (rec_hash(&rec) != 0) {
100 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
101 "tdb_check: Bad hash table hash value %llu\n",
102 (long long)rec_hash(&rec));
107 return check_hash_tree(tdb, off,
108 TDB_SUBLEVEL_HASH_BITS-TDB_HASH_GROUP_BITS,
109 hprefix, hprefix_bits,
110 used, num_used, num_found);
113 static int off_cmp(const tdb_off_t *a, const tdb_off_t *b)
115 /* Can overflow an int. */
121 static uint64_t get_bits(uint64_t h, unsigned num, unsigned *used)
125 return (h >> (64 - *used)) & ((1U << num) - 1);
128 static bool check_hash_tree(struct tdb_context *tdb,
129 tdb_off_t off, unsigned int group_bits,
131 unsigned hprefix_bits,
137 const tdb_off_t *hash;
138 struct tdb_used_record rec;
140 hash = tdb_access_read(tdb, off,
142 << (group_bits + TDB_HASH_GROUP_BITS),
147 for (g = 0; g < (1 << group_bits); g++) {
148 const tdb_off_t *group = hash + (g << TDB_HASH_GROUP_BITS);
149 for (b = 0; b < (1 << TDB_HASH_GROUP_BITS); b++) {
150 unsigned int bucket, i, used_bits;
156 off = group[b] & TDB_OFF_MASK;
157 p = asearch(&off, used, num_used, off_cmp);
159 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
160 "tdb_check: Invalid offset %llu "
165 /* Mark it invalid. */
169 if (is_subhash(group[b])) {
172 << (group_bits + TDB_HASH_GROUP_BITS))
173 + g * (1 << TDB_HASH_GROUP_BITS) + b;
175 if (!check_hash_record(tdb,
176 group[b] & TDB_OFF_MASK,
180 + TDB_HASH_GROUP_BITS,
181 used, num_used, num_found))
187 /* Does it belong here at all? */
188 h = hash_record(tdb, off);
190 if (get_bits(h, hprefix_bits, &used_bits) != hprefix
192 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
193 "check: bad hash placement"
194 " 0x%llx vs 0x%llx\n",
195 (long long)h, (long long)hprefix);
199 /* Does it belong in this group? */
200 if (get_bits(h, group_bits, &used_bits) != g) {
201 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
202 "check: bad group %llu vs %u\n",
207 /* Are bucket bits correct? */
208 bucket = group[b] & TDB_OFF_HASH_GROUP_MASK;
209 if (get_bits(h, TDB_HASH_GROUP_BITS, &used_bits)
211 used_bits -= TDB_HASH_GROUP_BITS;
212 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
213 "check: bad bucket %u vs %u\n",
214 (unsigned)get_bits(h,
221 /* There must not be any zero entries between
222 * the bucket it belongs in and this one! */
225 i = (i + 1) % (1 << TDB_HASH_GROUP_BITS)) {
227 tdb->log(tdb, TDB_DEBUG_ERROR,
229 "check: bad group placement"
236 if (tdb_read_convert(tdb, off, &rec, sizeof(rec)) == -1)
239 /* Bottom bits must match header. */
240 if ((h & ((1 << 5)-1)) != rec_hash(&rec)) {
241 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
242 "tdb_check: Bad hash magic at"
243 " offset %llu (0x%llx vs 0x%llx)\n",
246 (long long)rec_hash(&rec));
251 tdb_access_release(tdb, hash);
255 tdb_access_release(tdb, hash);
259 static bool check_hash(struct tdb_context *tdb,
263 size_t num_found = 0;
265 if (!check_hash_tree(tdb, offsetof(struct tdb_header, hashtable),
266 TDB_TOPLEVEL_HASH_BITS-TDB_HASH_GROUP_BITS,
267 0, 0, used, num_used, &num_found))
270 if (num_found != num_used) {
271 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
272 "tdb_check: Not all entries are in hash\n");
278 static bool check_free(struct tdb_context *tdb,
280 const struct tdb_free_record *frec,
282 tdb_off_t zone_off, unsigned int bucket)
284 if (frec_magic(frec) != TDB_FREE_MAGIC) {
285 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
286 "tdb_check: offset %llu bad magic 0x%llx\n",
287 (long long)off, (long long)frec->magic_and_meta);
290 if (tdb->methods->oob(tdb, off
291 + frec->data_len+sizeof(struct tdb_used_record),
294 if (off < zone_off || off >= zone_off + (1ULL<<frec_zone_bits(frec))) {
295 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
296 "tdb_check: offset %llu outside zone %llu-%llu\n",
299 (long long)zone_off + (1ULL<<frec_zone_bits(frec)));
302 if (size_to_bucket(frec_zone_bits(frec), frec->data_len) != bucket) {
303 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
304 "tdb_check: offset %llu in wrong bucket %u vs %u\n",
307 size_to_bucket(frec_zone_bits(frec), frec->data_len));
310 if (prev != frec->prev) {
311 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
312 "tdb_check: offset %llu bad prev %llu vs %llu\n",
314 (long long)prev, (long long)frec->prev);
320 static tdb_len_t check_free_list(struct tdb_context *tdb,
326 struct free_zone_header zhdr;
330 if (tdb_read_convert(tdb, zone_off, &zhdr, sizeof(zhdr)) == -1)
333 for (i = 0; i <= BUCKETS_FOR_ZONE(zhdr.zone_bits); i++) {
334 tdb_off_t off, prev = 0, *p;
335 struct tdb_free_record f;
337 h = bucket_off(zone_off, i);
338 for (off = tdb_read_off(tdb, h); off; off = f.next) {
339 if (off == TDB_OFF_ERR)
341 if (tdb_read_convert(tdb, off, &f, sizeof(f)))
343 if (!check_free(tdb, off, &f, prev, zone_off, i))
346 /* FIXME: Check hash bits */
347 p = asearch(&off, free, num_free, off_cmp);
349 tdb->log(tdb, TDB_DEBUG_ERROR,
351 "tdb_check: Invalid offset"
352 " %llu in free table\n",
356 /* Mark it invalid. */
362 return 1ULL << zhdr.zone_bits;
365 static tdb_off_t check_zone(struct tdb_context *tdb, tdb_off_t zone_off,
366 tdb_off_t **used, size_t *num_used,
367 tdb_off_t **free, size_t *num_free,
368 unsigned int *max_zone_bits)
370 struct free_zone_header zhdr;
371 tdb_off_t off, hdrlen;
374 if (tdb_read_convert(tdb, zone_off, &zhdr, sizeof(zhdr)) == -1)
377 if (zhdr.zone_bits < INITIAL_ZONE_BITS) {
378 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
379 "check: bad zone_bits %llu at zone %llu\n",
380 (long long)zhdr.zone_bits, (long long)zone_off);
384 /* Zone bits can only increase... */
385 if (zhdr.zone_bits > *max_zone_bits)
386 *max_zone_bits = zhdr.zone_bits;
387 else if (zhdr.zone_bits < *max_zone_bits) {
388 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
389 "check: small zone_bits %llu at zone %llu\n",
390 (long long)zhdr.zone_bits, (long long)zone_off);
394 /* Zone must be within file! */
395 if (tdb->methods->oob(tdb, zone_off + (1ULL << zhdr.zone_bits), false))
398 hdrlen = sizeof(zhdr)
399 + (BUCKETS_FOR_ZONE(zhdr.zone_bits) + 1) * sizeof(tdb_off_t);
400 for (off = zone_off + hdrlen;
401 off < zone_off + (1ULL << zhdr.zone_bits);
404 struct tdb_used_record u;
405 struct tdb_free_record f;
407 p = tdb_get(tdb, off, &pad, sizeof(pad));
410 if (frec_magic(&p->f) == TDB_FREE_MAGIC
411 || frec_magic(&p->f) == TDB_COALESCING_MAGIC) {
412 if (frec_zone_bits(&p->f) != zhdr.zone_bits) {
413 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
414 "tdb_check: Bad free zone bits %u"
416 frec_zone_bits(&p->f),
420 len = sizeof(p->u) + p->f.data_len;
421 if (off + len > zone_off + (1ULL << zhdr.zone_bits)) {
422 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
423 "tdb_check: free overlength %llu"
425 (long long)len, (long long)off);
428 /* This record is free! */
429 if (frec_magic(&p->f) == TDB_FREE_MAGIC
430 && !append(free, num_free, off))
433 uint64_t klen, dlen, extra;
435 /* This record is used! */
436 if (rec_magic(&p->u) != TDB_MAGIC) {
437 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
438 "tdb_check: Bad magic 0x%llx"
440 (long long)rec_magic(&p->u),
445 if (rec_zone_bits(&p->u) != zhdr.zone_bits) {
446 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
447 "tdb_check: Bad zone bits %u"
449 rec_zone_bits(&p->u),
454 if (!append(used, num_used, off))
457 klen = rec_key_length(&p->u);
458 dlen = rec_data_length(&p->u);
459 extra = rec_extra_padding(&p->u);
461 len = sizeof(p->u) + klen + dlen + extra;
462 if (off + len > zone_off + (1ULL << zhdr.zone_bits)) {
463 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
464 "tdb_check: used overlength %llu"
466 (long long)len, (long long)off);
470 if (len < sizeof(p->f)) {
471 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
472 "tdb_check: too short record %llu at"
474 (long long)len, (long long)off);
479 return 1ULL << zhdr.zone_bits;
482 /* FIXME: call check() function. */
483 int tdb_check(struct tdb_context *tdb,
484 int (*check)(TDB_DATA key, TDB_DATA data, void *private_data),
487 tdb_off_t *free = NULL, *used = NULL, off;
489 size_t num_free = 0, num_used = 0, num_found = 0;
490 unsigned max_zone_bits = INITIAL_ZONE_BITS;
493 if (tdb_allrecord_lock(tdb, F_RDLCK, TDB_LOCK_WAIT, false) != 0)
496 if (tdb_lock_expand(tdb, F_RDLCK) != 0) {
497 tdb_allrecord_unlock(tdb, F_RDLCK);
501 if (!check_header(tdb))
504 /* First we do a linear scan, checking all records. */
505 for (off = sizeof(struct tdb_header);
506 off < tdb->map_size - 1;
508 len = check_zone(tdb, off, &used, &num_used, &free, &num_free,
510 if (len == TDB_OFF_ERR)
515 if (tdb->methods->read(tdb, tdb->map_size - 1, &tailer, 1) == -1)
517 if (tailer != max_zone_bits) {
518 tdb->ecode = TDB_ERR_CORRUPT;
519 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
520 "tdb_check: Bad tailer value %u vs %u\n", tailer,
525 /* FIXME: Check key uniqueness? */
526 if (!check_hash(tdb, used, num_used))
529 for (off = sizeof(struct tdb_header);
530 off < tdb->map_size - 1;
532 len = check_free_list(tdb, off, free, num_free, &num_found);
533 if (len == TDB_OFF_ERR)
536 if (num_found != num_free) {
537 tdb->log(tdb, TDB_DEBUG_ERROR, tdb->log_priv,
538 "tdb_check: Not all entries are in free table\n");
542 tdb_allrecord_unlock(tdb, F_RDLCK);
543 tdb_unlock_expand(tdb, F_RDLCK);
547 tdb_allrecord_unlock(tdb, F_RDLCK);
548 tdb_unlock_expand(tdb, F_RDLCK);