git.ozlabs.org Git - ccan/blob - ccan/tdb2/hash.c

   1  /*
   2    Trivial Database 2: hash 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 <assert.h>
  20 #include <ccan/hash/hash.h>
  21
  22 static uint64_t jenkins_hash(const void *key, size_t length, uint64_t seed,
  23                              void *arg)
  24 {
  25         uint64_t ret;
  26         /* hash64_stable assumes lower bits are more important; they are a
  27          * slightly better hash.  We use the upper bits first, so swap them. */
  28         ret = hash64_stable((const unsigned char *)key, length, seed);
  29         return (ret >> 32) | (ret << 32);
  30 }
  31
  32 void tdb_hash_init(struct tdb_context *tdb)
  33 {
  34         tdb->khash = jenkins_hash;
  35         tdb->hash_priv = NULL;
  36 }
  37
  38 uint64_t tdb_hash(struct tdb_context *tdb, const void *ptr, size_t len)
  39 {
  40         return tdb->khash(ptr, len, tdb->hash_seed, tdb->hash_priv);
  41 }
  42
  43 uint64_t hash_record(struct tdb_context *tdb, tdb_off_t off)
  44 {
  45         const struct tdb_used_record *r;
  46         const void *key;
  47         uint64_t klen, hash;
  48
  49         r = tdb_access_read(tdb, off, sizeof(*r), true);
  50         if (!r)
  51                 /* FIXME */
  52                 return 0;
  53
  54         klen = rec_key_length(r);
  55         tdb_access_release(tdb, r);
  56
  57         key = tdb_access_read(tdb, off + sizeof(*r), klen, false);
  58         if (!key)
  59                 return 0;
  60
  61         hash = tdb_hash(tdb, key, klen);
  62         tdb_access_release(tdb, key);
  63         return hash;
  64 }
  65
  66 /* Get bits from a value. */
  67 static uint32_t bits(uint64_t val, unsigned start, unsigned num)
  68 {
  69         assert(num <= 32);
  70         return (val >> start) & ((1U << num) - 1);
  71 }
  72
  73 /* We take bits from the top: that way we can lock whole sections of the hash
  74  * by using lock ranges. */
  75 static uint32_t use_bits(struct hash_info *h, unsigned num)
  76 {
  77         h->hash_used += num;
  78         return bits(h->h, 64 - h->hash_used, num);
  79 }
  80
  81 /* Does entry match? */
  82 static bool match(struct tdb_context *tdb,
  83                   struct hash_info *h,
  84                   const struct tdb_data *key,
  85                   tdb_off_t val,
  86                   struct tdb_used_record *rec)
  87 {
  88         bool ret = false;
  89         const unsigned char *rkey;
  90         tdb_off_t off;
  91
  92         add_stat(tdb, compares, 1);
  93         /* Desired bucket must match. */
  94         if (h->home_bucket != (val & TDB_OFF_HASH_GROUP_MASK)) {
  95                 add_stat(tdb, compare_wrong_bucket, 1);
  96                 return ret;
  97         }
  98
  99         /* Top bits of offset == next bits of hash. */
 100         if (bits(val, TDB_OFF_HASH_EXTRA_BIT, TDB_OFF_UPPER_STEAL_EXTRA)
 101             != bits(h->h, 64 - h->hash_used - TDB_OFF_UPPER_STEAL_EXTRA,
 102                     TDB_OFF_UPPER_STEAL_EXTRA)) {
 103                 add_stat(tdb, compare_wrong_offsetbits, 1);
 104                 return ret;
 105         }
 106
 107         off = val & TDB_OFF_MASK;
 108         if (tdb_read_convert(tdb, off, rec, sizeof(*rec)) == -1)
 109                 return ret;
 110
 111         /* FIXME: check extra bits in header? */
 112         if (rec_key_length(rec) != key->dsize) {
 113                 add_stat(tdb, compare_wrong_keylen, 1);
 114                 return ret;
 115         }
 116
 117         rkey = tdb_access_read(tdb, off + sizeof(*rec), key->dsize, false);
 118         if (!rkey)
 119                 return ret;
 120         if (memcmp(rkey, key->dptr, key->dsize) == 0)
 121                 ret = true;
 122         else
 123                 add_stat(tdb, compare_wrong_keycmp, 1);
 124         tdb_access_release(tdb, rkey);
 125         return ret;
 126 }
 127
 128 static tdb_off_t hbucket_off(tdb_off_t group_start, unsigned bucket)
 129 {
 130         return group_start
 131                 + (bucket % (1 << TDB_HASH_GROUP_BITS)) * sizeof(tdb_off_t);
 132 }
 133
 134 bool is_subhash(tdb_off_t val)
 135 {
 136         return (val >> TDB_OFF_UPPER_STEAL_SUBHASH_BIT) & 1;
 137 }
 138
 139 /* FIXME: Guess the depth, don't over-lock! */
 140 static tdb_off_t hlock_range(tdb_off_t group, tdb_off_t *size)
 141 {
 142         *size = 1ULL << (64 - (TDB_TOPLEVEL_HASH_BITS - TDB_HASH_GROUP_BITS));
 143         return group << (64 - (TDB_TOPLEVEL_HASH_BITS - TDB_HASH_GROUP_BITS));
 144 }
 145
 146 /* This is the core routine which searches the hashtable for an entry.
 147  * On error, no locks are held and TDB_OFF_ERR is returned.
 148  * Otherwise, hinfo is filled in (and the optional tinfo).
 149  * If not found, the return value is 0.
 150  * If found, the return value is the offset, and *rec is the record. */
 151 tdb_off_t find_and_lock(struct tdb_context *tdb,
 152                         struct tdb_data key,
 153                         int ltype,
 154                         struct hash_info *h,
 155                         struct tdb_used_record *rec,
 156                         struct traverse_info *tinfo)
 157 {
 158         uint32_t i, group;
 159         tdb_off_t hashtable;
 160
 161         h->h = tdb_hash(tdb, key.dptr, key.dsize);
 162         h->hash_used = 0;
 163         group = use_bits(h, TDB_TOPLEVEL_HASH_BITS - TDB_HASH_GROUP_BITS);
 164         h->home_bucket = use_bits(h, TDB_HASH_GROUP_BITS);
 165
 166         h->hlock_start = hlock_range(group, &h->hlock_range);
 167         if (tdb_lock_hashes(tdb, h->hlock_start, h->hlock_range, ltype,
 168                             TDB_LOCK_WAIT))
 169                 return TDB_OFF_ERR;
 170
 171         hashtable = offsetof(struct tdb_header, hashtable);
 172         if (tinfo) {
 173                 tinfo->toplevel_group = group;
 174                 tinfo->num_levels = 1;
 175                 tinfo->levels[0].entry = 0;
 176                 tinfo->levels[0].hashtable = hashtable
 177                         + (group << TDB_HASH_GROUP_BITS) * sizeof(tdb_off_t);
 178                 tinfo->levels[0].total_buckets = 1 << TDB_HASH_GROUP_BITS;
 179         }
 180
 181         while (likely(h->hash_used < 64)) {
 182                 /* Read in the hash group. */
 183                 h->group_start = hashtable
 184                         + group * (sizeof(tdb_off_t) << TDB_HASH_GROUP_BITS);
 185
 186                 if (tdb_read_convert(tdb, h->group_start, &h->group,
 187                                      sizeof(h->group)) == -1)
 188                         goto fail;
 189
 190                 /* Pointer to another hash table?  Go down... */
 191                 if (is_subhash(h->group[h->home_bucket])) {
 192                         hashtable = (h->group[h->home_bucket] & TDB_OFF_MASK)
 193                                 + sizeof(struct tdb_used_record);
 194                         if (tinfo) {
 195                                 /* When we come back, use *next* bucket */
 196                                 tinfo->levels[tinfo->num_levels-1].entry
 197                                         += h->home_bucket + 1;
 198                         }
 199                         group = use_bits(h, TDB_SUBLEVEL_HASH_BITS
 200                                          - TDB_HASH_GROUP_BITS);
 201                         h->home_bucket = use_bits(h, TDB_HASH_GROUP_BITS);
 202                         if (tinfo) {
 203                                 tinfo->levels[tinfo->num_levels].hashtable
 204                                         = hashtable;
 205                                 tinfo->levels[tinfo->num_levels].total_buckets
 206                                         = 1 << TDB_SUBLEVEL_HASH_BITS;
 207                                 tinfo->levels[tinfo->num_levels].entry
 208                                         = group << TDB_HASH_GROUP_BITS;
 209                                 tinfo->num_levels++;
 210                         }
 211                         continue;
 212                 }
 213
 214                 /* It's in this group: search (until 0 or all searched) */
 215                 for (i = 0, h->found_bucket = h->home_bucket;
 216                      i < (1 << TDB_HASH_GROUP_BITS);
 217                      i++, h->found_bucket = ((h->found_bucket+1)
 218                                              % (1 << TDB_HASH_GROUP_BITS))) {
 219                         if (is_subhash(h->group[h->found_bucket]))
 220                                 continue;
 221
 222                         if (!h->group[h->found_bucket])
 223                                 break;
 224
 225                         if (match(tdb, h, &key, h->group[h->found_bucket],
 226                                   rec)) {
 227                                 if (tinfo) {
 228                                         tinfo->levels[tinfo->num_levels-1].entry
 229                                                 += h->found_bucket;
 230                                 }
 231                                 return h->group[h->found_bucket] & TDB_OFF_MASK;
 232                         }
 233                 }
 234                 /* Didn't find it: h indicates where it would go. */
 235                 return 0;
 236         }
 237
 238         /* FIXME: We hit the bottom.  Chain! */
 239         abort();
 240
 241 fail:
 242         tdb_unlock_hashes(tdb, h->hlock_start, h->hlock_range, ltype);
 243         return TDB_OFF_ERR;
 244 }
 245
 246 /* I wrote a simple test, expanding a hash to 2GB, for the following
 247  * cases:
 248  * 1) Expanding all the buckets at once,
 249  * 2) Expanding the bucket we wanted to place the new entry into.
 250  * 3) Expanding the most-populated bucket,
 251  *
 252  * I measured the worst/average/best density during this process.
 253  * 1) 3%/16%/30%
 254  * 2) 4%/20%/38%
 255  * 3) 6%/22%/41%
 256  *
 257  * So we figure out the busiest bucket for the moment.
 258  */
 259 static unsigned fullest_bucket(struct tdb_context *tdb,
 260                                const tdb_off_t *group,
 261                                unsigned new_bucket)
 262 {
 263         unsigned counts[1 << TDB_HASH_GROUP_BITS] = { 0 };
 264         unsigned int i, best_bucket;
 265
 266         /* Count the new entry. */
 267         counts[new_bucket]++;
 268         best_bucket = new_bucket;
 269
 270         for (i = 0; i < (1 << TDB_HASH_GROUP_BITS); i++) {
 271                 unsigned this_bucket;
 272
 273                 if (is_subhash(group[i]))
 274                         continue;
 275                 this_bucket = group[i] & TDB_OFF_HASH_GROUP_MASK;
 276                 if (++counts[this_bucket] > counts[best_bucket])
 277                         best_bucket = this_bucket;
 278         }
 279
 280         return best_bucket;
 281 }
 282
 283 static bool put_into_group(tdb_off_t *group,
 284                            unsigned bucket, tdb_off_t encoded)
 285 {
 286         unsigned int i;
 287
 288         for (i = 0; i < (1 << TDB_HASH_GROUP_BITS); i++) {
 289                 unsigned b = (bucket + i) % (1 << TDB_HASH_GROUP_BITS);
 290
 291                 if (group[b] == 0) {
 292                         group[b] = encoded;
 293                         return true;
 294                 }
 295         }
 296         return false;
 297 }
 298
 299 static void force_into_group(tdb_off_t *group,
 300                              unsigned bucket, tdb_off_t encoded)
 301 {
 302         if (!put_into_group(group, bucket, encoded))
 303                 abort();
 304 }
 305
 306 static tdb_off_t encode_offset(tdb_off_t new_off, struct hash_info *h)
 307 {
 308         return h->home_bucket
 309                 | new_off
 310                 | ((uint64_t)bits(h->h,
 311                                   64 - h->hash_used - TDB_OFF_UPPER_STEAL_EXTRA,
 312                                   TDB_OFF_UPPER_STEAL_EXTRA)
 313                    << TDB_OFF_HASH_EXTRA_BIT);
 314 }
 315
 316 /* Simply overwrite the hash entry we found before. */
 317 int replace_in_hash(struct tdb_context *tdb,
 318                     struct hash_info *h,
 319                     tdb_off_t new_off)
 320 {
 321         return tdb_write_off(tdb, hbucket_off(h->group_start, h->found_bucket),
 322                              encode_offset(new_off, h));
 323 }
 324
 325 /* Add into a newly created subhash. */
 326 static int add_to_subhash(struct tdb_context *tdb, tdb_off_t subhash,
 327                           unsigned hash_used, tdb_off_t val)
 328 {
 329         tdb_off_t off = (val & TDB_OFF_MASK), *group;
 330         struct hash_info h;
 331         unsigned int gnum;
 332
 333         h.hash_used = hash_used;
 334
 335         /* FIXME chain if hash_used == 64 */
 336         if (hash_used + TDB_SUBLEVEL_HASH_BITS > 64)
 337                 abort();
 338
 339         h.h = hash_record(tdb, off);
 340         gnum = use_bits(&h, TDB_SUBLEVEL_HASH_BITS-TDB_HASH_GROUP_BITS);
 341         h.group_start = subhash + sizeof(struct tdb_used_record)
 342                 + gnum * (sizeof(tdb_off_t) << TDB_HASH_GROUP_BITS);
 343         h.home_bucket = use_bits(&h, TDB_HASH_GROUP_BITS);
 344
 345         group = tdb_access_write(tdb, h.group_start,
 346                                  sizeof(*group) << TDB_HASH_GROUP_BITS, true);
 347         if (!group)
 348                 return -1;
 349         force_into_group(group, h.home_bucket, encode_offset(off, &h));
 350         return tdb_access_commit(tdb, group);
 351 }
 352
 353 static int expand_group(struct tdb_context *tdb, struct hash_info *h)
 354 {
 355         unsigned bucket, num_vals, i;
 356         tdb_off_t subhash;
 357         tdb_off_t vals[1 << TDB_HASH_GROUP_BITS];
 358
 359         /* Attach new empty subhash under fullest bucket. */
 360         bucket = fullest_bucket(tdb, h->group, h->home_bucket);
 361
 362         subhash = alloc(tdb, 0, sizeof(tdb_off_t) << TDB_SUBLEVEL_HASH_BITS,
 363                         0, false);
 364         if (subhash == TDB_OFF_ERR)
 365                 return -1;
 366
 367         add_stat(tdb, alloc_subhash, 1);
 368         if (zero_out(tdb, subhash + sizeof(struct tdb_used_record),
 369                      sizeof(tdb_off_t) << TDB_SUBLEVEL_HASH_BITS) == -1)
 370                 return -1;
 371
 372         /* Remove any which are destined for bucket or are in wrong place. */
 373         num_vals = 0;
 374         for (i = 0; i < (1 << TDB_HASH_GROUP_BITS); i++) {
 375                 unsigned home_bucket = h->group[i] & TDB_OFF_HASH_GROUP_MASK;
 376                 if (!h->group[i] || is_subhash(h->group[i]))
 377                         continue;
 378                 if (home_bucket == bucket || home_bucket != i) {
 379                         vals[num_vals++] = h->group[i];
 380                         h->group[i] = 0;
 381                 }
 382         }
 383         /* FIXME: This assert is valid, but we do this during unit test :( */
 384         /* assert(num_vals); */
 385
 386         /* Overwrite expanded bucket with subhash pointer. */
 387         h->group[bucket] = subhash | (1ULL << TDB_OFF_UPPER_STEAL_SUBHASH_BIT);
 388
 389         /* Put values back. */
 390         for (i = 0; i < num_vals; i++) {
 391                 unsigned this_bucket = vals[i] & TDB_OFF_HASH_GROUP_MASK;
 392
 393                 if (this_bucket == bucket) {
 394                         if (add_to_subhash(tdb, subhash, h->hash_used, vals[i]))
 395                                 return -1;
 396                 } else {
 397                         /* There should be room to put this back. */
 398                         force_into_group(h->group, this_bucket, vals[i]);
 399                 }
 400         }
 401         return 0;
 402 }
 403
 404 int delete_from_hash(struct tdb_context *tdb, struct hash_info *h)
 405 {
 406         unsigned int i, num_movers = 0;
 407         tdb_off_t movers[1 << TDB_HASH_GROUP_BITS];
 408
 409         h->group[h->found_bucket] = 0;
 410         for (i = 1; i < (1 << TDB_HASH_GROUP_BITS); i++) {
 411                 unsigned this_bucket;
 412
 413                 this_bucket = (h->found_bucket+i) % (1 << TDB_HASH_GROUP_BITS);
 414                 /* Empty bucket?  We're done. */
 415                 if (!h->group[this_bucket])
 416                         break;
 417
 418                 /* Ignore subhashes. */
 419                 if (is_subhash(h->group[this_bucket]))
 420                         continue;
 421
 422                 /* If this one is not happy where it is, we'll move it. */
 423                 if ((h->group[this_bucket] & TDB_OFF_HASH_GROUP_MASK)
 424                     != this_bucket) {
 425                         movers[num_movers++] = h->group[this_bucket];
 426                         h->group[this_bucket] = 0;
 427                 }
 428         }
 429
 430         /* Put back the ones we erased. */
 431         for (i = 0; i < num_movers; i++) {
 432                 force_into_group(h->group, movers[i] & TDB_OFF_HASH_GROUP_MASK,
 433                                  movers[i]);
 434         }
 435
 436         /* Now we write back the hash group */
 437         return tdb_write_convert(tdb, h->group_start,
 438                                  h->group, sizeof(h->group));
 439 }
 440
 441 int add_to_hash(struct tdb_context *tdb, struct hash_info *h, tdb_off_t new_off)
 442 {
 443         /* FIXME: chain! */
 444         if (h->hash_used >= 64)
 445                 abort();
 446
 447         /* We hit an empty bucket during search?  That's where it goes. */
 448         if (!h->group[h->found_bucket]) {
 449                 h->group[h->found_bucket] = encode_offset(new_off, h);
 450                 /* Write back the modified group. */
 451                 return tdb_write_convert(tdb, h->group_start,
 452                                          h->group, sizeof(h->group));
 453         }
 454
 455         /* We're full.  Expand. */
 456         if (expand_group(tdb, h) == -1)
 457                 return -1;
 458
 459         if (is_subhash(h->group[h->home_bucket])) {
 460                 /* We were expanded! */
 461                 tdb_off_t hashtable;
 462                 unsigned int gnum;
 463
 464                 /* Write back the modified group. */
 465                 if (tdb_write_convert(tdb, h->group_start, h->group,
 466                                       sizeof(h->group)))
 467                         return -1;
 468
 469                 /* Move hashinfo down a level. */
 470                 hashtable = (h->group[h->home_bucket] & TDB_OFF_MASK)
 471                         + sizeof(struct tdb_used_record);
 472                 gnum = use_bits(h,TDB_SUBLEVEL_HASH_BITS - TDB_HASH_GROUP_BITS);
 473                 h->home_bucket = use_bits(h, TDB_HASH_GROUP_BITS);
 474                 h->group_start = hashtable
 475                         + gnum * (sizeof(tdb_off_t) << TDB_HASH_GROUP_BITS);
 476                 if (tdb_read_convert(tdb, h->group_start, &h->group,
 477                                      sizeof(h->group)) == -1)
 478                         return -1;
 479         }
 480
 481         /* Expanding the group must have made room if it didn't choose this
 482          * bucket. */
 483         if (put_into_group(h->group, h->home_bucket, encode_offset(new_off, h)))
 484                 return tdb_write_convert(tdb, h->group_start,
 485                                          h->group, sizeof(h->group));
 486
 487         /* This can happen if all hashes in group (and us) dropped into same
 488          * group in subhash. */
 489         return add_to_hash(tdb, h, new_off);
 490 }
 491
 492 /* Traverse support: returns offset of record, or 0 or TDB_OFF_ERR. */
 493 static tdb_off_t iterate_hash(struct tdb_context *tdb,
 494                               struct traverse_info *tinfo)
 495 {
 496         tdb_off_t off, val;
 497         unsigned int i;
 498         struct traverse_level *tlevel;
 499
 500         tlevel = &tinfo->levels[tinfo->num_levels-1];
 501
 502 again:
 503         for (i = tdb_find_nonzero_off(tdb, tlevel->hashtable,
 504                                       tlevel->entry, tlevel->total_buckets);
 505              i != tlevel->total_buckets;
 506              i = tdb_find_nonzero_off(tdb, tlevel->hashtable,
 507                                       i+1, tlevel->total_buckets)) {
 508                 val = tdb_read_off(tdb, tlevel->hashtable+sizeof(tdb_off_t)*i);
 509                 if (unlikely(val == TDB_OFF_ERR))
 510                         return TDB_OFF_ERR;
 511
 512                 off = val & TDB_OFF_MASK;
 513
 514                 /* This makes the delete-all-in-traverse case work
 515                  * (and simplifies our logic a little). */
 516                 if (off == tinfo->prev)
 517                         continue;
 518
 519                 tlevel->entry = i;
 520
 521                 if (!is_subhash(val)) {
 522                         /* Found one. */
 523                         tinfo->prev = off;
 524                         return off;
 525                 }
 526
 527                 /* When we come back, we want the next one */
 528                 tlevel->entry++;
 529                 tinfo->num_levels++;
 530                 tlevel++;
 531                 tlevel->hashtable = off + sizeof(struct tdb_used_record);
 532                 tlevel->entry = 0;
 533                 tlevel->total_buckets = (1 << TDB_SUBLEVEL_HASH_BITS);
 534                 goto again;
 535         }
 536
 537         /* Nothing there? */
 538         if (tinfo->num_levels == 1)
 539                 return 0;
 540
 541         /* Go back up and keep searching. */
 542         tinfo->num_levels--;
 543         tlevel--;
 544         goto again;
 545 }
 546
 547 /* Return 1 if we find something, 0 if not, -1 on error. */
 548 int next_in_hash(struct tdb_context *tdb, int ltype,
 549                  struct traverse_info *tinfo,
 550                  TDB_DATA *kbuf, size_t *dlen)
 551 {
 552         const unsigned group_bits = TDB_TOPLEVEL_HASH_BITS-TDB_HASH_GROUP_BITS;
 553         tdb_off_t hlock_start, hlock_range, off;
 554
 555         while (tinfo->toplevel_group < (1 << group_bits)) {
 556                 hlock_start = (tdb_off_t)tinfo->toplevel_group
 557                         << (64 - group_bits);
 558                 hlock_range = 1ULL << group_bits;
 559                 if (tdb_lock_hashes(tdb, hlock_start, hlock_range, ltype,
 560                                     TDB_LOCK_WAIT) != 0)
 561                         return -1;
 562
 563                 off = iterate_hash(tdb, tinfo);
 564                 if (off) {
 565                         struct tdb_used_record rec;
 566
 567                         if (tdb_read_convert(tdb, off, &rec, sizeof(rec))) {
 568                                 tdb_unlock_hashes(tdb,
 569                                                   hlock_start, hlock_range,
 570                                                   ltype);
 571                                 return -1;
 572                         }
 573                         if (rec_magic(&rec) != TDB_MAGIC) {
 574                                 tdb_logerr(tdb, TDB_ERR_CORRUPT,
 575                                            TDB_DEBUG_FATAL,
 576                                            "next_in_hash:"
 577                                            " corrupt record at %llu",
 578                                            (long long)off);
 579                                 return -1;
 580                         }
 581
 582                         kbuf->dsize = rec_key_length(&rec);
 583
 584                         /* They want data as well? */
 585                         if (dlen) {
 586                                 *dlen = rec_data_length(&rec);
 587                                 kbuf->dptr = tdb_alloc_read(tdb,
 588                                                             off + sizeof(rec),
 589                                                             kbuf->dsize
 590                                                             + *dlen);
 591                         } else {
 592                                 kbuf->dptr = tdb_alloc_read(tdb,
 593                                                             off + sizeof(rec),
 594                                                             kbuf->dsize);
 595                         }
 596                         tdb_unlock_hashes(tdb, hlock_start, hlock_range, ltype);
 597                         return kbuf->dptr ? 1 : -1;
 598                 }
 599
 600                 tdb_unlock_hashes(tdb, hlock_start, hlock_range, ltype);
 601
 602                 tinfo->toplevel_group++;
 603                 tinfo->levels[0].hashtable
 604                         += (sizeof(tdb_off_t) << TDB_HASH_GROUP_BITS);
 605                 tinfo->levels[0].entry = 0;
 606         }
 607         return 0;
 608 }
 609
 610 /* Return 1 if we find something, 0 if not, -1 on error. */
 611 int first_in_hash(struct tdb_context *tdb, int ltype,
 612                   struct traverse_info *tinfo,
 613                   TDB_DATA *kbuf, size_t *dlen)
 614 {
 615         tinfo->prev = 0;
 616         tinfo->toplevel_group = 0;
 617         tinfo->num_levels = 1;
 618         tinfo->levels[0].hashtable = offsetof(struct tdb_header, hashtable);
 619         tinfo->levels[0].entry = 0;
 620         tinfo->levels[0].total_buckets = (1 << TDB_HASH_GROUP_BITS);
 621
 622         return next_in_hash(tdb, ltype, tinfo, kbuf, dlen);
 623 }
 624
 625 /* Even if the entry isn't in this hash bucket, you'd have to lock this
 626  * bucket to find it. */
 627 static int chainlock(struct tdb_context *tdb, const TDB_DATA *key,
 628                      int ltype, enum tdb_lock_flags waitflag,
 629                      const char *func)
 630 {
 631         int ret;
 632         uint64_t h = tdb_hash(tdb, key->dptr, key->dsize);
 633         tdb_off_t lockstart, locksize;
 634         unsigned int group, gbits;
 635
 636         gbits = TDB_TOPLEVEL_HASH_BITS - TDB_HASH_GROUP_BITS;
 637         group = bits(h, 64 - gbits, gbits);
 638
 639         lockstart = hlock_range(group, &locksize);
 640
 641         ret = tdb_lock_hashes(tdb, lockstart, locksize, ltype, waitflag);
 642         tdb_trace_1rec(tdb, func, *key);
 643         return ret;
 644 }
 645
 646 /* lock/unlock one hash chain. This is meant to be used to reduce
 647    contention - it cannot guarantee how many records will be locked */
 648 int tdb_chainlock(struct tdb_context *tdb, TDB_DATA key)
 649 {
 650         return chainlock(tdb, &key, F_WRLCK, TDB_LOCK_WAIT, "tdb_chainlock");
 651 }
 652
 653 int tdb_chainunlock(struct tdb_context *tdb, TDB_DATA key)
 654 {
 655         uint64_t h = tdb_hash(tdb, key.dptr, key.dsize);
 656         tdb_off_t lockstart, locksize;
 657         unsigned int group, gbits;
 658
 659         gbits = TDB_TOPLEVEL_HASH_BITS - TDB_HASH_GROUP_BITS;
 660         group = bits(h, 64 - gbits, gbits);
 661
 662         lockstart = hlock_range(group, &locksize);
 663
 664         tdb_trace_1rec(tdb, "tdb_chainunlock", key);
 665         return tdb_unlock_hashes(tdb, lockstart, locksize, F_WRLCK);
 666 }