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

   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/ilog/ilog.h>
  21 #include <time.h>
  22 #include <assert.h>
  23 #include <limits.h>
  24
  25 static unsigned fls64(uint64_t val)
  26 {
  27         return ilog64(val);
  28 }
  29
  30 /* In which bucket would we find a particular record size? (ignoring header) */
  31 unsigned int size_to_bucket(tdb_len_t data_len)
  32 {
  33         unsigned int bucket;
  34
  35         /* We can't have records smaller than this. */
  36         assert(data_len >= TDB_MIN_DATA_LEN);
  37
  38         /* Ignoring the header... */
  39         if (data_len - TDB_MIN_DATA_LEN <= 64) {
  40                 /* 0 in bucket 0, 8 in bucket 1... 64 in bucket 8. */
  41                 bucket = (data_len - TDB_MIN_DATA_LEN) / 8;
  42         } else {
  43                 /* After that we go power of 2. */
  44                 bucket = fls64(data_len - TDB_MIN_DATA_LEN) + 2;
  45         }
  46
  47         if (unlikely(bucket >= TDB_FREE_BUCKETS))
  48                 bucket = TDB_FREE_BUCKETS - 1;
  49         return bucket;
  50 }
  51
  52 tdb_off_t first_ftable(struct tdb_context *tdb)
  53 {
  54         return tdb_read_off(tdb, offsetof(struct tdb_header, free_table));
  55 }
  56
  57 tdb_off_t next_ftable(struct tdb_context *tdb, tdb_off_t ftable)
  58 {
  59         return tdb_read_off(tdb, ftable + offsetof(struct tdb_freetable,next));
  60 }
  61
  62 int tdb_ftable_init(struct tdb_context *tdb)
  63 {
  64         /* Use reservoir sampling algorithm to select a free list at random. */
  65         unsigned int rnd, max = 0, count = 0;
  66         tdb_off_t off;
  67
  68         tdb->ftable_off = off = first_ftable(tdb);
  69         tdb->ftable = 0;
  70
  71         while (off) {
  72                 if (TDB_OFF_IS_ERR(off)) {
  73                         tdb->ecode = off;
  74                         return -1;
  75                 }
  76
  77                 rnd = random();
  78                 if (rnd >= max) {
  79                         tdb->ftable_off = off;
  80                         tdb->ftable = count;
  81                         max = rnd;
  82                 }
  83
  84                 off = next_ftable(tdb, off);
  85                 count++;
  86         }
  87         return 0;
  88 }
  89
  90 /* Offset of a given bucket. */
  91 tdb_off_t bucket_off(tdb_off_t ftable_off, unsigned bucket)
  92 {
  93         return ftable_off + offsetof(struct tdb_freetable, buckets)
  94                 + bucket * sizeof(tdb_off_t);
  95 }
  96
  97 /* Returns free_buckets + 1, or list number to search, or -ve error. */
  98 static tdb_off_t find_free_head(struct tdb_context *tdb,
  99                                 tdb_off_t ftable_off,
 100                                 tdb_off_t bucket)
 101 {
 102         /* Speculatively search for a non-zero bucket. */
 103         return tdb_find_nonzero_off(tdb, bucket_off(ftable_off, 0),
 104                                     bucket, TDB_FREE_BUCKETS);
 105 }
 106
 107 /* Remove from free bucket. */
 108 static int remove_from_list(struct tdb_context *tdb,
 109                             tdb_off_t b_off, tdb_off_t r_off,
 110                             const struct tdb_free_record *r)
 111 {
 112         tdb_off_t off;
 113         enum TDB_ERROR ecode;
 114
 115         /* Front of list? */
 116         if (frec_prev(r) == 0) {
 117                 off = b_off;
 118         } else {
 119                 off = frec_prev(r) + offsetof(struct tdb_free_record, next);
 120         }
 121
 122 #ifdef CCAN_TDB2_DEBUG
 123         if (tdb_read_off(tdb, off) != r_off) {
 124                 tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
 125                          "remove_from_list: %llu bad prev in list %llu",
 126                          (long long)r_off, (long long)b_off);
 127                 return -1;
 128         }
 129 #endif
 130
 131         /* r->prev->next = r->next */
 132         ecode = tdb_write_off(tdb, off, r->next);
 133         if (ecode != TDB_SUCCESS) {
 134                 tdb->ecode = ecode;
 135                 return -1;
 136         }
 137
 138         if (r->next != 0) {
 139                 off = r->next + offsetof(struct tdb_free_record,magic_and_prev);
 140                 /* r->next->prev = r->prev */
 141
 142 #ifdef CCAN_TDB2_DEBUG
 143                 if (tdb_read_off(tdb, off) & TDB_OFF_MASK != r_off) {
 144                         tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
 145                                    "remove_from_list: %llu bad list %llu",
 146                                    (long long)r_off, (long long)b_off);
 147                         return -1;
 148                 }
 149 #endif
 150
 151                 ecode = tdb_write_off(tdb, off, r->magic_and_prev);
 152                 if (ecode != TDB_SUCCESS) {
 153                         tdb->ecode = ecode;
 154                         return -1;
 155                 }
 156         }
 157         return 0;
 158 }
 159
 160 /* Enqueue in this free bucket. */
 161 static int enqueue_in_free(struct tdb_context *tdb,
 162                            tdb_off_t b_off,
 163                            tdb_off_t off,
 164                            tdb_len_t len)
 165 {
 166         struct tdb_free_record new;
 167         enum TDB_ERROR ecode;
 168         uint64_t magic = (TDB_FREE_MAGIC << (64 - TDB_OFF_UPPER_STEAL));
 169
 170         /* We only need to set ftable_and_len; rest is set in enqueue_in_free */
 171         new.ftable_and_len = ((uint64_t)tdb->ftable << (64 - TDB_OFF_UPPER_STEAL))
 172                 | len;
 173         /* prev = 0. */
 174         new.magic_and_prev = magic;
 175
 176         /* new->next = head. */
 177         new.next = tdb_read_off(tdb, b_off);
 178         if (TDB_OFF_IS_ERR(new.next)) {
 179                 tdb->ecode = new.next;
 180                 return -1;
 181         }
 182
 183         if (new.next) {
 184 #ifdef CCAN_TDB2_DEBUG
 185                 if (tdb_read_off(tdb,
 186                                  new.next + offsetof(struct tdb_free_record,
 187                                                      magic_and_prev))
 188                     != magic) {
 189                         tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
 190                                    "enqueue_in_free: %llu bad head"
 191                                    " prev %llu",
 192                                    (long long)new.next, (long long)b_off);
 193                         return -1;
 194                 }
 195 #endif
 196                 /* next->prev = new. */
 197                 ecode = tdb_write_off(tdb, new.next
 198                                       + offsetof(struct tdb_free_record,
 199                                                  magic_and_prev),
 200                                       off | magic);
 201                 if (ecode != TDB_SUCCESS) {
 202                         tdb->ecode = ecode;
 203                         return -1;
 204                 }
 205         }
 206         /* head = new */
 207         ecode = tdb_write_off(tdb, b_off, off);
 208         if (ecode != TDB_SUCCESS) {
 209                 tdb->ecode = ecode;
 210                 return -1;
 211         }
 212
 213         ecode = tdb_write_convert(tdb, off, &new, sizeof(new));
 214         if (ecode != TDB_SUCCESS) {
 215                 tdb->ecode = ecode;
 216                 return -1;
 217         }
 218         return 0;
 219 }
 220
 221 /* List need not be locked. */
 222 int add_free_record(struct tdb_context *tdb,
 223                     tdb_off_t off, tdb_len_t len_with_header)
 224 {
 225         tdb_off_t b_off;
 226         tdb_len_t len;
 227         int ret;
 228         enum TDB_ERROR ecode;
 229
 230         assert(len_with_header >= sizeof(struct tdb_free_record));
 231
 232         len = len_with_header - sizeof(struct tdb_used_record);
 233
 234         b_off = bucket_off(tdb->ftable_off, size_to_bucket(len));
 235         ecode = tdb_lock_free_bucket(tdb, b_off, TDB_LOCK_WAIT);
 236         if (ecode != TDB_SUCCESS) {
 237                 tdb->ecode = ecode;
 238                 return -1;
 239         }
 240
 241         ret = enqueue_in_free(tdb, b_off, off, len);
 242         tdb_unlock_free_bucket(tdb, b_off);
 243         return ret;
 244 }
 245
 246 static size_t adjust_size(size_t keylen, size_t datalen)
 247 {
 248         size_t size = keylen + datalen;
 249
 250         if (size < TDB_MIN_DATA_LEN)
 251                 size = TDB_MIN_DATA_LEN;
 252
 253         /* Round to next uint64_t boundary. */
 254         return (size + (sizeof(uint64_t) - 1ULL)) & ~(sizeof(uint64_t) - 1ULL);
 255 }
 256
 257 /* If we have enough left over to be useful, split that off. */
 258 static size_t record_leftover(size_t keylen, size_t datalen,
 259                               bool want_extra, size_t total_len)
 260 {
 261         ssize_t leftover;
 262
 263         if (want_extra)
 264                 datalen += datalen / 2;
 265         leftover = total_len - adjust_size(keylen, datalen);
 266
 267         if (leftover < (ssize_t)sizeof(struct tdb_free_record))
 268                 return 0;
 269
 270         return leftover;
 271 }
 272
 273 static tdb_off_t ftable_offset(struct tdb_context *tdb, unsigned int ftable)
 274 {
 275         tdb_off_t off;
 276         unsigned int i;
 277
 278         if (likely(tdb->ftable == ftable))
 279                 return tdb->ftable_off;
 280
 281         off = first_ftable(tdb);
 282         for (i = 0; i < ftable; i++) {
 283                 if (TDB_OFF_IS_ERR(off)) {
 284                         tdb->ecode = off;
 285                         break;
 286                 }
 287                 off = next_ftable(tdb, off);
 288         }
 289         return off;
 290 }
 291
 292 /* Note: we unlock the current bucket if we coalesce or fail. */
 293 static int coalesce(struct tdb_context *tdb,
 294                     tdb_off_t off, tdb_off_t b_off, tdb_len_t data_len)
 295 {
 296         tdb_off_t end;
 297         struct tdb_free_record rec;
 298         enum TDB_ERROR ecode;
 299
 300         add_stat(tdb, alloc_coalesce_tried, 1);
 301         end = off + sizeof(struct tdb_used_record) + data_len;
 302
 303         while (end < tdb->map_size) {
 304                 const struct tdb_free_record *r;
 305                 tdb_off_t nb_off;
 306                 unsigned ftable, bucket;
 307
 308                 r = tdb_access_read(tdb, end, sizeof(*r), true);
 309                 if (TDB_PTR_IS_ERR(r)) {
 310                         tdb->ecode = TDB_PTR_ERR(r);
 311                         goto err;
 312                 }
 313
 314                 if (frec_magic(r) != TDB_FREE_MAGIC
 315                     || frec_ftable(r) == TDB_FTABLE_NONE) {
 316                         tdb_access_release(tdb, r);
 317                         break;
 318                 }
 319
 320                 ftable = frec_ftable(r);
 321                 bucket = size_to_bucket(frec_len(r));
 322                 nb_off = bucket_off(ftable_offset(tdb, ftable), bucket);
 323                 tdb_access_release(tdb, r);
 324
 325                 /* We may be violating lock order here, so best effort. */
 326                 if (tdb_lock_free_bucket(tdb, nb_off, TDB_LOCK_NOWAIT)
 327                     != TDB_SUCCESS) {
 328                         add_stat(tdb, alloc_coalesce_lockfail, 1);
 329                         break;
 330                 }
 331
 332                 /* Now we have lock, re-check. */
 333                 ecode = tdb_read_convert(tdb, end, &rec, sizeof(rec));
 334                 if (ecode != TDB_SUCCESS) {
 335                         tdb->ecode = ecode;
 336                         tdb_unlock_free_bucket(tdb, nb_off);
 337                         goto err;
 338                 }
 339
 340                 if (unlikely(frec_magic(&rec) != TDB_FREE_MAGIC)) {
 341                         add_stat(tdb, alloc_coalesce_race, 1);
 342                         tdb_unlock_free_bucket(tdb, nb_off);
 343                         break;
 344                 }
 345
 346                 if (unlikely(frec_ftable(&rec) != ftable)
 347                     || unlikely(size_to_bucket(frec_len(&rec)) != bucket)) {
 348                         add_stat(tdb, alloc_coalesce_race, 1);
 349                         tdb_unlock_free_bucket(tdb, nb_off);
 350                         break;
 351                 }
 352
 353                 if (remove_from_list(tdb, nb_off, end, &rec) == -1) {
 354                         tdb_unlock_free_bucket(tdb, nb_off);
 355                         goto err;
 356                 }
 357
 358                 end += sizeof(struct tdb_used_record) + frec_len(&rec);
 359                 tdb_unlock_free_bucket(tdb, nb_off);
 360                 add_stat(tdb, alloc_coalesce_num_merged, 1);
 361         }
 362
 363         /* Didn't find any adjacent free? */
 364         if (end == off + sizeof(struct tdb_used_record) + data_len)
 365                 return 0;
 366
 367         /* OK, expand initial record */
 368         ecode = tdb_read_convert(tdb, off, &rec, sizeof(rec));
 369         if (ecode != TDB_SUCCESS) {
 370                 tdb->ecode = ecode;
 371                 goto err;
 372         }
 373
 374         if (frec_len(&rec) != data_len) {
 375                 tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
 376                            "coalesce: expected data len %zu not %zu",
 377                            (size_t)data_len, (size_t)frec_len(&rec));
 378                 goto err;
 379         }
 380
 381         if (remove_from_list(tdb, b_off, off, &rec) == -1)
 382                 goto err;
 383
 384         /* We have to drop this to avoid deadlocks, so make sure record
 385          * doesn't get coalesced by someone else! */
 386         rec.ftable_and_len = (TDB_FTABLE_NONE << (64 - TDB_OFF_UPPER_STEAL))
 387                 | (end - off - sizeof(struct tdb_used_record));
 388         ecode = tdb_write_off(tdb, off + offsetof(struct tdb_free_record,
 389                                                   ftable_and_len),
 390                               rec.ftable_and_len);
 391         if (ecode != TDB_SUCCESS) {
 392                 tdb->ecode = ecode;
 393                 goto err;
 394         }
 395
 396         add_stat(tdb, alloc_coalesce_succeeded, 1);
 397         tdb_unlock_free_bucket(tdb, b_off);
 398
 399         if (add_free_record(tdb, off, end - off) == -1)
 400                 return -1;
 401         return 1;
 402
 403 err:
 404         /* To unify error paths, we *always* unlock bucket on error. */
 405         tdb_unlock_free_bucket(tdb, b_off);
 406         return -1;
 407 }
 408
 409 /* We need size bytes to put our key and data in. */
 410 static tdb_off_t lock_and_alloc(struct tdb_context *tdb,
 411                                 tdb_off_t ftable_off,
 412                                 tdb_off_t bucket,
 413                                 size_t keylen, size_t datalen,
 414                                 bool want_extra,
 415                                 unsigned magic,
 416                                 unsigned hashlow)
 417 {
 418         tdb_off_t off, b_off,best_off;
 419         struct tdb_free_record best = { 0 };
 420         double multiplier;
 421         size_t size = adjust_size(keylen, datalen);
 422         enum TDB_ERROR ecode;
 423
 424         add_stat(tdb, allocs, 1);
 425 again:
 426         b_off = bucket_off(ftable_off, bucket);
 427
 428         /* FIXME: Try non-blocking wait first, to measure contention. */
 429         /* Lock this bucket. */
 430         ecode = tdb_lock_free_bucket(tdb, b_off, TDB_LOCK_WAIT);
 431         if (ecode != TDB_SUCCESS) {
 432                 tdb->ecode = ecode;
 433                 return TDB_OFF_ERR;
 434         }
 435
 436         best.ftable_and_len = -1ULL;
 437         best_off = 0;
 438
 439         /* Get slack if we're after extra. */
 440         if (want_extra)
 441                 multiplier = 1.5;
 442         else
 443                 multiplier = 1.0;
 444
 445         /* Walk the list to see if any are large enough, getting less fussy
 446          * as we go. */
 447         off = tdb_read_off(tdb, b_off);
 448         if (TDB_OFF_IS_ERR(off)) {
 449                 tdb->ecode = off;
 450                 goto unlock_err;
 451         }
 452
 453         while (off) {
 454                 const struct tdb_free_record *r;
 455                 tdb_len_t len;
 456                 tdb_off_t next;
 457
 458                 r = tdb_access_read(tdb, off, sizeof(*r), true);
 459                 if (TDB_PTR_IS_ERR(r)) {
 460                         tdb->ecode = TDB_PTR_ERR(r);
 461                         goto unlock_err;
 462                 }
 463
 464                 if (frec_magic(r) != TDB_FREE_MAGIC) {
 465                         tdb_access_release(tdb, r);
 466                         tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
 467                                  "lock_and_alloc: %llu non-free 0x%llx",
 468                                  (long long)off, (long long)r->magic_and_prev);
 469                         goto unlock_err;
 470                 }
 471
 472                 if (frec_len(r) >= size && frec_len(r) < frec_len(&best)) {
 473                         best_off = off;
 474                         best = *r;
 475                 }
 476
 477                 if (frec_len(&best) < size * multiplier && best_off) {
 478                         tdb_access_release(tdb, r);
 479                         break;
 480                 }
 481
 482                 multiplier *= 1.01;
 483
 484                 next = r->next;
 485                 len = frec_len(r);
 486                 tdb_access_release(tdb, r);
 487
 488                 /* Since we're going slow anyway, try coalescing here. */
 489                 switch (coalesce(tdb, off, b_off, len)) {
 490                 case -1:
 491                         /* This has already unlocked on error. */
 492                         return -1;
 493                 case 1:
 494                         /* This has unlocked list, restart. */
 495                         goto again;
 496                 }
 497                 off = next;
 498         }
 499
 500         /* If we found anything at all, use it. */
 501         if (best_off) {
 502                 struct tdb_used_record rec;
 503                 size_t leftover;
 504
 505                 /* We're happy with this size: take it. */
 506                 if (remove_from_list(tdb, b_off, best_off, &best) != 0)
 507                         goto unlock_err;
 508
 509                 leftover = record_leftover(keylen, datalen, want_extra,
 510                                            frec_len(&best));
 511
 512                 assert(keylen + datalen + leftover <= frec_len(&best));
 513                 /* We need to mark non-free before we drop lock, otherwise
 514                  * coalesce() could try to merge it! */
 515                 if (set_header(tdb, &rec, magic, keylen, datalen,
 516                                frec_len(&best) - leftover, hashlow) != 0)
 517                         goto unlock_err;
 518
 519                 ecode = tdb_write_convert(tdb, best_off, &rec, sizeof(rec));
 520                 if (ecode != TDB_SUCCESS) {
 521                         tdb->ecode = ecode;
 522                         goto unlock_err;
 523                 }
 524
 525                 /* Bucket of leftover will be <= current bucket, so nested
 526                  * locking is allowed. */
 527                 if (leftover) {
 528                         add_stat(tdb, alloc_leftover, 1);
 529                         if (add_free_record(tdb,
 530                                             best_off + sizeof(rec)
 531                                             + frec_len(&best) - leftover,
 532                                             leftover))
 533                                 best_off = TDB_OFF_ERR;
 534                 }
 535                 tdb_unlock_free_bucket(tdb, b_off);
 536
 537                 return best_off;
 538         }
 539
 540         tdb_unlock_free_bucket(tdb, b_off);
 541         return 0;
 542
 543 unlock_err:
 544         tdb_unlock_free_bucket(tdb, b_off);
 545         return TDB_OFF_ERR;
 546 }
 547
 548 /* Get a free block from current free list, or 0 if none. */
 549 static tdb_off_t get_free(struct tdb_context *tdb,
 550                           size_t keylen, size_t datalen, bool want_extra,
 551                           unsigned magic, unsigned hashlow)
 552 {
 553         tdb_off_t off, ftable_off;
 554         tdb_off_t start_b, b, ftable;
 555         bool wrapped = false;
 556
 557         /* If they are growing, add 50% to get to higher bucket. */
 558         if (want_extra)
 559                 start_b = size_to_bucket(adjust_size(keylen,
 560                                                      datalen + datalen / 2));
 561         else
 562                 start_b = size_to_bucket(adjust_size(keylen, datalen));
 563
 564         ftable_off = tdb->ftable_off;
 565         ftable = tdb->ftable;
 566         while (!wrapped || ftable_off != tdb->ftable_off) {
 567                 /* Start at exact size bucket, and search up... */
 568                 for (b = find_free_head(tdb, ftable_off, start_b);
 569                      b < TDB_FREE_BUCKETS;
 570                      b = find_free_head(tdb, ftable_off, b + 1)) {
 571                         /* Try getting one from list. */
 572                         off = lock_and_alloc(tdb, ftable_off,
 573                                              b, keylen, datalen, want_extra,
 574                                              magic, hashlow);
 575                         if (off == TDB_OFF_ERR)
 576                                 return TDB_OFF_ERR;
 577                         if (off != 0) {
 578                                 if (b == start_b)
 579                                         add_stat(tdb, alloc_bucket_exact, 1);
 580                                 if (b == TDB_FREE_BUCKETS - 1)
 581                                         add_stat(tdb, alloc_bucket_max, 1);
 582                                 /* Worked?  Stay using this list. */
 583                                 tdb->ftable_off = ftable_off;
 584                                 tdb->ftable = ftable;
 585                                 return off;
 586                         }
 587                         /* Didn't work.  Try next bucket. */
 588                 }
 589
 590                 if (TDB_OFF_IS_ERR(b)) {
 591                         tdb->ecode = b;
 592                         return 0;
 593                 }
 594
 595                 /* Hmm, try next table. */
 596                 ftable_off = next_ftable(tdb, ftable_off);
 597                 if (TDB_OFF_IS_ERR(ftable_off)) {
 598                         tdb->ecode = ftable_off;
 599                         return 0;
 600                 }
 601                 ftable++;
 602
 603                 if (ftable_off == 0) {
 604                         wrapped = true;
 605                         ftable_off = first_ftable(tdb);
 606                         if (TDB_OFF_IS_ERR(ftable_off)) {
 607                                 tdb->ecode = ftable_off;
 608                                 return 0;
 609                         }
 610                         ftable = 0;
 611                 }
 612         }
 613
 614         return 0;
 615 }
 616
 617 int set_header(struct tdb_context *tdb,
 618                struct tdb_used_record *rec,
 619                unsigned magic, uint64_t keylen, uint64_t datalen,
 620                uint64_t actuallen, unsigned hashlow)
 621 {
 622         uint64_t keybits = (fls64(keylen) + 1) / 2;
 623
 624         /* Use bottom bits of hash, so it's independent of hash table size. */
 625         rec->magic_and_meta = (hashlow & ((1 << 11)-1))
 626                 | ((actuallen - (keylen + datalen)) << 11)
 627                 | (keybits << 43)
 628                 | ((uint64_t)magic << 48);
 629         rec->key_and_data_len = (keylen | (datalen << (keybits*2)));
 630
 631         /* Encoding can fail on big values. */
 632         if (rec_key_length(rec) != keylen
 633             || rec_data_length(rec) != datalen
 634             || rec_extra_padding(rec) != actuallen - (keylen + datalen)) {
 635                 tdb_logerr(tdb, TDB_ERR_IO, TDB_LOG_ERROR,
 636                          "Could not encode k=%llu,d=%llu,a=%llu",
 637                          (long long)keylen, (long long)datalen,
 638                          (long long)actuallen);
 639                 return -1;
 640         }
 641         return 0;
 642 }
 643
 644 /* Expand the database. */
 645 static int tdb_expand(struct tdb_context *tdb, tdb_len_t size)
 646 {
 647         uint64_t old_size;
 648         tdb_len_t wanted;
 649         enum TDB_ERROR ecode;
 650
 651         /* We need room for the record header too. */
 652         wanted = sizeof(struct tdb_used_record) + size;
 653
 654         /* Need to hold a hash lock to expand DB: transactions rely on it. */
 655         if (!(tdb->flags & TDB_NOLOCK)
 656             && !tdb->allrecord_lock.count && !tdb_has_hash_locks(tdb)) {
 657                 tdb_logerr(tdb, TDB_ERR_LOCK, TDB_LOG_ERROR,
 658                            "tdb_expand: must hold lock during expand");
 659                 return -1;
 660         }
 661
 662         /* always make room for at least 100 more records, and at
 663            least 25% more space. */
 664         if (size * TDB_EXTENSION_FACTOR > tdb->map_size / 4)
 665                 wanted = size * TDB_EXTENSION_FACTOR;
 666         else
 667                 wanted = tdb->map_size / 4;
 668         wanted = adjust_size(0, wanted);
 669
 670         /* Only one person can expand file at a time. */
 671         ecode = tdb_lock_expand(tdb, F_WRLCK);
 672         if (ecode != TDB_SUCCESS) {
 673                 tdb->ecode = ecode;
 674                 return -1;
 675         }
 676
 677         /* Someone else may have expanded the file, so retry. */
 678         old_size = tdb->map_size;
 679         tdb->methods->oob(tdb, tdb->map_size + 1, true);
 680         if (tdb->map_size != old_size) {
 681                 tdb_unlock_expand(tdb, F_WRLCK);
 682                 return 0;
 683         }
 684
 685         ecode = tdb->methods->expand_file(tdb, wanted);
 686         if (ecode != TDB_SUCCESS) {
 687                 tdb->ecode = ecode;
 688                 tdb_unlock_expand(tdb, F_WRLCK);
 689                 return -1;
 690         }
 691
 692         /* We need to drop this lock before adding free record. */
 693         tdb_unlock_expand(tdb, F_WRLCK);
 694
 695         add_stat(tdb, expands, 1);
 696         return add_free_record(tdb, old_size, wanted);
 697 }
 698
 699 /* This won't fail: it will expand the database if it has to. */
 700 tdb_off_t alloc(struct tdb_context *tdb, size_t keylen, size_t datalen,
 701                 uint64_t hash, unsigned magic, bool growing)
 702 {
 703         tdb_off_t off;
 704
 705         /* We can't hold pointers during this: we could unmap! */
 706         assert(!tdb->direct_access);
 707
 708         for (;;) {
 709                 off = get_free(tdb, keylen, datalen, growing, magic, hash);
 710                 if (likely(off != 0))
 711                         break;
 712
 713                 if (tdb_expand(tdb, adjust_size(keylen, datalen)))
 714                         return TDB_OFF_ERR;
 715         }
 716
 717         return off;
 718 }