root/lib/tdb/common/tdb.c

/* [<][>][^][v][top][bottom][index][help] */

DEFINITIONS

This source file includes following definitions.
  1. tdb_increment_seqnum_nonblock
  2. tdb_increment_seqnum
  3. tdb_key_compare
  4. tdb_find
  5. tdb_find_lock_hash
  6. tdb_update_hash
  7. tdb_fetch
  8. tdb_parse_record
  9. tdb_exists_hash
  10. tdb_exists
  11. tdb_do_delete
  12. tdb_count_dead
  13. tdb_purge_dead
  14. tdb_delete_hash
  15. tdb_delete
  16. tdb_find_dead
  17. tdb_store
  18. tdb_append
  19. tdb_name
  20. tdb_fd
  21. tdb_log_fn
  22. tdb_get_seqnum
  23. tdb_hash_size
  24. tdb_map_size
  25. tdb_get_flags
  26. tdb_add_flags
  27. tdb_remove_flags
  28. tdb_enable_seqnum
  29. tdb_free_region
  30. tdb_wipe_all
  31. repack_traverse
  32. tdb_repack

   1  /* 
   2    Unix SMB/CIFS implementation.
   3 
   4    trivial database library
   5 
   6    Copyright (C) Andrew Tridgell              1999-2005
   7    Copyright (C) Paul `Rusty' Russell              2000
   8    Copyright (C) Jeremy Allison                    2000-2003
   9    
  10      ** NOTE! The following LGPL license applies to the tdb
  11      ** library. This does NOT imply that all of Samba is released
  12      ** under the LGPL
  13    
  14    This library is free software; you can redistribute it and/or
  15    modify it under the terms of the GNU Lesser General Public
  16    License as published by the Free Software Foundation; either
  17    version 3 of the License, or (at your option) any later version.
  18 
  19    This library is distributed in the hope that it will be useful,
  20    but WITHOUT ANY WARRANTY; without even the implied warranty of
  21    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  22    Lesser General Public License for more details.
  23 
  24    You should have received a copy of the GNU Lesser General Public
  25    License along with this library; if not, see <http://www.gnu.org/licenses/>.
  26 */
  27 
  28 #include "tdb_private.h"
  29 
  30 TDB_DATA tdb_null;
  31 
  32 /*
  33   non-blocking increment of the tdb sequence number if the tdb has been opened using
  34   the TDB_SEQNUM flag
  35 */
  36 void tdb_increment_seqnum_nonblock(struct tdb_context *tdb)
     /* [<][>][^][v][top][bottom][index][help] */
  37 {
  38         tdb_off_t seqnum=0;
  39         
  40         if (!(tdb->flags & TDB_SEQNUM)) {
  41                 return;
  42         }
  43 
  44         /* we ignore errors from this, as we have no sane way of
  45            dealing with them.
  46         */
  47         tdb_ofs_read(tdb, TDB_SEQNUM_OFS, &seqnum);
  48         seqnum++;
  49         tdb_ofs_write(tdb, TDB_SEQNUM_OFS, &seqnum);
  50 }
  51 
  52 /*
  53   increment the tdb sequence number if the tdb has been opened using
  54   the TDB_SEQNUM flag
  55 */
  56 static void tdb_increment_seqnum(struct tdb_context *tdb)
     /* [<][>][^][v][top][bottom][index][help] */
  57 {
  58         if (!(tdb->flags & TDB_SEQNUM)) {
  59                 return;
  60         }
  61 
  62         if (tdb_brlock(tdb, TDB_SEQNUM_OFS, F_WRLCK, F_SETLKW, 1, 1) != 0) {
  63                 return;
  64         }
  65 
  66         tdb_increment_seqnum_nonblock(tdb);
  67 
  68         tdb_brlock(tdb, TDB_SEQNUM_OFS, F_UNLCK, F_SETLKW, 1, 1);
  69 }
  70 
  71 static int tdb_key_compare(TDB_DATA key, TDB_DATA data, void *private_data)
     /* [<][>][^][v][top][bottom][index][help] */
  72 {
  73         return memcmp(data.dptr, key.dptr, data.dsize);
  74 }
  75 
  76 /* Returns 0 on fail.  On success, return offset of record, and fills
  77    in rec */
  78 static tdb_off_t tdb_find(struct tdb_context *tdb, TDB_DATA key, uint32_t hash,
     /* [<][>][^][v][top][bottom][index][help] */
  79                         struct list_struct *r)
  80 {
  81         tdb_off_t rec_ptr;
  82         
  83         /* read in the hash top */
  84         if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
  85                 return 0;
  86 
  87         /* keep looking until we find the right record */
  88         while (rec_ptr) {
  89                 if (tdb_rec_read(tdb, rec_ptr, r) == -1)
  90                         return 0;
  91 
  92                 if (!TDB_DEAD(r) && hash==r->full_hash
  93                     && key.dsize==r->key_len
  94                     && tdb_parse_data(tdb, key, rec_ptr + sizeof(*r),
  95                                       r->key_len, tdb_key_compare,
  96                                       NULL) == 0) {
  97                         return rec_ptr;
  98                 }
  99                 /* detect tight infinite loop */
 100                 if (rec_ptr == r->next) {
 101                         TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_find: loop detected.\n"));
 102                         return TDB_ERRCODE(TDB_ERR_CORRUPT, 0);
 103                 }
 104                 rec_ptr = r->next;
 105         }
 106         return TDB_ERRCODE(TDB_ERR_NOEXIST, 0);
 107 }
 108 
 109 /* As tdb_find, but if you succeed, keep the lock */
 110 tdb_off_t tdb_find_lock_hash(struct tdb_context *tdb, TDB_DATA key, uint32_t hash, int locktype,
     /* [<][>][^][v][top][bottom][index][help] */
 111                            struct list_struct *rec)
 112 {
 113         uint32_t rec_ptr;
 114 
 115         if (tdb_lock(tdb, BUCKET(hash), locktype) == -1)
 116                 return 0;
 117         if (!(rec_ptr = tdb_find(tdb, key, hash, rec)))
 118                 tdb_unlock(tdb, BUCKET(hash), locktype);
 119         return rec_ptr;
 120 }
 121 
 122 
 123 /* update an entry in place - this only works if the new data size
 124    is <= the old data size and the key exists.
 125    on failure return -1.
 126 */
 127 static int tdb_update_hash(struct tdb_context *tdb, TDB_DATA key, uint32_t hash, TDB_DATA dbuf)
     /* [<][>][^][v][top][bottom][index][help] */
 128 {
 129         struct list_struct rec;
 130         tdb_off_t rec_ptr;
 131 
 132         /* find entry */
 133         if (!(rec_ptr = tdb_find(tdb, key, hash, &rec)))
 134                 return -1;
 135 
 136         /* must be long enough key, data and tailer */
 137         if (rec.rec_len < key.dsize + dbuf.dsize + sizeof(tdb_off_t)) {
 138                 tdb->ecode = TDB_SUCCESS; /* Not really an error */
 139                 return -1;
 140         }
 141 
 142         if (tdb->methods->tdb_write(tdb, rec_ptr + sizeof(rec) + rec.key_len,
 143                       dbuf.dptr, dbuf.dsize) == -1)
 144                 return -1;
 145 
 146         if (dbuf.dsize != rec.data_len) {
 147                 /* update size */
 148                 rec.data_len = dbuf.dsize;
 149                 return tdb_rec_write(tdb, rec_ptr, &rec);
 150         }
 151  
 152         return 0;
 153 }
 154 
 155 /* find an entry in the database given a key */
 156 /* If an entry doesn't exist tdb_err will be set to
 157  * TDB_ERR_NOEXIST. If a key has no data attached
 158  * then the TDB_DATA will have zero length but
 159  * a non-zero pointer
 160  */
 161 TDB_DATA tdb_fetch(struct tdb_context *tdb, TDB_DATA key)
     /* [<][>][^][v][top][bottom][index][help] */
 162 {
 163         tdb_off_t rec_ptr;
 164         struct list_struct rec;
 165         TDB_DATA ret;
 166         uint32_t hash;
 167 
 168         /* find which hash bucket it is in */
 169         hash = tdb->hash_fn(&key);
 170         if (!(rec_ptr = tdb_find_lock_hash(tdb,key,hash,F_RDLCK,&rec)))
 171                 return tdb_null;
 172 
 173         ret.dptr = tdb_alloc_read(tdb, rec_ptr + sizeof(rec) + rec.key_len,
 174                                   rec.data_len);
 175         ret.dsize = rec.data_len;
 176         tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
 177         return ret;
 178 }
 179 
 180 /*
 181  * Find an entry in the database and hand the record's data to a parsing
 182  * function. The parsing function is executed under the chain read lock, so it
 183  * should be fast and should not block on other syscalls.
 184  *
 185  * DONT CALL OTHER TDB CALLS FROM THE PARSER, THIS MIGHT LEAD TO SEGFAULTS.
 186  *
 187  * For mmapped tdb's that do not have a transaction open it points the parsing
 188  * function directly at the mmap area, it avoids the malloc/memcpy in this
 189  * case. If a transaction is open or no mmap is available, it has to do
 190  * malloc/read/parse/free.
 191  *
 192  * This is interesting for all readers of potentially large data structures in
 193  * the tdb records, ldb indexes being one example.
 194  */
 195 
 196 int tdb_parse_record(struct tdb_context *tdb, TDB_DATA key,
     /* [<][>][^][v][top][bottom][index][help] */
 197                      int (*parser)(TDB_DATA key, TDB_DATA data,
 198                                    void *private_data),
 199                      void *private_data)
 200 {
 201         tdb_off_t rec_ptr;
 202         struct list_struct rec;
 203         int ret;
 204         uint32_t hash;
 205 
 206         /* find which hash bucket it is in */
 207         hash = tdb->hash_fn(&key);
 208 
 209         if (!(rec_ptr = tdb_find_lock_hash(tdb,key,hash,F_RDLCK,&rec))) {
 210                 return TDB_ERRCODE(TDB_ERR_NOEXIST, 0);
 211         }
 212 
 213         ret = tdb_parse_data(tdb, key, rec_ptr + sizeof(rec) + rec.key_len,
 214                              rec.data_len, parser, private_data);
 215 
 216         tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
 217 
 218         return ret;
 219 }
 220 
 221 /* check if an entry in the database exists 
 222 
 223    note that 1 is returned if the key is found and 0 is returned if not found
 224    this doesn't match the conventions in the rest of this module, but is
 225    compatible with gdbm
 226 */
 227 static int tdb_exists_hash(struct tdb_context *tdb, TDB_DATA key, uint32_t hash)
     /* [<][>][^][v][top][bottom][index][help] */
 228 {
 229         struct list_struct rec;
 230         
 231         if (tdb_find_lock_hash(tdb, key, hash, F_RDLCK, &rec) == 0)
 232                 return 0;
 233         tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
 234         return 1;
 235 }
 236 
 237 int tdb_exists(struct tdb_context *tdb, TDB_DATA key)
     /* [<][>][^][v][top][bottom][index][help] */
 238 {
 239         uint32_t hash = tdb->hash_fn(&key);
 240         return tdb_exists_hash(tdb, key, hash);
 241 }
 242 
 243 /* actually delete an entry in the database given the offset */
 244 int tdb_do_delete(struct tdb_context *tdb, tdb_off_t rec_ptr, struct list_struct *rec)
     /* [<][>][^][v][top][bottom][index][help] */
 245 {
 246         tdb_off_t last_ptr, i;
 247         struct list_struct lastrec;
 248 
 249         if (tdb->read_only || tdb->traverse_read) return -1;
 250 
 251         if (((tdb->traverse_write != 0) && (!TDB_DEAD(rec))) ||
 252             tdb_write_lock_record(tdb, rec_ptr) == -1) {
 253                 /* Someone traversing here: mark it as dead */
 254                 rec->magic = TDB_DEAD_MAGIC;
 255                 return tdb_rec_write(tdb, rec_ptr, rec);
 256         }
 257         if (tdb_write_unlock_record(tdb, rec_ptr) != 0)
 258                 return -1;
 259 
 260         /* find previous record in hash chain */
 261         if (tdb_ofs_read(tdb, TDB_HASH_TOP(rec->full_hash), &i) == -1)
 262                 return -1;
 263         for (last_ptr = 0; i != rec_ptr; last_ptr = i, i = lastrec.next)
 264                 if (tdb_rec_read(tdb, i, &lastrec) == -1)
 265                         return -1;
 266 
 267         /* unlink it: next ptr is at start of record. */
 268         if (last_ptr == 0)
 269                 last_ptr = TDB_HASH_TOP(rec->full_hash);
 270         if (tdb_ofs_write(tdb, last_ptr, &rec->next) == -1)
 271                 return -1;
 272 
 273         /* recover the space */
 274         if (tdb_free(tdb, rec_ptr, rec) == -1)
 275                 return -1;
 276         return 0;
 277 }
 278 
 279 static int tdb_count_dead(struct tdb_context *tdb, uint32_t hash)
     /* [<][>][^][v][top][bottom][index][help] */
 280 {
 281         int res = 0;
 282         tdb_off_t rec_ptr;
 283         struct list_struct rec;
 284         
 285         /* read in the hash top */
 286         if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
 287                 return 0;
 288 
 289         while (rec_ptr) {
 290                 if (tdb_rec_read(tdb, rec_ptr, &rec) == -1)
 291                         return 0;
 292 
 293                 if (rec.magic == TDB_DEAD_MAGIC) {
 294                         res += 1;
 295                 }
 296                 rec_ptr = rec.next;
 297         }
 298         return res;
 299 }
 300 
 301 /*
 302  * Purge all DEAD records from a hash chain
 303  */
 304 static int tdb_purge_dead(struct tdb_context *tdb, uint32_t hash)
     /* [<][>][^][v][top][bottom][index][help] */
 305 {
 306         int res = -1;
 307         struct list_struct rec;
 308         tdb_off_t rec_ptr;
 309 
 310         if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
 311                 return -1;
 312         }
 313         
 314         /* read in the hash top */
 315         if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
 316                 goto fail;
 317 
 318         while (rec_ptr) {
 319                 tdb_off_t next;
 320 
 321                 if (tdb_rec_read(tdb, rec_ptr, &rec) == -1) {
 322                         goto fail;
 323                 }
 324 
 325                 next = rec.next;
 326 
 327                 if (rec.magic == TDB_DEAD_MAGIC
 328                     && tdb_do_delete(tdb, rec_ptr, &rec) == -1) {
 329                         goto fail;
 330                 }
 331                 rec_ptr = next;
 332         }
 333         res = 0;
 334  fail:
 335         tdb_unlock(tdb, -1, F_WRLCK);
 336         return res;
 337 }
 338 
 339 /* delete an entry in the database given a key */
 340 static int tdb_delete_hash(struct tdb_context *tdb, TDB_DATA key, uint32_t hash)
     /* [<][>][^][v][top][bottom][index][help] */
 341 {
 342         tdb_off_t rec_ptr;
 343         struct list_struct rec;
 344         int ret;
 345 
 346         if (tdb->max_dead_records != 0) {
 347 
 348                 /*
 349                  * Allow for some dead records per hash chain, mainly for
 350                  * tdb's with a very high create/delete rate like locking.tdb.
 351                  */
 352 
 353                 if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
 354                         return -1;
 355 
 356                 if (tdb_count_dead(tdb, hash) >= tdb->max_dead_records) {
 357                         /*
 358                          * Don't let the per-chain freelist grow too large,
 359                          * delete all existing dead records
 360                          */
 361                         tdb_purge_dead(tdb, hash);
 362                 }
 363 
 364                 if (!(rec_ptr = tdb_find(tdb, key, hash, &rec))) {
 365                         tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
 366                         return -1;
 367                 }
 368 
 369                 /*
 370                  * Just mark the record as dead.
 371                  */
 372                 rec.magic = TDB_DEAD_MAGIC;
 373                 ret = tdb_rec_write(tdb, rec_ptr, &rec);
 374         }
 375         else {
 376                 if (!(rec_ptr = tdb_find_lock_hash(tdb, key, hash, F_WRLCK,
 377                                                    &rec)))
 378                         return -1;
 379 
 380                 ret = tdb_do_delete(tdb, rec_ptr, &rec);
 381         }
 382 
 383         if (ret == 0) {
 384                 tdb_increment_seqnum(tdb);
 385         }
 386 
 387         if (tdb_unlock(tdb, BUCKET(rec.full_hash), F_WRLCK) != 0)
 388                 TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_delete: WARNING tdb_unlock failed!\n"));
 389         return ret;
 390 }
 391 
 392 int tdb_delete(struct tdb_context *tdb, TDB_DATA key)
     /* [<][>][^][v][top][bottom][index][help] */
 393 {
 394         uint32_t hash = tdb->hash_fn(&key);
 395         return tdb_delete_hash(tdb, key, hash);
 396 }
 397 
 398 /*
 399  * See if we have a dead record around with enough space
 400  */
 401 static tdb_off_t tdb_find_dead(struct tdb_context *tdb, uint32_t hash,
     /* [<][>][^][v][top][bottom][index][help] */
 402                                struct list_struct *r, tdb_len_t length)
 403 {
 404         tdb_off_t rec_ptr;
 405         
 406         /* read in the hash top */
 407         if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1)
 408                 return 0;
 409 
 410         /* keep looking until we find the right record */
 411         while (rec_ptr) {
 412                 if (tdb_rec_read(tdb, rec_ptr, r) == -1)
 413                         return 0;
 414 
 415                 if (TDB_DEAD(r) && r->rec_len >= length) {
 416                         /*
 417                          * First fit for simple coding, TODO: change to best
 418                          * fit
 419                          */
 420                         return rec_ptr;
 421                 }
 422                 rec_ptr = r->next;
 423         }
 424         return 0;
 425 }
 426 
 427 /* store an element in the database, replacing any existing element
 428    with the same key 
 429 
 430    return 0 on success, -1 on failure
 431 */
 432 int tdb_store(struct tdb_context *tdb, TDB_DATA key, TDB_DATA dbuf, int flag)
     /* [<][>][^][v][top][bottom][index][help] */
 433 {
 434         struct list_struct rec;
 435         uint32_t hash;
 436         tdb_off_t rec_ptr;
 437         char *p = NULL;
 438         int ret = -1;
 439 
 440         if (tdb->read_only || tdb->traverse_read) {
 441                 tdb->ecode = TDB_ERR_RDONLY;
 442                 return -1;
 443         }
 444 
 445         /* find which hash bucket it is in */
 446         hash = tdb->hash_fn(&key);
 447         if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
 448                 return -1;
 449 
 450         /* check for it existing, on insert. */
 451         if (flag == TDB_INSERT) {
 452                 if (tdb_exists_hash(tdb, key, hash)) {
 453                         tdb->ecode = TDB_ERR_EXISTS;
 454                         goto fail;
 455                 }
 456         } else {
 457                 /* first try in-place update, on modify or replace. */
 458                 if (tdb_update_hash(tdb, key, hash, dbuf) == 0) {
 459                         goto done;
 460                 }
 461                 if (tdb->ecode == TDB_ERR_NOEXIST &&
 462                     flag == TDB_MODIFY) {
 463                         /* if the record doesn't exist and we are in TDB_MODIFY mode then
 464                          we should fail the store */
 465                         goto fail;
 466                 }
 467         }
 468         /* reset the error code potentially set by the tdb_update() */
 469         tdb->ecode = TDB_SUCCESS;
 470 
 471         /* delete any existing record - if it doesn't exist we don't
 472            care.  Doing this first reduces fragmentation, and avoids
 473            coalescing with `allocated' block before it's updated. */
 474         if (flag != TDB_INSERT)
 475                 tdb_delete_hash(tdb, key, hash);
 476 
 477         /* Copy key+value *before* allocating free space in case malloc
 478            fails and we are left with a dead spot in the tdb. */
 479 
 480         if (!(p = (char *)malloc(key.dsize + dbuf.dsize))) {
 481                 tdb->ecode = TDB_ERR_OOM;
 482                 goto fail;
 483         }
 484 
 485         memcpy(p, key.dptr, key.dsize);
 486         if (dbuf.dsize)
 487                 memcpy(p+key.dsize, dbuf.dptr, dbuf.dsize);
 488 
 489         if (tdb->max_dead_records != 0) {
 490                 /*
 491                  * Allow for some dead records per hash chain, look if we can
 492                  * find one that can hold the new record. We need enough space
 493                  * for key, data and tailer. If we find one, we don't have to
 494                  * consult the central freelist.
 495                  */
 496                 rec_ptr = tdb_find_dead(
 497                         tdb, hash, &rec,
 498                         key.dsize + dbuf.dsize + sizeof(tdb_off_t));
 499 
 500                 if (rec_ptr != 0) {
 501                         rec.key_len = key.dsize;
 502                         rec.data_len = dbuf.dsize;
 503                         rec.full_hash = hash;
 504                         rec.magic = TDB_MAGIC;
 505                         if (tdb_rec_write(tdb, rec_ptr, &rec) == -1
 506                             || tdb->methods->tdb_write(
 507                                     tdb, rec_ptr + sizeof(rec),
 508                                     p, key.dsize + dbuf.dsize) == -1) {
 509                                 goto fail;
 510                         }
 511                         goto done;
 512                 }
 513         }
 514 
 515         /*
 516          * We have to allocate some space from the freelist, so this means we
 517          * have to lock it. Use the chance to purge all the DEAD records from
 518          * the hash chain under the freelist lock.
 519          */
 520 
 521         if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
 522                 goto fail;
 523         }
 524 
 525         if ((tdb->max_dead_records != 0)
 526             && (tdb_purge_dead(tdb, hash) == -1)) {
 527                 tdb_unlock(tdb, -1, F_WRLCK);
 528                 goto fail;
 529         }
 530 
 531         /* we have to allocate some space */
 532         rec_ptr = tdb_allocate(tdb, key.dsize + dbuf.dsize, &rec);
 533 
 534         tdb_unlock(tdb, -1, F_WRLCK);
 535 
 536         if (rec_ptr == 0) {
 537                 goto fail;
 538         }
 539 
 540         /* Read hash top into next ptr */
 541         if (tdb_ofs_read(tdb, TDB_HASH_TOP(hash), &rec.next) == -1)
 542                 goto fail;
 543 
 544         rec.key_len = key.dsize;
 545         rec.data_len = dbuf.dsize;
 546         rec.full_hash = hash;
 547         rec.magic = TDB_MAGIC;
 548 
 549         /* write out and point the top of the hash chain at it */
 550         if (tdb_rec_write(tdb, rec_ptr, &rec) == -1
 551             || tdb->methods->tdb_write(tdb, rec_ptr+sizeof(rec), p, key.dsize+dbuf.dsize)==-1
 552             || tdb_ofs_write(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1) {
 553                 /* Need to tdb_unallocate() here */
 554                 goto fail;
 555         }
 556 
 557  done:
 558         ret = 0;
 559  fail:
 560         if (ret == 0) {
 561                 tdb_increment_seqnum(tdb);
 562         }
 563 
 564         SAFE_FREE(p); 
 565         tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
 566         return ret;
 567 }
 568 
 569 
 570 /* Append to an entry. Create if not exist. */
 571 int tdb_append(struct tdb_context *tdb, TDB_DATA key, TDB_DATA new_dbuf)
     /* [<][>][^][v][top][bottom][index][help] */
 572 {
 573         uint32_t hash;
 574         TDB_DATA dbuf;
 575         int ret = -1;
 576 
 577         /* find which hash bucket it is in */
 578         hash = tdb->hash_fn(&key);
 579         if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1)
 580                 return -1;
 581 
 582         dbuf = tdb_fetch(tdb, key);
 583 
 584         if (dbuf.dptr == NULL) {
 585                 dbuf.dptr = (unsigned char *)malloc(new_dbuf.dsize);
 586         } else {
 587                 unsigned char *new_dptr = (unsigned char *)realloc(dbuf.dptr,
 588                                                      dbuf.dsize + new_dbuf.dsize);
 589                 if (new_dptr == NULL) {
 590                         free(dbuf.dptr);
 591                 }
 592                 dbuf.dptr = new_dptr;
 593         }
 594 
 595         if (dbuf.dptr == NULL) {
 596                 tdb->ecode = TDB_ERR_OOM;
 597                 goto failed;
 598         }
 599 
 600         memcpy(dbuf.dptr + dbuf.dsize, new_dbuf.dptr, new_dbuf.dsize);
 601         dbuf.dsize += new_dbuf.dsize;
 602 
 603         ret = tdb_store(tdb, key, dbuf, 0);
 604         
 605 failed:
 606         tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
 607         SAFE_FREE(dbuf.dptr);
 608         return ret;
 609 }
 610 
 611 
 612 /*
 613   return the name of the current tdb file
 614   useful for external logging functions
 615 */
 616 const char *tdb_name(struct tdb_context *tdb)
     /* [<][>][^][v][top][bottom][index][help] */
 617 {
 618         return tdb->name;
 619 }
 620 
 621 /*
 622   return the underlying file descriptor being used by tdb, or -1
 623   useful for external routines that want to check the device/inode
 624   of the fd
 625 */
 626 int tdb_fd(struct tdb_context *tdb)
     /* [<][>][^][v][top][bottom][index][help] */
 627 {
 628         return tdb->fd;
 629 }
 630 
 631 /*
 632   return the current logging function
 633   useful for external tdb routines that wish to log tdb errors
 634 */
 635 tdb_log_func tdb_log_fn(struct tdb_context *tdb)
     /* [<][>][^][v][top][bottom][index][help] */
 636 {
 637         return tdb->log.log_fn;
 638 }
 639 
 640 
 641 /*
 642   get the tdb sequence number. Only makes sense if the writers opened
 643   with TDB_SEQNUM set. Note that this sequence number will wrap quite
 644   quickly, so it should only be used for a 'has something changed'
 645   test, not for code that relies on the count of the number of changes
 646   made. If you want a counter then use a tdb record.
 647 
 648   The aim of this sequence number is to allow for a very lightweight
 649   test of a possible tdb change.
 650 */
 651 int tdb_get_seqnum(struct tdb_context *tdb)
     /* [<][>][^][v][top][bottom][index][help] */
 652 {
 653         tdb_off_t seqnum=0;
 654 
 655         tdb_ofs_read(tdb, TDB_SEQNUM_OFS, &seqnum);
 656         return seqnum;
 657 }
 658 
 659 int tdb_hash_size(struct tdb_context *tdb)
     /* [<][>][^][v][top][bottom][index][help] */
 660 {
 661         return tdb->header.hash_size;
 662 }
 663 
 664 size_t tdb_map_size(struct tdb_context *tdb)
     /* [<][>][^][v][top][bottom][index][help] */
 665 {
 666         return tdb->map_size;
 667 }
 668 
 669 int tdb_get_flags(struct tdb_context *tdb)
     /* [<][>][^][v][top][bottom][index][help] */
 670 {
 671         return tdb->flags;
 672 }
 673 
 674 void tdb_add_flags(struct tdb_context *tdb, unsigned flags)
     /* [<][>][^][v][top][bottom][index][help] */
 675 {
 676         tdb->flags |= flags;
 677 }
 678 
 679 void tdb_remove_flags(struct tdb_context *tdb, unsigned flags)
     /* [<][>][^][v][top][bottom][index][help] */
 680 {
 681         tdb->flags &= ~flags;
 682 }
 683 
 684 
 685 /*
 686   enable sequence number handling on an open tdb
 687 */
 688 void tdb_enable_seqnum(struct tdb_context *tdb)
     /* [<][>][^][v][top][bottom][index][help] */
 689 {
 690         tdb->flags |= TDB_SEQNUM;
 691 }
 692 
 693 
 694 /*
 695   add a region of the file to the freelist. Length is the size of the region in bytes, 
 696   which includes the free list header that needs to be added
 697  */
 698 static int tdb_free_region(struct tdb_context *tdb, tdb_off_t offset, ssize_t length)
     /* [<][>][^][v][top][bottom][index][help] */
 699 {
 700         struct list_struct rec;
 701         if (length <= sizeof(rec)) {
 702                 /* the region is not worth adding */
 703                 return 0;
 704         }
 705         if (length + offset > tdb->map_size) {
 706                 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_free_region: adding region beyond end of file\n"));
 707                 return -1;              
 708         }
 709         memset(&rec,'\0',sizeof(rec));
 710         rec.rec_len = length - sizeof(rec);
 711         if (tdb_free(tdb, offset, &rec) == -1) {
 712                 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_free_region: failed to add free record\n"));
 713                 return -1;
 714         }
 715         return 0;
 716 }
 717 
 718 /*
 719   wipe the entire database, deleting all records. This can be done
 720   very fast by using a global lock. The entire data portion of the
 721   file becomes a single entry in the freelist.
 722 
 723   This code carefully steps around the recovery area, leaving it alone
 724  */
 725 int tdb_wipe_all(struct tdb_context *tdb)
     /* [<][>][^][v][top][bottom][index][help] */
 726 {
 727         int i;
 728         tdb_off_t offset = 0;
 729         ssize_t data_len;
 730         tdb_off_t recovery_head;
 731         tdb_len_t recovery_size = 0;
 732 
 733         if (tdb_lockall(tdb) != 0) {
 734                 return -1;
 735         }
 736 
 737         /* see if the tdb has a recovery area, and remember its size
 738            if so. We don't want to lose this as otherwise each
 739            tdb_wipe_all() in a transaction will increase the size of
 740            the tdb by the size of the recovery area */
 741         if (tdb_ofs_read(tdb, TDB_RECOVERY_HEAD, &recovery_head) == -1) {
 742                 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_wipe_all: failed to read recovery head\n"));
 743                 goto failed;
 744         }
 745 
 746         if (recovery_head != 0) {
 747                 struct list_struct rec;
 748                 if (tdb->methods->tdb_read(tdb, recovery_head, &rec, sizeof(rec), DOCONV()) == -1) {
 749                         TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_wipe_all: failed to read recovery record\n"));
 750                         return -1;
 751                 }       
 752                 recovery_size = rec.rec_len + sizeof(rec);
 753         }
 754 
 755         /* wipe the hashes */
 756         for (i=0;i<tdb->header.hash_size;i++) {
 757                 if (tdb_ofs_write(tdb, TDB_HASH_TOP(i), &offset) == -1) {
 758                         TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_wipe_all: failed to write hash %d\n", i));
 759                         goto failed;
 760                 }
 761         }
 762 
 763         /* wipe the freelist */
 764         if (tdb_ofs_write(tdb, FREELIST_TOP, &offset) == -1) {
 765                 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_wipe_all: failed to write freelist\n"));
 766                 goto failed;
 767         }
 768 
 769         /* add all the rest of the file to the freelist, possibly leaving a gap 
 770            for the recovery area */
 771         if (recovery_size == 0) {
 772                 /* the simple case - the whole file can be used as a freelist */
 773                 data_len = (tdb->map_size - TDB_DATA_START(tdb->header.hash_size));
 774                 if (tdb_free_region(tdb, TDB_DATA_START(tdb->header.hash_size), data_len) != 0) {
 775                         goto failed;
 776                 }
 777         } else {
 778                 /* we need to add two freelist entries - one on either
 779                    side of the recovery area 
 780 
 781                    Note that we cannot shift the recovery area during
 782                    this operation. Only the transaction.c code may
 783                    move the recovery area or we risk subtle data
 784                    corruption
 785                 */
 786                 data_len = (recovery_head - TDB_DATA_START(tdb->header.hash_size));
 787                 if (tdb_free_region(tdb, TDB_DATA_START(tdb->header.hash_size), data_len) != 0) {
 788                         goto failed;
 789                 }
 790                 /* and the 2nd free list entry after the recovery area - if any */
 791                 data_len = tdb->map_size - (recovery_head+recovery_size);
 792                 if (tdb_free_region(tdb, recovery_head+recovery_size, data_len) != 0) {
 793                         goto failed;
 794                 }
 795         }
 796 
 797         if (tdb_unlockall(tdb) != 0) {
 798                 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_wipe_all: failed to unlock\n"));
 799                 goto failed;
 800         }
 801 
 802         return 0;
 803 
 804 failed:
 805         tdb_unlockall(tdb);
 806         return -1;
 807 }
 808 
 809 struct traverse_state {
 810         bool error;
 811         struct tdb_context *dest_db;
 812 };
 813 
 814 /*
 815   traverse function for repacking
 816  */
 817 static int repack_traverse(struct tdb_context *tdb, TDB_DATA key, TDB_DATA data, void *private)
     /* [<][>][^][v][top][bottom][index][help] */
 818 {
 819         struct traverse_state *state = (struct traverse_state *)private;
 820         if (tdb_store(state->dest_db, key, data, TDB_INSERT) != 0) {
 821                 state->error = true;
 822                 return -1;
 823         }
 824         return 0;
 825 }
 826 
 827 /*
 828   repack a tdb
 829  */
 830 int tdb_repack(struct tdb_context *tdb)
     /* [<][>][^][v][top][bottom][index][help] */
 831 {
 832         struct tdb_context *tmp_db;
 833         struct traverse_state state;
 834 
 835         if (tdb_transaction_start(tdb) != 0) {
 836                 TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Failed to start transaction\n"));
 837                 return -1;
 838         }
 839 
 840         tmp_db = tdb_open("tmpdb", tdb_hash_size(tdb), TDB_INTERNAL, O_RDWR|O_CREAT, 0);
 841         if (tmp_db == NULL) {
 842                 TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Failed to create tmp_db\n"));
 843                 tdb_transaction_cancel(tdb);
 844                 return -1;
 845         }
 846 
 847         state.error = false;
 848         state.dest_db = tmp_db;
 849 
 850         if (tdb_traverse_read(tdb, repack_traverse, &state) == -1) {
 851                 TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Failed to traverse copying out\n"));
 852                 tdb_transaction_cancel(tdb);
 853                 tdb_close(tmp_db);
 854                 return -1;              
 855         }
 856 
 857         if (state.error) {
 858                 TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Error during traversal\n"));
 859                 tdb_transaction_cancel(tdb);
 860                 tdb_close(tmp_db);
 861                 return -1;
 862         }
 863 
 864         if (tdb_wipe_all(tdb) != 0) {
 865                 TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Failed to wipe database\n"));
 866                 tdb_transaction_cancel(tdb);
 867                 tdb_close(tmp_db);
 868                 return -1;
 869         }
 870 
 871         state.error = false;
 872         state.dest_db = tdb;
 873 
 874         if (tdb_traverse_read(tmp_db, repack_traverse, &state) == -1) {
 875                 TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Failed to traverse copying back\n"));
 876                 tdb_transaction_cancel(tdb);
 877                 tdb_close(tmp_db);
 878                 return -1;              
 879         }
 880 
 881         if (state.error) {
 882                 TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Error during second traversal\n"));
 883                 tdb_transaction_cancel(tdb);
 884                 tdb_close(tmp_db);
 885                 return -1;
 886         }
 887 
 888         tdb_close(tmp_db);
 889 
 890         if (tdb_transaction_commit(tdb) != 0) {
 891                 TDB_LOG((tdb, TDB_DEBUG_FATAL, __location__ " Failed to commit\n"));
 892                 return -1;
 893         }
 894 
 895         return 0;
 896 }

/* [<][>][^][v][top][bottom][index][help] */