summaryrefslogtreecommitdiff
path: root/source3/tdb/tdb.c
blob: 97c8e00ae685961014a0fe26b0910fc9dddf6002 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
/* 
   Unix SMB/Netbios implementation.
   Version 3.0
   Samba database functions
   Copyright (C) Andrew Tridgell              1999-2000
   Copyright (C) Luke Kenneth Casson Leighton      2000
   Copyright (C) Paul `Rusty' Russell		   2000
   Copyright (C) Jeremy Allison			   2000
   
   This program is free software; you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
   the Free Software Foundation; either version 2 of the License, or
   (at your option) any later version.
   
   This program is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   GNU General Public License for more details.
   
   You should have received a copy of the GNU General Public License
   along with this program; if not, write to the Free Software
   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#ifdef STANDALONE
#if HAVE_CONFIG_H
#include <config.h>
#endif

#include <stdlib.h>
#include <stdio.h>
#include <fcntl.h>
#include <unistd.h>
#include <string.h>
#include <fcntl.h>
#include <errno.h>
#include <sys/mman.h>
#include <sys/stat.h>
#include "tdb.h"
#include "spinlock.h"
#else
#include "includes.h"
#endif

#define TDB_MAGIC_FOOD "TDB file\n"
#define TDB_VERSION (0x26011967 + 6)
#define TDB_MAGIC (0x26011999U)
#define TDB_FREE_MAGIC (~TDB_MAGIC)
#define TDB_DEAD_MAGIC (0xFEE1DEAD)
#define TDB_ALIGNMENT 4
#define MIN_REC_SIZE (2*sizeof(struct list_struct) + TDB_ALIGNMENT)
#define DEFAULT_HASH_SIZE 131
#define TDB_PAGE_SIZE 0x2000
#define FREELIST_TOP (sizeof(struct tdb_header))
#define TDB_ALIGN(x,a) (((x) + (a)-1) & ~((a)-1))
#define TDB_BYTEREV(x) (((((x)&0xff)<<24)|((x)&0xFF00)<<8)|(((x)>>8)&0xFF00)|((x)>>24))
#define TDB_DEAD(r) ((r)->magic == TDB_DEAD_MAGIC)
#define TDB_BAD_MAGIC(r) ((r)->magic != TDB_MAGIC && !TDB_DEAD(r))
#define TDB_HASH_TOP(hash) (FREELIST_TOP + (BUCKET(hash)+1)*sizeof(tdb_off))
#define TDB_LOG(x) (tdb->log_fn?((tdb->log_fn x),0) : 0)

/* lock offsets */
#define GLOBAL_LOCK 0
#define ACTIVE_LOCK 4

#ifndef MAP_FILE
#define MAP_FILE 0
#endif

#ifndef MAP_FAILED
#define MAP_FAILED ((void *)-1)
#endif

#define BUCKET(hash) ((hash) % tdb->header.hash_size)
TDB_DATA tdb_null;

/* all contexts, to ensure no double-opens (fcntl locks don't nest!) */
static TDB_CONTEXT *tdbs = NULL;

static void tdb_munmap(TDB_CONTEXT *tdb)
{
#ifdef HAVE_MMAP
	if (tdb->map_ptr) munmap(tdb->map_ptr, tdb->map_size);
#endif
	tdb->map_ptr = NULL;
}

static void tdb_mmap(TDB_CONTEXT *tdb)
{
#ifdef HAVE_MMAP
	if (!(tdb->flags & TDB_NOMMAP)) {
		tdb->map_ptr = mmap(NULL, tdb->map_size, 
				    PROT_READ|(tdb->read_only? 0:PROT_WRITE), 
				    MAP_SHARED|MAP_FILE, tdb->fd, 0);

		/*
		 * NB. When mmap fails it returns MAP_FAILED *NOT* NULL !!!!
		 */

		if (tdb->map_ptr == MAP_FAILED) {
			tdb->map_ptr = NULL;
			TDB_LOG((tdb, 2, "tdb_mmap failed for size %d (%s)\n", 
				 tdb->map_size, strerror(errno)));
		}
	}
#else
	tdb->map_ptr = NULL;
#endif
}

/* Endian conversion: we only ever deal with 4 byte quantities */
static void *convert(void *buf, u32 size)
{
	u32 i, *p = buf;
	for (i = 0; i < size / 4; i++) p[i] = TDB_BYTEREV(p[i]);
	return buf;
}
#define DOCONV() (tdb->flags & TDB_CONVERT)
#define CONVERT(x) (DOCONV() ? convert(&x, sizeof(x)) : &x)

/* the body of the database is made of one list_struct for the free space
   plus a separate data list for each hash value */
struct list_struct {
	tdb_off next; /* offset of the next record in the list */
	tdb_len rec_len; /* total byte length of record */
	tdb_len key_len; /* byte length of key */
	tdb_len data_len; /* byte length of data */
	u32 full_hash; /* the full 32 bit hash of the key */
	u32 magic;   /* try to catch errors */
	/* the following union is implied:
	   union {
              char record[rec_len];
	      struct {
	        char key[key_len];
		char data[data_len];
	      }
	      u32 totalsize; (tailer)
           } */
};

/* a byte range locking function - return 0 on success
   this functions locks/unlocks 1 byte at the specified offset */
static int tdb_brlock(TDB_CONTEXT *tdb, tdb_off offset, 
		      int rw_type, int lck_type, int probe)
{
	struct flock fl;

        if (tdb->flags & TDB_NOLOCK) return 0;
	if (tdb->read_only) return -1;

	fl.l_type = rw_type;
	fl.l_whence = SEEK_SET;
	fl.l_start = offset;
	fl.l_len = 1;
	fl.l_pid = 0;

	if (fcntl(tdb->fd,lck_type,&fl)) {
		if (!probe) {
			TDB_LOG((tdb, 5,"tdb_brlock failed at offset %d rw_type=%d lck_type=%d\n", 
				 offset, rw_type, lck_type));
		}
		return TDB_ERRCODE(TDB_ERR_LOCK, -1);
	}
	return 0;
}

/* lock a list in the database. list -1 is the alloc list */
static int tdb_lock(TDB_CONTEXT *tdb, int list, int ltype)
{
	if (list < -1 || list >= (int)tdb->header.hash_size) {
		TDB_LOG((tdb, 0,"tdb_lock: invalid list %d for ltype=%d\n", 
			   list, ltype));
		return -1;
	}
	if (tdb->flags & TDB_NOLOCK) return 0;

	/* Since fcntl locks don't nest, we do a lock for the first one,
	   and simply bump the count for future ones */
	if (tdb->locked[list+1].count == 0) {
		if (tdb->header.rwlocks) {
			if (tdb_spinlock(tdb, list, ltype)) {
				TDB_LOG((tdb, 0, "tdb_lock spinlock on list ltype=%d\n", 
					   list, ltype));
				return -1;
			}
		} else if (tdb_brlock(tdb,FREELIST_TOP+4*list,ltype,F_SETLKW, 0)) {
			TDB_LOG((tdb, 0,"tdb_lock failed on list %d ltype=%d (%s)\n", 
					   list, ltype, strerror(errno)));
			return -1;
		}
		tdb->locked[list+1].ltype = ltype;
	}
	tdb->locked[list+1].count++;
	return 0;
}

/* unlock the database: returns void because it's too late for errors. */
static void tdb_unlock(TDB_CONTEXT *tdb, int list, int ltype)
{
	if (tdb->flags & TDB_NOLOCK) return;

	/* Sanity checks */
	if (list < -1 || list >= (int)tdb->header.hash_size) return;
	if (tdb->locked[list+1].count==0) return;

	if (tdb->locked[list+1].count == 1) {
		/* Down to last nested lock: unlock underneath */
		if (tdb->header.rwlocks) tdb_spinunlock(tdb, list, ltype);
		else tdb_brlock(tdb, FREELIST_TOP+4*list, F_UNLCK, F_SETLKW, 0);
	}
	tdb->locked[list+1].count--;
}

/* This is based on the hash agorithm from gdbm */
static u32 tdb_hash(TDB_DATA *key)
{
	u32 value;	/* Used to compute the hash value.  */
	u32   i;	/* Used to cycle through random values. */

	/* Set the initial value from the key size. */
	for (value = 0x238F13AF * key->dsize, i=0; i < key->dsize; i++)
		value = (value + (key->dptr[i] << (i*5 % 24)));

	return (1103515243 * value + 12345);  
}

/* check for an out of bounds access - if it is out of bounds then
   see if the database has been expanded by someone else and expand
   if necessary 
   note that "len" is the minimum length needed for the db
*/
static int tdb_oob(TDB_CONTEXT *tdb, tdb_off len, int probe)
{
	struct stat st;
	if (len <= tdb->map_size) return 0;
	if (tdb->flags & TDB_INTERNAL) return 0;

	fstat(tdb->fd, &st);
	if (st.st_size < (size_t)len) {
		if (!probe) {
			TDB_LOG((tdb, 0,"tdb_oob len %d beyond eof at %d\n",
				 (int)len, (int)st.st_size));
		}
		return TDB_ERRCODE(TDB_ERR_IO, -1);
	}

	/* Unmap, update size, remap */
	tdb_munmap(tdb);
	tdb->map_size = st.st_size;
	tdb_mmap(tdb);
	return 0;
}

/* write a lump of data at a specified offset */
static int tdb_write(TDB_CONTEXT *tdb, tdb_off off, void *buf, tdb_len len)
{
	if (tdb_oob(tdb, off + len, 0) != 0) return -1;

	if (tdb->map_ptr) memcpy(off + (char *)tdb->map_ptr, buf, len);
	else if (lseek(tdb->fd, off, SEEK_SET) != off
		 || write(tdb->fd, buf, len) != (ssize_t)len) {
		TDB_LOG((tdb, 0,"tdb_write failed at %d len=%d (%s)\n",
			   off, len, strerror(errno)));
		return TDB_ERRCODE(TDB_ERR_IO, -1);
	}
	return 0;
}

/* read a lump of data at a specified offset, maybe convert */
static int tdb_read(TDB_CONTEXT *tdb,tdb_off off,void *buf,tdb_len len,int cv)
{
	if (tdb_oob(tdb, off + len, 0) != 0) return -1;

	if (tdb->map_ptr) memcpy(buf, off + (char *)tdb->map_ptr, len);
	else if (lseek(tdb->fd, off, SEEK_SET) != off
		 || read(tdb->fd, buf, len) != (ssize_t)len) {
		TDB_LOG((tdb, 0,"tdb_read failed at %d len=%d (%s)\n",
			   off, len, strerror(errno)));
		return TDB_ERRCODE(TDB_ERR_IO, -1);
	}
	if (cv) convert(buf, len);
	return 0;
}

/* read a lump of data, allocating the space for it */
static char *tdb_alloc_read(TDB_CONTEXT *tdb, tdb_off offset, tdb_len len)
{
	char *buf;

	if (!(buf = malloc(len))) {
		TDB_LOG((tdb, 0,"tdb_alloc_read malloc failed len=%d (%s)\n",
			   len, strerror(errno)));
		return TDB_ERRCODE(TDB_ERR_OOM, buf);
	}
	if (tdb_read(tdb, offset, buf, len, 0) == -1) {
		free(buf);
		return NULL;
	}
	return buf;
}

/* read/write a tdb_off */
static int ofs_read(TDB_CONTEXT *tdb, tdb_off offset, tdb_off *d)
{
	return tdb_read(tdb, offset, (char*)d, sizeof(*d), DOCONV());
}
static int ofs_write(TDB_CONTEXT *tdb, tdb_off offset, tdb_off *d)
{
	tdb_off off = *d;
	return tdb_write(tdb, offset, CONVERT(off), sizeof(*d));
}

/* read/write a record */
static int rec_read(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
{
	if (tdb_read(tdb, offset, rec, sizeof(*rec),DOCONV()) == -1) return -1;
	if (TDB_BAD_MAGIC(rec)) {
		TDB_LOG((tdb, 0,"rec_read bad magic 0x%x at offset=%d\n", rec->magic, offset));
		return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
	}
	return tdb_oob(tdb, rec->next+sizeof(*rec), 0);
}
static int rec_write(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
{
	struct list_struct r = *rec;
	return tdb_write(tdb, offset, CONVERT(r), sizeof(r));
}

/* read a freelist record and check for simple errors */
static int rec_free_read(TDB_CONTEXT *tdb, tdb_off off, struct list_struct *rec)
{
	if (tdb_read(tdb, off, rec, sizeof(*rec),DOCONV()) == -1) return -1;
	if (rec->magic != TDB_FREE_MAGIC) {
		TDB_LOG((tdb, 0,"rec_free_read bad magic 0x%x at offset=%d\n", 
			   rec->magic, off));
		return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
	}
	if (tdb_oob(tdb, rec->next+sizeof(*rec), 0) != 0) return -1;
	return 0;
}

/* update a record tailer (must hold allocation lock) */
static int update_tailer(TDB_CONTEXT *tdb, tdb_off offset,
			 const struct list_struct *rec)
{
	tdb_off totalsize;

	/* Offset of tailer from record header */
	totalsize = sizeof(*rec) + rec->rec_len;
	return ofs_write(tdb, offset + totalsize - sizeof(tdb_off),
			 &totalsize);
}

#ifdef TDB_DEBUG
static tdb_off tdb_dump_record(TDB_CONTEXT *tdb, tdb_off offset)
{
	struct list_struct rec;
	tdb_off tailer_ofs, tailer;

	if (tdb_read(tdb, offset, (char *)&rec, sizeof(rec), DOCONV()) == -1) {
		printf("ERROR: failed to read record at %u\n", offset);
		return 0;
	}

	printf(" rec: offset=%u next=%d rec_len=%d key_len=%d data_len=%d full_hash=0x%x magic=0x%x\n",
	       offset, rec.next, rec.rec_len, rec.key_len, rec.data_len, rec.full_hash, rec.magic);

	tailer_ofs = offset + sizeof(rec) + rec.rec_len - sizeof(tdb_off);
	if (ofs_read(tdb, tailer_ofs, &tailer) == -1) {
		printf("ERROR: failed to read tailer at %u\n", tailer_ofs);
		return rec.next;
	}

	if (tailer != rec.rec_len + sizeof(rec)) {
		printf("ERROR: tailer does not match record! tailer=%u totalsize=%u\n", tailer, rec.rec_len + sizeof(rec));
	}
	return rec.next;
}

static void tdb_dump_chain(TDB_CONTEXT *tdb, int i)
{
	tdb_off rec_ptr, top;

	top = TDB_HASH_TOP(i);

	tdb_lock(tdb, i, F_WRLCK);

	if (ofs_read(tdb, top, &rec_ptr) == -1) {
		tdb_unlock(tdb, i, F_WRLCK);
		return;
	}

	if (rec_ptr) printf("hash=%d\n", i);

	while (rec_ptr) {
		rec_ptr = tdb_dump_record(tdb, rec_ptr);
	}
	tdb_unlock(tdb, i, F_WRLCK);
}

void tdb_dump_all(TDB_CONTEXT *tdb)
{
	tdb_off off;
	int i;
	for (i=0;i<tdb->header.hash_size;i++) {
		tdb_dump_chain(tdb, i);
	}
	printf("freelist:\n");
	tdb_dump_chain(tdb, -1);
}

void tdb_printfreelist(TDB_CONTEXT *tdb)
{
	long total_free = 0;
	tdb_off offset, rec_ptr, last_ptr;
	struct list_struct rec, lastrec, newrec;

	tdb_lock(tdb, -1, F_WRLCK);

	last_ptr = 0;
	offset = FREELIST_TOP;

	/* read in the freelist top */
	if (ofs_read(tdb, offset, &rec_ptr) == -1) {
		return;
	}

	printf("freelist top=[0x%08x]\n", rec_ptr );
	while (rec_ptr) {
		if (tdb_read(tdb, rec_ptr, (char *)&rec, sizeof(rec), DOCONV()) == -1) {
			return;
		}

		if (rec.magic != TDB_FREE_MAGIC) {
			printf("bad magic 0x%08x in free list\n", rec.magic);
			return;
		}

		printf("entry offset=[0x%08x], rec.rec_len = [0x%08x (%d)]\n", rec.next, rec.rec_len, rec.rec_len );
		total_free += rec.rec_len;

		/* move to the next record */
		rec_ptr = rec.next;
	}
	printf("total rec_len = [0x%08x (%d)]\n", total_free, total_free );

	tdb_unlock(tdb, -1, F_WRLCK);
}
#endif

/* Remove an element from the freelist.  Must have alloc lock. */
static int remove_from_freelist(TDB_CONTEXT *tdb, tdb_off off, tdb_off next)
{
	tdb_off last_ptr, i;

	/* read in the freelist top */
	last_ptr = FREELIST_TOP;
	while (ofs_read(tdb, last_ptr, &i) != -1 && i != 0) {
		if (i == off) {
			/* We've found it! */
			return ofs_write(tdb, last_ptr, &next);
		}
		/* Follow chain (next offset is at start of record) */
		last_ptr = i;
	}
	TDB_LOG((tdb, 0,"remove_from_freelist: not on list at off=%d\n", off));
	return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
}

/* Add an element into the freelist. Merge adjacent records if
   neccessary. */
static int tdb_free(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
{
	tdb_off right, left;

	/* Allocation and tailer lock */
	if (tdb_lock(tdb, -1, F_WRLCK) != 0) return -1;

	/* set an initial tailer, so if we fail we don't leave a bogus record */
	update_tailer(tdb, offset, rec);

	/* Look right first (I'm an Australian, dammit) */
	right = offset + sizeof(*rec) + rec->rec_len;
	if (right + sizeof(*rec) <= tdb->map_size) {
		struct list_struct r;

		if (tdb_read(tdb, right, &r, sizeof(r), DOCONV()) == -1) {
			TDB_LOG((tdb, 0, "tdb_free: right read failed at %u\n", right));
			goto left;
		}

		/* If it's free, expand to include it. */
		if (r.magic == TDB_FREE_MAGIC) {
			if (remove_from_freelist(tdb, right, r.next) == -1) {
				TDB_LOG((tdb, 0, "tdb_free: right free failed at %u\n", right));
				goto left;
			}
			rec->rec_len += sizeof(r) + r.rec_len;
		}
	}

left:
	/* Look left */
	left = offset - sizeof(tdb_off);
	if (left > TDB_HASH_TOP(tdb->header.hash_size-1)) {
		struct list_struct l;
		tdb_off leftsize;

		/* Read in tailer and jump back to header */
		if (ofs_read(tdb, left, &leftsize) == -1) {
			TDB_LOG((tdb, 0, "tdb_free: left offset read failed at %u\n", left));
			goto update;
		}
		left = offset - leftsize;

		/* Now read in record */
		if (tdb_read(tdb, left, &l, sizeof(l), DOCONV()) == -1) {
			TDB_LOG((tdb, 0, "tdb_free: left read failed at %u (%u)\n", left, leftsize));
			goto update;
		}

		/* If it's free, expand to include it. */
		if (l.magic == TDB_FREE_MAGIC) {
			if (remove_from_freelist(tdb, left, l.next) == -1) {
				TDB_LOG((tdb, 0, "tdb_free: left free failed at %u\n", left));
				goto update;
			} else {
				offset = left;
				rec->rec_len += leftsize;
			}
		}
	}

update:
	if (update_tailer(tdb, offset, rec) == -1) {
		TDB_LOG((tdb, 0, "tdb_free: update_tailer failed at %u\n", offset));
		goto fail;
	}

	/* Now, prepend to free list */
	rec->magic = TDB_FREE_MAGIC;

	if (ofs_read(tdb, FREELIST_TOP, &rec->next) == -1 ||
	    rec_write(tdb, offset, rec) == -1 ||
	    ofs_write(tdb, FREELIST_TOP, &offset) == -1) {
		TDB_LOG((tdb, 0, "tdb_free record write failed at offset=%d\n", offset));
		goto fail;
	}

	/* And we're done. */
	tdb_unlock(tdb, -1, F_WRLCK);
	return 0;

 fail:
	tdb_unlock(tdb, -1, F_WRLCK);
	return -1;
}


/* expand a file.  we prefer to use ftruncate, as that is what posix
  says to use for mmap expansion */
static int expand_file(TDB_CONTEXT *tdb, tdb_off size, tdb_off addition)
{
	char buf[1024];
#if HAVE_FTRUNCATE_EXTEND
	if (ftruncate(tdb->fd, size+addition) != 0) {
		TDB_LOG((tdb, 0, "expand_file ftruncate to %d failed (%s)\n", 
			   size+addition, strerror(errno)));
		return -1;
	}
#else
	char b = 0;
	if (lseek(tdb->fd, (size+addition) - 1, SEEK_SET) != (size+addition) - 1 || 
	    write(tdb->fd, &b, 1) != 1) {
		TDB_LOG((tdb, 0, "expand_file to %d failed (%s)\n", 
			   size+addition, strerror(errno)));
		return -1;
	}
#endif
	/* now fill the file with something. This ensures that the file isn't sparse, which would be
	   very bad if we ran out of disk. This must be done with write, not via mmap */
	memset(buf, 0x42, sizeof(buf));
	if (lseek(tdb->fd, size, SEEK_SET) != size) return -1;
	while (addition) {
		int n = addition>sizeof(buf)?sizeof(buf):addition;
		int ret = write(tdb->fd, buf, n);
		if (ret != n) {
			TDB_LOG((tdb, 0, "expand_file write of %d failed (%s)\n", 
				   n, strerror(errno)));
			return -1;
		}
		addition -= n;
	}
	return 0;
}


/* expand the database at least size bytes by expanding the underlying
   file and doing the mmap again if necessary */
static int tdb_expand(TDB_CONTEXT *tdb, tdb_off size)
{
	struct list_struct rec;
	tdb_off offset;

	if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
		TDB_LOG((tdb, 0, "lock failed in tdb_expand\n"));
		return 0;
	}

	/* must know about any previous expansions by another process */
	tdb_oob(tdb, tdb->map_size + 1, 1);

	/* always make room for at least 10 more records, and round
           the database up to a multiple of TDB_PAGE_SIZE */
	size = TDB_ALIGN(tdb->map_size + size*10, TDB_PAGE_SIZE) - tdb->map_size;

	if (!(tdb->flags & TDB_INTERNAL))
		tdb_munmap(tdb);

	/*
	 * We must ensure the file is unmapped before doing this
	 * to ensure consistency with systems like OpenBSD where
	 * writes and mmaps are not consistent.
	 */

	/* expand the file itself */
	if (!(tdb->flags & TDB_INTERNAL)) {
		if (expand_file(tdb, tdb->map_size, size) != 0) goto fail;
	}

	tdb->map_size += size;

	if (tdb->flags & TDB_INTERNAL)
		tdb->map_ptr = realloc(tdb->map_ptr, tdb->map_size);

	/*
	 * We must ensure the file is remapped before adding the space
	 * to ensure consistency with systems like OpenBSD where
	 * writes and mmaps are not consistent.
	 */

	/* We're ok if the mmap fails as we'll fallback to read/write */
	tdb_mmap(tdb);

	/* form a new freelist record */
	memset(&rec,'\0',sizeof(rec));
	rec.rec_len = size - sizeof(rec);

	/* link it into the free list */
	offset = tdb->map_size - size;
	if (tdb_free(tdb, offset, &rec) == -1) goto fail;

	tdb_unlock(tdb, -1, F_WRLCK);
	return 0;
 fail:
	tdb_unlock(tdb, -1, F_WRLCK);
	return -1;
}

/* allocate some space from the free list. The offset returned points
   to a unconnected list_struct within the database with room for at
   least length bytes of total data

   0 is returned if the space could not be allocated
 */
static tdb_off tdb_allocate(TDB_CONTEXT *tdb, tdb_len length,
			    struct list_struct *rec)
{
	tdb_off rec_ptr, last_ptr, newrec_ptr;
	struct list_struct newrec;

	if (tdb_lock(tdb, -1, F_WRLCK) == -1) return 0;

	/* Extra bytes required for tailer */
	length += sizeof(tdb_off);

 again:
	last_ptr = FREELIST_TOP;

	/* read in the freelist top */
	if (ofs_read(tdb, FREELIST_TOP, &rec_ptr) == -1) goto fail;

	/* keep looking until we find a freelist record big enough */
	while (rec_ptr) {
		if (rec_free_read(tdb, rec_ptr, rec) == -1) goto fail;

		if (rec->rec_len >= length) {
			/* found it - now possibly split it up  */
			if (rec->rec_len > length + MIN_REC_SIZE) {
				/* Length of left piece */
				length = TDB_ALIGN(length, TDB_ALIGNMENT);

				/* Right piece to go on free list */
				newrec.rec_len = rec->rec_len
					- (sizeof(*rec) + length);
				newrec_ptr = rec_ptr + sizeof(*rec) + length;

				/* And left record is shortened */
				rec->rec_len = length;
			} else
				newrec_ptr = 0;

			/* Remove allocated record from the free list */
			if (ofs_write(tdb, last_ptr, &rec->next) == -1)
				goto fail;

			/* Update header: do this before we drop alloc
                           lock, otherwise tdb_free() might try to
                           merge with us, thinking we're free.
                           (Thanks Jeremy Allison). */
			rec->magic = TDB_MAGIC;
			if (rec_write(tdb, rec_ptr, rec) == -1)
				goto fail;

			/* Did we create new block? */
			if (newrec_ptr) {
				/* Update allocated record tailer (we
                                   shortened it). */
				if (update_tailer(tdb, rec_ptr, rec) == -1)
					goto fail;

				/* Free new record */
				if (tdb_free(tdb, newrec_ptr, &newrec) == -1)
					goto fail;
			}

			/* all done - return the new record offset */
			tdb_unlock(tdb, -1, F_WRLCK);
			return rec_ptr;
		}
		/* move to the next record */
		last_ptr = rec_ptr;
		rec_ptr = rec->next;
	}
	/* we didn't find enough space. See if we can expand the
	   database and if we can then try again */
	if (tdb_expand(tdb, length + sizeof(*rec)) == 0) goto again;
 fail:
	tdb_unlock(tdb, -1, F_WRLCK);
	return 0;
}

/* initialise a new database with a specified hash size */
static int tdb_new_database(TDB_CONTEXT *tdb, int hash_size)
{
	struct tdb_header *newdb;
	int size, ret;

	/* We make it up in memory, then write it out if not internal */
	size = sizeof(struct tdb_header) + (hash_size+1)*sizeof(tdb_off);
	if (!(newdb = calloc(size, 1))) return TDB_ERRCODE(TDB_ERR_OOM, -1);

        /* Fill in the header */
        newdb->version = TDB_VERSION;
        newdb->hash_size = hash_size;
#ifdef USE_SPINLOCKS
	newdb->rwlocks = size;
#endif
	if (tdb->flags & TDB_INTERNAL) {
		tdb->map_size = size;
		tdb->map_ptr = (char *)newdb;
		memcpy(&tdb->header, newdb, sizeof(tdb->header));
		/* Convert the `ondisk' version if asked. */
		CONVERT(*newdb);
		return 0;
	}
	lseek(tdb->fd, 0, SEEK_SET);
	ftruncate(tdb->fd, 0);
	/* This creates an endian-converted header, as if read from disk */
	CONVERT(*newdb);
	memcpy(&tdb->header, newdb, sizeof(tdb->header));
	/* Don't endian-convert the magic food! */
	memcpy(newdb->magic_food, TDB_MAGIC_FOOD, strlen(TDB_MAGIC_FOOD)+1);
	if (write(tdb->fd, newdb, size) != size) ret = -1;
	else ret = tdb_create_rwlocks(tdb->fd, hash_size);

	free(newdb);
	return ret;
}

/* Returns 0 on fail.  On success, return offset of record, and fills
   in rec */
static tdb_off tdb_find(TDB_CONTEXT *tdb, TDB_DATA key, u32 hash,
			struct list_struct *r)
{
	tdb_off rec_ptr;
	
	/* read in the hash top */
	if (ofs_read(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1) return 0;

	/* keep looking until we find the right record */
	while (rec_ptr) {
		if (rec_read(tdb, rec_ptr, r) == -1) return 0;

		if (!TDB_DEAD(r) && hash==r->full_hash && key.dsize==r->key_len) {
			char *k;
			/* a very likely hit - read the key */
			k = tdb_alloc_read(tdb, rec_ptr + sizeof(*r), 
					   r->key_len);
			if (!k) return 0;

			if (memcmp(key.dptr, k, key.dsize) == 0) {
				free(k);
				return rec_ptr;
			}
			free(k);
		}
		rec_ptr = r->next;
	}
	return TDB_ERRCODE(TDB_ERR_NOEXIST, 0);
}

/* If they do lockkeys, check that this hash is one they locked */
static int tdb_keylocked(TDB_CONTEXT *tdb, u32 hash)
{
	u32 i;
	if (!tdb->lockedkeys) return 1;
	for (i = 0; i < tdb->lockedkeys[0]; i++)
		if (tdb->lockedkeys[i+1] == hash) return 1;
	return TDB_ERRCODE(TDB_ERR_NOLOCK, 0);
}

/* As tdb_find, but if you succeed, keep the lock */
static tdb_off tdb_find_lock(TDB_CONTEXT *tdb, TDB_DATA key, int locktype,
			     struct list_struct *rec)
{
	u32 hash, rec_ptr;

	hash = tdb_hash(&key);
	if (!tdb_keylocked(tdb, hash)) return 0;
	if (tdb_lock(tdb, BUCKET(hash), locktype) == -1) return 0;
	if (!(rec_ptr = tdb_find(tdb, key, hash, rec)))
		tdb_unlock(tdb, BUCKET(hash), locktype);
	return rec_ptr;
}

enum TDB_ERROR tdb_error(TDB_CONTEXT *tdb)
{
	return tdb->ecode;
}

static struct tdb_errname {
	enum TDB_ERROR ecode; const char *estring;
} emap[] = { {TDB_SUCCESS, "Success"},
	     {TDB_ERR_CORRUPT, "Corrupt database"},
	     {TDB_ERR_IO, "IO Error"},
	     {TDB_ERR_LOCK, "Locking error"},
	     {TDB_ERR_OOM, "Out of memory"},
	     {TDB_ERR_EXISTS, "Record exists"},
	     {TDB_ERR_NOLOCK, "Lock exists on other keys"},
	     {TDB_ERR_NOEXIST, "Record does not exist"} };

/* Error string for the last tdb error */
const char *tdb_errorstr(TDB_CONTEXT *tdb)
{
	u32 i;
	for (i = 0; i < sizeof(emap) / sizeof(struct tdb_errname); i++)
		if (tdb->ecode == emap[i].ecode) return emap[i].estring;
	return "Invalid error code";
}

/* update an entry in place - this only works if the new data size
   is <= the old data size and the key exists.
   on failure return -1
*/
static int tdb_update(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf)
{
	struct list_struct rec;
	tdb_off rec_ptr;
	int ret = -1;

	/* find entry */
	if (!(rec_ptr = tdb_find_lock(tdb, key, F_WRLCK, &rec))) return -1;

	/* must be long enough key, data and tailer */
	if (rec.rec_len < key.dsize + dbuf.dsize + sizeof(tdb_off)) goto out;

	if (tdb_write(tdb, rec_ptr + sizeof(rec) + rec.key_len,
		      dbuf.dptr, dbuf.dsize) == -1)
		goto out;

	if (dbuf.dsize != rec.data_len) {
		/* update size */
		rec.data_len = dbuf.dsize;
		ret = rec_write(tdb, rec_ptr, &rec);
	}
	else ret = 0;
 out:
	tdb_unlock(tdb, BUCKET(rec.full_hash), F_WRLCK);
	return ret;
}

/* find an entry in the database given a key */
TDB_DATA tdb_fetch(TDB_CONTEXT *tdb, TDB_DATA key)
{
	tdb_off rec_ptr;
	struct list_struct rec;
	TDB_DATA ret;

	/* find which hash bucket it is in */
	if (!(rec_ptr = tdb_find_lock(tdb,key,F_RDLCK,&rec))) return tdb_null;

	ret.dptr = tdb_alloc_read(tdb, rec_ptr + sizeof(rec) + rec.key_len,
				  rec.data_len);
	ret.dsize = rec.data_len;
	tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
	return ret;
}

/* check if an entry in the database exists 

   note that 1 is returned if the key is found and 0 is returned if not found
   this doesn't match the conventions in the rest of this module, but is
   compatible with gdbm
*/
int tdb_exists(TDB_CONTEXT *tdb, TDB_DATA key)
{
	struct list_struct rec;
	
	if (tdb_find_lock(tdb, key, F_RDLCK, &rec) == 0) return 0;
	tdb_unlock(tdb, BUCKET(rec.full_hash), F_RDLCK);
	return 1;
}

/* record lock stops delete underneath */
static int lock_record(TDB_CONTEXT *tdb, tdb_off off)
{
	return off ? tdb_brlock(tdb, off, F_RDLCK, F_SETLKW, 0) : 0;
}
/* write locks override our own fcntl readlocks, so check it here */
static int write_lock_record(TDB_CONTEXT *tdb, tdb_off off)
{
	struct tdb_traverse_lock *i;
	for (i = &tdb->travlocks; i; i = i->next) if (i->off == off) return -1;
	return tdb_brlock(tdb, off, F_WRLCK, F_SETLK, 1);
}
static int write_unlock_record(TDB_CONTEXT *tdb, tdb_off off)
{
	return tdb_brlock(tdb, off, F_UNLCK, F_SETLK, 0);
}
/* fcntl locks don't stack: avoid unlocking someone else's */
static int unlock_record(TDB_CONTEXT *tdb, tdb_off off)
{
	struct tdb_traverse_lock *i;
	u32 count = 0;

	if (off == 0) return 0;
	for (i = &tdb->travlocks; i; i = i->next) if (i->off == off) count++;
	return (count == 1 ? tdb_brlock(tdb, off, F_UNLCK, F_SETLKW, 0) : 0);
}

/* actually delete an entry in the database given the offset */
static int do_delete(TDB_CONTEXT *tdb, tdb_off rec_ptr, struct list_struct*rec)
{
	tdb_off last_ptr, i;
	struct list_struct lastrec;

	if (tdb->read_only) return -1;

	if (write_lock_record(tdb, rec_ptr) == -1) {
		/* Someone traversing here: mark it as dead */
		rec->magic = TDB_DEAD_MAGIC;
		return rec_write(tdb, rec_ptr, rec);
	}
	write_unlock_record(tdb, rec_ptr);

	/* find previous record in hash chain */
	if (ofs_read(tdb, TDB_HASH_TOP(rec->full_hash), &i) == -1) return -1;
	for (last_ptr = 0; i != rec_ptr; last_ptr = i, i = lastrec.next)
		if (rec_read(tdb, i, &lastrec) == -1) return -1;

	/* unlink it: next ptr is at start of record. */
	if (last_ptr == 0) last_ptr = TDB_HASH_TOP(rec->full_hash);
	if (ofs_write(tdb, last_ptr, &rec->next) == -1) return -1;

	/* recover the space */
	if (tdb_free(tdb, rec_ptr, rec) == -1) return -1;
	return 0;
}

/* Uses traverse lock: 0 = finish, -1 = error, other = record offset */
static int tdb_next_lock(TDB_CONTEXT *tdb, struct tdb_traverse_lock *tlock,
			 struct list_struct *rec)
{
	int want_next = (tlock->off != 0);

	/* No traversal allows if you've called tdb_lockkeys() */
	if (tdb->lockedkeys) return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);

	/* Lock each chain from the start one. */
	for (; tlock->hash < tdb->header.hash_size; tlock->hash++) {
		if (tdb_lock(tdb, tlock->hash, F_WRLCK) == -1) return -1;

		/* No previous record?  Start at top of chain. */
		if (!tlock->off) {
			if (ofs_read(tdb, TDB_HASH_TOP(tlock->hash),
				     &tlock->off) == -1)
				goto fail;
		} else {
			/* Otherwise unlock the previous record. */
			unlock_record(tdb, tlock->off);
		}

		if (want_next) {
			/* We have offset of old record: grab next */
			if (rec_read(tdb, tlock->off, rec) == -1) goto fail;
			tlock->off = rec->next;
		}

		/* Iterate through chain */
		while( tlock->off) {
			tdb_off current;
			if (rec_read(tdb, tlock->off, rec) == -1) goto fail;
			if (!TDB_DEAD(rec)) {
				/* Woohoo: we found one! */
				lock_record(tdb, tlock->off);
				return tlock->off;
			}
			/* Try to clean dead ones from old traverses */
			current = tlock->off;
			tlock->off = rec->next;
			do_delete(tdb, current, rec);
		}
		tdb_unlock(tdb, tlock->hash, F_WRLCK);
		want_next = 0;
	}
	/* We finished iteration without finding anything */
	return TDB_ERRCODE(TDB_SUCCESS, 0);

 fail:
	tlock->off = 0;
	tdb_unlock(tdb, tlock->hash, F_WRLCK);
	return -1;
}

/* traverse the entire database - calling fn(tdb, key, data) on each element.
   return -1 on error or the record count traversed
   if fn is NULL then it is not called
   a non-zero return value from fn() indicates that the traversal should stop
  */
int tdb_traverse(TDB_CONTEXT *tdb, tdb_traverse_func fn, void *state)
{
	TDB_DATA key, dbuf;
	struct list_struct rec;
	struct tdb_traverse_lock tl = { NULL, 0, 0 };
	int ret, count = 0;

	/* This was in the initializaton, above, but the IRIX compiler
	 * did not like it.  crh
	 */
	tl.next = tdb->travlocks.next;

	/* fcntl locks don't stack: beware traverse inside traverse */
	tdb->travlocks.next = &tl;

	/* tdb_next_lock places locks on the record returned, and its chain */
	while ((ret = tdb_next_lock(tdb, &tl, &rec)) > 0) {
		count++;
		/* now read the full record */
		key.dptr = tdb_alloc_read(tdb, tl.off + sizeof(rec), 
					  rec.key_len + rec.data_len);
		if (!key.dptr) {
			tdb_unlock(tdb, tl.hash, F_WRLCK);
			unlock_record(tdb, tl.off);
			tdb->travlocks.next = tl.next;
			return -1;
		}
		key.dsize = rec.key_len;
		dbuf.dptr = key.dptr + rec.key_len;
		dbuf.dsize = rec.data_len;

		/* Drop chain lock, call out */
		tdb_unlock(tdb, tl.hash, F_WRLCK);
		if (fn && fn(tdb, key, dbuf, state)) {
			/* They want us to terminate traversal */
			unlock_record(tdb, tl.off);
			tdb->travlocks.next = tl.next;
			free(key.dptr);
			return count;
		}
		free(key.dptr);
	}
	tdb->travlocks.next = tl.next;
	if (ret < 0)
		return -1;
	else
		return count;
}

/* find the first entry in the database and return its key */
TDB_DATA tdb_firstkey(TDB_CONTEXT *tdb)
{
	TDB_DATA key;
	struct list_struct rec;

	/* release any old lock */
	unlock_record(tdb, tdb->travlocks.off);
	tdb->travlocks.off = tdb->travlocks.hash = 0;

	if (tdb_next_lock(tdb, &tdb->travlocks, &rec) <= 0) return tdb_null;
	/* now read the key */
	key.dsize = rec.key_len;
	key.dptr =tdb_alloc_read(tdb,tdb->travlocks.off+sizeof(rec),key.dsize);
	tdb_unlock(tdb, BUCKET(tdb->travlocks.hash), F_WRLCK);
	return key;
}

/* find the next entry in the database, returning its key */
TDB_DATA tdb_nextkey(TDB_CONTEXT *tdb, TDB_DATA oldkey)
{
	u32 oldhash;
	TDB_DATA key = tdb_null;
	struct list_struct rec;
	char *k = NULL;

	/* Is locked key the old key?  If so, traverse will be reliable. */
	if (tdb->travlocks.off) {
		if (tdb_lock(tdb,tdb->travlocks.hash,F_WRLCK)) return tdb_null;
		if (rec_read(tdb, tdb->travlocks.off, &rec) == -1
		    || !(k = tdb_alloc_read(tdb,tdb->travlocks.off+sizeof(rec),
					    rec.key_len))
		    || memcmp(k, oldkey.dptr, oldkey.dsize) != 0) {
			/* No, it wasn't: unlock it and start from scratch */
			unlock_record(tdb, tdb->travlocks.off);
			tdb_unlock(tdb, tdb->travlocks.hash, F_WRLCK);
			tdb->travlocks.off = 0;
		}

		if (k)
			free(k);
	}

	if (!tdb->travlocks.off) {
		/* No previous element: do normal find, and lock record */
		tdb->travlocks.off = tdb_find_lock(tdb, oldkey, F_WRLCK, &rec);
		if (!tdb->travlocks.off) return tdb_null;
		tdb->travlocks.hash = BUCKET(rec.full_hash);
		lock_record(tdb, tdb->travlocks.off);
	}
	oldhash = tdb->travlocks.hash;

	/* Grab next record: locks chain and returned record,
	   unlocks old record */
	if (tdb_next_lock(tdb, &tdb->travlocks, &rec) > 0) {
		key.dsize = rec.key_len;
		key.dptr = tdb_alloc_read(tdb, tdb->travlocks.off+sizeof(rec),
					  key.dsize);
		/* Unlock the chain of this new record */
		tdb_unlock(tdb, tdb->travlocks.hash, F_WRLCK);
	}
	/* Unlock the chain of old record */
	tdb_unlock(tdb, BUCKET(oldhash), F_WRLCK);
	return key;
}

/* delete an entry in the database given a key */
int tdb_delete(TDB_CONTEXT *tdb, TDB_DATA key)
{
	tdb_off rec_ptr;
	struct list_struct rec;
	int ret;

	if (!(rec_ptr = tdb_find_lock(tdb, key, F_WRLCK, &rec))) return -1;
	ret = do_delete(tdb, rec_ptr, &rec);
	tdb_unlock(tdb, BUCKET(rec.full_hash), F_WRLCK);
	return ret;
}

/* store an element in the database, replacing any existing element
   with the same key 

   return 0 on success, -1 on failure
*/
int tdb_store(TDB_CONTEXT *tdb, TDB_DATA key, TDB_DATA dbuf, int flag)
{
	struct list_struct rec;
	u32 hash;
	tdb_off rec_ptr;
	char *p = NULL;
	int ret = 0;

	/* find which hash bucket it is in */
	hash = tdb_hash(&key);
	if (!tdb_keylocked(tdb, hash)) return -1;
	if (tdb_lock(tdb, BUCKET(hash), F_WRLCK) == -1) return -1;

	/* check for it existing, on insert. */
	if (flag == TDB_INSERT) {
		if (tdb_exists(tdb, key)) {
			tdb->ecode = TDB_ERR_EXISTS;
			goto fail;
		}
	} else {
		/* first try in-place update, on modify or replace. */
		if (tdb_update(tdb, key, dbuf) == 0) goto out;
		if (flag == TDB_MODIFY && tdb->ecode == TDB_ERR_NOEXIST)
			goto fail;
	}
	/* reset the error code potentially set by the tdb_update() */
	tdb->ecode = TDB_SUCCESS;

	/* delete any existing record - if it doesn't exist we don't
           care.  Doing this first reduces fragmentation, and avoids
           coalescing with `allocated' block before it's updated. */
	if (flag != TDB_INSERT) tdb_delete(tdb, key);

	/* Copy key+value *before* allocating free space in case malloc
	   fails and we are left with a dead spot in the tdb. */

	if (!(p = (char *)malloc(key.dsize + dbuf.dsize))) {
		tdb->ecode = TDB_ERR_OOM;
		goto fail;
	}

	memcpy(p, key.dptr, key.dsize);
	memcpy(p+key.dsize, dbuf.dptr, dbuf.dsize);

	/* now we're into insert / modify / replace of a record which
	 * we know could not be optimised by an in-place store (for
	 * various reasons).  */
	if (!(rec_ptr = tdb_allocate(tdb, key.dsize + dbuf.dsize, &rec)))
		goto fail;

	/* Read hash top into next ptr */
	if (ofs_read(tdb, TDB_HASH_TOP(hash), &rec.next) == -1) goto fail;

	rec.key_len = key.dsize;
	rec.data_len = dbuf.dsize;
	rec.full_hash = hash;
	rec.magic = TDB_MAGIC;

	/* write out and point the top of the hash chain at it */
	if (rec_write(tdb, rec_ptr, &rec) == -1
	    || tdb_write(tdb, rec_ptr+sizeof(rec), p, key.dsize+dbuf.dsize)==-1
	    || ofs_write(tdb, TDB_HASH_TOP(hash), &rec_ptr) == -1) {
	fail:
		/* Need to tdb_unallocate() here */
		ret = -1;
	}
 out:
	if (p)
		free(p); 
	tdb_unlock(tdb, BUCKET(hash), F_WRLCK);
	return ret;
}

/* open the database, creating it if necessary 

   The open_flags and mode are passed straight to the open call on the
   database file. A flags value of O_WRONLY is invalid. The hash size
   is advisory, use zero for a default value.

   return is NULL on error */
TDB_CONTEXT *tdb_open(char *name, int hash_size, int tdb_flags,
		      int open_flags, mode_t mode)
{
	TDB_CONTEXT tdb, *ret, *i;
	struct stat st;
	int rev = 0, locked;

	memset(&tdb, 0, sizeof(tdb));
	tdb.fd = -1;
	tdb.name = NULL;
	tdb.map_ptr = NULL;
	tdb.lockedkeys = NULL;
	tdb.flags = tdb_flags;

	if ((open_flags & O_ACCMODE) == O_WRONLY) goto fail;
	if (hash_size == 0) hash_size = DEFAULT_HASH_SIZE;
	if ((open_flags & O_ACCMODE) == O_RDONLY) {
		tdb.read_only = 1;
		/* read only databases don't do locking */
		tdb.flags |= TDB_NOLOCK;
	}

	/* internal databases don't mmap or lock, and start off cleared */
	if (tdb.flags & TDB_INTERNAL) {
		tdb.flags |= (TDB_NOLOCK | TDB_NOMMAP);
		tdb.flags &= ~TDB_CLEAR_IF_FIRST;
		tdb_new_database(&tdb, hash_size);
		goto internal;
	}

	if ((tdb.fd = open(name, open_flags, mode)) == -1) goto fail;

	/* ensure there is only one process initialising at once */
	tdb_brlock(&tdb, GLOBAL_LOCK, F_WRLCK, F_SETLKW, 0);
	
	/* we need to zero database if we are the only one with it open */
	if ((locked = (tdb_brlock(&tdb, ACTIVE_LOCK, F_WRLCK, F_SETLK, 0) == 0))
	    && (tdb_flags & TDB_CLEAR_IF_FIRST)) {
		open_flags |= O_CREAT;
		ftruncate(tdb.fd, 0);
	}

	if (read(tdb.fd, &tdb.header, sizeof(tdb.header)) != sizeof(tdb.header)
	    || strcmp(tdb.header.magic_food, TDB_MAGIC_FOOD) != 0
	    || (tdb.header.version != TDB_VERSION
		&& !(rev = (tdb.header.version==TDB_BYTEREV(TDB_VERSION))))) {
		/* its not a valid database - possibly initialise it */
		if (!(open_flags & O_CREAT)
		    || tdb_new_database(&tdb, hash_size) == -1) goto fail;
		rev = (tdb.flags & TDB_CONVERT);
	}
	if (!rev) tdb.flags &= ~TDB_CONVERT;
	else {
		tdb.flags |= TDB_CONVERT;
		convert(&tdb.header, sizeof(tdb.header));
	}
	fstat(tdb.fd, &st);
	/* Is it already in the open list?  If so, fail. */
	for (i = tdbs; i; i = i->next) {
		if (i->device == st.st_dev && i->inode == st.st_ino) {
			errno = EBUSY;
			close(tdb.fd);
			return NULL;
		}
	}

	/* map the database and fill in the return structure */
	tdb.name = (char *)strdup(name);
	tdb.map_size = st.st_size;
	tdb.device = st.st_dev;
	tdb.inode = st.st_ino;
        tdb.locked = calloc(tdb.header.hash_size+1, sizeof(tdb.locked[0]));
        if (!tdb.locked) goto fail;
	tdb_mmap(&tdb);
	if (locked) {
		tdb_clear_spinlocks(&tdb);
		tdb_brlock(&tdb, ACTIVE_LOCK, F_UNLCK, F_SETLK, 0);
	}
	/* leave this lock in place to indicate it's in use */
	tdb_brlock(&tdb, ACTIVE_LOCK, F_RDLCK, F_SETLKW, 0);

 internal:
	if (!(ret = malloc(sizeof(tdb)))) goto fail;
	*ret = tdb;
	tdb_brlock(&tdb, GLOBAL_LOCK, F_UNLCK, F_SETLKW, 0);
	ret->next = tdbs;
	tdbs = ret;
	return ret;

 fail:
        if (tdb.name) free(tdb.name);
	if (tdb.fd != -1) close(tdb.fd);
	tdb_munmap(&tdb);
	return NULL;
}

/* close a database */
int tdb_close(TDB_CONTEXT *tdb)
{
	TDB_CONTEXT **i;
	int ret = 0;

	if (tdb->map_ptr) {
		if (tdb->flags & TDB_INTERNAL) free(tdb->map_ptr);
		else tdb_munmap(tdb);
	}
	if (tdb->name) free(tdb->name);
	if (tdb->fd != -1) {
		ret = close(tdb->fd);
	}
	if (tdb->locked) free(tdb->locked);
	if (tdb->lockedkeys) free(tdb->lockedkeys);

	/* Remove from contexts list */
	for (i = &tdbs; *i; i = &(*i)->next) {
		if (*i == tdb) {
			*i = tdb->next;
			break;
		}
	}

	memset(tdb, 0, sizeof(*tdb));
	free(tdb);

	return ret;
}

/* lock/unlock entire database */
int tdb_lockall(TDB_CONTEXT *tdb)
{
	u32 i;

	/* There are no locks on read-only dbs */
	if (tdb->read_only) return TDB_ERRCODE(TDB_ERR_LOCK, -1);
	if (tdb->lockedkeys) return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
	for (i = 0; i < tdb->header.hash_size; i++) 
		if (tdb_lock(tdb, i, F_WRLCK))
			break;

	/* If error, release locks we have... */
	if (i < tdb->header.hash_size) {
		u32 j;

		for ( j = 0; j < i; j++)
			tdb_unlock(tdb, j, F_WRLCK);
		return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
	}

	return 0;
}
void tdb_unlockall(TDB_CONTEXT *tdb)
{
	u32 i;
	for (i=0; i < tdb->header.hash_size; i++) tdb_unlock(tdb, i, F_WRLCK);
}

int tdb_lockkeys(TDB_CONTEXT *tdb, u32 number, TDB_DATA keys[])
{
	u32 i, j, hash;

	/* Can't lock more keys if already locked */
	if (tdb->lockedkeys) return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
	if (!(tdb->lockedkeys = malloc(sizeof(u32) * (number+1))))
		return TDB_ERRCODE(TDB_ERR_OOM, -1);
	/* First number in array is # keys */
	tdb->lockedkeys[0] = number;

	/* Insertion sort by bucket */
	for (i = 0; i < number; i++) {
		hash = tdb_hash(&keys[i]);
		for (j = 0;
		     j < i && BUCKET(tdb->lockedkeys[j+1]) < BUCKET(hash);
		     j++);
		memmove(&tdb->lockedkeys[j+2], &tdb->lockedkeys[j+1],
			sizeof(u32) * (i-j));
		tdb->lockedkeys[j+1] = hash;
	}
	/* Finally, lock in order */
	for (i = 0; i < number; i++)
		if (tdb_lock(tdb, i, F_WRLCK))
			break;

	/* If error, release locks we have... */
	if (i < number) {
		for ( j = 0; j < i; j++)
			tdb_unlock(tdb, j, F_WRLCK);
		free(tdb->lockedkeys);
		tdb->lockedkeys = NULL;
		return TDB_ERRCODE(TDB_ERR_NOLOCK, -1);
	}
	return 0;
}

/* Unlock the keys previously locked by tdb_lockkeys() */
void tdb_unlockkeys(TDB_CONTEXT *tdb)
{
	u32 i;
	for (i = 0; i < tdb->lockedkeys[0]; i++)
		tdb_unlock(tdb, tdb->lockedkeys[i+1], F_WRLCK);
	free(tdb->lockedkeys);
	tdb->lockedkeys = NULL;
}

/* lock/unlock one hash chain. This is meant to be used to reduce
   contention - it cannot guarantee how many records will be locked */
int tdb_chainlock(TDB_CONTEXT *tdb, TDB_DATA key)
{
	return tdb_lock(tdb, BUCKET(tdb_hash(&key)), F_WRLCK);
}
void tdb_chainunlock(TDB_CONTEXT *tdb, TDB_DATA key)
{
	tdb_unlock(tdb, BUCKET(tdb_hash(&key)), F_WRLCK);
}


/* register a loging function */
void tdb_logging_function(TDB_CONTEXT *tdb, void (*fn)(TDB_CONTEXT *, int , const char *, ...))
{
	tdb->log_fn = fn;
}