/ Check-in [e71fb0fb]
Login

Many hyperlinks are disabled.
Use anonymous login to enable hyperlinks.

Overview
Comment:Move the asynchronous IO code from src/test_async.c to ext/async/. Refactor it to be a standalone module and to support windows. (CVS 6539)
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | trunk
Files: files | file ages | folders
SHA1: e71fb0fb8d83b4453c3c1e84606bf58d04926809
User & Date: danielk1977 2009-04-23 14:58:40
Context
2009-04-23
18:41
Remove some incorrect async tests; (CVS 6540) check-in: 03af25b3 user: shane tags: trunk
14:58
Move the asynchronous IO code from src/test_async.c to ext/async/. Refactor it to be a standalone module and to support windows. (CVS 6539) check-in: e71fb0fb user: danielk1977 tags: trunk
13:22
Rework the column-cache mechanism to be more robust (and more correct). The column-alias cache is currently disabled, (CVS 6538) check-in: dd4d67a6 user: drh tags: trunk
Changes
Hide Diffs Unified Diffs Show Whitespace Changes Patch

Added ext/async/sqlite3async.c.





















































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
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
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
/*
** 2005 December 14
**
** The author disclaims copyright to this source code.  In place of
** a legal notice, here is a blessing:
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
**
** $Id: sqlite3async.c,v 1.1 2009/04/23 14:58:40 danielk1977 Exp $
**
** This file contains an example implementation of an asynchronous IO 
** backend for SQLite.
**
** WHAT IS ASYNCHRONOUS I/O?
**
** With asynchronous I/O, write requests are handled by a separate thread
** running in the background.  This means that the thread that initiates
** a database write does not have to wait for (sometimes slow) disk I/O
** to occur.  The write seems to happen very quickly, though in reality
** it is happening at its usual slow pace in the background.
**
** Asynchronous I/O appears to give better responsiveness, but at a price.
** You lose the Durable property.  With the default I/O backend of SQLite,
** once a write completes, you know that the information you wrote is
** safely on disk.  With the asynchronous I/O, this is not the case.  If
** your program crashes or if a power loss occurs after the database
** write but before the asynchronous write thread has completed, then the
** database change might never make it to disk and the next user of the
** database might not see your change.
**
** You lose Durability with asynchronous I/O, but you still retain the
** other parts of ACID:  Atomic,  Consistent, and Isolated.  Many
** appliations get along fine without the Durablity.
**
** HOW IT WORKS
**
** Asynchronous I/O works by creating a special SQLite "vfs" structure
** and registering it with sqlite3_vfs_register(). When files opened via 
** this vfs are written to (using sqlite3OsWrite()), the data is not 
** written directly to disk, but is placed in the "write-queue" to be
** handled by the background thread.
**
** When files opened with the asynchronous vfs are read from 
** (using sqlite3OsRead()), the data is read from the file on 
** disk and the write-queue, so that from the point of view of
** the vfs reader the OsWrite() appears to have already completed.
**
** The special vfs is registered (and unregistered) by calls to 
** function asyncEnable() (see below).
**
** LIMITATIONS
**
** This demonstration code is deliberately kept simple in order to keep
** the main ideas clear and easy to understand.  Real applications that
** want to do asynchronous I/O might want to add additional capabilities.
** For example, in this demonstration if writes are happening at a steady
** stream that exceeds the I/O capability of the background writer thread,
** the queue of pending write operations will grow without bound until we
** run out of memory.  Users of this technique may want to keep track of
** the quantity of pending writes and stop accepting new write requests
** when the buffer gets to be too big.
**
** LOCKING + CONCURRENCY
**
** Multiple connections from within a single process that use this
** implementation of asynchronous IO may access a single database
** file concurrently. From the point of view of the user, if all
** connections are from within a single process, there is no difference
** between the concurrency offered by "normal" SQLite and SQLite
** using the asynchronous backend.
**
** If connections from within multiple processes may access the
** database file, the ENABLE_FILE_LOCKING symbol (see below) must be
** defined. If it is not defined, then no locks are established on 
** the database file. In this case, if multiple processes access 
** the database file, corruption will quickly result.
**
** If ENABLE_FILE_LOCKING is defined (the default), then connections 
** from within multiple processes may access a single database file 
** without risking corruption. However concurrency is reduced as
** follows:
**
**   * When a connection using asynchronous IO begins a database
**     transaction, the database is locked immediately. However the
**     lock is not released until after all relevant operations
**     in the write-queue have been flushed to disk. This means
**     (for example) that the database may remain locked for some 
**     time after a "COMMIT" or "ROLLBACK" is issued.
**
**   * If an application using asynchronous IO executes transactions
**     in quick succession, other database users may be effectively
**     locked out of the database. This is because when a BEGIN
**     is executed, a database lock is established immediately. But
**     when the corresponding COMMIT or ROLLBACK occurs, the lock
**     is not released until the relevant part of the write-queue 
**     has been flushed through. As a result, if a COMMIT is followed
**     by a BEGIN before the write-queue is flushed through, the database 
**     is never unlocked,preventing other processes from accessing 
**     the database.
**
** Defining ENABLE_FILE_LOCKING when using an NFS or other remote 
** file-system may slow things down, as synchronous round-trips to the 
** server may be required to establish database file locks.
*/

#if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_ASYNCIO)

#include "sqlite3async.h"

#define ENABLE_FILE_LOCKING

#ifndef SQLITE_AMALGAMATION
# include "sqliteInt.h"
# include <assert.h>
# include <string.h>
#endif

/* Useful macros used in several places */
#define MIN(x,y) ((x)<(y)?(x):(y))
#define MAX(x,y) ((x)>(y)?(x):(y))

/* Forward references */
typedef struct AsyncWrite AsyncWrite;
typedef struct AsyncFile AsyncFile;
typedef struct AsyncFileData AsyncFileData;
typedef struct AsyncFileLock AsyncFileLock;
typedef struct AsyncLock AsyncLock;

/* Enable for debugging */
static int sqlite3async_trace = 0;
# define ASYNC_TRACE(X) if( sqlite3async_trace ) asyncTrace X
static void asyncTrace(const char *zFormat, ...){
  char *z;
  va_list ap;
  va_start(ap, zFormat);
  z = sqlite3_vmprintf(zFormat, ap);
  va_end(ap);
  fprintf(stderr, "[%d] %s", 0 /* (int)pthread_self() */, z);
  sqlite3_free(z);
}

/*
** THREAD SAFETY NOTES
**
** Basic rules:
**
**     * Both read and write access to the global write-op queue must be 
**       protected by the async.queueMutex. As are the async.ioError and
**       async.nFile variables.
**
**     * The async.pLock list and all AsyncLock and AsyncFileLock
**       structures must be protected by the async.lockMutex mutex.
**
**     * The file handles from the underlying system are not assumed to 
**       be thread safe.
**
**     * See the last two paragraphs under "The Writer Thread" for
**       an assumption to do with file-handle synchronization by the Os.
**
** Deadlock prevention:
**
**     There are three mutex used by the system: the "writer" mutex, 
**     the "queue" mutex and the "lock" mutex. Rules are:
**
**     * It is illegal to block on the writer mutex when any other mutex
**       are held, and 
**
**     * It is illegal to block on the queue mutex when the lock mutex
**       is held.
**
**     i.e. mutex's must be grabbed in the order "writer", "queue", "lock".
**
** File system operations (invoked by SQLite thread):
**
**     xOpen
**     xDelete
**     xFileExists
**
** File handle operations (invoked by SQLite thread):
**
**         asyncWrite, asyncClose, asyncTruncate, asyncSync 
**    
**     The operations above add an entry to the global write-op list. They
**     prepare the entry, acquire the async.queueMutex momentarily while
**     list pointers are  manipulated to insert the new entry, then release
**     the mutex and signal the writer thread to wake up in case it happens
**     to be asleep.
**
**    
**         asyncRead, asyncFileSize.
**
**     Read operations. Both of these read from both the underlying file
**     first then adjust their result based on pending writes in the 
**     write-op queue.   So async.queueMutex is held for the duration
**     of these operations to prevent other threads from changing the
**     queue in mid operation.
**    
**
**         asyncLock, asyncUnlock, asyncCheckReservedLock
**    
**     These primitives implement in-process locking using a hash table
**     on the file name.  Files are locked correctly for connections coming
**     from the same process.  But other processes cannot see these locks
**     and will therefore not honor them.
**
**
** The writer thread:
**
**     The async.writerMutex is used to make sure only there is only
**     a single writer thread running at a time.
**
**     Inside the writer thread is a loop that works like this:
**
**         WHILE (write-op list is not empty)
**             Do IO operation at head of write-op list
**             Remove entry from head of write-op list
**         END WHILE
**
**     The async.queueMutex is always held during the <write-op list is 
**     not empty> test, and when the entry is removed from the head
**     of the write-op list. Sometimes it is held for the interim
**     period (while the IO is performed), and sometimes it is
**     relinquished. It is relinquished if (a) the IO op is an
**     ASYNC_CLOSE or (b) when the file handle was opened, two of
**     the underlying systems handles were opened on the same
**     file-system entry.
**
**     If condition (b) above is true, then one file-handle 
**     (AsyncFile.pBaseRead) is used exclusively by sqlite threads to read the
**     file, the other (AsyncFile.pBaseWrite) by sqlite3_async_flush() 
**     threads to perform write() operations. This means that read 
**     operations are not blocked by asynchronous writes (although 
**     asynchronous writes may still be blocked by reads).
**
**     This assumes that the OS keeps two handles open on the same file
**     properly in sync. That is, any read operation that starts after a
**     write operation on the same file system entry has completed returns
**     data consistent with the write. We also assume that if one thread 
**     reads a file while another is writing it all bytes other than the
**     ones actually being written contain valid data.
**
**     If the above assumptions are not true, set the preprocessor symbol
**     SQLITE_ASYNC_TWO_FILEHANDLES to 0.
*/


#ifndef NDEBUG
# define TESTONLY( X ) X
#else
# define TESTONLY( X )
#endif

/*
** There are two definitions of the following functions. One for pthreads
** compatible systems and one for Win32. These functions isolate the OS
** specific code required by each platform.
**
** The system uses three mutexes and a single condition variable. To
** block on a mutex, async_mutex_enter() is called. The parameter passed
** to async_mutex_enter(), which must be one of ASYNC_MUTEX_LOCK,
** ASYNC_MUTEX_QUEUE or ASYNC_MUTEX_WRITER, identifies which of the three
** mutexes to lock. Similarly, to unlock a mutex, async_mutex_leave() is
** called with a parameter identifying the mutex being unlocked. Mutexes
** are not recursive - it is an error to call async_mutex_enter() to
** lock a mutex that is already locked, or to call async_mutex_leave()
** to unlock a mutex that is not currently locked.
**
** The async_cond_wait() and async_cond_signal() functions are modelled
** on the pthreads functions with similar names. The first parameter to
** both functions is always ASYNC_COND_QUEUE. When async_cond_wait()
** is called the mutex identified by the second parameter must be held.
** The mutex is unlocked, and the calling thread simultaneously begins 
** waiting for the condition variable to be signalled by another thread.
** After another thread signals the condition variable, the calling
** thread stops waiting, locks mutex eMutex and returns. The 
** async_cond_signal() function is used to signal the condition variable. 
** It is assumed that the mutex used by the thread calling async_cond_wait() 
** is held by the caller of async_cond_signal() (otherwise there would be 
** a race condition).
**
** It is guaranteed that no other thread will call async_cond_wait() when
** there is already a thread waiting on the condition variable.
**
** The async_sched_yield() function is called to suggest to the operating
** system that it would be a good time to shift the current thread off the
** CPU. The system will still work if this function is not implemented
** (it is not currently implemented for win32), but it might be marginally
** more efficient if it is.
*/
static void async_mutex_enter(int eMutex);
static void async_mutex_leave(int eMutex);
static void async_cond_wait(int eCond, int eMutex);
static void async_cond_signal(int eCond);
static void async_sched_yield(void);

/*
** There are also two definitions of the following. async_os_initialize()
** is called when the asynchronous VFS is first installed, and os_shutdown()
** is called when it is uninstalled (from within sqlite3async_shutdown()).
**
** For pthreads builds, both of these functions are no-ops. For win32,
** they provide an opportunity to initialize and finalize the required
** mutex and condition variables.
**
** If async_os_initialize() returns other than zero, then the initialization
** fails and SQLITE_ERROR is returned to the user.
*/
static int async_os_initialize(void);
static void async_os_shutdown(void);

/* Values for use as the 'eMutex' argument of the above functions. The
** integer values assigned to these constants are important for assert()
** statements that verify that mutexes are locked in the correct order.
** Specifically, it is unsafe to try to lock mutex N while holding a lock 
** on mutex M if (M<=N).
*/
#define ASYNC_MUTEX_LOCK    0
#define ASYNC_MUTEX_QUEUE   1
#define ASYNC_MUTEX_WRITER  2

/* Values for use as the 'eCond' argument of the above functions. */
#define ASYNC_COND_QUEUE    0

/*************************************************************************
** Start of OS specific code.
*/
#if SQLITE_OS_WIN || defined(_WIN32) || defined(WIN32) || defined(__CYGWIN__) || defined(__MINGW32__) || defined(__BORLANDC__)

/* The following block contains the win32 specific code. */

#define mutex_held(X) (GetCurrentThreadId()==primitives.aHolder[X])

static struct AsyncPrimitives {
  int isInit;
  DWORD aHolder[3];
  CRITICAL_SECTION aMutex[3];
  HANDLE aCond[1];
} primitives = { 0 };

static int async_os_initialize(void){
  if( !primitives.isInit ){
    primitives.aCond[0] = CreateEvent(NULL, TRUE, FALSE, 0);
    if( primitives.aCond[0]==NULL ){
      return 1;
    }
    InitializeCriticalSection(&primitives.aMutex[0]);
    InitializeCriticalSection(&primitives.aMutex[1]);
    InitializeCriticalSection(&primitives.aMutex[2]);
    primitives.isInit = 1;
  }
  return 0;
}
static void async_os_shutdown(void){
  if( primitives.isInit ){
    DeleteCriticalSection(&primitives.aMutex[0]);
    DeleteCriticalSection(&primitives.aMutex[1]);
    DeleteCriticalSection(&primitives.aMutex[2]);
    CloseHandle(primitives.aCond[0]);
    primitives.isInit = 0;
  }
}

/* The following block contains the Win32 specific code. */
static void async_mutex_enter(int eMutex){
  assert( eMutex==0 || eMutex==1 || eMutex==2 );
  assert( eMutex!=2 || (!mutex_held(0) && !mutex_held(1) && !mutex_held(2)) );
  assert( eMutex!=1 || (!mutex_held(0) && !mutex_held(1)) );
  assert( eMutex!=0 || (!mutex_held(0)) );
  EnterCriticalSection(&primitives.aMutex[eMutex]);
  TESTONLY( primitives.aHolder[eMutex] = GetCurrentThreadId(); )
}
static void async_mutex_leave(int eMutex){
  assert( eMutex==0 || eMutex==1 || eMutex==2 );
  assert( mutex_held(eMutex) );
  TESTONLY( primitives.aHolder[eMutex] = 0; )
  LeaveCriticalSection(&primitives.aMutex[eMutex]);
}
static void async_cond_wait(int eCond, int eMutex){
  ResetEvent(primitives.aCond[eCond]);
  async_mutex_leave(eMutex);
  WaitForSingleObject(primitives.aCond[eCond], INFINITE);
  async_mutex_enter(eMutex);
}
static void async_cond_signal(int eCond){
  assert( mutex_held(ASYNC_MUTEX_QUEUE) );
  SetEvent(primitives.aCond[eCond]);
}
static void async_sched_yield(void){
  /* Todo: Find out if win32 offers anything like sched_yield() */
}
#else

/* The following block contains the pthreads specific code. */
#include <pthread.h>
#include <sched.h>

#define mutex_held(X) pthread_equal(primitives.aHolder[X], pthread_self())

static int  async_os_initialize(void) {return 0;}
static void async_os_shutdown(void) {}

static struct AsyncPrimitives {
  pthread_mutex_t aMutex[3];
  pthread_cond_t aCond[1];
  pthread_t aHolder[3];
} primitives = {
  { PTHREAD_MUTEX_INITIALIZER, 
    PTHREAD_MUTEX_INITIALIZER, 
    PTHREAD_MUTEX_INITIALIZER
  } , {
    PTHREAD_COND_INITIALIZER
  } , { 0, 0, 0 }
};

static void async_mutex_enter(int eMutex){
  assert( eMutex==0 || eMutex==1 || eMutex==2 );
  assert( eMutex!=2 || (!mutex_held(0) && !mutex_held(1) && !mutex_held(2)) );
  assert( eMutex!=1 || (!mutex_held(0) && !mutex_held(1)) );
  assert( eMutex!=0 || (!mutex_held(0)) );
  pthread_mutex_lock(&primitives.aMutex[eMutex]);
  TESTONLY( primitives.aHolder[eMutex] = pthread_self(); )
}
static void async_mutex_leave(int eMutex){
  assert( eMutex==0 || eMutex==1 || eMutex==2 );
  assert( mutex_held(eMutex) );
  TESTONLY( primitives.aHolder[eMutex] = 0; )
  pthread_mutex_unlock(&primitives.aMutex[eMutex]);
}
static void async_cond_wait(int eCond, int eMutex){
  assert( eMutex==0 || eMutex==1 || eMutex==2 );
  assert( mutex_held(eMutex) );
  TESTONLY( primitives.aHolder[eMutex] = 0; )
  pthread_cond_wait(&primitives.aCond[eCond], &primitives.aMutex[eMutex]);
  TESTONLY( primitives.aHolder[eMutex] = pthread_self(); )
}
static void async_cond_signal(int eCond){
  assert( mutex_held(ASYNC_MUTEX_QUEUE) );
  pthread_cond_signal(&primitives.aCond[eCond]);
}
static void async_sched_yield(void){
  sched_yield();
}
#endif
/*
** End of OS specific code.
*************************************************************************/

#define assert_mutex_is_held(X) assert( mutex_held(X) )


#ifndef SQLITE_ASYNC_TWO_FILEHANDLES
/* #define SQLITE_ASYNC_TWO_FILEHANDLES 0 */
#define SQLITE_ASYNC_TWO_FILEHANDLES 1
#endif

/*
** State information is held in the static variable "async" defined
** as the following structure.
**
** Both async.ioError and async.nFile are protected by async.queueMutex.
*/
static struct TestAsyncStaticData {
  AsyncWrite *pQueueFirst;     /* Next write operation to be processed */
  AsyncWrite *pQueueLast;      /* Last write operation on the list */
  AsyncLock *pLock;            /* Linked list of all AsyncLock structures */
  volatile int ioDelay;        /* Extra delay between write operations */
  volatile int eHalt;          /* One of the SQLITEASYNC_HALT_XXX values */
  int ioError;                 /* True if an IO error has occurred */
  int nFile;                   /* Number of open files (from sqlite pov) */
} async = { 0,0,0,0,0,0,0 };

/* Possible values of AsyncWrite.op */
#define ASYNC_NOOP          0
#define ASYNC_WRITE         1
#define ASYNC_SYNC          2
#define ASYNC_TRUNCATE      3
#define ASYNC_CLOSE         4
#define ASYNC_DELETE        5
#define ASYNC_OPENEXCLUSIVE 6
#define ASYNC_UNLOCK        7

/* Names of opcodes.  Used for debugging only.
** Make sure these stay in sync with the macros above!
*/
static const char *azOpcodeName[] = {
  "NOOP", "WRITE", "SYNC", "TRUNCATE", "CLOSE", "DELETE", "OPENEX", "UNLOCK"
};

/*
** Entries on the write-op queue are instances of the AsyncWrite
** structure, defined here.
**
** The interpretation of the iOffset and nByte variables varies depending 
** on the value of AsyncWrite.op:
**
** ASYNC_NOOP:
**     No values used.
**
** ASYNC_WRITE:
**     iOffset -> Offset in file to write to.
**     nByte   -> Number of bytes of data to write (pointed to by zBuf).
**
** ASYNC_SYNC:
**     nByte   -> flags to pass to sqlite3OsSync().
**
** ASYNC_TRUNCATE:
**     iOffset -> Size to truncate file to.
**     nByte   -> Unused.
**
** ASYNC_CLOSE:
**     iOffset -> Unused.
**     nByte   -> Unused.
**
** ASYNC_DELETE:
**     iOffset -> Contains the "syncDir" flag.
**     nByte   -> Number of bytes of zBuf points to (file name).
**
** ASYNC_OPENEXCLUSIVE:
**     iOffset -> Value of "delflag".
**     nByte   -> Number of bytes of zBuf points to (file name).
**
** ASYNC_UNLOCK:
**     nByte   -> Argument to sqlite3OsUnlock().
**
**
** For an ASYNC_WRITE operation, zBuf points to the data to write to the file. 
** This space is sqlite3_malloc()d along with the AsyncWrite structure in a
** single blob, so is deleted when sqlite3_free() is called on the parent 
** structure.
*/
struct AsyncWrite {
  AsyncFileData *pFileData;    /* File to write data to or sync */
  int op;                      /* One of ASYNC_xxx etc. */
  sqlite_int64 iOffset;        /* See above */
  int nByte;          /* See above */
  char *zBuf;         /* Data to write to file (or NULL if op!=ASYNC_WRITE) */
  AsyncWrite *pNext;  /* Next write operation (to any file) */
};

/*
** An instance of this structure is created for each distinct open file 
** (i.e. if two handles are opened on the one file, only one of these
** structures is allocated) and stored in the async.aLock hash table. The
** keys for async.aLock are the full pathnames of the opened files.
**
** AsyncLock.pList points to the head of a linked list of AsyncFileLock
** structures, one for each handle currently open on the file.
**
** If the opened file is not a main-database (the SQLITE_OPEN_MAIN_DB is
** not passed to the sqlite3OsOpen() call), or if ENABLE_FILE_LOCKING is 
** not defined at compile time, variables AsyncLock.pFile and 
** AsyncLock.eLock are never used. Otherwise, pFile is a file handle
** opened on the file in question and used to obtain the file-system 
** locks required by database connections within this process.
**
** See comments above the asyncLock() function for more details on 
** the implementation of database locking used by this backend.
*/
struct AsyncLock {
  char *zFile;
  int nFile;
  sqlite3_file *pFile;
  int eLock;
  AsyncFileLock *pList;
  AsyncLock *pNext;           /* Next in linked list headed by async.pLock */
};

/*
** An instance of the following structure is allocated along with each
** AsyncFileData structure (see AsyncFileData.lock), but is only used if the
** file was opened with the SQLITE_OPEN_MAIN_DB.
*/
struct AsyncFileLock {
  int eLock;                /* Internally visible lock state (sqlite pov) */
  int eAsyncLock;           /* Lock-state with write-queue unlock */
  AsyncFileLock *pNext;
};

/* 
** The AsyncFile structure is a subclass of sqlite3_file used for 
** asynchronous IO. 
**
** All of the actual data for the structure is stored in the structure
** pointed to by AsyncFile.pData, which is allocated as part of the
** sqlite3OsOpen() using sqlite3_malloc(). The reason for this is that the
** lifetime of the AsyncFile structure is ended by the caller after OsClose()
** is called, but the data in AsyncFileData may be required by the
** writer thread after that point.
*/
struct AsyncFile {
  sqlite3_io_methods *pMethod;
  AsyncFileData *pData;
};
struct AsyncFileData {
  char *zName;               /* Underlying OS filename - used for debugging */
  int nName;                 /* Number of characters in zName */
  sqlite3_file *pBaseRead;   /* Read handle to the underlying Os file */
  sqlite3_file *pBaseWrite;  /* Write handle to the underlying Os file */
  AsyncFileLock lock;        /* Lock state for this handle */
  AsyncLock *pLock;          /* AsyncLock object for this file system entry */
  AsyncWrite closeOp;        /* Preallocated close operation */
};

/*
** Add an entry to the end of the global write-op list. pWrite should point 
** to an AsyncWrite structure allocated using sqlite3_malloc().  The writer
** thread will call sqlite3_free() to free the structure after the specified
** operation has been completed.
**
** Once an AsyncWrite structure has been added to the list, it becomes the
** property of the writer thread and must not be read or modified by the
** caller.  
*/
static void addAsyncWrite(AsyncWrite *pWrite){
  /* We must hold the queue mutex in order to modify the queue pointers */
  if( pWrite->op!=ASYNC_UNLOCK ){
    async_mutex_enter(ASYNC_MUTEX_QUEUE);
  }

  /* Add the record to the end of the write-op queue */
  assert( !pWrite->pNext );
  if( async.pQueueLast ){
    assert( async.pQueueFirst );
    async.pQueueLast->pNext = pWrite;
  }else{
    async.pQueueFirst = pWrite;
  }
  async.pQueueLast = pWrite;
  ASYNC_TRACE(("PUSH %p (%s %s %d)\n", pWrite, azOpcodeName[pWrite->op],
         pWrite->pFileData ? pWrite->pFileData->zName : "-", pWrite->iOffset));

  if( pWrite->op==ASYNC_CLOSE ){
    async.nFile--;
  }

  /* The writer thread might have been idle because there was nothing
  ** on the write-op queue for it to do.  So wake it up. */
  async_cond_signal(ASYNC_COND_QUEUE);

  /* Drop the queue mutex */
  if( pWrite->op!=ASYNC_UNLOCK ){
    async_mutex_leave(ASYNC_MUTEX_QUEUE);
  }
}

/*
** Increment async.nFile in a thread-safe manner.
*/
static void incrOpenFileCount(void){
  /* We must hold the queue mutex in order to modify async.nFile */
  async_mutex_enter(ASYNC_MUTEX_QUEUE);
  if( async.nFile==0 ){
    async.ioError = SQLITE_OK;
  }
  async.nFile++;
  async_mutex_leave(ASYNC_MUTEX_QUEUE);
}

/*
** This is a utility function to allocate and populate a new AsyncWrite
** structure and insert it (via addAsyncWrite() ) into the global list.
*/
static int addNewAsyncWrite(
  AsyncFileData *pFileData, 
  int op, 
  sqlite3_int64 iOffset, 
  int nByte,
  const char *zByte
){
  AsyncWrite *p;
  if( op!=ASYNC_CLOSE && async.ioError ){
    return async.ioError;
  }
  p = sqlite3_malloc(sizeof(AsyncWrite) + (zByte?nByte:0));
  if( !p ){
    /* The upper layer does not expect operations like OsWrite() to
    ** return SQLITE_NOMEM. This is partly because under normal conditions
    ** SQLite is required to do rollback without calling malloc(). So
    ** if malloc() fails here, treat it as an I/O error. The above
    ** layer knows how to handle that.
    */
    return SQLITE_IOERR;
  }
  p->op = op;
  p->iOffset = iOffset;
  p->nByte = nByte;
  p->pFileData = pFileData;
  p->pNext = 0;
  if( zByte ){
    p->zBuf = (char *)&p[1];
    memcpy(p->zBuf, zByte, nByte);
  }else{
    p->zBuf = 0;
  }
  addAsyncWrite(p);
  return SQLITE_OK;
}

/*
** Close the file. This just adds an entry to the write-op list, the file is
** not actually closed.
*/
static int asyncClose(sqlite3_file *pFile){
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;

  /* Unlock the file, if it is locked */
  async_mutex_enter(ASYNC_MUTEX_LOCK);
  p->lock.eLock = 0;
  async_mutex_leave(ASYNC_MUTEX_LOCK);

  addAsyncWrite(&p->closeOp);
  return SQLITE_OK;
}

/*
** Implementation of sqlite3OsWrite() for asynchronous files. Instead of 
** writing to the underlying file, this function adds an entry to the end of
** the global AsyncWrite list. Either SQLITE_OK or SQLITE_NOMEM may be
** returned.
*/
static int asyncWrite(
  sqlite3_file *pFile, 
  const void *pBuf, 
  int amt, 
  sqlite3_int64 iOff
){
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;
  return addNewAsyncWrite(p, ASYNC_WRITE, iOff, amt, pBuf);
}

/*
** Read data from the file. First we read from the filesystem, then adjust 
** the contents of the buffer based on ASYNC_WRITE operations in the 
** write-op queue.
**
** This method holds the mutex from start to finish.
*/
static int asyncRead(
  sqlite3_file *pFile, 
  void *zOut, 
  int iAmt, 
  sqlite3_int64 iOffset
){
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;
  int rc = SQLITE_OK;
  sqlite3_int64 filesize;
  int nRead;
  sqlite3_file *pBase = p->pBaseRead;

  /* Grab the write queue mutex for the duration of the call */
  async_mutex_enter(ASYNC_MUTEX_QUEUE);

  /* If an I/O error has previously occurred in this virtual file 
  ** system, then all subsequent operations fail.
  */
  if( async.ioError!=SQLITE_OK ){
    rc = async.ioError;
    goto asyncread_out;
  }

  if( pBase->pMethods ){
    rc = pBase->pMethods->xFileSize(pBase, &filesize);
    if( rc!=SQLITE_OK ){
      goto asyncread_out;
    }
    nRead = MIN(filesize - iOffset, iAmt);
    if( nRead>0 ){
      rc = pBase->pMethods->xRead(pBase, zOut, nRead, iOffset);
      ASYNC_TRACE(("READ %s %d bytes at %d\n", p->zName, nRead, iOffset));
    }
  }

  if( rc==SQLITE_OK ){
    AsyncWrite *pWrite;
    char *zName = p->zName;

    for(pWrite=async.pQueueFirst; pWrite; pWrite = pWrite->pNext){
      if( pWrite->op==ASYNC_WRITE && (
        (pWrite->pFileData==p) ||
        (zName && pWrite->pFileData->zName==zName)
      )){
        int iBeginOut = (pWrite->iOffset-iOffset);
        int iBeginIn = -iBeginOut;
        int nCopy;

        if( iBeginIn<0 ) iBeginIn = 0;
        if( iBeginOut<0 ) iBeginOut = 0;
        nCopy = MIN(pWrite->nByte-iBeginIn, iAmt-iBeginOut);

        if( nCopy>0 ){
          memcpy(&((char *)zOut)[iBeginOut], &pWrite->zBuf[iBeginIn], nCopy);
          ASYNC_TRACE(("OVERREAD %d bytes at %d\n", nCopy, iBeginOut+iOffset));
        }
      }
    }
  }

asyncread_out:
  async_mutex_leave(ASYNC_MUTEX_QUEUE);
  return rc;
}

/*
** Truncate the file to nByte bytes in length. This just adds an entry to 
** the write-op list, no IO actually takes place.
*/
static int asyncTruncate(sqlite3_file *pFile, sqlite3_int64 nByte){
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;
  return addNewAsyncWrite(p, ASYNC_TRUNCATE, nByte, 0, 0);
}

/*
** Sync the file. This just adds an entry to the write-op list, the 
** sync() is done later by sqlite3_async_flush().
*/
static int asyncSync(sqlite3_file *pFile, int flags){
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;
  return addNewAsyncWrite(p, ASYNC_SYNC, 0, flags, 0);
}

/*
** Read the size of the file. First we read the size of the file system 
** entry, then adjust for any ASYNC_WRITE or ASYNC_TRUNCATE operations 
** currently in the write-op list. 
**
** This method holds the mutex from start to finish.
*/
int asyncFileSize(sqlite3_file *pFile, sqlite3_int64 *piSize){
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;
  int rc = SQLITE_OK;
  sqlite3_int64 s = 0;
  sqlite3_file *pBase;

  async_mutex_enter(ASYNC_MUTEX_QUEUE);

  /* Read the filesystem size from the base file. If pBaseRead is NULL, this
  ** means the file hasn't been opened yet. In this case all relevant data 
  ** must be in the write-op queue anyway, so we can omit reading from the
  ** file-system.
  */
  pBase = p->pBaseRead;
  if( pBase->pMethods ){
    rc = pBase->pMethods->xFileSize(pBase, &s);
  }

  if( rc==SQLITE_OK ){
    AsyncWrite *pWrite;
    for(pWrite=async.pQueueFirst; pWrite; pWrite = pWrite->pNext){
      if( pWrite->op==ASYNC_DELETE 
       && p->zName 
       && strcmp(p->zName, pWrite->zBuf)==0 
      ){
        s = 0;
      }else if( pWrite->pFileData && (
          (pWrite->pFileData==p) 
       || (p->zName && pWrite->pFileData->zName==p->zName) 
      )){
        switch( pWrite->op ){
          case ASYNC_WRITE:
            s = MAX(pWrite->iOffset + (sqlite3_int64)(pWrite->nByte), s);
            break;
          case ASYNC_TRUNCATE:
            s = MIN(s, pWrite->iOffset);
            break;
        }
      }
    }
    *piSize = s;
  }
  async_mutex_leave(ASYNC_MUTEX_QUEUE);
  return rc;
}

/*
** Lock or unlock the actual file-system entry.
*/
static int getFileLock(AsyncLock *pLock){
  int rc = SQLITE_OK;
  AsyncFileLock *pIter;
  int eRequired = 0;

  if( pLock->pFile ){
    for(pIter=pLock->pList; pIter; pIter=pIter->pNext){
      assert(pIter->eAsyncLock>=pIter->eLock);
      if( pIter->eAsyncLock>eRequired ){
        eRequired = pIter->eAsyncLock;
        assert(eRequired>=0 && eRequired<=SQLITE_LOCK_EXCLUSIVE);
      }
    }

    if( eRequired>pLock->eLock ){
      rc = pLock->pFile->pMethods->xLock(pLock->pFile, eRequired);
      if( rc==SQLITE_OK ){
        pLock->eLock = eRequired;
      }
    }
    else if( eRequired<pLock->eLock && eRequired<=SQLITE_LOCK_SHARED ){
      rc = pLock->pFile->pMethods->xUnlock(pLock->pFile, eRequired);
      if( rc==SQLITE_OK ){
        pLock->eLock = eRequired;
      }
    }
  }

  return rc;
}

/*
** Return the AsyncLock structure from the global async.pLock list 
** associated with the file-system entry identified by path zName 
** (a string of nName bytes). If no such structure exists, return 0.
*/
static AsyncLock *findLock(const char *zName, int nName){
  AsyncLock *p = async.pLock;
  while( p && (p->nFile!=nName || memcmp(p->zFile, zName, nName)) ){
    p = p->pNext;
  }
  return p;
}

/*
** The following two methods - asyncLock() and asyncUnlock() - are used
** to obtain and release locks on database files opened with the
** asynchronous backend.
*/
static int asyncLock(sqlite3_file *pFile, int eLock){
  int rc = SQLITE_OK;
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;

  if( p->zName ){
    async_mutex_enter(ASYNC_MUTEX_LOCK);
    if( p->lock.eLock<eLock ){
      AsyncLock *pLock = p->pLock;
      AsyncFileLock *pIter;
      assert(pLock && pLock->pList);
      for(pIter=pLock->pList; pIter; pIter=pIter->pNext){
        if( pIter!=&p->lock && (
          (eLock==SQLITE_LOCK_EXCLUSIVE && pIter->eLock>=SQLITE_LOCK_SHARED) ||
          (eLock==SQLITE_LOCK_PENDING && pIter->eLock>=SQLITE_LOCK_RESERVED) ||
          (eLock==SQLITE_LOCK_RESERVED && pIter->eLock>=SQLITE_LOCK_RESERVED) ||
          (eLock==SQLITE_LOCK_SHARED && pIter->eLock>=SQLITE_LOCK_PENDING)
        )){
          rc = SQLITE_BUSY;
        }
      }
      if( rc==SQLITE_OK ){
        p->lock.eLock = eLock;
        p->lock.eAsyncLock = MAX(p->lock.eAsyncLock, eLock);
      }
      assert(p->lock.eAsyncLock>=p->lock.eLock);
      if( rc==SQLITE_OK ){
        rc = getFileLock(pLock);
      }
    }
    async_mutex_leave(ASYNC_MUTEX_LOCK);
  }

  ASYNC_TRACE(("LOCK %d (%s) rc=%d\n", eLock, p->zName, rc));
  return rc;
}
static int asyncUnlock(sqlite3_file *pFile, int eLock){
  int rc = SQLITE_OK;
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;
  if( p->zName ){
    AsyncFileLock *pLock = &p->lock;
    async_mutex_enter(ASYNC_MUTEX_QUEUE);
    async_mutex_enter(ASYNC_MUTEX_LOCK);
    pLock->eLock = MIN(pLock->eLock, eLock);
    rc = addNewAsyncWrite(p, ASYNC_UNLOCK, 0, eLock, 0);
    async_mutex_leave(ASYNC_MUTEX_LOCK);
    async_mutex_leave(ASYNC_MUTEX_QUEUE);
  }
  return rc;
}

/*
** This function is called when the pager layer first opens a database file
** and is checking for a hot-journal.
*/
static int asyncCheckReservedLock(sqlite3_file *pFile, int *pResOut){
  int ret = 0;
  AsyncFileLock *pIter;
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;

  async_mutex_enter(ASYNC_MUTEX_LOCK);
  for(pIter=p->pLock->pList; pIter; pIter=pIter->pNext){
    if( pIter->eLock>=SQLITE_LOCK_RESERVED ){
      ret = 1;
    }
  }
  async_mutex_leave(ASYNC_MUTEX_LOCK);

  ASYNC_TRACE(("CHECK-LOCK %d (%s)\n", ret, p->zName));
  *pResOut = ret;
  return SQLITE_OK;
}

/* 
** sqlite3_file_control() implementation.
*/
static int asyncFileControl(sqlite3_file *id, int op, void *pArg){
  switch( op ){
    case SQLITE_FCNTL_LOCKSTATE: {
      async_mutex_enter(ASYNC_MUTEX_LOCK);
      *(int*)pArg = ((AsyncFile*)id)->pData->lock.eLock;
      async_mutex_leave(ASYNC_MUTEX_LOCK);
      return SQLITE_OK;
    }
  }
  return SQLITE_ERROR;
}

/* 
** Return the device characteristics and sector-size of the device. It
** is not tricky to implement these correctly, as this backend might 
** not have an open file handle at this point.
*/
static int asyncSectorSize(sqlite3_file *pFile){
  return 512;
}
static int asyncDeviceCharacteristics(sqlite3_file *pFile){
  return 0;
}

static int unlinkAsyncFile(AsyncFileData *pData){
  AsyncFileLock **ppIter;
  int rc = SQLITE_OK;

  if( pData->zName ){
    AsyncLock *pLock = pData->pLock;
    for(ppIter=&pLock->pList; *ppIter; ppIter=&((*ppIter)->pNext)){
      if( (*ppIter)==&pData->lock ){
        *ppIter = pData->lock.pNext;
        break;
      }
    }
    if( !pLock->pList ){
      AsyncLock **pp;
      if( pLock->pFile ){
        pLock->pFile->pMethods->xClose(pLock->pFile);
      }
      for(pp=&async.pLock; *pp!=pLock; pp=&((*pp)->pNext));
      *pp = pLock->pNext;
      sqlite3_free(pLock);
    }else{
      rc = getFileLock(pLock);
    }
  }

  return rc;
}

/*
** The parameter passed to this function is a copy of a 'flags' parameter
** passed to this modules xOpen() method. This function returns true
** if the file should be opened asynchronously, or false if it should
** be opened immediately.
**
** If the file is to be opened asynchronously, then asyncOpen() will add
** an entry to the event queue and the file will not actually be opened
** until the event is processed. Otherwise, the file is opened directly
** by the caller.
*/
static int doAsynchronousOpen(int flags){
  return (flags&SQLITE_OPEN_CREATE) && (
      (flags&SQLITE_OPEN_MAIN_JOURNAL) ||
      (flags&SQLITE_OPEN_TEMP_JOURNAL) ||
      (flags&SQLITE_OPEN_DELETEONCLOSE)
  );
}

/*
** Open a file.
*/
static int asyncOpen(
  sqlite3_vfs *pAsyncVfs,
  const char *zName,
  sqlite3_file *pFile,
  int flags,
  int *pOutFlags
){
  static sqlite3_io_methods async_methods = {
    1,                               /* iVersion */
    asyncClose,                      /* xClose */
    asyncRead,                       /* xRead */
    asyncWrite,                      /* xWrite */
    asyncTruncate,                   /* xTruncate */
    asyncSync,                       /* xSync */
    asyncFileSize,                   /* xFileSize */
    asyncLock,                       /* xLock */
    asyncUnlock,                     /* xUnlock */
    asyncCheckReservedLock,          /* xCheckReservedLock */
    asyncFileControl,                /* xFileControl */
    asyncSectorSize,                 /* xSectorSize */
    asyncDeviceCharacteristics       /* xDeviceCharacteristics */
  };

  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  AsyncFile *p = (AsyncFile *)pFile;
  int nName = 0;
  int rc = SQLITE_OK;
  int nByte;
  AsyncFileData *pData;
  AsyncLock *pLock = 0;
  char *z;
  int isAsyncOpen = doAsynchronousOpen(flags);

  /* If zName is NULL, then the upper layer is requesting an anonymous file */
  if( zName ){
    nName = strlen(zName)+1;
  }

  nByte = (
    sizeof(AsyncFileData) +        /* AsyncFileData structure */
    2 * pVfs->szOsFile +           /* AsyncFileData.pBaseRead and pBaseWrite */
    nName                          /* AsyncFileData.zName */
  ); 
  z = sqlite3_malloc(nByte);
  if( !z ){
    return SQLITE_NOMEM;
  }
  memset(z, 0, nByte);
  pData = (AsyncFileData*)z;
  z += sizeof(pData[0]);
  pData->pBaseRead = (sqlite3_file*)z;
  z += pVfs->szOsFile;
  pData->pBaseWrite = (sqlite3_file*)z;
  pData->closeOp.pFileData = pData;
  pData->closeOp.op = ASYNC_CLOSE;

  if( zName ){
    z += pVfs->szOsFile;
    pData->zName = z;
    pData->nName = nName;
    memcpy(pData->zName, zName, nName);
  }

  if( !isAsyncOpen ){
    int flagsout;
    rc = pVfs->xOpen(pVfs, pData->zName, pData->pBaseRead, flags, &flagsout);
    if( rc==SQLITE_OK && (flagsout&SQLITE_OPEN_READWRITE) ){
      rc = pVfs->xOpen(pVfs, pData->zName, pData->pBaseWrite, flags, 0);
    }
    if( pOutFlags ){
      *pOutFlags = flagsout;
    }
  }

  async_mutex_enter(ASYNC_MUTEX_LOCK);

  if( zName && rc==SQLITE_OK ){
    pLock = findLock(pData->zName, pData->nName);
    if( !pLock ){
      int nByte = pVfs->szOsFile + sizeof(AsyncLock) + pData->nName + 1; 
      pLock = (AsyncLock *)sqlite3_malloc(nByte);
      if( pLock ){
        memset(pLock, 0, nByte);
#ifdef ENABLE_FILE_LOCKING
        if( flags&SQLITE_OPEN_MAIN_DB ){
          pLock->pFile = (sqlite3_file *)&pLock[1];
          rc = pVfs->xOpen(pVfs, pData->zName, pLock->pFile, flags, 0);
          if( rc!=SQLITE_OK ){
            sqlite3_free(pLock);
            pLock = 0;
          }
        }
#endif
        if( pLock ){
          pLock->nFile = pData->nName;
          pLock->zFile = &((char *)(&pLock[1]))[pVfs->szOsFile];
          memcpy(pLock->zFile, pData->zName, pLock->nFile);
          pLock->pNext = async.pLock;
          async.pLock = pLock;
        }
      }else{
        rc = SQLITE_NOMEM;
      }
    }
  }

  if( rc==SQLITE_OK ){
    p->pMethod = &async_methods;
    p->pData = pData;

    /* Link AsyncFileData.lock into the linked list of 
    ** AsyncFileLock structures for this file.
    */
    if( zName ){
      pData->lock.pNext = pLock->pList;
      pLock->pList = &pData->lock;
      pData->zName = pLock->zFile;
    }
  }else{
    if( pData->pBaseRead->pMethods ){
      pData->pBaseRead->pMethods->xClose(pData->pBaseRead);
    }
    if( pData->pBaseWrite->pMethods ){
      pData->pBaseWrite->pMethods->xClose(pData->pBaseWrite);
    }
    sqlite3_free(pData);
  }

  async_mutex_leave(ASYNC_MUTEX_LOCK);

  if( rc==SQLITE_OK ){
    incrOpenFileCount();
    pData->pLock = pLock;
  }

  if( rc==SQLITE_OK && isAsyncOpen ){
    rc = addNewAsyncWrite(pData, ASYNC_OPENEXCLUSIVE, (sqlite3_int64)flags,0,0);
    if( rc==SQLITE_OK ){
      if( pOutFlags ) *pOutFlags = flags;
    }else{
      async_mutex_enter(ASYNC_MUTEX_LOCK);
      unlinkAsyncFile(pData);
      async_mutex_leave(ASYNC_MUTEX_LOCK);
      sqlite3_free(pData);
    }
  }
  if( rc!=SQLITE_OK ){
    p->pMethod = 0;
  }
  return rc;
}

/*
** Implementation of sqlite3OsDelete. Add an entry to the end of the 
** write-op queue to perform the delete.
*/
static int asyncDelete(sqlite3_vfs *pAsyncVfs, const char *z, int syncDir){
  return addNewAsyncWrite(0, ASYNC_DELETE, syncDir, strlen(z)+1, z);
}

/*
** Implementation of sqlite3OsAccess. This method holds the mutex from
** start to finish.
*/
static int asyncAccess(
  sqlite3_vfs *pAsyncVfs, 
  const char *zName, 
  int flags,
  int *pResOut
){
  int rc;
  int ret;
  AsyncWrite *p;
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;

  assert(flags==SQLITE_ACCESS_READWRITE 
      || flags==SQLITE_ACCESS_READ 
      || flags==SQLITE_ACCESS_EXISTS 
  );

  async_mutex_enter(ASYNC_MUTEX_QUEUE);
  rc = pVfs->xAccess(pVfs, zName, flags, &ret);
  if( rc==SQLITE_OK && flags==SQLITE_ACCESS_EXISTS ){
    for(p=async.pQueueFirst; p; p = p->pNext){
      if( p->op==ASYNC_DELETE && 0==strcmp(p->zBuf, zName) ){
        ret = 0;
      }else if( p->op==ASYNC_OPENEXCLUSIVE 
             && p->pFileData->zName
             && 0==strcmp(p->pFileData->zName, zName) 
      ){
        ret = 1;
      }
    }
  }
  ASYNC_TRACE(("ACCESS(%s): %s = %d\n", 
    flags==SQLITE_ACCESS_READWRITE?"read-write":
    flags==SQLITE_ACCESS_READ?"read":"exists"
    , zName, ret)
  );
  async_mutex_leave(ASYNC_MUTEX_QUEUE);
  *pResOut = ret;
  return rc;
}

/*
** Fill in zPathOut with the full path to the file identified by zPath.
*/
static int asyncFullPathname(
  sqlite3_vfs *pAsyncVfs, 
  const char *zPath, 
  int nPathOut,
  char *zPathOut
){
  int rc;
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  rc = pVfs->xFullPathname(pVfs, zPath, nPathOut, zPathOut);

  /* Because of the way intra-process file locking works, this backend
  ** needs to return a canonical path. The following block assumes the
  ** file-system uses unix style paths. 
  */
  if( rc==SQLITE_OK ){
    int i, j;
    int n = nPathOut;
    char *z = zPathOut;
    while( n>1 && z[n-1]=='/' ){ n--; }
    for(i=j=0; i<n; i++){
      if( z[i]=='/' ){
        if( z[i+1]=='/' ) continue;
        if( z[i+1]=='.' && i+2<n && z[i+2]=='/' ){
          i += 1;
          continue;
        }
        if( z[i+1]=='.' && i+3<n && z[i+2]=='.' && z[i+3]=='/' ){
          while( j>0 && z[j-1]!='/' ){ j--; }
          if( j>0 ){ j--; }
          i += 2;
          continue;
        }
      }
      z[j++] = z[i];
    }
    z[j] = 0;
  }

  return rc;
}
static void *asyncDlOpen(sqlite3_vfs *pAsyncVfs, const char *zPath){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  return pVfs->xDlOpen(pVfs, zPath);
}
static void asyncDlError(sqlite3_vfs *pAsyncVfs, int nByte, char *zErrMsg){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  pVfs->xDlError(pVfs, nByte, zErrMsg);
}
static void (*asyncDlSym(
  sqlite3_vfs *pAsyncVfs, 
  void *pHandle, 
  const char *zSymbol
))(void){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  return pVfs->xDlSym(pVfs, pHandle, zSymbol);
}
static void asyncDlClose(sqlite3_vfs *pAsyncVfs, void *pHandle){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  pVfs->xDlClose(pVfs, pHandle);
}
static int asyncRandomness(sqlite3_vfs *pAsyncVfs, int nByte, char *zBufOut){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  return pVfs->xRandomness(pVfs, nByte, zBufOut);
}
static int asyncSleep(sqlite3_vfs *pAsyncVfs, int nMicro){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  return pVfs->xSleep(pVfs, nMicro);
}
static int asyncCurrentTime(sqlite3_vfs *pAsyncVfs, double *pTimeOut){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  return pVfs->xCurrentTime(pVfs, pTimeOut);
}

static sqlite3_vfs async_vfs = {
  1,                    /* iVersion */
  sizeof(AsyncFile),    /* szOsFile */
  0,                    /* mxPathname */
  0,                    /* pNext */
  SQLITEASYNC_VFSNAME,  /* zName */
  0,                    /* pAppData */
  asyncOpen,            /* xOpen */
  asyncDelete,          /* xDelete */
  asyncAccess,          /* xAccess */
  asyncFullPathname,    /* xFullPathname */
  asyncDlOpen,          /* xDlOpen */
  asyncDlError,         /* xDlError */
  asyncDlSym,           /* xDlSym */
  asyncDlClose,         /* xDlClose */
  asyncRandomness,      /* xDlError */
  asyncSleep,           /* xDlSym */
  asyncCurrentTime      /* xDlClose */
};

/* 
** This procedure runs in a separate thread, reading messages off of the
** write queue and processing them one by one.  
**
** If async.writerHaltNow is true, then this procedure exits
** after processing a single message.
**
** If async.writerHaltWhenIdle is true, then this procedure exits when
** the write queue is empty.
**
** If both of the above variables are false, this procedure runs
** indefinately, waiting for operations to be added to the write queue
** and processing them in the order in which they arrive.
**
** An artifical delay of async.ioDelay milliseconds is inserted before
** each write operation in order to simulate the effect of a slow disk.
**
** Only one instance of this procedure may be running at a time.
*/
static void asyncWriterThread(void){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)(async_vfs.pAppData);
  AsyncWrite *p = 0;
  int rc = SQLITE_OK;
  int holdingMutex = 0;

  async_mutex_enter(ASYNC_MUTEX_WRITER);

  while( async.eHalt!=SQLITEASYNC_HALT_NOW ){
    int doNotFree = 0;
    sqlite3_file *pBase = 0;

    if( !holdingMutex ){
      async_mutex_enter(ASYNC_MUTEX_QUEUE);
    }
    while( (p = async.pQueueFirst)==0 ){
      if( async.eHalt!=SQLITEASYNC_HALT_NEVER ){
        async_mutex_leave(ASYNC_MUTEX_QUEUE);
        break;
      }else{
        ASYNC_TRACE(("IDLE\n"));
        async_cond_wait(ASYNC_COND_QUEUE, ASYNC_MUTEX_QUEUE);
        ASYNC_TRACE(("WAKEUP\n"));
      }
    }
    if( p==0 ) break;
    holdingMutex = 1;

    /* Right now this thread is holding the mutex on the write-op queue.
    ** Variable 'p' points to the first entry in the write-op queue. In
    ** the general case, we hold on to the mutex for the entire body of
    ** the loop. 
    **
    ** However in the cases enumerated below, we relinquish the mutex,
    ** perform the IO, and then re-request the mutex before removing 'p' from
    ** the head of the write-op queue. The idea is to increase concurrency with
    ** sqlite threads.
    **
    **     * An ASYNC_CLOSE operation.
    **     * An ASYNC_OPENEXCLUSIVE operation. For this one, we relinquish 
    **       the mutex, call the underlying xOpenExclusive() function, then
    **       re-aquire the mutex before seting the AsyncFile.pBaseRead 
    **       variable.
    **     * ASYNC_SYNC and ASYNC_WRITE operations, if 
    **       SQLITE_ASYNC_TWO_FILEHANDLES was set at compile time and two
    **       file-handles are open for the particular file being "synced".
    */
    if( async.ioError!=SQLITE_OK && p->op!=ASYNC_CLOSE ){
      p->op = ASYNC_NOOP;
    }
    if( p->pFileData ){
      pBase = p->pFileData->pBaseWrite;
      if( 
        p->op==ASYNC_CLOSE || 
        p->op==ASYNC_OPENEXCLUSIVE ||
        (pBase->pMethods && (p->op==ASYNC_SYNC || p->op==ASYNC_WRITE) ) 
      ){
        async_mutex_leave(ASYNC_MUTEX_QUEUE);
        holdingMutex = 0;
      }
      if( !pBase->pMethods ){
        pBase = p->pFileData->pBaseRead;
      }
    }

    switch( p->op ){
      case ASYNC_NOOP:
        break;

      case ASYNC_WRITE:
        assert( pBase );
        ASYNC_TRACE(("WRITE %s %d bytes at %d\n",
                p->pFileData->zName, p->nByte, p->iOffset));
        rc = pBase->pMethods->xWrite(pBase, (void *)(p->zBuf), p->nByte, p->iOffset);
        break;

      case ASYNC_SYNC:
        assert( pBase );
        ASYNC_TRACE(("SYNC %s\n", p->pFileData->zName));
        rc = pBase->pMethods->xSync(pBase, p->nByte);
        break;

      case ASYNC_TRUNCATE:
        assert( pBase );
        ASYNC_TRACE(("TRUNCATE %s to %d bytes\n", 
                p->pFileData->zName, p->iOffset));
        rc = pBase->pMethods->xTruncate(pBase, p->iOffset);
        break;

      case ASYNC_CLOSE: {
        AsyncFileData *pData = p->pFileData;
        ASYNC_TRACE(("CLOSE %s\n", p->pFileData->zName));
        if( pData->pBaseWrite->pMethods ){
          pData->pBaseWrite->pMethods->xClose(pData->pBaseWrite);
        }
        if( pData->pBaseRead->pMethods ){
          pData->pBaseRead->pMethods->xClose(pData->pBaseRead);
        }

        /* Unlink AsyncFileData.lock from the linked list of AsyncFileLock 
        ** structures for this file. Obtain the async.lockMutex mutex 
        ** before doing so.
        */
        async_mutex_enter(ASYNC_MUTEX_LOCK);
        rc = unlinkAsyncFile(pData);
        async_mutex_leave(ASYNC_MUTEX_LOCK);

        if( !holdingMutex ){
          async_mutex_enter(ASYNC_MUTEX_QUEUE);
          holdingMutex = 1;
        }
        assert_mutex_is_held(ASYNC_MUTEX_QUEUE);
        async.pQueueFirst = p->pNext;
        sqlite3_free(pData);
        doNotFree = 1;
        break;
      }

      case ASYNC_UNLOCK: {
        AsyncWrite *pIter;
        AsyncFileData *pData = p->pFileData;
        int eLock = p->nByte;

        /* When a file is locked by SQLite using the async backend, it is 
        ** locked within the 'real' file-system synchronously. When it is
        ** unlocked, an ASYNC_UNLOCK event is added to the write-queue to
        ** unlock the file asynchronously. The design of the async backend
        ** requires that the 'real' file-system file be locked from the
        ** time that SQLite first locks it (and probably reads from it)
        ** until all asynchronous write events that were scheduled before
        ** SQLite unlocked the file have been processed.
        **
        ** This is more complex if SQLite locks and unlocks the file multiple
        ** times in quick succession. For example, if SQLite does: 
        ** 
        **   lock, write, unlock, lock, write, unlock
        **
        ** Each "lock" operation locks the file immediately. Each "write" 
        ** and "unlock" operation adds an event to the event queue. If the
        ** second "lock" operation is performed before the first "unlock"
        ** operation has been processed asynchronously, then the first
        ** "unlock" cannot be safely processed as is, since this would mean
        ** the file was unlocked when the second "write" operation is
        ** processed. To work around this, when processing an ASYNC_UNLOCK
        ** operation, SQLite:
        **
        **   1) Unlocks the file to the minimum of the argument passed to
        **      the xUnlock() call and the current lock from SQLite's point
        **      of view, and
        **
        **   2) Only unlocks the file at all if this event is the last
        **      ASYNC_UNLOCK event on this file in the write-queue.
        */ 
        assert( holdingMutex==1 );
        assert( async.pQueueFirst==p );
        for(pIter=async.pQueueFirst->pNext; pIter; pIter=pIter->pNext){
          if( pIter->pFileData==pData && pIter->op==ASYNC_UNLOCK ) break;
        }
        if( !pIter ){
          async_mutex_enter(ASYNC_MUTEX_LOCK);
          pData->lock.eAsyncLock = MIN(
              pData->lock.eAsyncLock, MAX(pData->lock.eLock, eLock)
          );
          assert(pData->lock.eAsyncLock>=pData->lock.eLock);
          rc = getFileLock(pData->pLock);
          async_mutex_leave(ASYNC_MUTEX_LOCK);
        }
        break;
      }

      case ASYNC_DELETE:
        ASYNC_TRACE(("DELETE %s\n", p->zBuf));
        rc = pVfs->xDelete(pVfs, p->zBuf, (int)p->iOffset);
        break;

      case ASYNC_OPENEXCLUSIVE: {
        int flags = (int)p->iOffset;
        AsyncFileData *pData = p->pFileData;
        ASYNC_TRACE(("OPEN %s flags=%d\n", p->zBuf, (int)p->iOffset));
        assert(pData->pBaseRead->pMethods==0 && pData->pBaseWrite->pMethods==0);
        rc = pVfs->xOpen(pVfs, pData->zName, pData->pBaseRead, flags, 0);
        assert( holdingMutex==0 );
        async_mutex_enter(ASYNC_MUTEX_QUEUE);
        holdingMutex = 1;
        break;
      }

      default: assert(!"Illegal value for AsyncWrite.op");
    }

    /* If we didn't hang on to the mutex during the IO op, obtain it now
    ** so that the AsyncWrite structure can be safely removed from the 
    ** global write-op queue.
    */
    if( !holdingMutex ){
      async_mutex_enter(ASYNC_MUTEX_QUEUE);
      holdingMutex = 1;
    }
    /* ASYNC_TRACE(("UNLINK %p\n", p)); */
    if( p==async.pQueueLast ){
      async.pQueueLast = 0;
    }
    if( !doNotFree ){
      assert_mutex_is_held(ASYNC_MUTEX_QUEUE);
      async.pQueueFirst = p->pNext;
      sqlite3_free(p);
    }
    assert( holdingMutex );

    /* An IO error has occurred. We cannot report the error back to the
    ** connection that requested the I/O since the error happened 
    ** asynchronously.  The connection has already moved on.  There 
    ** really is nobody to report the error to.
    **
    ** The file for which the error occurred may have been a database or
    ** journal file. Regardless, none of the currently queued operations
    ** associated with the same database should now be performed. Nor should
    ** any subsequently requested IO on either a database or journal file 
    ** handle for the same database be accepted until the main database
    ** file handle has been closed and reopened.
    **
    ** Furthermore, no further IO should be queued or performed on any file
    ** handle associated with a database that may have been part of a 
    ** multi-file transaction that included the database associated with 
    ** the IO error (i.e. a database ATTACHed to the same handle at some 
    ** point in time).
    */
    if( rc!=SQLITE_OK ){
      async.ioError = rc;
    }

    if( async.ioError && !async.pQueueFirst ){
      async_mutex_enter(ASYNC_MUTEX_LOCK);
      if( 0==async.pLock ){
        async.ioError = SQLITE_OK;
      }
      async_mutex_leave(ASYNC_MUTEX_LOCK);
    }

    /* Drop the queue mutex before continuing to the next write operation
    ** in order to give other threads a chance to work with the write queue.
    */
    if( !async.pQueueFirst || !async.ioError ){
      async_mutex_leave(ASYNC_MUTEX_QUEUE);
      holdingMutex = 0;
      if( async.ioDelay>0 ){
        pVfs->xSleep(pVfs, async.ioDelay);
      }else{
        async_sched_yield();
      }
    }
  }
  
  async_mutex_leave(ASYNC_MUTEX_WRITER);
  return;
}

/*
** Install the asynchronous VFS.
*/ 
int sqlite3async_initialize(const char *zParent, int isDefault){
  int rc = SQLITE_OK;
  if( async_vfs.pAppData==0 ){
    sqlite3_vfs *pParent = sqlite3_vfs_find(zParent);
    if( !pParent || async_os_initialize() ){
      rc = SQLITE_ERROR;
    }else if( SQLITE_OK!=(rc = sqlite3_vfs_register(&async_vfs, isDefault)) ){
      async_os_shutdown();
    }else{
      async_vfs.pAppData = (void *)pParent;
      async_vfs.mxPathname = ((sqlite3_vfs *)async_vfs.pAppData)->mxPathname;
    }
  }
  return rc;
}

/*
** Uninstall the asynchronous VFS.
*/
void sqlite3async_shutdown(void){
  if( async_vfs.pAppData ){
    async_os_shutdown();
    sqlite3_vfs_unregister((sqlite3_vfs *)&async_vfs);
    async_vfs.pAppData = 0;
  }
}

/*
** Process events on the write-queue.
*/
void sqlite3async_run(void){
  asyncWriterThread();
}

/*
** Control/configure the asynchronous IO system.
*/
int sqlite3async_control(int op, ...){
  va_list ap;
  va_start(ap, op);
  switch( op ){
    case SQLITEASYNC_HALT: {
      int eWhen = va_arg(ap, int);
      if( eWhen!=SQLITEASYNC_HALT_NEVER
       && eWhen!=SQLITEASYNC_HALT_NOW
       && eWhen!=SQLITEASYNC_HALT_IDLE
      ){
        return SQLITE_ERROR;
      }
      async.eHalt = eWhen;
      async_mutex_enter(ASYNC_MUTEX_QUEUE);
      async_cond_signal(ASYNC_COND_QUEUE);
      async_mutex_leave(ASYNC_MUTEX_QUEUE);
      break;
    }

    case SQLITEASYNC_DELAY: {
      int iDelay = va_arg(ap, int);
      async.ioDelay = iDelay;
      break;
    }
      
    case SQLITEASYNC_GET_HALT: {
      int *peWhen = va_arg(ap, int *);
      *peWhen = async.eHalt;
      break;
    }
    case SQLITEASYNC_GET_DELAY: {
      int *piDelay = va_arg(ap, int *);
      *piDelay = async.ioDelay;
      break;
    }

    default:
      return SQLITE_ERROR;
  }
  return SQLITE_OK;
}

#endif /* !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_ASYNCIO) */

Added ext/async/sqlite3async.h.

























































































>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
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

#ifndef __SQLITEASYNC_H_
#define __SQLITEASYNC_H_ 1

#define SQLITEASYNC_VFSNAME "sqlite3async"

/*
** Install the asynchronous IO VFS.
*/ 
int sqlite3async_initialize(const char *zParent, int isDefault);

/*
** Uninstall the asynchronous IO VFS.
*/ 
void sqlite3async_shutdown();

/*
** Process events on the write-queue.
*/
void sqlite3async_run();

/*
** Control/configure the asynchronous IO system.
*/
int sqlite3async_control(int op, ...);

/*
** Values that can be used as the first argument to sqlite3async_control().
*/
#define SQLITEASYNC_HALT       1
#define SQLITEASYNC_DELAY      2
#define SQLITEASYNC_GET_HALT   3
#define SQLITEASYNC_GET_DELAY  4

/*
** If the first argument to sqlite3async_control() is SQLITEASYNC_HALT,
** the second argument should be one of the following.
*/
#define SQLITEASYNC_HALT_NEVER 0       /* Never halt (default value) */
#define SQLITEASYNC_HALT_NOW   1       /* Halt as soon as possible */
#define SQLITEASYNC_HALT_IDLE  2       /* Halt when write-queue is empty */

#endif        /* ifndef __SQLITEASYNC_H_ */

Changes to main.mk.

42
43
44
45
46
47
48

49
50
51
52
53
54
55
...
251
252
253
254
255
256
257
258

259
260
261
262
263
264
265
# build the SQLite library and testing tools.
################################################################################

# This is how we compile
#
TCCX =  $(TCC) $(OPTS) -I. -I$(TOP)/src -I$(TOP) 
TCCX += -I$(TOP)/ext/rtree -I$(TOP)/ext/icu -I$(TOP)/ext/fts3


# Object files for the SQLite library.
#
LIBOBJ+= alter.o analyze.o attach.o auth.o \
         backup.o bitvec.o btmutex.o btree.o build.o \
         callback.o complete.o date.o delete.o expr.o fault.o \
         fts3.o fts3_expr.o fts3_hash.o fts3_icu.o fts3_porter.o \
................................................................................
  $(TOP)/src/os_os2.c $(TOP)/src/os_unix.c $(TOP)/src/os_win.c                 \
  $(TOP)/src/pager.c $(TOP)/src/pragma.c $(TOP)/src/prepare.c                  \
  $(TOP)/src/printf.c $(TOP)/src/random.c $(TOP)/src/pcache.c                  \
  $(TOP)/src/pcache1.c $(TOP)/src/select.c $(TOP)/src/tokenize.c               \
  $(TOP)/src/utf.c $(TOP)/src/util.c $(TOP)/src/vdbeapi.c $(TOP)/src/vdbeaux.c \
  $(TOP)/src/vdbe.c $(TOP)/src/vdbemem.c $(TOP)/src/where.c parse.c            \
  $(TOP)/ext/fts3/fts3.c $(TOP)/ext/fts3/fts3_expr.c                           \
  $(TOP)/ext/fts3/fts3_tokenizer.c 


# Header files used by all library source files.
#
HDR = \
   $(TOP)/src/btree.h \
   $(TOP)/src/btreeInt.h \
   $(TOP)/src/hash.h \







>







 







|
>







42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
...
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
# build the SQLite library and testing tools.
################################################################################

# This is how we compile
#
TCCX =  $(TCC) $(OPTS) -I. -I$(TOP)/src -I$(TOP) 
TCCX += -I$(TOP)/ext/rtree -I$(TOP)/ext/icu -I$(TOP)/ext/fts3
TCCX += -I$(TOP)/ext/async

# Object files for the SQLite library.
#
LIBOBJ+= alter.o analyze.o attach.o auth.o \
         backup.o bitvec.o btmutex.o btree.o build.o \
         callback.o complete.o date.o delete.o expr.o fault.o \
         fts3.o fts3_expr.o fts3_hash.o fts3_icu.o fts3_porter.o \
................................................................................
  $(TOP)/src/os_os2.c $(TOP)/src/os_unix.c $(TOP)/src/os_win.c                 \
  $(TOP)/src/pager.c $(TOP)/src/pragma.c $(TOP)/src/prepare.c                  \
  $(TOP)/src/printf.c $(TOP)/src/random.c $(TOP)/src/pcache.c                  \
  $(TOP)/src/pcache1.c $(TOP)/src/select.c $(TOP)/src/tokenize.c               \
  $(TOP)/src/utf.c $(TOP)/src/util.c $(TOP)/src/vdbeapi.c $(TOP)/src/vdbeaux.c \
  $(TOP)/src/vdbe.c $(TOP)/src/vdbemem.c $(TOP)/src/where.c parse.c            \
  $(TOP)/ext/fts3/fts3.c $(TOP)/ext/fts3/fts3_expr.c                           \
  $(TOP)/ext/fts3/fts3_tokenizer.c                                             \
  $(TOP)/ext/async/sqlite3async.c

# Header files used by all library source files.
#
HDR = \
   $(TOP)/src/btree.h \
   $(TOP)/src/btreeInt.h \
   $(TOP)/src/hash.h \

Changes to src/test_async.c.

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
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678

1679
1680
1681
1682
1683
1684
1685
....
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702





1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738

1739
1740
1741
1742
1743
1744
1745
....
1748
1749
1750
1751
1752
1753
1754

1755
1756
1757
1758
1759
1760
1761
1762
1763
1764

1765






1766
1767








1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779



1780


1781
1782
1783


1784
1785
1786
1787
1788
1789

1790

1791
1792
1793
1794
1795
1796
1797
....
1802
1803
1804
1805
1806
1807
1808
1809
1810






1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851

**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
**
** $Id: test_async.c,v 1.58 2009/04/21 18:20:45 danielk1977 Exp $
**
** This file contains an example implementation of an asynchronous IO 
** backend for SQLite.
**
** WHAT IS ASYNCHRONOUS I/O?
**
** With asynchronous I/O, write requests are handled by a separate thread
** running in the background.  This means that the thread that initiates
** a database write does not have to wait for (sometimes slow) disk I/O
** to occur.  The write seems to happen very quickly, though in reality
** it is happening at its usual slow pace in the background.
**
** Asynchronous I/O appears to give better responsiveness, but at a price.
** You lose the Durable property.  With the default I/O backend of SQLite,
** once a write completes, you know that the information you wrote is
** safely on disk.  With the asynchronous I/O, this is not the case.  If
** your program crashes or if a power loss occurs after the database
** write but before the asynchronous write thread has completed, then the
** database change might never make it to disk and the next user of the
** database might not see your change.
**
** You lose Durability with asynchronous I/O, but you still retain the
** other parts of ACID:  Atomic,  Consistent, and Isolated.  Many
** appliations get along fine without the Durablity.
**
** HOW IT WORKS
**
** Asynchronous I/O works by creating a special SQLite "vfs" structure
** and registering it with sqlite3_vfs_register(). When files opened via 
** this vfs are written to (using sqlite3OsWrite()), the data is not 
** written directly to disk, but is placed in the "write-queue" to be
** handled by the background thread.
**
** When files opened with the asynchronous vfs are read from 
** (using sqlite3OsRead()), the data is read from the file on 
** disk and the write-queue, so that from the point of view of
** the vfs reader the OsWrite() appears to have already completed.
**
** The special vfs is registered (and unregistered) by calls to 
** function asyncEnable() (see below).
**
** LIMITATIONS
**
** This demonstration code is deliberately kept simple in order to keep
** the main ideas clear and easy to understand.  Real applications that
** want to do asynchronous I/O might want to add additional capabilities.
** For example, in this demonstration if writes are happening at a steady
** stream that exceeds the I/O capability of the background writer thread,
** the queue of pending write operations will grow without bound until we
** run out of memory.  Users of this technique may want to keep track of
** the quantity of pending writes and stop accepting new write requests
** when the buffer gets to be too big.
**
** LOCKING + CONCURRENCY
**
** Multiple connections from within a single process that use this
** implementation of asynchronous IO may access a single database
** file concurrently. From the point of view of the user, if all
** connections are from within a single process, there is no difference
** between the concurrency offered by "normal" SQLite and SQLite
** using the asynchronous backend.
**
** If connections from within multiple processes may access the
** database file, the ENABLE_FILE_LOCKING symbol (see below) must be
** defined. If it is not defined, then no locks are established on 
** the database file. In this case, if multiple processes access 
** the database file, corruption will quickly result.
**
** If ENABLE_FILE_LOCKING is defined (the default), then connections 
** from within multiple processes may access a single database file 
** without risking corruption. However concurrency is reduced as
** follows:
**
**   * When a connection using asynchronous IO begins a database
**     transaction, the database is locked immediately. However the
**     lock is not released until after all relevant operations
**     in the write-queue have been flushed to disk. This means
**     (for example) that the database may remain locked for some 
**     time after a "COMMIT" or "ROLLBACK" is issued.
**
**   * If an application using asynchronous IO executes transactions
**     in quick succession, other database users may be effectively
**     locked out of the database. This is because when a BEGIN
**     is executed, a database lock is established immediately. But
**     when the corresponding COMMIT or ROLLBACK occurs, the lock
**     is not released until the relevant part of the write-queue 
**     has been flushed through. As a result, if a COMMIT is followed
**     by a BEGIN before the write-queue is flushed through, the database 
**     is never unlocked,preventing other processes from accessing 
**     the database.
**
** Defining ENABLE_FILE_LOCKING when using an NFS or other remote 
** file-system may slow things down, as synchronous round-trips to the 
** server may be required to establish database file locks.
*/
#define ENABLE_FILE_LOCKING

#ifndef SQLITE_AMALGAMATION
# include "sqliteInt.h"
# include <assert.h>
# include <string.h>
#endif

#include <tcl.h>

/*
** This test uses pthreads and hence only works on unix and with
** a threadsafe build of SQLite.
*/
#if SQLITE_OS_UNIX && SQLITE_THREADSAFE


/*
** This demo uses pthreads.  If you do not have a pthreads implementation
** for your operating system, you will need to recode the threading 
** logic.
*/

#include <pthread.h>
#include <sched.h>

/* Useful macros used in several places */
#define MIN(x,y) ((x)<(y)?(x):(y))
#define MAX(x,y) ((x)>(y)?(x):(y))

/* Forward references */
typedef struct AsyncWrite AsyncWrite;
typedef struct AsyncFile AsyncFile;
typedef struct AsyncFileData AsyncFileData;
typedef struct AsyncFileLock AsyncFileLock;
typedef struct AsyncLock AsyncLock;

/* Enable for debugging */
static int sqlite3async_trace = 0;
# define ASYNC_TRACE(X) if( sqlite3async_trace ) asyncTrace X
static void asyncTrace(const char *zFormat, ...){
  char *z;
  va_list ap;
  va_start(ap, zFormat);
  z = sqlite3_vmprintf(zFormat, ap);
  va_end(ap);
  fprintf(stderr, "[%d] %s", (int)pthread_self(), z);
  sqlite3_free(z);
}

/*
** THREAD SAFETY NOTES
**
** Basic rules:
**
**     * Both read and write access to the global write-op queue must be 
**       protected by the async.queueMutex. As are the async.ioError and
**       async.nFile variables.
**
**     * The async.pLock list and all AsyncLock and AsyncFileLock
**       structures must be protected by the async.lockMutex mutex.
**
**     * The file handles from the underlying system are not assumed to 
**       be thread safe.
**
**     * See the last two paragraphs under "The Writer Thread" for
**       an assumption to do with file-handle synchronization by the Os.
**
** Deadlock prevention:
**
**     There are three mutex used by the system: the "writer" mutex, 
**     the "queue" mutex and the "lock" mutex. Rules are:
**
**     * It is illegal to block on the writer mutex when any other mutex
**       are held, and 
**
**     * It is illegal to block on the queue mutex when the lock mutex
**       is held.
**
**     i.e. mutex's must be grabbed in the order "writer", "queue", "lock".
**
** File system operations (invoked by SQLite thread):
**
**     xOpen
**     xDelete
**     xFileExists
**
** File handle operations (invoked by SQLite thread):
**
**         asyncWrite, asyncClose, asyncTruncate, asyncSync 
**    
**     The operations above add an entry to the global write-op list. They
**     prepare the entry, acquire the async.queueMutex momentarily while
**     list pointers are  manipulated to insert the new entry, then release
**     the mutex and signal the writer thread to wake up in case it happens
**     to be asleep.
**
**    
**         asyncRead, asyncFileSize.
**
**     Read operations. Both of these read from both the underlying file
**     first then adjust their result based on pending writes in the 
**     write-op queue.   So async.queueMutex is held for the duration
**     of these operations to prevent other threads from changing the
**     queue in mid operation.
**    
**
**         asyncLock, asyncUnlock, asyncCheckReservedLock
**    
**     These primitives implement in-process locking using a hash table
**     on the file name.  Files are locked correctly for connections coming
**     from the same process.  But other processes cannot see these locks
**     and will therefore not honor them.
**
**
** The writer thread:
**
**     The async.writerMutex is used to make sure only there is only
**     a single writer thread running at a time.
**
**     Inside the writer thread is a loop that works like this:
**
**         WHILE (write-op list is not empty)
**             Do IO operation at head of write-op list
**             Remove entry from head of write-op list
**         END WHILE
**
**     The async.queueMutex is always held during the <write-op list is 
**     not empty> test, and when the entry is removed from the head
**     of the write-op list. Sometimes it is held for the interim
**     period (while the IO is performed), and sometimes it is
**     relinquished. It is relinquished if (a) the IO op is an
**     ASYNC_CLOSE or (b) when the file handle was opened, two of
**     the underlying systems handles were opened on the same
**     file-system entry.
**
**     If condition (b) above is true, then one file-handle 
**     (AsyncFile.pBaseRead) is used exclusively by sqlite threads to read the
**     file, the other (AsyncFile.pBaseWrite) by sqlite3_async_flush() 
**     threads to perform write() operations. This means that read 
**     operations are not blocked by asynchronous writes (although 
**     asynchronous writes may still be blocked by reads).
**
**     This assumes that the OS keeps two handles open on the same file
**     properly in sync. That is, any read operation that starts after a
**     write operation on the same file system entry has completed returns
**     data consistent with the write. We also assume that if one thread 
**     reads a file while another is writing it all bytes other than the
**     ones actually being written contain valid data.
**
**     If the above assumptions are not true, set the preprocessor symbol
**     SQLITE_ASYNC_TWO_FILEHANDLES to 0.
*/

#ifndef SQLITE_ASYNC_TWO_FILEHANDLES
/* #define SQLITE_ASYNC_TWO_FILEHANDLES 0 */
#define SQLITE_ASYNC_TWO_FILEHANDLES 1
#endif

/*
** State information is held in the static variable "async" defined
** as the following structure.
**
** Both async.ioError and async.nFile are protected by async.queueMutex.
*/
static struct TestAsyncStaticData {
  pthread_mutex_t lockMutex;   /* For access to aLock hash table */
  pthread_mutex_t queueMutex;  /* Mutex for access to write operation queue */
  pthread_mutex_t writerMutex; /* Prevents multiple writer threads */
  pthread_cond_t queueSignal;  /* For waking up sleeping writer thread */
  pthread_cond_t emptySignal;  /* Notify when the write queue is empty */
  AsyncWrite *pQueueFirst;     /* Next write operation to be processed */
  AsyncWrite *pQueueLast;      /* Last write operation on the list */
  AsyncLock *pLock;            /* Linked list of all AsyncLock structures */
  volatile int ioDelay;             /* Extra delay between write operations */
  volatile int writerHaltWhenIdle;  /* Writer thread halts when queue empty */
  volatile int writerHaltNow;       /* Writer thread halts after next op */
  int ioError;                 /* True if an IO error has occurred */
  int nFile;                   /* Number of open files (from sqlite pov) */
} async = {
  PTHREAD_MUTEX_INITIALIZER,
  PTHREAD_MUTEX_INITIALIZER,
  PTHREAD_MUTEX_INITIALIZER,
  PTHREAD_COND_INITIALIZER,
  PTHREAD_COND_INITIALIZER,
};

/* Possible values of AsyncWrite.op */
#define ASYNC_NOOP          0
#define ASYNC_WRITE         1
#define ASYNC_SYNC          2
#define ASYNC_TRUNCATE      3
#define ASYNC_CLOSE         4
#define ASYNC_DELETE        5
#define ASYNC_OPENEXCLUSIVE 6
#define ASYNC_UNLOCK        7

/* Names of opcodes.  Used for debugging only.
** Make sure these stay in sync with the macros above!
*/
static const char *azOpcodeName[] = {
  "NOOP", "WRITE", "SYNC", "TRUNCATE", "CLOSE", "DELETE", "OPENEX", "UNLOCK"
};

/*
** Entries on the write-op queue are instances of the AsyncWrite
** structure, defined here.
**
** The interpretation of the iOffset and nByte variables varies depending 
** on the value of AsyncWrite.op:
**
** ASYNC_NOOP:
**     No values used.
**
** ASYNC_WRITE:
**     iOffset -> Offset in file to write to.
**     nByte   -> Number of bytes of data to write (pointed to by zBuf).
**
** ASYNC_SYNC:
**     nByte   -> flags to pass to sqlite3OsSync().
**
** ASYNC_TRUNCATE:
**     iOffset -> Size to truncate file to.
**     nByte   -> Unused.
**
** ASYNC_CLOSE:
**     iOffset -> Unused.
**     nByte   -> Unused.
**
** ASYNC_DELETE:
**     iOffset -> Contains the "syncDir" flag.
**     nByte   -> Number of bytes of zBuf points to (file name).
**
** ASYNC_OPENEXCLUSIVE:
**     iOffset -> Value of "delflag".
**     nByte   -> Number of bytes of zBuf points to (file name).
**
** ASYNC_UNLOCK:
**     nByte   -> Argument to sqlite3OsUnlock().
**
**
** For an ASYNC_WRITE operation, zBuf points to the data to write to the file. 
** This space is sqlite3_malloc()d along with the AsyncWrite structure in a
** single blob, so is deleted when sqlite3_free() is called on the parent 
** structure.
*/
struct AsyncWrite {
  AsyncFileData *pFileData;    /* File to write data to or sync */
  int op;                      /* One of ASYNC_xxx etc. */
  sqlite_int64 iOffset;        /* See above */
  int nByte;          /* See above */
  char *zBuf;         /* Data to write to file (or NULL if op!=ASYNC_WRITE) */
  AsyncWrite *pNext;  /* Next write operation (to any file) */
};

/*
** An instance of this structure is created for each distinct open file 
** (i.e. if two handles are opened on the one file, only one of these
** structures is allocated) and stored in the async.aLock hash table. The
** keys for async.aLock are the full pathnames of the opened files.
**
** AsyncLock.pList points to the head of a linked list of AsyncFileLock
** structures, one for each handle currently open on the file.
**
** If the opened file is not a main-database (the SQLITE_OPEN_MAIN_DB is
** not passed to the sqlite3OsOpen() call), or if ENABLE_FILE_LOCKING is 
** not defined at compile time, variables AsyncLock.pFile and 
** AsyncLock.eLock are never used. Otherwise, pFile is a file handle
** opened on the file in question and used to obtain the file-system 
** locks required by database connections within this process.
**
** See comments above the asyncLock() function for more details on 
** the implementation of database locking used by this backend.
*/
struct AsyncLock {
  char *zFile;
  int nFile;
  sqlite3_file *pFile;
  int eLock;
  AsyncFileLock *pList;
  AsyncLock *pNext;           /* Next in linked list headed by async.pLock */
};

/*
** An instance of the following structure is allocated along with each
** AsyncFileData structure (see AsyncFileData.lock), but is only used if the
** file was opened with the SQLITE_OPEN_MAIN_DB.
*/
struct AsyncFileLock {
  int eLock;                /* Internally visible lock state (sqlite pov) */
  int eAsyncLock;           /* Lock-state with write-queue unlock */
  AsyncFileLock *pNext;
};

/* 
** The AsyncFile structure is a subclass of sqlite3_file used for 
** asynchronous IO. 
**
** All of the actual data for the structure is stored in the structure
** pointed to by AsyncFile.pData, which is allocated as part of the
** sqlite3OsOpen() using sqlite3_malloc(). The reason for this is that the
** lifetime of the AsyncFile structure is ended by the caller after OsClose()
** is called, but the data in AsyncFileData may be required by the
** writer thread after that point.
*/
struct AsyncFile {
  sqlite3_io_methods *pMethod;
  AsyncFileData *pData;
};
struct AsyncFileData {
  char *zName;               /* Underlying OS filename - used for debugging */
  int nName;                 /* Number of characters in zName */
  sqlite3_file *pBaseRead;   /* Read handle to the underlying Os file */
  sqlite3_file *pBaseWrite;  /* Write handle to the underlying Os file */
  AsyncFileLock lock;        /* Lock state for this handle */
  AsyncLock *pLock;          /* AsyncLock object for this file system entry */
  AsyncWrite closeOp;        /* Preallocated close operation */
};

/*
** The following async_XXX functions are debugging wrappers around the
** corresponding pthread_XXX functions:
**
**     pthread_mutex_lock();
**     pthread_mutex_unlock();
**     pthread_mutex_trylock();
**     pthread_cond_wait();
**
** It is illegal to pass any mutex other than those stored in the
** following global variables of these functions.
**
**     async.queueMutex
**     async.writerMutex
**     async.lockMutex
**
** If NDEBUG is defined, these wrappers do nothing except call the 
** corresponding pthreads function. If NDEBUG is not defined, then the
** following variables are used to store the thread-id (as returned
** by pthread_self()) currently holding the mutex, or 0 otherwise:
**
**     asyncdebug.queueMutexHolder
**     asyncdebug.writerMutexHolder
**     asyncdebug.lockMutexHolder
**
** These variables are used by some assert() statements that verify
** the statements made in the "Deadlock Prevention" notes earlier
** in this file.
*/
#ifndef NDEBUG

static struct TestAsyncDebugData {
  pthread_t lockMutexHolder;
  pthread_t queueMutexHolder;
  pthread_t writerMutexHolder;
} asyncdebug = {0, 0, 0};

/*
** Wrapper around pthread_mutex_lock(). Checks that we have not violated
** the anti-deadlock rules (see "Deadlock prevention" above).
*/
static int async_mutex_lock(pthread_mutex_t *pMutex){
  int iIdx;
  int rc;
  pthread_mutex_t *aMutex = (pthread_mutex_t *)(&async);
  pthread_t *aHolder = (pthread_t *)(&asyncdebug);

  /* The code in this 'ifndef NDEBUG' block depends on a certain alignment
   * of the variables in TestAsyncStaticData and TestAsyncDebugData. The
   * following assert() statements check that this has not been changed.
   *
   * Really, these only need to be run once at startup time.
   */
  assert(&(aMutex[0])==&async.lockMutex);
  assert(&(aMutex[1])==&async.queueMutex);
  assert(&(aMutex[2])==&async.writerMutex);
  assert(&(aHolder[0])==&asyncdebug.lockMutexHolder);
  assert(&(aHolder[1])==&asyncdebug.queueMutexHolder);
  assert(&(aHolder[2])==&asyncdebug.writerMutexHolder);

  assert( pthread_self()!=0 );
  for(iIdx=0; iIdx<3; iIdx++){
    if( pMutex==&aMutex[iIdx] ) break;

    /* This is the key assert(). Here we are checking that if the caller
     * is trying to block on async.writerMutex, neither of the other two
     * mutex are held. If the caller is trying to block on async.queueMutex,
     * lockMutex is not held.
     */
    assert(!pthread_equal(aHolder[iIdx], pthread_self()));
  }
  assert(iIdx<3);

  rc = pthread_mutex_lock(pMutex);
  if( rc==0 ){
    assert(aHolder[iIdx]==0);
    aHolder[iIdx] = pthread_self();
  }
  return rc;
}

/*
** Wrapper around pthread_mutex_unlock().
*/
static int async_mutex_unlock(pthread_mutex_t *pMutex){
  int iIdx;
  int rc;
  pthread_mutex_t *aMutex = (pthread_mutex_t *)(&async);
  pthread_t *aHolder = (pthread_t *)(&asyncdebug);

  for(iIdx=0; iIdx<3; iIdx++){
    if( pMutex==&aMutex[iIdx] ) break;
  }
  assert(iIdx<3);

  assert(pthread_equal(aHolder[iIdx], pthread_self()));
  aHolder[iIdx] = 0;
  rc = pthread_mutex_unlock(pMutex);
  assert(rc==0);

  return 0;
}

/*
** Wrapper around pthread_mutex_trylock().
*/
static int async_mutex_trylock(pthread_mutex_t *pMutex){
  int iIdx;
  int rc;
  pthread_mutex_t *aMutex = (pthread_mutex_t *)(&async);
  pthread_t *aHolder = (pthread_t *)(&asyncdebug);

  for(iIdx=0; iIdx<3; iIdx++){
    if( pMutex==&aMutex[iIdx] ) break;
  }
  assert(iIdx<3);

  rc = pthread_mutex_trylock(pMutex);
  if( rc==0 ){
    assert(aHolder[iIdx]==0);
    aHolder[iIdx] = pthread_self();
  }
  return rc;
}

/*
** Wrapper around pthread_cond_wait().
*/
static int async_cond_wait(pthread_cond_t *pCond, pthread_mutex_t *pMutex){
  int iIdx;
  int rc;
  pthread_mutex_t *aMutex = (pthread_mutex_t *)(&async);
  pthread_t *aHolder = (pthread_t *)(&asyncdebug);

  for(iIdx=0; iIdx<3; iIdx++){
    if( pMutex==&aMutex[iIdx] ) break;
  }
  assert(iIdx<3);

  assert(pthread_equal(aHolder[iIdx],pthread_self()));
  aHolder[iIdx] = 0;
  rc = pthread_cond_wait(pCond, pMutex);
  if( rc==0 ){
    aHolder[iIdx] = pthread_self();
  }
  return rc;
}

/*
** Assert that the mutex is held by the current thread.
*/
static void assert_mutex_is_held(pthread_mutex_t *pMutex){
  int iIdx;
  pthread_mutex_t *aMutex = (pthread_mutex_t *)(&async);
  pthread_t *aHolder = (pthread_t *)(&asyncdebug);

  for(iIdx=0; iIdx<3; iIdx++){
    if( pMutex==&aMutex[iIdx] ) break;
  }
  assert(iIdx<3);
  assert( aHolder[iIdx]==pthread_self() );
}

/* Call our async_XX wrappers instead of selected pthread_XX functions */
#define pthread_mutex_lock    async_mutex_lock
#define pthread_mutex_unlock  async_mutex_unlock
#define pthread_mutex_trylock async_mutex_trylock
#define pthread_cond_wait     async_cond_wait

#else    /* if defined(NDEBUG) */

#define assert_mutex_is_held(X)    /* A no-op when not debugging */

#endif   /* !defined(NDEBUG) */

/*
** Add an entry to the end of the global write-op list. pWrite should point 
** to an AsyncWrite structure allocated using sqlite3_malloc().  The writer
** thread will call sqlite3_free() to free the structure after the specified
** operation has been completed.
**
** Once an AsyncWrite structure has been added to the list, it becomes the
** property of the writer thread and must not be read or modified by the
** caller.  
*/
static void addAsyncWrite(AsyncWrite *pWrite){
  /* We must hold the queue mutex in order to modify the queue pointers */
  if( pWrite->op!=ASYNC_UNLOCK ){
    pthread_mutex_lock(&async.queueMutex);
  }

  /* Add the record to the end of the write-op queue */
  assert( !pWrite->pNext );
  if( async.pQueueLast ){
    assert( async.pQueueFirst );
    async.pQueueLast->pNext = pWrite;
  }else{
    async.pQueueFirst = pWrite;
  }
  async.pQueueLast = pWrite;
  ASYNC_TRACE(("PUSH %p (%s %s %d)\n", pWrite, azOpcodeName[pWrite->op],
         pWrite->pFileData ? pWrite->pFileData->zName : "-", pWrite->iOffset));

  if( pWrite->op==ASYNC_CLOSE ){
    async.nFile--;
  }

  /* Drop the queue mutex */
  if( pWrite->op!=ASYNC_UNLOCK ){
    pthread_mutex_unlock(&async.queueMutex);
  }

  /* The writer thread might have been idle because there was nothing
  ** on the write-op queue for it to do.  So wake it up. */
  pthread_cond_signal(&async.queueSignal);
}

/*
** Increment async.nFile in a thread-safe manner.
*/
static void incrOpenFileCount(void){
  /* We must hold the queue mutex in order to modify async.nFile */
  pthread_mutex_lock(&async.queueMutex);
  if( async.nFile==0 ){
    async.ioError = SQLITE_OK;
  }
  async.nFile++;
  pthread_mutex_unlock(&async.queueMutex);
}

/*
** This is a utility function to allocate and populate a new AsyncWrite
** structure and insert it (via addAsyncWrite() ) into the global list.
*/
static int addNewAsyncWrite(
  AsyncFileData *pFileData, 
  int op, 
  sqlite3_int64 iOffset, 
  int nByte,
  const char *zByte
){
  AsyncWrite *p;
  if( op!=ASYNC_CLOSE && async.ioError ){
    return async.ioError;
  }
  p = sqlite3_malloc(sizeof(AsyncWrite) + (zByte?nByte:0));
  if( !p ){
    /* The upper layer does not expect operations like OsWrite() to
    ** return SQLITE_NOMEM. This is partly because under normal conditions
    ** SQLite is required to do rollback without calling malloc(). So
    ** if malloc() fails here, treat it as an I/O error. The above
    ** layer knows how to handle that.
    */
    return SQLITE_IOERR;
  }
  p->op = op;
  p->iOffset = iOffset;
  p->nByte = nByte;
  p->pFileData = pFileData;
  p->pNext = 0;
  if( zByte ){
    p->zBuf = (char *)&p[1];
    memcpy(p->zBuf, zByte, nByte);
  }else{
    p->zBuf = 0;
  }
  addAsyncWrite(p);
  return SQLITE_OK;
}

/*
** Close the file. This just adds an entry to the write-op list, the file is
** not actually closed.
*/
static int asyncClose(sqlite3_file *pFile){
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;

  /* Unlock the file, if it is locked */
  pthread_mutex_lock(&async.lockMutex);
  p->lock.eLock = 0;
  pthread_mutex_unlock(&async.lockMutex);

  addAsyncWrite(&p->closeOp);
  return SQLITE_OK;
}

/*
** Implementation of sqlite3OsWrite() for asynchronous files. Instead of 
** writing to the underlying file, this function adds an entry to the end of
** the global AsyncWrite list. Either SQLITE_OK or SQLITE_NOMEM may be
** returned.
*/
static int asyncWrite(
  sqlite3_file *pFile, 
  const void *pBuf, 
  int amt, 
  sqlite3_int64 iOff
){
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;
  return addNewAsyncWrite(p, ASYNC_WRITE, iOff, amt, pBuf);
}

/*
** Read data from the file. First we read from the filesystem, then adjust 
** the contents of the buffer based on ASYNC_WRITE operations in the 
** write-op queue.
**
** This method holds the mutex from start to finish.
*/
static int asyncRead(
  sqlite3_file *pFile, 
  void *zOut, 
  int iAmt, 
  sqlite3_int64 iOffset
){
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;
  int rc = SQLITE_OK;
  sqlite3_int64 filesize;
  int nRead;
  sqlite3_file *pBase = p->pBaseRead;

  /* Grab the write queue mutex for the duration of the call */
  pthread_mutex_lock(&async.queueMutex);

  /* If an I/O error has previously occurred in this virtual file 
  ** system, then all subsequent operations fail.
  */
  if( async.ioError!=SQLITE_OK ){
    rc = async.ioError;
    goto asyncread_out;
  }

  if( pBase->pMethods ){
    rc = pBase->pMethods->xFileSize(pBase, &filesize);
    if( rc!=SQLITE_OK ){
      goto asyncread_out;
    }
    nRead = MIN(filesize - iOffset, iAmt);
    if( nRead>0 ){
      rc = pBase->pMethods->xRead(pBase, zOut, nRead, iOffset);
      ASYNC_TRACE(("READ %s %d bytes at %d\n", p->zName, nRead, iOffset));
    }
  }

  if( rc==SQLITE_OK ){
    AsyncWrite *pWrite;
    char *zName = p->zName;

    for(pWrite=async.pQueueFirst; pWrite; pWrite = pWrite->pNext){
      if( pWrite->op==ASYNC_WRITE && (
        (pWrite->pFileData==p) ||
        (zName && pWrite->pFileData->zName==zName)
      )){
        int iBeginOut = (pWrite->iOffset-iOffset);
        int iBeginIn = -iBeginOut;
        int nCopy;

        if( iBeginIn<0 ) iBeginIn = 0;
        if( iBeginOut<0 ) iBeginOut = 0;
        nCopy = MIN(pWrite->nByte-iBeginIn, iAmt-iBeginOut);

        if( nCopy>0 ){
          memcpy(&((char *)zOut)[iBeginOut], &pWrite->zBuf[iBeginIn], nCopy);
          ASYNC_TRACE(("OVERREAD %d bytes at %d\n", nCopy, iBeginOut+iOffset));
        }
      }
    }
  }

asyncread_out:
  pthread_mutex_unlock(&async.queueMutex);
  return rc;
}

/*
** Truncate the file to nByte bytes in length. This just adds an entry to 
** the write-op list, no IO actually takes place.
*/
static int asyncTruncate(sqlite3_file *pFile, sqlite3_int64 nByte){
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;
  return addNewAsyncWrite(p, ASYNC_TRUNCATE, nByte, 0, 0);
}

/*
** Sync the file. This just adds an entry to the write-op list, the 
** sync() is done later by sqlite3_async_flush().
*/
static int asyncSync(sqlite3_file *pFile, int flags){
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;
  return addNewAsyncWrite(p, ASYNC_SYNC, 0, flags, 0);
}

/*
** Read the size of the file. First we read the size of the file system 
** entry, then adjust for any ASYNC_WRITE or ASYNC_TRUNCATE operations 
** currently in the write-op list. 
**
** This method holds the mutex from start to finish.
*/
int asyncFileSize(sqlite3_file *pFile, sqlite3_int64 *piSize){
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;
  int rc = SQLITE_OK;
  sqlite3_int64 s = 0;
  sqlite3_file *pBase;

  pthread_mutex_lock(&async.queueMutex);

  /* Read the filesystem size from the base file. If pBaseRead is NULL, this
  ** means the file hasn't been opened yet. In this case all relevant data 
  ** must be in the write-op queue anyway, so we can omit reading from the
  ** file-system.
  */
  pBase = p->pBaseRead;
  if( pBase->pMethods ){
    rc = pBase->pMethods->xFileSize(pBase, &s);
  }

  if( rc==SQLITE_OK ){
    AsyncWrite *pWrite;
    for(pWrite=async.pQueueFirst; pWrite; pWrite = pWrite->pNext){
      if( pWrite->op==ASYNC_DELETE 
       && p->zName 
       && strcmp(p->zName, pWrite->zBuf)==0 
      ){
        s = 0;
      }else if( pWrite->pFileData && (
          (pWrite->pFileData==p) 
       || (p->zName && pWrite->pFileData->zName==p->zName) 
      )){
        switch( pWrite->op ){
          case ASYNC_WRITE:
            s = MAX(pWrite->iOffset + (sqlite3_int64)(pWrite->nByte), s);
            break;
          case ASYNC_TRUNCATE:
            s = MIN(s, pWrite->iOffset);
            break;
        }
      }
    }
    *piSize = s;
  }
  pthread_mutex_unlock(&async.queueMutex);
  return rc;
}

/*
** Lock or unlock the actual file-system entry.
*/
static int getFileLock(AsyncLock *pLock){
  int rc = SQLITE_OK;
  AsyncFileLock *pIter;
  int eRequired = 0;

  if( pLock->pFile ){
    for(pIter=pLock->pList; pIter; pIter=pIter->pNext){
      assert(pIter->eAsyncLock>=pIter->eLock);
      if( pIter->eAsyncLock>eRequired ){
        eRequired = pIter->eAsyncLock;
        assert(eRequired>=0 && eRequired<=SQLITE_LOCK_EXCLUSIVE);
      }
    }

    if( eRequired>pLock->eLock ){
      rc = pLock->pFile->pMethods->xLock(pLock->pFile, eRequired);
      if( rc==SQLITE_OK ){
        pLock->eLock = eRequired;
      }
    }
    else if( eRequired<pLock->eLock && eRequired<=SQLITE_LOCK_SHARED ){
      rc = pLock->pFile->pMethods->xUnlock(pLock->pFile, eRequired);
      if( rc==SQLITE_OK ){
        pLock->eLock = eRequired;
      }
    }
  }

  return rc;
}

/*
** Return the AsyncLock structure from the global async.pLock list 
** associated with the file-system entry identified by path zName 
** (a string of nName bytes). If no such structure exists, return 0.
*/
static AsyncLock *findLock(const char *zName, int nName){
  AsyncLock *p = async.pLock;
  while( p && (p->nFile!=nName || memcmp(p->zFile, zName, nName)) ){
    p = p->pNext;
  }
  return p;
}

/*
** The following two methods - asyncLock() and asyncUnlock() - are used
** to obtain and release locks on database files opened with the
** asynchronous backend.
*/
static int asyncLock(sqlite3_file *pFile, int eLock){
  int rc = SQLITE_OK;
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;

  if( p->zName ){
    pthread_mutex_lock(&async.lockMutex);
    if( p->lock.eLock<eLock ){
      AsyncLock *pLock = p->pLock;
      AsyncFileLock *pIter;
      assert(pLock && pLock->pList);
      for(pIter=pLock->pList; pIter; pIter=pIter->pNext){
        if( pIter!=&p->lock && (
          (eLock==SQLITE_LOCK_EXCLUSIVE && pIter->eLock>=SQLITE_LOCK_SHARED) ||
          (eLock==SQLITE_LOCK_PENDING && pIter->eLock>=SQLITE_LOCK_RESERVED) ||
          (eLock==SQLITE_LOCK_RESERVED && pIter->eLock>=SQLITE_LOCK_RESERVED) ||
          (eLock==SQLITE_LOCK_SHARED && pIter->eLock>=SQLITE_LOCK_PENDING)
        )){
          rc = SQLITE_BUSY;
        }
      }
      if( rc==SQLITE_OK ){
        p->lock.eLock = eLock;
        p->lock.eAsyncLock = MAX(p->lock.eAsyncLock, eLock);
      }
      assert(p->lock.eAsyncLock>=p->lock.eLock);
      if( rc==SQLITE_OK ){
        rc = getFileLock(pLock);
      }
    }
    pthread_mutex_unlock(&async.lockMutex);
  }

  ASYNC_TRACE(("LOCK %d (%s) rc=%d\n", eLock, p->zName, rc));
  return rc;
}
static int asyncUnlock(sqlite3_file *pFile, int eLock){
  int rc = SQLITE_OK;
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;
  if( p->zName ){
    AsyncFileLock *pLock = &p->lock;
    pthread_mutex_lock(&async.queueMutex);
    pthread_mutex_lock(&async.lockMutex);
    pLock->eLock = MIN(pLock->eLock, eLock);
    rc = addNewAsyncWrite(p, ASYNC_UNLOCK, 0, eLock, 0);
    pthread_mutex_unlock(&async.lockMutex);
    pthread_mutex_unlock(&async.queueMutex);
  }
  return rc;
}

/*
** This function is called when the pager layer first opens a database file
** and is checking for a hot-journal.
*/
static int asyncCheckReservedLock(sqlite3_file *pFile, int *pResOut){
  int ret = 0;
  AsyncFileLock *pIter;
  AsyncFileData *p = ((AsyncFile *)pFile)->pData;

  pthread_mutex_lock(&async.lockMutex);
  for(pIter=p->pLock->pList; pIter; pIter=pIter->pNext){
    if( pIter->eLock>=SQLITE_LOCK_RESERVED ){
      ret = 1;
    }
  }
  pthread_mutex_unlock(&async.lockMutex);

  ASYNC_TRACE(("CHECK-LOCK %d (%s)\n", ret, p->zName));
  *pResOut = ret;
  return SQLITE_OK;
}

/* 
** sqlite3_file_control() implementation.
*/
static int asyncFileControl(sqlite3_file *id, int op, void *pArg){
  switch( op ){
    case SQLITE_FCNTL_LOCKSTATE: {
      pthread_mutex_lock(&async.lockMutex);
      *(int*)pArg = ((AsyncFile*)id)->pData->lock.eLock;
      pthread_mutex_unlock(&async.lockMutex);
      return SQLITE_OK;
    }
  }
  return SQLITE_ERROR;
}

/* 
** Return the device characteristics and sector-size of the device. It
** is not tricky to implement these correctly, as this backend might 
** not have an open file handle at this point.
*/
static int asyncSectorSize(sqlite3_file *pFile){
  return 512;
}
static int asyncDeviceCharacteristics(sqlite3_file *pFile){
  return 0;
}

static int unlinkAsyncFile(AsyncFileData *pData){
  AsyncFileLock **ppIter;
  int rc = SQLITE_OK;

  if( pData->zName ){
    AsyncLock *pLock = pData->pLock;
    for(ppIter=&pLock->pList; *ppIter; ppIter=&((*ppIter)->pNext)){
      if( (*ppIter)==&pData->lock ){
        *ppIter = pData->lock.pNext;
        break;
      }
    }
    if( !pLock->pList ){
      AsyncLock **pp;
      if( pLock->pFile ){
        pLock->pFile->pMethods->xClose(pLock->pFile);
      }
      for(pp=&async.pLock; *pp!=pLock; pp=&((*pp)->pNext));
      *pp = pLock->pNext;
      sqlite3_free(pLock);
    }else{
      rc = getFileLock(pLock);
    }
  }

  return rc;
}

/*
** The parameter passed to this function is a copy of a 'flags' parameter
** passed to this modules xOpen() method. This function returns true
** if the file should be opened asynchronously, or false if it should
** be opened immediately.
**
** If the file is to be opened asynchronously, then asyncOpen() will add
** an entry to the event queue and the file will not actually be opened
** until the event is processed. Otherwise, the file is opened directly
** by the caller.
*/
static int doAsynchronousOpen(int flags){
  return (flags&SQLITE_OPEN_CREATE) && (
      (flags&SQLITE_OPEN_MAIN_JOURNAL) ||
      (flags&SQLITE_OPEN_TEMP_JOURNAL) ||
      (flags&SQLITE_OPEN_DELETEONCLOSE)
  );
}

/*
** Open a file.
*/
static int asyncOpen(
  sqlite3_vfs *pAsyncVfs,
  const char *zName,
  sqlite3_file *pFile,
  int flags,
  int *pOutFlags
){
  static sqlite3_io_methods async_methods = {
    1,                               /* iVersion */
    asyncClose,                      /* xClose */
    asyncRead,                       /* xRead */
    asyncWrite,                      /* xWrite */
    asyncTruncate,                   /* xTruncate */
    asyncSync,                       /* xSync */
    asyncFileSize,                   /* xFileSize */
    asyncLock,                       /* xLock */
    asyncUnlock,                     /* xUnlock */
    asyncCheckReservedLock,          /* xCheckReservedLock */
    asyncFileControl,                /* xFileControl */
    asyncSectorSize,                 /* xSectorSize */
    asyncDeviceCharacteristics       /* xDeviceCharacteristics */
  };

  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  AsyncFile *p = (AsyncFile *)pFile;
  int nName = 0;
  int rc = SQLITE_OK;
  int nByte;
  AsyncFileData *pData;
  AsyncLock *pLock = 0;
  char *z;
  int isAsyncOpen = doAsynchronousOpen(flags);

  /* If zName is NULL, then the upper layer is requesting an anonymous file */
  if( zName ){
    nName = strlen(zName)+1;
  }

  nByte = (
    sizeof(AsyncFileData) +        /* AsyncFileData structure */
    2 * pVfs->szOsFile +           /* AsyncFileData.pBaseRead and pBaseWrite */
    nName                          /* AsyncFileData.zName */
  ); 
  z = sqlite3_malloc(nByte);
  if( !z ){
    return SQLITE_NOMEM;
  }
  memset(z, 0, nByte);
  pData = (AsyncFileData*)z;
  z += sizeof(pData[0]);
  pData->pBaseRead = (sqlite3_file*)z;
  z += pVfs->szOsFile;
  pData->pBaseWrite = (sqlite3_file*)z;
  pData->closeOp.pFileData = pData;
  pData->closeOp.op = ASYNC_CLOSE;

  if( zName ){
    z += pVfs->szOsFile;
    pData->zName = z;
    pData->nName = nName;
    memcpy(pData->zName, zName, nName);
  }

  if( !isAsyncOpen ){
    int flagsout;
    rc = pVfs->xOpen(pVfs, pData->zName, pData->pBaseRead, flags, &flagsout);
    if( rc==SQLITE_OK && (flagsout&SQLITE_OPEN_READWRITE) ){
      rc = pVfs->xOpen(pVfs, pData->zName, pData->pBaseWrite, flags, 0);
    }
    if( pOutFlags ){
      *pOutFlags = flagsout;
    }
  }

  pthread_mutex_lock(&async.lockMutex);

  if( zName && rc==SQLITE_OK ){
    pLock = findLock(pData->zName, pData->nName);
    if( !pLock ){
      int nByte = pVfs->szOsFile + sizeof(AsyncLock) + pData->nName + 1; 
      pLock = (AsyncLock *)sqlite3_malloc(nByte);
      if( pLock ){
        memset(pLock, 0, nByte);
#ifdef ENABLE_FILE_LOCKING
        if( flags&SQLITE_OPEN_MAIN_DB ){
          pLock->pFile = (sqlite3_file *)&pLock[1];
          rc = pVfs->xOpen(pVfs, pData->zName, pLock->pFile, flags, 0);
          if( rc!=SQLITE_OK ){
            sqlite3_free(pLock);
            pLock = 0;
          }
        }
#endif
        if( pLock ){
          pLock->nFile = pData->nName;
          pLock->zFile = &((char *)(&pLock[1]))[pVfs->szOsFile];
          memcpy(pLock->zFile, pData->zName, pLock->nFile);
          pLock->pNext = async.pLock;
          async.pLock = pLock;
        }
      }else{
        rc = SQLITE_NOMEM;
      }
    }
  }

  if( rc==SQLITE_OK ){
    p->pMethod = &async_methods;
    p->pData = pData;

    /* Link AsyncFileData.lock into the linked list of 
    ** AsyncFileLock structures for this file.
    */
    if( zName ){
      pData->lock.pNext = pLock->pList;
      pLock->pList = &pData->lock;
      pData->zName = pLock->zFile;
    }
  }else{
    if( pData->pBaseRead->pMethods ){
      pData->pBaseRead->pMethods->xClose(pData->pBaseRead);
    }
    if( pData->pBaseWrite->pMethods ){
      pData->pBaseWrite->pMethods->xClose(pData->pBaseWrite);
    }
    sqlite3_free(pData);
  }

  pthread_mutex_unlock(&async.lockMutex);

  if( rc==SQLITE_OK ){
    incrOpenFileCount();
    pData->pLock = pLock;
  }

  if( rc==SQLITE_OK && isAsyncOpen ){
    rc = addNewAsyncWrite(pData, ASYNC_OPENEXCLUSIVE, (sqlite3_int64)flags,0,0);
    if( rc==SQLITE_OK ){
      if( pOutFlags ) *pOutFlags = flags;
    }else{
      pthread_mutex_lock(&async.lockMutex);
      unlinkAsyncFile(pData);
      pthread_mutex_unlock(&async.lockMutex);
      sqlite3_free(pData);
    }
  }
  if( rc!=SQLITE_OK ){
    p->pMethod = 0;
  }
  return rc;
}

/*
** Implementation of sqlite3OsDelete. Add an entry to the end of the 
** write-op queue to perform the delete.
*/
static int asyncDelete(sqlite3_vfs *pAsyncVfs, const char *z, int syncDir){
  return addNewAsyncWrite(0, ASYNC_DELETE, syncDir, strlen(z)+1, z);
}

/*
** Implementation of sqlite3OsAccess. This method holds the mutex from
** start to finish.
*/
static int asyncAccess(
  sqlite3_vfs *pAsyncVfs, 
  const char *zName, 
  int flags,
  int *pResOut
){
  int rc;
  int ret;
  AsyncWrite *p;
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;

  assert(flags==SQLITE_ACCESS_READWRITE 
      || flags==SQLITE_ACCESS_READ 
      || flags==SQLITE_ACCESS_EXISTS 
  );

  pthread_mutex_lock(&async.queueMutex);
  rc = pVfs->xAccess(pVfs, zName, flags, &ret);
  if( rc==SQLITE_OK && flags==SQLITE_ACCESS_EXISTS ){
    for(p=async.pQueueFirst; p; p = p->pNext){
      if( p->op==ASYNC_DELETE && 0==strcmp(p->zBuf, zName) ){
        ret = 0;
      }else if( p->op==ASYNC_OPENEXCLUSIVE 
             && p->pFileData->zName
             && 0==strcmp(p->pFileData->zName, zName) 
      ){
        ret = 1;
      }
    }
  }
  ASYNC_TRACE(("ACCESS(%s): %s = %d\n", 
    flags==SQLITE_ACCESS_READWRITE?"read-write":
    flags==SQLITE_ACCESS_READ?"read":"exists"
    , zName, ret)
  );
  pthread_mutex_unlock(&async.queueMutex);
  *pResOut = ret;
  return rc;
}

/*
** Fill in zPathOut with the full path to the file identified by zPath.
*/
static int asyncFullPathname(
  sqlite3_vfs *pAsyncVfs, 
  const char *zPath, 
  int nPathOut,
  char *zPathOut
){
  int rc;
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  rc = pVfs->xFullPathname(pVfs, zPath, nPathOut, zPathOut);

  /* Because of the way intra-process file locking works, this backend
  ** needs to return a canonical path. The following block assumes the
  ** file-system uses unix style paths. 
  */
  if( rc==SQLITE_OK ){
    int i, j;
    int n = nPathOut;
    char *z = zPathOut;
    while( n>1 && z[n-1]=='/' ){ n--; }
    for(i=j=0; i<n; i++){
      if( z[i]=='/' ){
        if( z[i+1]=='/' ) continue;
        if( z[i+1]=='.' && i+2<n && z[i+2]=='/' ){
          i += 1;
          continue;
        }
        if( z[i+1]=='.' && i+3<n && z[i+2]=='.' && z[i+3]=='/' ){
          while( j>0 && z[j-1]!='/' ){ j--; }
          if( j>0 ){ j--; }
          i += 2;
          continue;
        }
      }
      z[j++] = z[i];
    }
    z[j] = 0;
  }

  return rc;
}
static void *asyncDlOpen(sqlite3_vfs *pAsyncVfs, const char *zPath){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  return pVfs->xDlOpen(pVfs, zPath);
}
static void asyncDlError(sqlite3_vfs *pAsyncVfs, int nByte, char *zErrMsg){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  pVfs->xDlError(pVfs, nByte, zErrMsg);
}
static void (*asyncDlSym(
  sqlite3_vfs *pAsyncVfs, 
  void *pHandle, 
  const char *zSymbol
))(void){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  return pVfs->xDlSym(pVfs, pHandle, zSymbol);
}
static void asyncDlClose(sqlite3_vfs *pAsyncVfs, void *pHandle){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  pVfs->xDlClose(pVfs, pHandle);
}
static int asyncRandomness(sqlite3_vfs *pAsyncVfs, int nByte, char *zBufOut){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  return pVfs->xRandomness(pVfs, nByte, zBufOut);
}
static int asyncSleep(sqlite3_vfs *pAsyncVfs, int nMicro){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  return pVfs->xSleep(pVfs, nMicro);
}
static int asyncCurrentTime(sqlite3_vfs *pAsyncVfs, double *pTimeOut){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)pAsyncVfs->pAppData;
  return pVfs->xCurrentTime(pVfs, pTimeOut);
}

static sqlite3_vfs async_vfs = {
  1,                    /* iVersion */
  sizeof(AsyncFile),    /* szOsFile */
  0,                    /* mxPathname */
  0,                    /* pNext */
  "async",              /* zName */
  0,                    /* pAppData */
  asyncOpen,            /* xOpen */
  asyncDelete,          /* xDelete */
  asyncAccess,          /* xAccess */
  asyncFullPathname,    /* xFullPathname */
  asyncDlOpen,          /* xDlOpen */
  asyncDlError,         /* xDlError */
  asyncDlSym,           /* xDlSym */
  asyncDlClose,         /* xDlClose */
  asyncRandomness,      /* xDlError */
  asyncSleep,           /* xDlSym */
  asyncCurrentTime      /* xDlClose */
};


/*
** Call this routine to enable or disable the
** asynchronous IO features implemented in this file. 
**
** This routine is not even remotely threadsafe.  Do not call
** this routine while any SQLite database connections are open.
*/
static void asyncEnable(int enable){
  if( enable ){
    if( !async_vfs.pAppData ){
      async_vfs.pAppData = (void *)sqlite3_vfs_find(0);
      async_vfs.mxPathname = ((sqlite3_vfs *)async_vfs.pAppData)->mxPathname;
      sqlite3_vfs_register(&async_vfs, 1);
    }
  }else{
    if( async_vfs.pAppData ){
      sqlite3_vfs_unregister(&async_vfs);
      async_vfs.pAppData = 0;
    }
  }
}

/* 
** This procedure runs in a separate thread, reading messages off of the
** write queue and processing them one by one.  
**
** If async.writerHaltNow is true, then this procedure exits
** after processing a single message.
**
** If async.writerHaltWhenIdle is true, then this procedure exits when
** the write queue is empty.
**
** If both of the above variables are false, this procedure runs
** indefinately, waiting for operations to be added to the write queue
** and processing them in the order in which they arrive.
**
** An artifical delay of async.ioDelay milliseconds is inserted before
** each write operation in order to simulate the effect of a slow disk.
**
** Only one instance of this procedure may be running at a time.
*/
static void *asyncWriterThread(void *pIsStarted){
  sqlite3_vfs *pVfs = (sqlite3_vfs *)(async_vfs.pAppData);
  AsyncWrite *p = 0;
  int rc = SQLITE_OK;
  int holdingMutex = 0;

  if( pthread_mutex_trylock(&async.writerMutex) ){
    return 0;
  }
  (*(int *)pIsStarted) = 1;
  while( async.writerHaltNow==0 ){
    int doNotFree = 0;
    sqlite3_file *pBase = 0;

    if( !holdingMutex ){
      pthread_mutex_lock(&async.queueMutex);
    }
    while( (p = async.pQueueFirst)==0 ){
      pthread_cond_broadcast(&async.emptySignal);
      if( async.writerHaltWhenIdle ){
        pthread_mutex_unlock(&async.queueMutex);
        break;
      }else{
        ASYNC_TRACE(("IDLE\n"));
        pthread_cond_wait(&async.queueSignal, &async.queueMutex);
        ASYNC_TRACE(("WAKEUP\n"));
      }
    }
    if( p==0 ) break;
    holdingMutex = 1;

    /* Right now this thread is holding the mutex on the write-op queue.
    ** Variable 'p' points to the first entry in the write-op queue. In
    ** the general case, we hold on to the mutex for the entire body of
    ** the loop. 
    **
    ** However in the cases enumerated below, we relinquish the mutex,
    ** perform the IO, and then re-request the mutex before removing 'p' from
    ** the head of the write-op queue. The idea is to increase concurrency with
    ** sqlite threads.
    **
    **     * An ASYNC_CLOSE operation.
    **     * An ASYNC_OPENEXCLUSIVE operation. For this one, we relinquish 
    **       the mutex, call the underlying xOpenExclusive() function, then
    **       re-aquire the mutex before seting the AsyncFile.pBaseRead 
    **       variable.
    **     * ASYNC_SYNC and ASYNC_WRITE operations, if 
    **       SQLITE_ASYNC_TWO_FILEHANDLES was set at compile time and two
    **       file-handles are open for the particular file being "synced".
    */
    if( async.ioError!=SQLITE_OK && p->op!=ASYNC_CLOSE ){
      p->op = ASYNC_NOOP;
    }
    if( p->pFileData ){
      pBase = p->pFileData->pBaseWrite;
      if( 
        p->op==ASYNC_CLOSE || 
        p->op==ASYNC_OPENEXCLUSIVE ||
        (pBase->pMethods && (p->op==ASYNC_SYNC || p->op==ASYNC_WRITE) ) 
      ){
        pthread_mutex_unlock(&async.queueMutex);
        holdingMutex = 0;
      }
      if( !pBase->pMethods ){
        pBase = p->pFileData->pBaseRead;
      }
    }

    switch( p->op ){
      case ASYNC_NOOP:
        break;

      case ASYNC_WRITE:
        assert( pBase );
        ASYNC_TRACE(("WRITE %s %d bytes at %d\n",
                p->pFileData->zName, p->nByte, p->iOffset));
        rc = pBase->pMethods->xWrite(pBase, (void *)(p->zBuf), p->nByte, p->iOffset);
        break;

      case ASYNC_SYNC:
        assert( pBase );
        ASYNC_TRACE(("SYNC %s\n", p->pFileData->zName));
        rc = pBase->pMethods->xSync(pBase, p->nByte);
        break;

      case ASYNC_TRUNCATE:
        assert( pBase );
        ASYNC_TRACE(("TRUNCATE %s to %d bytes\n", 
                p->pFileData->zName, p->iOffset));
        rc = pBase->pMethods->xTruncate(pBase, p->iOffset);
        break;

      case ASYNC_CLOSE: {
        AsyncFileData *pData = p->pFileData;
        ASYNC_TRACE(("CLOSE %s\n", p->pFileData->zName));
        if( pData->pBaseWrite->pMethods ){
          pData->pBaseWrite->pMethods->xClose(pData->pBaseWrite);
        }
        if( pData->pBaseRead->pMethods ){
          pData->pBaseRead->pMethods->xClose(pData->pBaseRead);
        }

        /* Unlink AsyncFileData.lock from the linked list of AsyncFileLock 
        ** structures for this file. Obtain the async.lockMutex mutex 
        ** before doing so.
        */
        pthread_mutex_lock(&async.lockMutex);
        rc = unlinkAsyncFile(pData);
        pthread_mutex_unlock(&async.lockMutex);

        if( !holdingMutex ){
          pthread_mutex_lock(&async.queueMutex);
          holdingMutex = 1;
        }
        assert_mutex_is_held(&async.queueMutex);
        async.pQueueFirst = p->pNext;
        sqlite3_free(pData);
        doNotFree = 1;
        break;
      }

      case ASYNC_UNLOCK: {
        AsyncWrite *pIter;
        AsyncFileData *pData = p->pFileData;
        int eLock = p->nByte;

        /* When a file is locked by SQLite using the async backend, it is 
        ** locked within the 'real' file-system synchronously. When it is
        ** unlocked, an ASYNC_UNLOCK event is added to the write-queue to
        ** unlock the file asynchronously. The design of the async backend
        ** requires that the 'real' file-system file be locked from the
        ** time that SQLite first locks it (and probably reads from it)
        ** until all asynchronous write events that were scheduled before
        ** SQLite unlocked the file have been processed.
        **
        ** This is more complex if SQLite locks and unlocks the file multiple
        ** times in quick succession. For example, if SQLite does: 
        ** 
        **   lock, write, unlock, lock, write, unlock
        **
        ** Each "lock" operation locks the file immediately. Each "write" 
        ** and "unlock" operation adds an event to the event queue. If the
        ** second "lock" operation is performed before the first "unlock"
        ** operation has been processed asynchronously, then the first
        ** "unlock" cannot be safely processed as is, since this would mean
        ** the file was unlocked when the second "write" operation is
        ** processed. To work around this, when processing an ASYNC_UNLOCK
        ** operation, SQLite:
        **
        **   1) Unlocks the file to the minimum of the argument passed to
        **      the xUnlock() call and the current lock from SQLite's point
        **      of view, and
        **
        **   2) Only unlocks the file at all if this event is the last
        **      ASYNC_UNLOCK event on this file in the write-queue.
        */ 
        assert( holdingMutex==1 );
        assert( async.pQueueFirst==p );
        for(pIter=async.pQueueFirst->pNext; pIter; pIter=pIter->pNext){
          if( pIter->pFileData==pData && pIter->op==ASYNC_UNLOCK ) break;
        }
        if( !pIter ){
          pthread_mutex_lock(&async.lockMutex);
          pData->lock.eAsyncLock = MIN(
              pData->lock.eAsyncLock, MAX(pData->lock.eLock, eLock)
          );
          assert(pData->lock.eAsyncLock>=pData->lock.eLock);
          rc = getFileLock(pData->pLock);
          pthread_mutex_unlock(&async.lockMutex);
        }
        break;
      }

      case ASYNC_DELETE:
        ASYNC_TRACE(("DELETE %s\n", p->zBuf));
        rc = pVfs->xDelete(pVfs, p->zBuf, (int)p->iOffset);
        break;

      case ASYNC_OPENEXCLUSIVE: {
        int flags = (int)p->iOffset;
        AsyncFileData *pData = p->pFileData;
        ASYNC_TRACE(("OPEN %s flags=%d\n", p->zBuf, (int)p->iOffset));
        assert(pData->pBaseRead->pMethods==0 && pData->pBaseWrite->pMethods==0);
        rc = pVfs->xOpen(pVfs, pData->zName, pData->pBaseRead, flags, 0);
        assert( holdingMutex==0 );
        pthread_mutex_lock(&async.queueMutex);
        holdingMutex = 1;
        break;
      }

      default: assert(!"Illegal value for AsyncWrite.op");
    }

    /* If we didn't hang on to the mutex during the IO op, obtain it now
    ** so that the AsyncWrite structure can be safely removed from the 
    ** global write-op queue.
    */
    if( !holdingMutex ){
      pthread_mutex_lock(&async.queueMutex);
      holdingMutex = 1;
    }
    /* ASYNC_TRACE(("UNLINK %p\n", p)); */
    if( p==async.pQueueLast ){
      async.pQueueLast = 0;
    }
    if( !doNotFree ){
      assert_mutex_is_held(&async.queueMutex);
      async.pQueueFirst = p->pNext;
      sqlite3_free(p);
    }
    assert( holdingMutex );

    /* An IO error has occurred. We cannot report the error back to the
    ** connection that requested the I/O since the error happened 
    ** asynchronously.  The connection has already moved on.  There 
    ** really is nobody to report the error to.
    **
    ** The file for which the error occurred may have been a database or
    ** journal file. Regardless, none of the currently queued operations
    ** associated with the same database should now be performed. Nor should
    ** any subsequently requested IO on either a database or journal file 
    ** handle for the same database be accepted until the main database
    ** file handle has been closed and reopened.
    **
    ** Furthermore, no further IO should be queued or performed on any file
    ** handle associated with a database that may have been part of a 
    ** multi-file transaction that included the database associated with 
    ** the IO error (i.e. a database ATTACHed to the same handle at some 
    ** point in time).
    */
    if( rc!=SQLITE_OK ){
      async.ioError = rc;
    }

    if( async.ioError && !async.pQueueFirst ){
      pthread_mutex_lock(&async.lockMutex);
      if( 0==async.pLock ){
        async.ioError = SQLITE_OK;
      }
      pthread_mutex_unlock(&async.lockMutex);
    }

    /* Drop the queue mutex before continuing to the next write operation
    ** in order to give other threads a chance to work with the write queue.
    */
    if( !async.pQueueFirst || !async.ioError ){
      pthread_mutex_unlock(&async.queueMutex);
      holdingMutex = 0;
      if( async.ioDelay>0 ){
        pVfs->xSleep(pVfs, async.ioDelay);
      }else{
        sched_yield();
      }
    }
  }
  
  pthread_mutex_unlock(&async.writerMutex);
  return 0;
}

/**************************************************************************
** The remaining code defines a Tcl interface for testing the asynchronous
** IO implementation in this file.
**
** To adapt the code to a non-TCL environment, delete or comment out
** the code that follows.
*/


/*
** sqlite3async_enable ?YES/NO?
**
** Enable or disable the asynchronous I/O backend.  This command is
** not thread-safe.  Do not call it while any database connections
** are open.
................................................................................
  Tcl_Obj *CONST objv[]
){
  if( objc!=1 && objc!=2 ){
    Tcl_WrongNumArgs(interp, 1, objv, "?YES/NO?");
    return TCL_ERROR;
  }
  if( objc==1 ){
    Tcl_SetObjResult(interp, Tcl_NewBooleanObj(async_vfs.pAppData!=0));
  }else{
    int en;
    if( Tcl_GetBooleanFromObj(interp, objv[1], &en) ) return TCL_ERROR;
    asyncEnable(en);





  }
  return TCL_OK;
}

/*
** sqlite3async_halt  "now"|"idle"|"never"
**
** Set the conditions at which the writer thread will halt.
*/
static int testAsyncHalt(
  void * clientData,
  Tcl_Interp *interp,
  int objc,
  Tcl_Obj *CONST objv[]
){
  const char *zCond;
  if( objc!=2 ){
    Tcl_WrongNumArgs(interp, 1, objv, "\"now\"|\"idle\"|\"never\"");
    return TCL_ERROR;
  }
  zCond = Tcl_GetString(objv[1]);
  if( strcmp(zCond, "now")==0 ){
    async.writerHaltNow = 1;
    pthread_cond_broadcast(&async.queueSignal);
  }else if( strcmp(zCond, "idle")==0 ){
    async.writerHaltWhenIdle = 1;
    async.writerHaltNow = 0;
    pthread_cond_broadcast(&async.queueSignal);
  }else if( strcmp(zCond, "never")==0 ){
    async.writerHaltWhenIdle = 0;
    async.writerHaltNow = 0;
  }else{
    Tcl_AppendResult(interp, 
      "should be one of: \"now\", \"idle\", or \"never\"", (char*)0);
    return TCL_ERROR;
  }

  return TCL_OK;
}

/*
** sqlite3async_delay ?MS?
**
** Query or set the number of milliseconds of delay in the writer
................................................................................
*/
static int testAsyncDelay(
  void * clientData,
  Tcl_Interp *interp,
  int objc,
  Tcl_Obj *CONST objv[]
){

  if( objc!=1 && objc!=2 ){
    Tcl_WrongNumArgs(interp, 1, objv, "?MS?");
    return TCL_ERROR;
  }
  if( objc==1 ){
    Tcl_SetObjResult(interp, Tcl_NewIntObj(async.ioDelay));
  }else{
    int ioDelay;
    if( Tcl_GetIntFromObj(interp, objv[1], &ioDelay) ) return TCL_ERROR;
    async.ioDelay = ioDelay;

  }






  return TCL_OK;
}









/*
** sqlite3async_start
**
** Start a new writer thread.
*/
static int testAsyncStart(
  void * clientData,
  Tcl_Interp *interp,
  int objc,
  Tcl_Obj *CONST objv[]
){



  pthread_t x;


  int rc;
  volatile int isStarted = 0;
  rc = pthread_create(&x, 0, asyncWriterThread, (void *)&isStarted);


  if( rc ){
    Tcl_AppendResult(interp, "failed to create the thread", 0);
    return TCL_ERROR;
  }
  pthread_detach(x);
  while( isStarted==0 ){

    sched_yield();

  }
  return TCL_OK;
}

/*
** sqlite3async_wait
**
................................................................................
*/
static int testAsyncWait(
  void * clientData,
  Tcl_Interp *interp,
  int objc,
  Tcl_Obj *CONST objv[]
){
  int cnt = 10;
  if( async.writerHaltNow==0 && async.writerHaltWhenIdle==0 ){






    Tcl_AppendResult(interp, "would block forever", (char*)0);
    return TCL_ERROR;
  }

  while( cnt-- && !pthread_mutex_trylock(&async.writerMutex) ){
    pthread_mutex_unlock(&async.writerMutex);
    sched_yield();
  }
  if( cnt>=0 ){
    ASYNC_TRACE(("WAIT\n"));
    pthread_mutex_lock(&async.queueMutex);
    pthread_cond_broadcast(&async.queueSignal);
    pthread_mutex_unlock(&async.queueMutex);
    pthread_mutex_lock(&async.writerMutex);
    pthread_mutex_unlock(&async.writerMutex);
  }else{
    ASYNC_TRACE(("NO-WAIT\n"));
  }
  return TCL_OK;
}


#endif  /* SQLITE_OS_UNIX and SQLITE_THREADSAFE */

/*
** This routine registers the custom TCL commands defined in this
** module.  This should be the only procedure visible from outside
** of this module.
*/
int Sqlitetestasync_Init(Tcl_Interp *interp){
#if SQLITE_OS_UNIX && SQLITE_THREADSAFE
  Tcl_CreateObjCommand(interp,"sqlite3async_enable",testAsyncEnable,0,0);
  Tcl_CreateObjCommand(interp,"sqlite3async_halt",testAsyncHalt,0,0);
  Tcl_CreateObjCommand(interp,"sqlite3async_delay",testAsyncDelay,0,0);
  Tcl_CreateObjCommand(interp,"sqlite3async_start",testAsyncStart,0,0);
  Tcl_CreateObjCommand(interp,"sqlite3async_wait",testAsyncWait,0,0);
  Tcl_LinkVar(interp, "sqlite3async_trace",
      (char*)&sqlite3async_trace, TCL_LINK_INT);
#endif  /* SQLITE_OS_UNIX and SQLITE_THREADSAFE */
  return TCL_OK;
}








|

|
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<

<

<
<
<
<
<
>


<
<
<
<
<
>

<
<
<
<
<
>
|
|

<
<
<

<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
|
<
<
<
<
<
<
<
<
<
<
<
<
<
<
>

<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
<
>







 







|

|
|
|
>
>
>
>
>





|









|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
>







 







>




|
<
<
<
|
<
>
|
>
>
>
>
>
>


>
>
>
>
>
>
>
>












>
>
>
|
>
>

<
<
>
>
|
<


<

>

>







 







|
|
>
>
>
>
>
>




|
|
<
<
<
<
<
<
<
<
<
<
<
<



<
|







|





<
<
|


>
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
..
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
...
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
...
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
**
**    May you do good and not evil.
**    May you find forgiveness for yourself and forgive others.
**    May you share freely, never taking more than you give.
**
*************************************************************************
**
** $Id: test_async.c,v 1.59 2009/04/23 14:58:40 danielk1977 Exp $
**
** This file contains a binding of the asynchronous IO extension interface
** (defined in ext/async/sqlite3async.h) to Tcl.



























































































*/







#define TCL_THREADS 
#include <tcl.h>






#ifdef SQLITE_ENABLE_ASYNCIO






#include "sqlite3async.h"
#include "sqlite3.h"
#include <assert.h>












































































































































struct TestAsyncGlobal {



























































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































  int isInstalled;                     /* True when async VFS is installed */














} testasync_g = { 0 };





















































































































































































































































































































TCL_DECLARE_MUTEX(testasync_g_writerMutex);

/*
** sqlite3async_enable ?YES/NO?
**
** Enable or disable the asynchronous I/O backend.  This command is
** not thread-safe.  Do not call it while any database connections
** are open.
................................................................................
  Tcl_Obj *CONST objv[]
){
  if( objc!=1 && objc!=2 ){
    Tcl_WrongNumArgs(interp, 1, objv, "?YES/NO?");
    return TCL_ERROR;
  }
  if( objc==1 ){
    Tcl_SetObjResult(interp, Tcl_NewIntObj(testasync_g.isInstalled));
  }else{
    int enable;
    if( Tcl_GetBooleanFromObj(interp, objv[1], &enable) ) return TCL_ERROR;
    if( enable ){
      sqlite3async_initialize(0, 1);
    }else{
      sqlite3async_shutdown();
    }
    testasync_g.isInstalled = enable;
  }
  return TCL_OK;
}

/*
** sqlite3async_halt  ?"now"|"idle"|"never"?
**
** Set the conditions at which the writer thread will halt.
*/
static int testAsyncHalt(
  void * clientData,
  Tcl_Interp *interp,
  int objc,
  Tcl_Obj *CONST objv[]
){
  int eWhen;
  const char *azConstant[] = { "never", "now", "idle", 0 };

  assert( SQLITEASYNC_HALT_NEVER==0 );
  assert( SQLITEASYNC_HALT_NOW==1 );
  assert( SQLITEASYNC_HALT_IDLE==2 );

  if( objc!=1 && objc!=2 ){
    Tcl_WrongNumArgs(interp, 1, objv, "?OPTION?");
    return TCL_ERROR;
  }
  if( objc==2 ){
    if( Tcl_GetIndexFromObj(interp, objv[1], azConstant, "option", 0, &eWhen) ){
      return TCL_ERROR;
    }
    sqlite3async_control(SQLITEASYNC_HALT, eWhen);
  }

  /* Always return the current value of the 'halt' option. */
  sqlite3async_control(SQLITEASYNC_GET_HALT, &eWhen);
  Tcl_SetObjResult(interp, Tcl_NewStringObj(azConstant[eWhen], -1));

  return TCL_OK;
}

/*
** sqlite3async_delay ?MS?
**
** Query or set the number of milliseconds of delay in the writer
................................................................................
*/
static int testAsyncDelay(
  void * clientData,
  Tcl_Interp *interp,
  int objc,
  Tcl_Obj *CONST objv[]
){
  int iMs;
  if( objc!=1 && objc!=2 ){
    Tcl_WrongNumArgs(interp, 1, objv, "?MS?");
    return TCL_ERROR;
  }
  if( objc==2 ){



    if( Tcl_GetIntFromObj(interp, objv[1], &iMs) ){

      return TCL_ERROR;
    }
    sqlite3async_control(SQLITEASYNC_DELAY, iMs);
  }

  /* Always return the current value of the 'delay' option. */
  sqlite3async_control(SQLITEASYNC_GET_DELAY, &iMs);
  Tcl_SetObjResult(interp, Tcl_NewIntObj(iMs));
  return TCL_OK;
}

static Tcl_ThreadCreateType tclWriterThread(ClientData pIsStarted){
  Tcl_MutexLock(&testasync_g_writerMutex);
  *((int *)pIsStarted) = 1;
  sqlite3async_run();
  Tcl_MutexUnlock(&testasync_g_writerMutex);
  TCL_THREAD_CREATE_RETURN;
}

/*
** sqlite3async_start
**
** Start a new writer thread.
*/
static int testAsyncStart(
  void * clientData,
  Tcl_Interp *interp,
  int objc,
  Tcl_Obj *CONST objv[]
){
  volatile int isStarted = 0;
  ClientData threadData = (ClientData)&isStarted;

  Tcl_ThreadId x;
  const int nStack = TCL_THREAD_STACK_DEFAULT;
  const int flags = TCL_THREAD_NOFLAGS;
  int rc;



  rc = Tcl_CreateThread(&x, tclWriterThread, threadData, nStack, flags);
  if( rc!=TCL_OK ){

    return TCL_ERROR;
  }

  while( isStarted==0 ){
#if 0
    sched_yield();
#endif
  }
  return TCL_OK;
}

/*
** sqlite3async_wait
**
................................................................................
*/
static int testAsyncWait(
  void * clientData,
  Tcl_Interp *interp,
  int objc,
  Tcl_Obj *CONST objv[]
){
  int eCond;
  if( objc!=1 ){
    Tcl_WrongNumArgs(interp, 1, objv, "");
    return TCL_ERROR;
  }

  sqlite3async_control(SQLITEASYNC_GET_HALT, &eCond);
  if( eCond==SQLITEASYNC_HALT_NEVER ){
    Tcl_AppendResult(interp, "would block forever", (char*)0);
    return TCL_ERROR;
  }

  Tcl_MutexLock(&testasync_g_writerMutex);
  Tcl_MutexUnlock(&testasync_g_writerMutex);












  return TCL_OK;
}


#endif  /* SQLITE_ENABLE_ASYNCIO */

/*
** This routine registers the custom TCL commands defined in this
** module.  This should be the only procedure visible from outside
** of this module.
*/
int Sqlitetestasync_Init(Tcl_Interp *interp){
#if SQLITE_ENABLE_ASYNCIO
  Tcl_CreateObjCommand(interp,"sqlite3async_enable",testAsyncEnable,0,0);
  Tcl_CreateObjCommand(interp,"sqlite3async_halt",testAsyncHalt,0,0);
  Tcl_CreateObjCommand(interp,"sqlite3async_delay",testAsyncDelay,0,0);
  Tcl_CreateObjCommand(interp,"sqlite3async_start",testAsyncStart,0,0);
  Tcl_CreateObjCommand(interp,"sqlite3async_wait",testAsyncWait,0,0);


#endif  /* SQLITE_ENABLE_ASYNCIO */
  return TCL_OK;
}

Changes to test/insert3.test.

7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
...
162
163
164
165
166
167
168
169
170

171


172
173
174
175
176
177
178
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
# This file implements regression tests for SQLite library.  The
# focus of this file is testing corner cases of the INSERT statement.
#
# $Id: insert3.test,v 1.8 2008/12/23 10:37:47 danielk1977 Exp $

set testdir [file dirname $argv0]
source $testdir/tester.tcl

# All the tests in this file require trigger support
#
ifcapable {trigger} {
................................................................................
    execsql {
      CREATE TABLE t6(x,y DEFAULT 4.3, z DEFAULT x'6869');
      INSERT INTO t6 DEFAULT VALUES;
      SELECT * FROM t6;
    }
  } {{} 4.3 hi}
}
db close


file delete -force test.db


sqlite3 db test.db

#-------------------------------------------------------------------------
# While developing tests for a different feature (savepoint) the following
# sequence was found to cause an assert() in btree.c to fail. These
# tests are included to ensure that that bug is fixed.
#







|







 







<

>
|
>
>







7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
...
162
163
164
165
166
167
168

169
170
171
172
173
174
175
176
177
178
179
180
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
# This file implements regression tests for SQLite library.  The
# focus of this file is testing corner cases of the INSERT statement.
#
# $Id: insert3.test,v 1.9 2009/04/23 14:58:40 danielk1977 Exp $

set testdir [file dirname $argv0]
source $testdir/tester.tcl

# All the tests in this file require trigger support
#
ifcapable {trigger} {
................................................................................
    execsql {
      CREATE TABLE t6(x,y DEFAULT 4.3, z DEFAULT x'6869');
      INSERT INTO t6 DEFAULT VALUES;
      SELECT * FROM t6;
    }
  } {{} 4.3 hi}
}


foreach tab [db eval {SELECT name FROM sqlite_master WHERE type = 'table'}] {
  db eval "DROP TABLE $tab"
}
db close
sqlite3 db test.db

#-------------------------------------------------------------------------
# While developing tests for a different feature (savepoint) the following
# sequence was found to cause an assert() in btree.c to fail. These
# tests are included to ensure that that bug is fixed.
#

Changes to test/mutex1.test.

5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
..
93
94
95
96
97
98
99

100
101
102
103
104
105
106
107
108
...
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
#
#    May you do good and not evil.
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# $Id: mutex1.test,v 1.19 2009/03/24 15:08:10 drh Exp $

set testdir [file dirname $argv0]
source $testdir/tester.tcl

ifcapable !mutex {
  finish_test
  return
................................................................................
# Tests mutex1-2.* test the three thread-safety related modes that
# can be selected using sqlite3_config:
#
#   * Serialized mode,
#   * Multi-threaded mode,
#   * Single-threaded mode.
#

set enable_shared_cache [sqlite3_enable_shared_cache 1]
ifcapable threadsafe {
  foreach {mode mutexes} {
    singlethread {}
    multithread  {fast static_lru static_master static_mem static_open static_prng }
    serialized  {fast recursive static_lru static_master static_mem static_open static_prng}
  } {

    do_test mutex1.2.$mode.1 {
................................................................................
        if {$key ne "total" && $value > 0} {
          lappend res $key
        }
      }
      lsort $res
    } [lsort $mutexes]
  }
}
sqlite3_enable_shared_cache $enable_shared_cache

# Open and use a connection in "nomutex" mode. Test that no recursive
# mutexes are obtained.
ifcapable threadsafe {
  do_test mutex1.3.1 {
    catch {db close}
    clear_mutex_counters
    sqlite3 db test.db -nomutex 1
    execsql { SELECT * FROM abc }
  } {1 2 3 1 2 3 1 2 3}
  do_test mutex1.3.2 {







|







 







>
|
<







 







<




<







5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
..
93
94
95
96
97
98
99
100
101

102
103
104
105
106
107
108
...
129
130
131
132
133
134
135

136
137
138
139

140
141
142
143
144
145
146
#
#    May you do good and not evil.
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
#
# $Id: mutex1.test,v 1.20 2009/04/23 14:58:40 danielk1977 Exp $

set testdir [file dirname $argv0]
source $testdir/tester.tcl

ifcapable !mutex {
  finish_test
  return
................................................................................
# Tests mutex1-2.* test the three thread-safety related modes that
# can be selected using sqlite3_config:
#
#   * Serialized mode,
#   * Multi-threaded mode,
#   * Single-threaded mode.
#
ifcapable threadsafe&&shared_cache {
  set enable_shared_cache [sqlite3_enable_shared_cache 1]

  foreach {mode mutexes} {
    singlethread {}
    multithread  {fast static_lru static_master static_mem static_open static_prng }
    serialized  {fast recursive static_lru static_master static_mem static_open static_prng}
  } {

    do_test mutex1.2.$mode.1 {
................................................................................
        if {$key ne "total" && $value > 0} {
          lappend res $key
        }
      }
      lsort $res
    } [lsort $mutexes]
  }

  sqlite3_enable_shared_cache $enable_shared_cache

  # Open and use a connection in "nomutex" mode. Test that no recursive
  # mutexes are obtained.

  do_test mutex1.3.1 {
    catch {db close}
    clear_mutex_counters
    sqlite3 db test.db -nomutex 1
    execsql { SELECT * FROM abc }
  } {1 2 3 1 2 3 1 2 3}
  do_test mutex1.3.2 {

Changes to test/select1.test.

7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
....
1018
1019
1020
1021
1022
1023
1024



1025
1026
1027

1028
1029
1030
1031
1032
1033
1034
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
# This file implements regression tests for SQLite library.  The
# focus of this file is testing the SELECT statement.
#
# $Id: select1.test,v 1.67 2009/04/10 15:38:43 drh Exp $

set testdir [file dirname $argv0]
source $testdir/tester.tcl

# Try to select on a non-existant table.
#
do_test select1-1.1 {
................................................................................
      SELECT count(
        (SELECT a FROM abc WHERE a = NULL AND b >= upper.c) 
      ) FROM abc AS upper;
    }
  } {0}
}




db close
file delete -force test.db
sqlite3 db test.db

do_test select1-14.1 {
  execsql { 
    SELECT * FROM sqlite_master WHERE rowid>10; 
    SELECT * FROM sqlite_master WHERE rowid=10;
    SELECT * FROM sqlite_master WHERE rowid<10;
    SELECT * FROM sqlite_master WHERE rowid<=10;
    SELECT * FROM sqlite_master WHERE rowid>=10;







|







 







>
>
>

<

>







7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
....
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028

1029
1030
1031
1032
1033
1034
1035
1036
1037
#    May you find forgiveness for yourself and forgive others.
#    May you share freely, never taking more than you give.
#
#***********************************************************************
# This file implements regression tests for SQLite library.  The
# focus of this file is testing the SELECT statement.
#
# $Id: select1.test,v 1.68 2009/04/23 14:58:40 danielk1977 Exp $

set testdir [file dirname $argv0]
source $testdir/tester.tcl

# Try to select on a non-existant table.
#
do_test select1-1.1 {
................................................................................
      SELECT count(
        (SELECT a FROM abc WHERE a = NULL AND b >= upper.c) 
      ) FROM abc AS upper;
    }
  } {0}
}

foreach tab [db eval {SELECT name FROM sqlite_master WHERE type = 'table'}] {
  db eval "DROP TABLE $tab"
}
db close

sqlite3 db test.db

do_test select1-14.1 {
  execsql { 
    SELECT * FROM sqlite_master WHERE rowid>10; 
    SELECT * FROM sqlite_master WHERE rowid=10;
    SELECT * FROM sqlite_master WHERE rowid<10;
    SELECT * FROM sqlite_master WHERE rowid<=10;
    SELECT * FROM sqlite_master WHERE rowid>=10;