summaryrefslogtreecommitdiffstats
path: root/queue.c
blob: acfdefb34e83fa29ed8e764a4e3a7fc7bf7a7e6a (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
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
1739
1740
1741
1742
1743
1744
1745
1746
1747
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
1798
1799
1800
1801
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
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
// TODO: "preforked" worker threads
// TODO: do an if(debug) in dbgrintf - performanc ein release build!
// TODO: peekmsg() on first entry, with new/inprogress/deleted entry, destruction in 
// call consumer state. Facilitates retaining messages in queue until action could
// be called!
/* queue.c
 *
 * This file implements the queue object and its several queueing methods.
 * 
 * File begun on 2008-01-03 by RGerhards
 *
 * Copyright 2008 Rainer Gerhards and Adiscon GmbH.
 *
 * This file is part of rsyslog.
 *
 * Rsyslog is free software: you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation, either version 3 of the License, or
 * (at your option) any later version.
 *
 * Rsyslog is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with Rsyslog.  If not, see <http://www.gnu.org/licenses/>.
 *
 * A copy of the GPL can be found in the file "COPYING" in this distribution.
 */
#include "config.h"

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <assert.h>
#include <signal.h>
#include <pthread.h>
#include <fcntl.h>
#include <unistd.h>
#include <errno.h>

#include "rsyslog.h"
#include "syslogd.h"
#include "queue.h"
#include "stringbuf.h"
#include "srUtils.h"
#include "obj.h"

/* static data */
DEFobjStaticHelpers

/* forward-definitions */
rsRetVal queueChkPersist(queue_t *pThis);
static void *queueWorker(void *arg);
static rsRetVal queueGetQueueSize(queue_t *pThis, int *piQueueSize);
static rsRetVal queueChkWrkThrdChanges(queue_t *pThis);

/* methods */

/* send a command to a specific active thread. If the thread is not
 * active, the command is not sent.
 */
static inline rsRetVal
queueTellActWrkThrd(queue_t *pThis, int iIdx, qWrkCmd_t tCmd)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, queue);
	assert(iIdx >= 0 && iIdx <= pThis->iNumWorkerThreads);

	if(pThis->pWrkThrds[iIdx].tCurrCmd >= eWRKTHRD_RUN_INIT) {
		dbgprintf("Queue 0x%lx: sending command %d to thread %d\n", queueGetID(pThis), tCmd, iIdx);
		pThis->pWrkThrds[iIdx].tCurrCmd = tCmd;
	} else {
		dbgprintf("Queue 0x%lx: command %d NOT sent to inactive thread %d\n", queueGetID(pThis), tCmd, iIdx);
	}

	return iRet;
}

/* send a command to a specific thread
 * TODO: check if we can run into trouble with inactive threads
 */
static inline rsRetVal
queueTellWrkThrd(queue_t *pThis, int iIdx, qWrkCmd_t tCmd)
{
	DEFiRet;

	dbgprintf("Queue 0x%lx: sending command %d to thread %d\n", queueGetID(pThis), tCmd, iIdx);
	ISOBJ_TYPE_assert(pThis, queue);
	assert(iIdx >= 0 && iIdx <= pThis->iNumWorkerThreads);

	pThis->pWrkThrds[iIdx].tCurrCmd = tCmd;

	return iRet;
}


/* join a specific worker thread
 */
static inline rsRetVal
queueJoinWrkThrd(queue_t *pThis, int iIdx)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, queue);
	assert(iIdx >= 0 && iIdx <= pThis->iNumWorkerThreads);
	assert(pThis->pWrkThrds[iIdx].tCurrCmd != eWRKTHRD_STOPPED);

	dbgprintf("Queue 0x%lx: thread %d state %d, waiting for exit\n", queueGetID(pThis), iIdx,
		  pThis->pWrkThrds[iIdx].tCurrCmd);
	pthread_join(pThis->pWrkThrds[iIdx].thrdID, NULL);
	pThis->pWrkThrds[iIdx].tCurrCmd = eWRKTHRD_STOPPED; /* back to virgin... */
	dbgprintf("Queue 0x%lx: thread %d state %d, has exited\n", queueGetID(pThis), iIdx,
		  pThis->pWrkThrds[iIdx].tCurrCmd);

	return iRet;
}


/* Starts a worker thread (on a specific index [i]!)
 */
static inline rsRetVal
queueStrtWrkThrd(queue_t *pThis, int i)
{
	DEFiRet;
	int iState;

	ISOBJ_TYPE_assert(pThis, queue);
	assert(i >= 0 && i <= pThis->iNumWorkerThreads);
	assert(pThis->pWrkThrds[i].tCurrCmd < eWRKTHRD_RUN_INIT);

	queueTellWrkThrd(pThis, i, eWRKTHRD_RUN_INIT);
	iState = pthread_create(&(pThis->pWrkThrds[i].thrdID), NULL, queueWorker, (void*) pThis);
	dbgprintf("Queue 0x%lx: Worker thread %x, index %d started with state %d.\n",
		  (unsigned long) pThis, (unsigned) pThis->pWrkThrds[i].thrdID, i, iState);

	return iRet;
}


/* Starts a *new* worker thread. Function searches itself for a free index spot. It must only
 * be called when we have less than max workers active. Pending wrkr thread requests MUST have
 * been processed before calling this function. -- rgerhards, 2008-01-16
 */
static inline rsRetVal
queueStrtNewWrkThrd(queue_t *pThis)
{
	DEFiRet;
	int i;
	int iStartingUp;
	int iState;

	ISOBJ_TYPE_assert(pThis, queue);

	/* find free spot in thread table. If we find at least one worker that is in initializiation,
	 * we do NOT start a new one. Let's give the other one a chance, first.
	 */
	iStartingUp = -1;
	for(i = 1 ; i <= pThis->iNumWorkerThreads ; ++i)
		if(pThis->pWrkThrds[i].tCurrCmd == eWRKTHRD_STOPPED) {
			break;
		} else if(pThis->pWrkThrds[i].tCurrCmd == eWRKTHRD_RUN_INIT) {
			iStartingUp = i;
			break;
		}

dbgprintf("after thrd search: i %d, iStartingUp %d\n", i, iStartingUp);
	if(iStartingUp > -1)
		ABORT_FINALIZE(RS_RET_ALREADY_STARTING);

	assert(i <= pThis->iNumWorkerThreads); /* now there must be a free spot, else something is really wrong! */

	queueTellWrkThrd(pThis, i, eWRKTHRD_RUN_INIT);
	iState = pthread_create(&(pThis->pWrkThrds[i].thrdID), NULL, queueWorker, (void*) pThis);
	dbgprintf("Queue 0x%lx: Worker thread %x, index %d started with state %d.\n",
		  (unsigned long) pThis, (unsigned) pThis->pWrkThrds[i].thrdID, i, iState);
	/* we try to give the starting worker a little boost. It won't help much as we still
 	 * hold the queue's mutex, but at least it has a chance to start on a single-CPU system.
 	 */
	pthread_yield();

finalize_it:
	return iRet;
}


/* send a command to all active worker threads. A start index can be
 * given. Usually, this is 0 or 1. Thread 0 is reserved to disk-assisted
 * mode and this start index take care of the special handling it needs to
 * receive. -- rgerhards, 2008-01-16
 */
static inline rsRetVal
queueTellActWrkThrds(queue_t *pThis, int iStartIdx, qWrkCmd_t tCmd)
{
	DEFiRet;
	int i;

	ISOBJ_TYPE_assert(pThis, queue);
	assert(iStartIdx == 0 || iStartIdx == 1);

	/* tell the workers our request */
	for(i = iStartIdx ; i <= pThis->iNumWorkerThreads ; ++i)
		if(pThis->pWrkThrds[i].tCurrCmd >= eWRKTHRD_TERMINATED)
			queueTellActWrkThrd(pThis, i, tCmd);

	return iRet;
}


/* This once was used to start all regular worker threads. Now, we have
 * dynamic grow of the worker thread pool, based on needs. This function is
 * still preserved, but it now does not start all but only worker 1, which
 * is always present.
 * rgerhards, 2008-01-16
 */
static inline rsRetVal
queueStrtAllWrkThrds(queue_t *pThis)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, queue);
	assert(pThis->pWrkThrds[1].tCurrCmd < eWRKTHRD_RUN_INIT);
	//iRet = queueStrtWrkThrd(pThis, 1);

	return iRet;
}


/* compute an absolute time timeout suitable for calls to pthread_cond_timedwait()
 * rgerhards, 2008-01-14
 */
static rsRetVal
queueTimeoutComp(struct timespec *pt, int iTimeout)
{
	assert(pt != NULL);
	/* compute timeout */
	clock_gettime(CLOCK_REALTIME, pt);
	pt->tv_nsec += (iTimeout % 1000) * 1000000; /* think INTEGER arithmetic! */
	if(pt->tv_nsec > 999999999) { /* overrun? */
		pt->tv_nsec -= 1000000000;
		++pt->tv_sec;
	}
	pt->tv_sec += iTimeout / 1000;
	return RS_RET_OK; /* so far, this is static... */
}


/* wake up all worker threads. Param bWithDAWrk tells if the DA worker
 * is to be awaken, too. It needs special handling because it waits on
 * two different conditions depending on processing state.
 * rgerhards, 2008-01-16
 */
static inline rsRetVal
queueWakeupWrkThrds(queue_t *pThis, int bWithDAWrk)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, queue);

	pthread_cond_broadcast(pThis->notEmpty);
	if(bWithDAWrk && pThis->qRunsDA != QRUNS_REGULAR) {
		/* if running disk-assisted, workers may wait on that condition, too */
		pthread_cond_broadcast(&pThis->condDA);
	}

	return iRet;
}


/* This function Checks if (another) worker threads needs to be started. It
 * must be called while the caller holds a lock on the queue mutex. So it must not
 * do anything that either reaquires the mutex or forces somebody else to aquire
 * it (that would lead to a deadlock).
 * rgerhards, 2008-01-16
 */
static inline rsRetVal
queueChkAndStrtWrk(queue_t *pThis)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, queue);

	/* process any pending thread requests */
	queueChkWrkThrdChanges(pThis);

	/* check if we need to start up another worker (only in regular mode) */
	if(pThis->qRunsDA == QRUNS_REGULAR) {
		if(pThis->iCurNumWrkThrd < pThis->iNumWorkerThreads) {
dbgprintf("Queue %p: less than max workers are running, qsize %d, workers %d\n",
	   pThis, pThis->iQueueSize, pThis->iCurNumWrkThrd);
			/* check if we satisfy the min nbr of messages per worker to start a new one */
			if(pThis->iCurNumWrkThrd == 0 ||
			   pThis->iQueueSize / pThis->iCurNumWrkThrd > pThis->iMinMsgsPerWrkr) {
				dbgprintf("Queue 0x%lx: high activity - starting additional worker thread.\n",
					  queueGetID(pThis));
				queueStrtNewWrkThrd(pThis);
			}
		}
	}
	
	return iRet;
}


/* --------------- code for disk-assisted (DA) queue modes -------------------- */


/* Destruct DA queue. This is the last part of DA-to-normal-mode
 * transistion. This is called asynchronously and some time quite a
 * while after the actual transistion. The key point is that we need to
 * do it at some later time, because we need to destruct the DA queue. That,
 * however, can not be done in a thread that has been signalled 
 * This is to be called when we revert back to our own queue.
 * rgerhards, 2008-01-15
 */
static inline rsRetVal
queueTurnOffDAMode(queue_t *pThis)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, queue);
	assert(pThis->qRunsDA != QRUNS_REGULAR);

	/* if we need to pull any data that we still need from the (child) disk queue,
	 * now would be the time to do so. At present, we do not need this, but I'd like to
	 * keep that comment if future need arises.
	 */

	queueStrtAllWrkThrds(pThis);	/* restore our regular worker threads */
	pThis->qRunsDA = QRUNS_REGULAR; /* tell the world we are back in non-DA mode */

	/* we destruct the queue object, which will also shutdown the queue worker. As the queue is empty,
	 * this will be quick.
	 */
	queueDestruct(pThis->pqDA); /* and now we are ready to destruct the DA queue */
	pThis->pqDA = NULL;

	/* now free the remaining resources */
	pthread_mutex_destroy(&pThis->mutDA);
	pthread_cond_destroy(&pThis->condDA);

	queueTellWrkThrd(pThis, 0, eWRKTHRD_SHUTDOWN_IMMEDIATE);/* finally, tell ourselves to shutdown */
	dbgprintf("Queue 0x%lx: disk-assistance has been turned off, disk queue was empty (iRet %d)\n",
		  queueGetID(pThis), iRet);

	return iRet;
}

/* check if we had any worker thread changes and, if so, act
 * on them. At a minimum, terminated threads are harvested (joined).
 * This function MUST NEVER block on the queue mutex!
 */
static rsRetVal
queueChkWrkThrdChanges(queue_t *pThis)
{
	DEFiRet;
	int i;

	ISOBJ_TYPE_assert(pThis, queue);

	if(pThis->bThrdStateChanged == 0)
		FINALIZE;

	/* go through all threads (including DA thread) */
	for(i = 0 ; i <= pThis->iNumWorkerThreads ; ++i) {
		switch(pThis->pWrkThrds[i].tCurrCmd) {
			case eWRKTHRD_TERMINATED:
				queueJoinWrkThrd(pThis, i);
				break;
			/* these cases just to satisfy the compiler, we do not act an them: */
			case eWRKTHRD_STOPPED:
			case eWRKTHRD_RUN_INIT:
			case eWRKTHRD_RUNNING:
			case eWRKTHRD_SHUTDOWN:
			case eWRKTHRD_SHUTDOWN_IMMEDIATE:
				/* DO NOTHING */
				break;
		}
	}

finalize_it:
	return iRet;
}


/* check if we run in disk-assisted mode and record that
 * setting for easy (and quick!) access in the future. This
 * function must only be called from constructors and only
 * from those that support disk-assisted modes (aka memory-
 * based queue drivers).
 * rgerhards, 2008-01-14
 */
static rsRetVal
queueChkIsDA(queue_t *pThis)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, queue);
	if(pThis->pszFilePrefix != NULL) {
		pThis->bIsDA = 1;
		dbgprintf("Queue 0x%lx: is disk-assisted, disk will be used on demand\n", queueGetID(pThis));
	} else {
		dbgprintf("Queue 0x%lx: is NOT disk-assisted\n", queueGetID(pThis));
	}

	return iRet;
}


/* This is a special consumer to feed the disk-queue in disk-assited mode.
 * When active, our own queue more or less acts as a memory buffer to the disk.
 * So this consumer just needs to drain the memory queue and submit entries
 * to the disk queue. The disk queue will then call the actual consumer from
 * the app point of view (we chain two queues here).
 * This function must also handle the LowWaterMark situation, at which it is
 * switched back to in-memory queueing.
 * rgerhards, 2008-01-14
 */
static inline rsRetVal
queueDAConsumer(queue_t *pThis, int iMyThrdIndx, int iQueueSize, void *pUsr)
{
	DEFiRet;
	int iCancelStateSave;
	int iSizeDAQueue;

	ISOBJ_TYPE_assert(pThis, queue);
	ISOBJ_assert(pUsr);
	assert(pThis->qRunsDA != QRUNS_REGULAR);

dbgprintf("Queue %p/w%d: queueDAConsumer, queue size %d\n", pThis, iMyThrdIndx, iQueueSize);
	CHKiRet(queueEnqObj(pThis->pqDA, pUsr));

	/* We check if we reached the low water mark (but only if we are not in shutdown mode)
	 * Note that the child queue now in almost all cases is non-empty, because we just enqueued
	 * a message.
	 */
	if(iQueueSize <= pThis->iLowWtrMrk && pThis->pWrkThrds[iMyThrdIndx].tCurrCmd == eWRKTHRD_RUNNING) {
		dbgprintf("Queue 0x%lx/w%d: %d entries - passed low water mark in DA mode, sleeping\n",
			  queueGetID(pThis), iMyThrdIndx, iQueueSize);
		pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &iCancelStateSave);
dbgprintf("pre mutex lock (think about CLEANUP!)\n");
		pthread_mutex_lock(&pThis->mutDA);
dbgprintf("mutex locked (think about CLEANUP!)\n");
		/* wait for either passing the high water mark or the child disk queue drain */
		pthread_cond_wait(&pThis->condDA, &pThis->mutDA);
dbgprintf("condition returned\n");
		pthread_mutex_unlock(&pThis->mutDA);
dbgprintf("mutex unlocked (think about CLEANUP!)\n");
		pthread_setcancelstate(iCancelStateSave, NULL);
	}

	/* now check if the DA queue is empty. If so, we can turn off DA mode. Note that we must
	 * use queueGetQueueSize() in order to avoid a race on child iQueueSize. -- rgerhards, 2008-01-16
	 */
	CHKiRet(queueGetQueueSize(pThis->pqDA, &iSizeDAQueue));

dbgprintf("Queue %p/w%d: DA queue size now %d\n", pThis, iMyThrdIndx, iSizeDAQueue);
	if(iSizeDAQueue == 0) {
		CHKiRet(queueTurnOffDAMode(pThis)); /* this also unlocks the mutex! */
	}

finalize_it:
dbgprintf("DAConsumer returns with iRet %d\n", iRet);
	return iRet;
}


/* Start disk-assisted queue mode. All internal settings are changed. This is supposed
 * to be called from the DA worker, which must have been started before. The most important
 * chore of this function is to create the DA queue object. If that function fails,
 * the DA worker should return with an appropriate state, which in turn should lead to
 * a re-set to non-DA mode in the Enq process. The queue mutex must be locked when this
 * function is called, else a race on pThis->qRunsDA may happen.
 * rgerhards, 2008-01-15
 */
static rsRetVal
queueStrtDA(queue_t *pThis)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, queue);

	/* set up sync objects */
	pthread_mutex_init(&pThis->mutDA, NULL);
	pthread_cond_init(&pThis->condDA, NULL);

	/* create message queue */
	CHKiRet(queueConstruct(&pThis->pqDA, QUEUETYPE_DISK , 1, 0, pThis->pConsumer));

	/* as the created queue is the same object class, we take the
	 * liberty to access its properties directly.
	 */
	pThis->pqDA->condSignalOnEmpty = &pThis->condDA;
	pThis->pqDA->mutSignalOnEmpty = &pThis->mutDA;
	pThis->pqDA->condSignalOnEmpty2 = pThis->notEmpty;
	pThis->pqDA->bSignalOnEmpty = 2;

	CHKiRet(queueSetMaxFileSize(pThis->pqDA, pThis->iMaxFileSize));
	CHKiRet(queueSetFilePrefix(pThis->pqDA, pThis->pszFilePrefix, pThis->lenFilePrefix));
	CHKiRet(queueSetiPersistUpdCnt(pThis->pqDA, pThis->iPersistUpdCnt));
	CHKiRet(queueSettoActShutdown(pThis->pqDA, pThis->toActShutdown));
	CHKiRet(queueSettoEnq(pThis->pqDA, pThis->toEnq));
	CHKiRet(queueSetiHighWtrMrk(pThis->pqDA, 0));
	CHKiRet(queueSetiDiscardMrk(pThis->pqDA, 0));
	if(pThis->toQShutdown == 0) {
		CHKiRet(queueSettoQShutdown(pThis->pqDA, 0)); /* if the user really wants... */
	} else {
		/* we use the shortest possible shutdown (0 is endless!) because when we run on disk AND
		 * have an obviously large backlog, we can't finish it in any case. So there is no point
		 * in holding shutdown longer than necessary. -- rgerhards, 2008-01-15
		 */
		CHKiRet(queueSettoQShutdown(pThis->pqDA, 1));
	}

	iRet = queueStart(pThis->pqDA);
	/* file not found is expected, that means it is no previous QIF available */
	if(iRet != RS_RET_OK && iRet != RS_RET_FILE_NOT_FOUND)
		FINALIZE; /* something is wrong */

	/* tell our fellow workers to shut down
	 * NOTE: we do NOT join them by intension! If we did, we would hold draining
	 * the queue until some potentially long-running actions are finished. Having
	 * the ability to immediatly drain the queue was the primary intension of
	 * reserving worker thread 0 for DA queues. So if we would join the other
	 * workers, we would screw up and do against our design goal.
	 */
	CHKiRet(queueTellActWrkThrds(pThis, 1, eWRKTHRD_SHUTDOWN_IMMEDIATE));

	/* as we are right now starting DA mode because we are so busy, it is
	 * extremely unlikely that any worker is sleeping on empty queue. HOWEVER,
	 * we want to be on the safe side, and so we awake anyone that is waiting
	 * on one. So even if the scheduler plays badly with us, things should be
	 * quite well. -- rgerhards, 2008-01-15
	 */
	queueWakeupWrkThrds(pThis, 0); /* awake all workers, but not ourselves ;) */

	pThis->qRunsDA = QRUNS_DA;	/* we are now in DA mode! */

	dbgprintf("Queue 0x%lx: is now running in disk assisted mode, disk queue 0x%lx\n",
		  queueGetID(pThis), queueGetID(pThis->pqDA));

finalize_it:
	if(iRet != RS_RET_OK) {
		if(pThis->pqDA != NULL) {
			queueDestruct(pThis->pqDA);
			pThis->pqDA = NULL;
		}
		dbgprintf("Queue 0x%lx: error %d creating disk queue - giving up.\n",
		  	  queueGetID(pThis), iRet);
		pThis->bIsDA = 0;
	}

	return iRet;
}


/* initiate DA mode
 * rgerhards, 2008-01-16
 */
static inline rsRetVal
queueInitDA(queue_t *pThis)
{
	DEFiRet;

	/* indicate we now run in DA mode - this is reset by the DA worker if it fails */
	pThis->qRunsDA = QRUNS_DA_INIT;

	/* now we must start our DA worker thread - it does the rest of the initialization */
	iRet = queueStrtWrkThrd(pThis, 0);

	return iRet;
}


/* check if we need to start disk assisted mode and send some signals to
 * keep it running if we are already in it.
 * rgerhards, 2008-01-14
 */
static rsRetVal
queueChkStrtDA(queue_t *pThis)
{
	DEFiRet;
	int iCancelStateSave;

	ISOBJ_TYPE_assert(pThis, queue);

	/* if we do not hit the high water mark, we have nothing to do */
	if(pThis->iQueueSize != pThis->iHighWtrMrk)
		ABORT_FINALIZE(RS_RET_OK);

	if(pThis->qRunsDA != QRUNS_REGULAR) {
		/* then we need to signal that we are at the high water mark again. If that happens
		 * on our way down the queue, that doesn't matter, because then nobody is waiting
		 * on the condition variable.
		 */
		dbgprintf("Queue 0x%lx: %d entries - passed high water mark in DA mode, send notify\n",
			  queueGetID(pThis), pThis->iQueueSize);
		pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &iCancelStateSave);
		pthread_mutex_lock(&pThis->mutDA);
		pthread_cond_signal(&pThis->condDA);
		pthread_mutex_unlock(&pThis->mutDA);
		pthread_setcancelstate(iCancelStateSave, NULL);
		queueChkWrkThrdChanges(pThis); /* the queue mode may have changed while we waited, so check! */
	}

	/* we need to re-check if we run disk-assisted, because that status may have changed
	 * in our high water mark processing.
	 */
	if(pThis->qRunsDA == QRUNS_REGULAR) {
		/* if we reach this point, we are NOT currently running in DA mode. */
		dbgprintf("Queue 0x%lx: %d entries - passed high water mark for disk-assisted mode, initiating...\n",
			  queueGetID(pThis), pThis->iQueueSize);

		queueInitDA(pThis); /* initiate DA mode */
	}

finalize_it:
	return iRet;
}


/* --------------- end code for disk-assisted queue modes -------------------- */


/* Now, we define type-specific handlers. The provide a generic functionality,
 * but for this specific type of queue. The mapping to these handlers happens during
 * queue construction. Later on, handlers are called by pointers present in the
 * queue instance object.
 */

/* -------------------- fixed array -------------------- */
static rsRetVal qConstructFixedArray(queue_t *pThis)
{
	DEFiRet;

	assert(pThis != NULL);

	if(pThis->iMaxQueueSize == 0)
		ABORT_FINALIZE(RS_RET_QSIZE_ZERO);

	if((pThis->tVars.farray.pBuf = malloc(sizeof(void *) * pThis->iMaxQueueSize)) == NULL) {
		ABORT_FINALIZE(RS_RET_OUT_OF_MEMORY);
	}

	pThis->tVars.farray.head = 0;
	pThis->tVars.farray.tail = 0;

	queueChkIsDA(pThis);

finalize_it:
	return iRet;
}


static rsRetVal qDestructFixedArray(queue_t *pThis)
{
	DEFiRet;
	
	assert(pThis != NULL);

	if(pThis->tVars.farray.pBuf != NULL)
		free(pThis->tVars.farray.pBuf);

	return iRet;
}

static rsRetVal qAddFixedArray(queue_t *pThis, void* in)
{
	DEFiRet;

	assert(pThis != NULL);
	pThis->tVars.farray.pBuf[pThis->tVars.farray.tail] = in;
	pThis->tVars.farray.tail++;
	if (pThis->tVars.farray.tail == pThis->iMaxQueueSize)
		pThis->tVars.farray.tail = 0;

	return iRet;
}

static rsRetVal qDelFixedArray(queue_t *pThis, void **out)
{
	DEFiRet;

	assert(pThis != NULL);
	*out = (void*) pThis->tVars.farray.pBuf[pThis->tVars.farray.head];

	pThis->tVars.farray.head++;
	if (pThis->tVars.farray.head == pThis->iMaxQueueSize)
		pThis->tVars.farray.head = 0;

	return iRet;
}


/* -------------------- linked list  -------------------- */
static rsRetVal qConstructLinkedList(queue_t *pThis)
{
	DEFiRet;

	assert(pThis != NULL);

	pThis->tVars.linklist.pRoot = 0;
	pThis->tVars.linklist.pLast = 0;

	queueChkIsDA(pThis);

	return iRet;
}


static rsRetVal qDestructLinkedList(queue_t __attribute__((unused)) *pThis)
{
	DEFiRet;
	
	/* with the linked list type, there is nothing to do here. The
	 * reason is that the Destructor is only called after all entries
	 * have bene taken off the queue. In this case, there is nothing
	 * dynamic left with the linked list.
	 */

	return iRet;
}

static rsRetVal qAddLinkedList(queue_t *pThis, void* pUsr)
{
	DEFiRet;
	qLinkedList_t *pEntry;

	assert(pThis != NULL);
	if((pEntry = (qLinkedList_t*) malloc(sizeof(qLinkedList_t))) == NULL) {
		ABORT_FINALIZE(RS_RET_OUT_OF_MEMORY);
	}

	pEntry->pNext = NULL;
	pEntry->pUsr = pUsr;

	if(pThis->tVars.linklist.pRoot == NULL) {
		pThis->tVars.linklist.pRoot = pThis->tVars.linklist.pLast = pEntry;
	} else {
		pThis->tVars.linklist.pLast->pNext = pEntry;
		pThis->tVars.linklist.pLast = pEntry;
	}

finalize_it:
	return iRet;
}

static rsRetVal qDelLinkedList(queue_t *pThis, void **ppUsr)
{
	DEFiRet;
	qLinkedList_t *pEntry;

	assert(pThis != NULL);
	assert(pThis->tVars.linklist.pRoot != NULL);
	
	pEntry = pThis->tVars.linklist.pRoot;
	*ppUsr = pEntry->pUsr;

	if(pThis->tVars.linklist.pRoot == pThis->tVars.linklist.pLast) {
		pThis->tVars.linklist.pRoot = NULL;
		pThis->tVars.linklist.pLast = NULL;
	} else {
		pThis->tVars.linklist.pRoot = pEntry->pNext;
	}
	free(pEntry);

	return iRet;
}


/* -------------------- disk  -------------------- */


/* This method checks if there is any persistent information on the
 * queue.
 */
#if 0
static rsRetVal 
queueTryLoadPersistedInfo(queue_t *pThis)
{
	DEFiRet;
	strm_t *psQIF = NULL;
	uchar pszQIFNam[MAXFNAME];
	size_t lenQIFNam;
	struct stat stat_buf;
}
#endif


static rsRetVal
queueLoadPersStrmInfoFixup(strm_t *pStrm, queue_t *pThis)
{
	DEFiRet;
	ISOBJ_TYPE_assert(pStrm, strm);
	ISOBJ_TYPE_assert(pThis, queue);
	CHKiRet(strmSetDir(pStrm, glblGetWorkDir(), strlen((char*)glblGetWorkDir())));
finalize_it:
	return iRet;
}


/* This method checks if we have a QIF file for the current queue (no matter of
 * queue mode). Returns RS_RET_OK if we have a QIF file or an error status otherwise.
 * rgerhards, 2008-01-15
 */
static rsRetVal 
queueHaveQIF(queue_t *pThis)
{
	DEFiRet;
	uchar pszQIFNam[MAXFNAME];
	size_t lenQIFNam;
	struct stat stat_buf;

	ISOBJ_TYPE_assert(pThis, queue);

	if(pThis->pszFilePrefix == NULL)
		ABORT_FINALIZE(RS_RET_ERR); // TODO: change code!

	/* Construct file name */
	lenQIFNam = snprintf((char*)pszQIFNam, sizeof(pszQIFNam) / sizeof(uchar), "%s/%s.qi",
			     (char*) glblGetWorkDir(), (char*)pThis->pszFilePrefix);

	/* check if the file exists */
dbgprintf("stat HaveQIF '%s'\n", pszQIFNam);
	if(stat((char*) pszQIFNam, &stat_buf) == -1) {
		if(errno == ENOENT) {
			dbgprintf("Queue 0x%lx: no .qi file found\n", queueGetID(pThis));
			ABORT_FINALIZE(RS_RET_FILE_NOT_FOUND);
		} else {
			dbgprintf("Queue 0x%lx: error %d trying to access .qi file\n", queueGetID(pThis), errno);
			ABORT_FINALIZE(RS_RET_IO_ERROR);
		}
	}
	/* If we reach this point, we have a .qi file */

finalize_it:
	return iRet;
}


/* The method loads the persistent queue information.
 * rgerhards, 2008-01-11
 */
static rsRetVal 
queueTryLoadPersistedInfo(queue_t *pThis)
{
	DEFiRet;
	strm_t *psQIF = NULL;
	uchar pszQIFNam[MAXFNAME];
	size_t lenQIFNam;
	struct stat stat_buf;

	ISOBJ_TYPE_assert(pThis, queue);

	/* Construct file name */
	lenQIFNam = snprintf((char*)pszQIFNam, sizeof(pszQIFNam) / sizeof(uchar), "%s/%s.qi",
			     (char*) glblGetWorkDir(), (char*)pThis->pszFilePrefix);

	/* check if the file exists */
dbgprintf("stat '%s'\n", pszQIFNam);
	if(stat((char*) pszQIFNam, &stat_buf) == -1) {
		if(errno == ENOENT) {
			dbgprintf("Queue 0x%lx: clean startup, no .qi file found\n", queueGetID(pThis));
			ABORT_FINALIZE(RS_RET_FILE_NOT_FOUND);
		} else {
			dbgprintf("Queue 0x%lx: error %d trying to access .qi file\n", queueGetID(pThis), errno);
			ABORT_FINALIZE(RS_RET_IO_ERROR);
		}
	}

	/* If we reach this point, we have a .qi file */

	CHKiRet(strmConstruct(&psQIF));
	CHKiRet(strmSetDir(psQIF, glblGetWorkDir(), strlen((char*)glblGetWorkDir())));
	CHKiRet(strmSettOperationsMode(psQIF, STREAMMODE_READ));
	CHKiRet(strmSetsType(psQIF, STREAMTYPE_FILE_SINGLE));
	CHKiRet(strmSetFName(psQIF, pszQIFNam, lenQIFNam));
	CHKiRet(strmConstructFinalize(psQIF));

	/* first, we try to read the property bag for ourselfs */
	CHKiRet(objDeserializePropBag((obj_t*) pThis, psQIF));
	
	/* and now the stream objects (some order as when persisted!) */
	CHKiRet(objDeserialize(&pThis->tVars.disk.pWrite, OBJstrm, psQIF,
			       (rsRetVal(*)(obj_t*,void*))queueLoadPersStrmInfoFixup, pThis));
	CHKiRet(objDeserialize(&pThis->tVars.disk.pRead, OBJstrm, psQIF,
			       (rsRetVal(*)(obj_t*,void*))queueLoadPersStrmInfoFixup, pThis));

	CHKiRet(strmSeekCurrOffs(pThis->tVars.disk.pWrite));
	CHKiRet(strmSeekCurrOffs(pThis->tVars.disk.pRead));

	/* OK, we could successfully read the file, so we now can request that it be
	 * deleted when we are done with the persisted information.
	 */
	pThis->bNeedDelQIF = 1;

finalize_it:
	if(psQIF != NULL)
		strmDestruct(psQIF);

	if(iRet != RS_RET_OK) {
		dbgprintf("Queue 0x%lx: error %d reading .qi file - can not start queue\n",
			  queueGetID(pThis), iRet);
	}

	return iRet;
}


/* disk queue constructor.
 * Note that we use a file limit of 10,000,000 files. That number should never pose a
 * problem. If so, I guess the user has a design issue... But of course, the code can
 * always be changed (though it would probably be more appropriate to increase the
 * allowed file size at this point - that should be a config setting...
 * rgerhards, 2008-01-10
 */
static rsRetVal qConstructDisk(queue_t *pThis)
{
	DEFiRet;
	int bRestarted = 0;

	assert(pThis != NULL);

	/* and now check if there is some persistent information that needs to be read in */
	iRet = queueTryLoadPersistedInfo(pThis);
	if(iRet == RS_RET_OK)
		bRestarted = 1;
	else if(iRet != RS_RET_FILE_NOT_FOUND)
			FINALIZE;

dbgprintf("qConstructDisk: bRestarted %d, iRet %d\n", bRestarted, iRet);
	if(bRestarted == 1) {
		;
	} else {
		CHKiRet(strmConstruct(&pThis->tVars.disk.pWrite));
		CHKiRet(strmSetDir(pThis->tVars.disk.pWrite, glblGetWorkDir(), strlen((char*)glblGetWorkDir())));
		CHKiRet(strmSetiMaxFiles(pThis->tVars.disk.pWrite, 10000000));
		CHKiRet(strmSettOperationsMode(pThis->tVars.disk.pWrite, STREAMMODE_WRITE));
		CHKiRet(strmSetsType(pThis->tVars.disk.pWrite, STREAMTYPE_FILE_CIRCULAR));
		CHKiRet(strmConstructFinalize(pThis->tVars.disk.pWrite));

		CHKiRet(strmConstruct(&pThis->tVars.disk.pRead));
		CHKiRet(strmSetbDeleteOnClose(pThis->tVars.disk.pRead, 1));
		CHKiRet(strmSetDir(pThis->tVars.disk.pRead, glblGetWorkDir(), strlen((char*)glblGetWorkDir())));
		CHKiRet(strmSetiMaxFiles(pThis->tVars.disk.pRead, 10000000));
		CHKiRet(strmSettOperationsMode(pThis->tVars.disk.pRead, STREAMMODE_READ));
		CHKiRet(strmSetsType(pThis->tVars.disk.pRead, STREAMTYPE_FILE_CIRCULAR));
		CHKiRet(strmConstructFinalize(pThis->tVars.disk.pRead));


		CHKiRet(strmSetFName(pThis->tVars.disk.pWrite, pThis->pszFilePrefix, pThis->lenFilePrefix));
		CHKiRet(strmSetFName(pThis->tVars.disk.pRead,  pThis->pszFilePrefix, pThis->lenFilePrefix));
	}

	/* now we set (and overwrite in case of a persisted restart) some parameters which
	 * should always reflect the current configuration variables. Be careful by doing so,
	 * for example file name generation must not be changed as that would break the
	 * ability to read existing queue files. -- rgerhards, 2008-01-12
	 */
CHKiRet(strmSetiMaxFileSize(pThis->tVars.disk.pWrite, pThis->iMaxFileSize));
CHKiRet(strmSetiMaxFileSize(pThis->tVars.disk.pRead, pThis->iMaxFileSize));

finalize_it:
	return iRet;
}


static rsRetVal qDestructDisk(queue_t *pThis)
{
	DEFiRet;
	
	assert(pThis != NULL);
	
	strmDestruct(pThis->tVars.disk.pWrite);
	strmDestruct(pThis->tVars.disk.pRead);

	if(pThis->pszSpoolDir != NULL)
		free(pThis->pszSpoolDir);

	return iRet;
}

static rsRetVal qAddDisk(queue_t *pThis, void* pUsr)
{
	DEFiRet;

	assert(pThis != NULL);

	CHKiRet((objSerialize(pUsr))(pUsr, pThis->tVars.disk.pWrite));
	CHKiRet(strmFlush(pThis->tVars.disk.pWrite));

finalize_it:
	return iRet;
}

static rsRetVal qDelDisk(queue_t *pThis, void **ppUsr)
{
	return objDeserialize(ppUsr, OBJMsg, pThis->tVars.disk.pRead, NULL, NULL);
}

/* -------------------- direct (no queueing) -------------------- */
static rsRetVal qConstructDirect(queue_t __attribute__((unused)) *pThis)
{
	return RS_RET_OK;
}


static rsRetVal qDestructDirect(queue_t __attribute__((unused)) *pThis)
{
	return RS_RET_OK;
}

static rsRetVal qAddDirect(queue_t *pThis, void* pUsr)
{
	DEFiRet;
	rsRetVal iRetLocal;

	assert(pThis != NULL);

	/* TODO: calling the consumer should go into its own function! -- rgerhards, 2008-01-05*/
	iRetLocal = pThis->pConsumer(pUsr);
	if(iRetLocal != RS_RET_OK)
		dbgprintf("Queue 0x%lx: Consumer returned iRet %d\n",
			  queueGetID(pThis), iRetLocal);
	--pThis->iQueueSize; /* this is kind of a hack, but its the smartest thing we can do given
			      * the somewhat astonishing fact that this queue type does not actually
			      * queue anything ;)
			      */

	return iRet;
}

static rsRetVal qDelDirect(queue_t __attribute__((unused)) *pThis, __attribute__((unused)) void **out)
{
	return RS_RET_OK;
}


/* --------------- end type-specific handlers -------------------- */


/* generic code to add a queue entry */
static rsRetVal
queueAdd(queue_t *pThis, void *pUsr)
{
	DEFiRet;

	assert(pThis != NULL);
	CHKiRet(pThis->qAdd(pThis, pUsr));

	++pThis->iQueueSize;

	dbgprintf("Queue 0x%lx: entry added, size now %d entries\n", queueGetID(pThis), pThis->iQueueSize);

finalize_it:
	return iRet;
}


/* generic code to remove a queue entry */
static rsRetVal
queueDel(queue_t *pThis, void *pUsr)
{
	DEFiRet;

	assert(pThis != NULL);

	/* we do NOT abort if we encounter an error, because otherwise the queue
	 * will not be decremented, what will most probably result in an endless loop.
	 * If we decrement, however, we may lose a message. But that is better than
	 * losing the whole process because it loops... -- rgerhards, 2008-01-03
	 */
	iRet = pThis->qDel(pThis, pUsr);
	--pThis->iQueueSize;

	dbgprintf("Queue 0x%lx: entry deleted, state %d, size now %d entries\n",
		  queueGetID(pThis), iRet, pThis->iQueueSize);

	return iRet;
}


/* Send a shutdown command to all workers and awake them. This function
 * does NOT wait for them to terminate. Set bIncludeDAWRk to send the
 * termination command to the DA worker, too (else this does not happen).
 * rgerhards, 2008-01-16
 */
static inline rsRetVal
queueWrkThrdReqTrm(queue_t *pThis, qWrkCmd_t tShutdownCmd, int bIncludeDAWrk)
{
	DEFiRet;

	if(bIncludeDAWrk) {
		queueTellActWrkThrds(pThis, 0, tShutdownCmd);/* first tell the workers our request */
		queueWakeupWrkThrds(pThis, 1); /* awake all workers including DA-worker */
	} else {
		queueTellActWrkThrds(pThis, 1, tShutdownCmd);/* first tell the workers our request */
		queueWakeupWrkThrds(pThis, 0); /* awake all workers but not DA-worker */
	}

	return iRet;
}


/* Send a shutdown command to all workers and see if they terminate.
 * A timeout may be specified.
 * rgerhards, 2008-01-14
 */
static rsRetVal
queueWrkThrdTrm(queue_t *pThis, qWrkCmd_t tShutdownCmd, long iTimeout)
{
	DEFiRet;
	int bTimedOut;
	struct timespec t;

	queueTellActWrkThrds(pThis, 0, tShutdownCmd);/* first tell the workers our request */
	queueWakeupWrkThrds(pThis, 1); /* awake all workers including DA-worker */
	queueTimeoutComp(&t, iTimeout);/* get timeout */
		
	/* and wait for their termination */
	pthread_mutex_lock(pThis->mut);
	bTimedOut = 0;
	while(pThis->iCurNumWrkThrd > 0 && !bTimedOut) {
		dbgprintf("Queue 0x%lx: waiting %ldms on worker thread termination, %d still running\n",
			   queueGetID(pThis), iTimeout, pThis->iCurNumWrkThrd);

		if(pthread_cond_timedwait(&pThis->condThrdTrm, pThis->mut, &t) != 0) {
			dbgprintf("Queue 0x%lx: timeout waiting on worker thread termination\n", queueGetID(pThis));
			bTimedOut = 1;	/* we exit the loop on timeout */
		}
	}
	pthread_mutex_unlock(pThis->mut);

	if(bTimedOut)
		iRet = RS_RET_TIMED_OUT;

	return iRet;
}


/* Unconditionally cancel all running worker threads.
 * rgerhards, 2008-01-14
 */
static rsRetVal
queueWrkThrdCancel(queue_t *pThis)
{
	DEFiRet;
	int i;
	// TODO: we need to implement peek(), without it (today!) we lose one message upon
	// worker cancellation! -- rgerhards, 2008-01-14

	/* awake the workers one more time, just to be sure */
	queueWakeupWrkThrds(pThis, 1); /* awake all workers including DA-worker */

	/* first tell the workers our request */
	for(i = 0 ; i <= pThis->iNumWorkerThreads ; ++i)
		if(pThis->pWrkThrds[i].tCurrCmd >= eWRKTHRD_TERMINATED) {
			dbgprintf("Queue 0x%lx: canceling worker thread %d\n", queueGetID(pThis), i);
			pthread_cancel(pThis->pWrkThrds[i].thrdID);
		}

	return iRet;
}


/* Worker thread management function carried out when the main
 * worker is about to terminate.
 */
static rsRetVal queueShutdownWorkers(queue_t *pThis)
{
	DEFiRet;
	int i;

	assert(pThis != NULL);

	dbgprintf("Queue 0x%lx: initiating worker thread shutdown sequence\n", (unsigned long) pThis);

	/* even if the timeout count is set to 0 (run endless), we still call the queueWrkThrdTrm(). This
	 * is necessary so that all threads get sent the termination command. With a timeout of 0, however,
	 * the function returns immediate with RS_RET_TIMED_OUT. We catch that state and accept it as
	 * good.
	 */
	iRet = queueWrkThrdTrm(pThis, eWRKTHRD_SHUTDOWN, pThis->toQShutdown);
	if(iRet == RS_RET_TIMED_OUT) {
		if(pThis->toQShutdown == 0) {
			iRet = RS_RET_OK;
		} else {
			/* OK, we now need to try force the shutdown */
			dbgprintf("Queue 0x%lx: regular worker shutdown timed out, now trying immediate\n",
				  queueGetID(pThis));
			iRet = queueWrkThrdTrm(pThis, eWRKTHRD_SHUTDOWN_IMMEDIATE, pThis->toActShutdown);
		}
	}

	if(iRet != RS_RET_OK) { /* this is true on actual error on first try or timeout and error on second */
		/* still didn't work out - so we now need to cancel the workers */
		dbgprintf("Queue 0x%lx: worker threads could not be shutdown, now canceling them\n", (unsigned long) pThis);
		iRet = queueWrkThrdCancel(pThis);
	}
	
	/* finally join the threads
	 * In case of a cancellation, this may actually take some time. This is also
	 * needed to clean up the thread descriptors, even with a regular termination.
	 * And, most importantly, this is needed if we have an indifitite termination
	 * time set (timeout == 0)! -- rgerhards, 2008-01-14
	 */
	for(i = 0 ; i <= pThis->iNumWorkerThreads ; ++i) {
		if(pThis->pWrkThrds[i].tCurrCmd != eWRKTHRD_STOPPED) {
			queueJoinWrkThrd(pThis, i);
		}
	}

	/* as we may have cancelled a thread, clean up our internal structure. All are
	 * terminated now. For simplicity, we simply overwrite the states.
	 */
	for(i = 0 ; i <= pThis->iNumWorkerThreads ; ++i) {
		if(pThis->pWrkThrds[i].tCurrCmd != eWRKTHRD_STOPPED) {
			pThis->pWrkThrds[i].tCurrCmd = eWRKTHRD_TERMINATED;
		}
	}

	dbgprintf("Queue 0x%lx: worker threads terminated, remaining queue size %d.\n",
		  queueGetID(pThis), pThis->iQueueSize);

	return iRet;
}


/* This is a helper for queueWorker() it either calls the configured
 * consumer or the DA-consumer (if in disk-assisted mode). It is 
 * protected by the queue mutex, but MUST release it as soon as possible.
 * Most importantly, it must release it before the consumer is called.
 * rgerhards, 2008-01-14
 */
static inline rsRetVal
queueWorkerChkAndCallConsumer(queue_t *pThis, int iMyThrdIndx, int iCancelStateSave)
{
	DEFiRet;
	rsRetVal iRetLocal;
	int iSeverity;
	int iQueueSize;
	void *pUsr;
	int qRunsDA;


	/* first check if we have still something to process */
	if(pThis->iQueueSize == 0 ||
	   (	(pThis->pWrkThrds[iMyThrdIndx].tCurrCmd != eWRKTHRD_RUNNING)
            && (pThis->pWrkThrds[iMyThrdIndx].tCurrCmd != eWRKTHRD_SHUTDOWN)
	   )) {
		pthread_mutex_unlock(pThis->mut);
		pthread_setcancelstate(iCancelStateSave, NULL);
		FINALIZE;
	}

	/* dequeue element (still protected from mutex) */
	iRet = queueDel(pThis, &pUsr);
	queueChkPersist(pThis); // when we support peek(), we must do this down after the del!
	qRunsDA = pThis->qRunsDA; /* do a local copy so that we prevent a race after mutex release */
	iQueueSize = pThis->iQueueSize; /* ... and the same for this property */
	pthread_mutex_unlock(pThis->mut);
	pthread_cond_signal(pThis->notFull);
	pthread_setcancelstate(iCancelStateSave, NULL);
	/* do actual processing (the lengthy part, runs in parallel)
	 * If we had a problem while dequeing, we do not call the consumer,
	 * but we otherwise ignore it. This is in the hopes that it will be
	 * self-healing. However, this is really not a good thing.
	 * rgerhards, 2008-01-03
	 */
	if(iRet != RS_RET_OK)
		FINALIZE;

	if(qRunsDA == QRUNS_DA) {
		queueDAConsumer(pThis, iMyThrdIndx, iQueueSize, pUsr);
	} else {
		/* we are running in normal, non-disk-assisted mode */
		/* do a quick check if we need to drain the queue */
		if(pThis->iDiscardMrk > 0 && iQueueSize >= pThis->iDiscardMrk) {
			iRetLocal = objGetSeverity(pUsr, &iSeverity);
			if(iRetLocal == RS_RET_OK && iSeverity >= pThis->iDiscardSeverity) {
				dbgprintf("Queue 0x%lx/w%d: dequeue/queue nearly full (%d entries), "
					  "discarded severity %d message\n",
					  queueGetID(pThis), iMyThrdIndx, iQueueSize, iSeverity);
				objDestruct(pUsr);
			}
		} else {
			dbgprintf("Queue 0x%lx/w%d: worker executes consumer...\n",
				   queueGetID(pThis), iMyThrdIndx);
			iRetLocal = pThis->pConsumer(pUsr);
			if(iRetLocal != RS_RET_OK) {
				dbgprintf("Queue 0x%lx/w%d: Consumer returned iRet %d\n",
					  queueGetID(pThis), iMyThrdIndx, iRetLocal);
			}
		}
	}

finalize_it:
	if(iRet != RS_RET_OK) {
		dbgprintf("Queue 0x%lx/w%d: error %d dequeueing element - ignoring, but strange things "
			  "may happen\n", queueGetID(pThis), iMyThrdIndx, iRet);
	}
dbgprintf("CallConsumer returns %d\n", iRet);
	return iRet;
}


/* Each queue has at least one associated worker (consumer) thread. It will pull
 * the message from the queue and pass it to a user-defined function.
 * This function was provided on construction. It MUST be thread-safe.
 * Worker thread 0 is always reserved for disk-assisted mode (if the queue
 * is not DA, this worker will be dormant). All other workers are for
 * regular operations mode. Workers are started and stopped as need arises.
 * rgerhards, 2008-01-15
 */
static void *
queueWorker(void *arg)
{
	queue_t *pThis = (queue_t*) arg;
	sigset_t sigSet;
	struct timespec t;
	int iMyThrdIndx;	/* index for this thread in queue thread table */
	int iCancelStateSave;

	ISOBJ_TYPE_assert(pThis, queue);

	sigfillset(&sigSet);
	pthread_sigmask(SIG_BLOCK, &sigSet, NULL);

	/* first find myself in the queue's thread table */
	for(iMyThrdIndx = 0 ; iMyThrdIndx <= pThis->iNumWorkerThreads ; ++iMyThrdIndx)
		if(pThis->pWrkThrds[iMyThrdIndx].thrdID == pthread_self())
			break;
	assert(pThis->pWrkThrds[iMyThrdIndx].thrdID == pthread_self());

	dbgprintf("Queue 0x%lx/w%d: worker thread startup.\n", queueGetID(pThis), iMyThrdIndx);

	/* do some one-time initialization */
	pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &iCancelStateSave);
	pthread_mutex_lock(pThis->mut);

	pThis->iCurNumWrkThrd++; /* tell the world there is one more worker */

	if(iMyThrdIndx == 0) { /* are we the DA worker? */
		if(queueStrtDA(pThis) != RS_RET_OK) { /* then fully initialize the DA queue! */
			/* if we could not init the DA queue, we have nothing to do, so shut down. */
			queueTellWrkThrd(pThis, 0, eWRKTHRD_SHUTDOWN_IMMEDIATE);
		}
	}
			
	/* finally change to RUNNING state. We need to check if we actually should still run,
	 * because someone may have requested us to shut down even before we got a chance to do
	 * our init. That would be a bad race... -- rgerhards, 2008-01-16
	 */
	if(pThis->pWrkThrds[iMyThrdIndx].tCurrCmd == eWRKTHRD_RUN_INIT)
		pThis->pWrkThrds[iMyThrdIndx].tCurrCmd = eWRKTHRD_RUNNING; /* we are running now! */

	pthread_mutex_unlock(pThis->mut);
	pthread_setcancelstate(iCancelStateSave, NULL);
	/* end one-time stuff */

	/* now we have our identity, on to real processing */
	while(pThis->pWrkThrds[iMyThrdIndx].tCurrCmd == eWRKTHRD_RUNNING
	      || (pThis->pWrkThrds[iMyThrdIndx].tCurrCmd == eWRKTHRD_SHUTDOWN && pThis->iQueueSize > 0)) {
		pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &iCancelStateSave);
		pthread_mutex_lock(pThis->mut);

		/* process any pending thread requests */
		queueChkWrkThrdChanges(pThis);

dbgprintf("Queue %p/w%d: pre empty queue, qsize %d\n", pThis, iMyThrdIndx, pThis->iQueueSize);
		while(pThis->iQueueSize == 0 && pThis->pWrkThrds[iMyThrdIndx].tCurrCmd == eWRKTHRD_RUNNING) {
			dbgprintf("Queue 0x%lx/w%d: queue EMPTY, waiting for next message.\n",
				  queueGetID(pThis), iMyThrdIndx);
			if(pThis->bSignalOnEmpty > 0) {
				/* we need to signal our parent queue that we are empty */
dbgprintf("Queue %p/w%d: signal parent we are empty\n", pThis, iMyThrdIndx);
				pthread_mutex_lock(pThis->mutSignalOnEmpty);
				pthread_cond_signal(pThis->condSignalOnEmpty);
				pthread_mutex_unlock(pThis->mutSignalOnEmpty);
dbgprintf("Queue %p/w%d: signaling parent empty done\n", pThis, iMyThrdIndx);
				/* we now need to re-check if we still shall continue to
				 * run. This is important because the parent may have changed our
				 * state. So we simply go back to the begin of the loop.
				 */
				//continue;
			}
			if(pThis->bSignalOnEmpty > 1) {
				/* no mutex associated with this condition, it's just a try (but needed
				 * to wakeup a parent worker if e.g. the queue was restarted from disk) */
				pthread_cond_signal(pThis->condSignalOnEmpty2);
			}
			/* If we arrive here, we have the regular case, where we can safely assume that
			 * iQueueSize and tCmd have not changed since the while().
			 */
dbgprintf("Queue %p/w%d: pre condwait ->notEmpty, worker shutdown %d\n", pThis, iMyThrdIndx, pThis->toWrkShutdown);
			if(pThis->toWrkShutdown == -1) {
dbgprintf("worker never times out!\n");
				/* never shut down any started worker */
				pthread_cond_wait(pThis->notEmpty, pThis->mut);
			} else {
				queueTimeoutComp(&t, pThis->toWrkShutdown);/* get absolute timeout */
				if(pthread_cond_timedwait (pThis->notEmpty, pThis->mut, &t) != 0) {
					dbgprintf("Queue 0x%lx/w%d: inactivity timeout, worker terminating...\n",
						  queueGetID(pThis), iMyThrdIndx);
					/* we use SHUTDOWN (and not SHUTDOWN_IMMEDIATE) so that the worker
					 * does not terminate if in the mean time a new message arrived.
					 */
					pThis->pWrkThrds[iMyThrdIndx].tCurrCmd = eWRKTHRD_SHUTDOWN;
				}
			}
dbgprintf("Queue %p/w%d: post condwait ->notEmpty\n", pThis, iMyThrdIndx);
		}

		queueWorkerChkAndCallConsumer(pThis, iMyThrdIndx, iCancelStateSave);

		/* Now make sure we can get canceled - it is not specified if pthread_setcancelstate() is
		 * a cancellation point in itself. As we run most of the time without cancel enabled, I fear
		 * we may never get cancelled if we do not create a cancellation point ourselfs.
		 */
		pthread_testcancel();
		/* We now yield to give the other threads a chance to obtain the mutex. If we do not
		 * do that, this thread may very well aquire the mutex again before another thread
		 * has even a chance to run. The reason is that mutex operations are free to be
		 * implemented in the quickest possible way (and they typically are!). That is, the
		 * mutex lock/unlock most probably just does an atomic memory swap and does not necessarily
		 * schedule other threads waiting on the same mutex. That can lead to the same thread
		 * aquiring the mutex ever and ever again while all others are starving for it. We
		 * have exactly seen this behaviour when we deliberately introduced a long-running
		 * test action which basically did a sleep. I understand that with real actions the
		 * likelihood of this starvation condition is very low - but it could still happen
		 * and would be very hard to debug. The yield() is a sure fix, its performance overhead
		 * should be well accepted given the above facts. -- rgerhards, 2008-01-10
		 */
		pthread_yield();
dbgprintf("Queue 0x%lx/w%d: end worker run, queue cmd currently %d\n", 
	  queueGetID(pThis), iMyThrdIndx, pThis->pWrkThrds[iMyThrdIndx].tCurrCmd);
		if(Debug && (pThis->pWrkThrds[iMyThrdIndx].tCurrCmd == eWRKTHRD_SHUTDOWN) && pThis->iQueueSize > 0)
			dbgprintf("Queue 0x%lx/w%d: worker does not yet terminate because it still has "
				  " %d messages to process.\n", queueGetID(pThis), iMyThrdIndx, pThis->iQueueSize);
	}

	/* indicate termination */
	pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &iCancelStateSave);
	pthread_mutex_lock(pThis->mut);
	pThis->iCurNumWrkThrd--;
	if(pThis->pWrkThrds[iMyThrdIndx].tCurrCmd == eWRKTHRD_SHUTDOWN ||
	   pThis->pWrkThrds[iMyThrdIndx].tCurrCmd == eWRKTHRD_SHUTDOWN_IMMEDIATE) {
	   	/* in shutdown case, we need to flag termination. All other commands
		 * have a meaning to the thread harvester, so we can not overwrite them
		 */
		pThis->pWrkThrds[iMyThrdIndx].tCurrCmd = eWRKTHRD_TERMINATED;
	}
	pThis->bThrdStateChanged = 1; /* indicate change, so harverster will be called */
	pthread_cond_signal(&pThis->condThrdTrm); /* important for shutdown situation */
	dbgprintf("Queue 0x%lx/w%d: thread terminates with %d entries left in queue, %d workers running.\n",
		  queueGetID(pThis), iMyThrdIndx, pThis->iQueueSize, pThis->iCurNumWrkThrd);
	pthread_mutex_unlock(pThis->mut);
	pthread_setcancelstate(iCancelStateSave, NULL);

	pthread_exit(0);
}


/* Constructor for the queue object
 * This constructs the data structure, but does not yet start the queue. That
 * is done by queueStart(). The reason is that we want to give the caller a chance
 * to modify some parameters before the queue is actually started.
 */
rsRetVal queueConstruct(queue_t **ppThis, queueType_t qType, int iWorkerThreads,
		        int iMaxQueueSize, rsRetVal (*pConsumer)(void*))
{
	DEFiRet;
	queue_t *pThis;

	assert(ppThis != NULL);
	assert(pConsumer != NULL);
	assert(iWorkerThreads >= 0);

	if((pThis = (queue_t *)calloc(1, sizeof(queue_t))) == NULL) {
		ABORT_FINALIZE(RS_RET_OUT_OF_MEMORY);
	}

	/* we have an object, so let's fill the properties */
	objConstructSetObjInfo(pThis);
	if((pThis->pszSpoolDir = (uchar*) strdup((char*)glblGetWorkDir())) == NULL)
		ABORT_FINALIZE(RS_RET_OUT_OF_MEMORY);

	pThis->lenSpoolDir = strlen((char*)pThis->pszSpoolDir);
	pThis->iMaxFileSize = 1024 * 1024; /* default is 1 MiB */
	pThis->iQueueSize = 0;
	pThis->iMaxQueueSize = iMaxQueueSize;
	pThis->pConsumer = pConsumer;
	pThis->mut = (pthread_mutex_t *) malloc (sizeof (pthread_mutex_t));
	pthread_mutex_init (pThis->mut, NULL);
	pThis->notFull = (pthread_cond_t *) malloc (sizeof (pthread_cond_t));
	pthread_cond_init (pThis->notFull, NULL);
	pThis->notEmpty = (pthread_cond_t *) malloc (sizeof (pthread_cond_t));
	pthread_cond_init (pThis->notEmpty, NULL);
	pThis->iNumWorkerThreads = iWorkerThreads;

	pThis->pszFilePrefix = NULL;
	pThis->qType = qType;

	/* set type-specific handlers and other very type-specific things (we can not totally hide it...) */
	switch(qType) {
		case QUEUETYPE_FIXED_ARRAY:
			pThis->qConstruct = qConstructFixedArray;
			pThis->qDestruct = qDestructFixedArray;
			pThis->qAdd = qAddFixedArray;
			pThis->qDel = qDelFixedArray;
			break;
		case QUEUETYPE_LINKEDLIST:
			pThis->qConstruct = qConstructLinkedList;
			pThis->qDestruct = qDestructLinkedList;
			pThis->qAdd = qAddLinkedList;
			pThis->qDel = qDelLinkedList;
			break;
		case QUEUETYPE_DISK:
			pThis->qConstruct = qConstructDisk;
			pThis->qDestruct = qDestructDisk;
			pThis->qAdd = qAddDisk;
			pThis->qDel = qDelDisk;
			/* special handling */
			pThis->iNumWorkerThreads = 1; /* we need exactly one worker */
			break;
		case QUEUETYPE_DIRECT:
			pThis->qConstruct = qConstructDirect;
			pThis->qDestruct = qDestructDirect;
			pThis->qAdd = qAddDirect;
			pThis->qDel = qDelDirect;
			break;
	}

finalize_it:
	OBJCONSTRUCT_CHECK_SUCCESS_AND_CLEANUP
	return iRet;
}


/* start up the queue - it must have been constructed and parameters defined
 * before.
 */
rsRetVal queueStart(queue_t *pThis) /* this is the ConstructionFinalizer */
{
	DEFiRet;
	rsRetVal iRetLocal;
	int bInitialized = 0; /* is queue already initialized? */

	assert(pThis != NULL);

	/* call type-specific constructor */
	CHKiRet(pThis->qConstruct(pThis)); /* this also sets bIsDA */

	dbgprintf("Queue 0x%lx: type %d, disk assisted %d, maxFileSz %ld starting\n", queueGetID(pThis), pThis->qType,
		   pThis->bIsDA, pThis->iMaxFileSize);

	if(pThis->qType == QUEUETYPE_DIRECT)
		FINALIZE;	/* with direct queues, we are already finished... */

	if((pThis->pWrkThrds = calloc(pThis->iNumWorkerThreads + 1, sizeof(qWrkThrd_t))) == NULL)
		ABORT_FINALIZE(RS_RET_OUT_OF_MEMORY);

	if(pThis->bIsDA) {
		/* If we are disk-assisted, we need to check if there is a QIF file
		 * which we need to load. -- rgerhards, 2008-01-15
		 */
		iRetLocal = queueHaveQIF(pThis);
		if(iRetLocal == RS_RET_OK) {
			dbgprintf("Queue 0x%lx: on-disk queue present, needs to be reloaded\n",
				  queueGetID(pThis));

			queueInitDA(pThis); /* initiate DA mode */
			bInitialized = 1; /* we are done */
		} else {
			// TODO: use logerror? -- rgerhards, 2008-01-16
			dbgprintf("Queue 0x%lx: error %d trying to access on-disk queue files, starting without them. "
			          "Some data may be lost\n", queueGetID(pThis), iRetLocal);
		}
	}

	if(!bInitialized) {
		dbgprintf("Queue 0x%lx: queue starts up without (loading) any disk state\n", queueGetID(pThis));
		/* worker 0 is reserved for disk-assisted mode, so do not start */
		queueTellWrkThrd(pThis, 0, eWRKTHRD_STOPPED);

		/* fire up the worker threads */
		queueStrtAllWrkThrds(pThis);
	}

finalize_it:
	return iRet;
}


/* persist the queue to disk. If we have something to persist, we first
 * save the information on the queue properties itself and then we call
 * the queue-type specific drivers.
 * rgerhards, 2008-01-10
 */
static rsRetVal queuePersist(queue_t *pThis)
{
	DEFiRet;
	strm_t *psQIF = NULL; /* Queue Info File */
	uchar pszQIFNam[MAXFNAME];
	size_t lenQIFNam;

	assert(pThis != NULL);

	if(pThis->qType != QUEUETYPE_DISK) {
		if(pThis->iQueueSize > 0)
			ABORT_FINALIZE(RS_RET_NOT_IMPLEMENTED); /* TODO: later... */
		else
			FINALIZE; /* if the queue is empty, we are happy and done... */
	}

	dbgprintf("Queue 0x%lx: persisting queue to disk, %d entries...\n", queueGetID(pThis), pThis->iQueueSize);
	/* Construct file name */
	lenQIFNam = snprintf((char*)pszQIFNam, sizeof(pszQIFNam) / sizeof(uchar), "%s/%s.qi",
		             (char*) glblGetWorkDir(), (char*)pThis->pszFilePrefix);

	if(pThis->iQueueSize == 0) {
		if(pThis->bNeedDelQIF) {
			unlink((char*)pszQIFNam);
			pThis->bNeedDelQIF = 0;
		}
		/* indicate spool file needs to be deleted */
		CHKiRet(strmSetbDeleteOnClose(pThis->tVars.disk.pRead, 1));
		FINALIZE; /* nothing left to do, so be happy */
	}

	CHKiRet(strmConstruct(&psQIF));
	CHKiRet(strmSetDir(psQIF, glblGetWorkDir(), strlen((char*)glblGetWorkDir())));
	CHKiRet(strmSettOperationsMode(psQIF, STREAMMODE_WRITE));
	CHKiRet(strmSetiAddtlOpenFlags(psQIF, O_TRUNC));
	CHKiRet(strmSetsType(psQIF, STREAMTYPE_FILE_SINGLE));
	CHKiRet(strmSetFName(psQIF, pszQIFNam, lenQIFNam));
	CHKiRet(strmConstructFinalize(psQIF));

	/* first, write the property bag for ourselfs
	 * And, surprisingly enough, we currently need to persist only the size of the
	 * queue. All the rest is re-created with then-current config parameters when the
	 * queue is re-created. Well, we'll also save the current queue type, just so that
	 * we know when somebody has changed the queue type... -- rgerhards, 2008-01-11
	 */
	CHKiRet(objBeginSerializePropBag(psQIF, (obj_t*) pThis));
	objSerializeSCALAR(psQIF, iQueueSize, INT);
	CHKiRet(objEndSerialize(psQIF));

	/* this is disk specific and must be moved to a function */
	CHKiRet(strmSerialize(pThis->tVars.disk.pWrite, psQIF));
	CHKiRet(strmSerialize(pThis->tVars.disk.pRead, psQIF));
	
	/* persist queue object itself */

	/* tell the input file object that it must not delete the file on close if the queue is non-empty */
	CHKiRet(strmSetbDeleteOnClose(pThis->tVars.disk.pRead, 0));

	/* we have persisted the queue object. So whenever it comes to an empty queue,
	 * we need to delete the QIF. Thus, we indicte that need.
	 */
	pThis->bNeedDelQIF = 1;

finalize_it:
	if(psQIF != NULL)
		strmDestruct(psQIF);

	return iRet;
}


/* check if we need to persist the current queue info. If an
 * error occurs, thus should be ignored by caller (but we still
 * abide to our regular call interface)...
 * rgerhards, 2008-01-13
 */
rsRetVal queueChkPersist(queue_t *pThis)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, queue);

	if(pThis->iPersistUpdCnt && ++pThis->iUpdsSincePersist >= pThis->iPersistUpdCnt) {
		queuePersist(pThis);
		pThis->iUpdsSincePersist = 0;
	}

	return iRet;
}


/* destructor for the queue object */
rsRetVal queueDestruct(queue_t *pThis)
{
	DEFiRet;

	assert(pThis != NULL);

	/* if running DA, tell the DA workers to shut down. This saves us some CPU cycles which
	 * we can use to persist the remaining in-memory data to disk quicker. -- rgerhads, 2008-01-16
	 * TODO: we actually need to change the queue to an "input-only" mode, that also prevents
	 * startup of the thread again further down in the process. None of that really hurts, so we
	 * leave it for the time being. -- rgerhards, 2008-01-16
	 */
	if(pThis->qRunsDA != QRUNS_REGULAR)
		queueWrkThrdReqTrm(pThis->pqDA, eWRKTHRD_SHUTDOWN_IMMEDIATE, 0);

	/* then, terminate our own worker threads */
	if(pThis->pWrkThrds != NULL) {
		queueShutdownWorkers(pThis);
		free(pThis->pWrkThrds);
		pThis->pWrkThrds = NULL;
	}

	/* of we have now data left in in-memory queues, this data will be lost if we do not
	 * persist it to a disk queue.
	 * TODO: implement code rgerhards, 2008-01-16
	 */

	/* if running DA, terminate disk queue */
	if(pThis->qRunsDA != QRUNS_REGULAR)
		queueDestruct(pThis->pqDA);

	/* persist the queue (we always do that - queuePersits() does cleanup it the queue is empty) */
	CHKiRet_Hdlr(queuePersist(pThis)) {
		dbgprintf("Queue 0x%lx: error %d persisting queue - data lost!\n", (unsigned long) pThis, iRet);
	}

	/* ... then free resources */
	pthread_mutex_destroy(pThis->mut);
	free(pThis->mut);
	pthread_cond_destroy(pThis->notFull);
	free(pThis->notFull);
	pthread_cond_destroy(pThis->notEmpty);
	free(pThis->notEmpty);
	/* type-specific destructor */
	iRet = pThis->qDestruct(pThis);

	if(pThis->pszFilePrefix != NULL)
		free(pThis->pszFilePrefix);

	/* and finally delete the queue objet itself */
	free(pThis);

	return iRet;
}


/* set the queue's file prefix
 * The passed-in string is duplicated. So if the caller does not need
 * it any longer, it must free it.
 * rgerhards, 2008-01-09
 */
rsRetVal
queueSetFilePrefix(queue_t *pThis, uchar *pszPrefix, size_t iLenPrefix)
{
	DEFiRet;

	if(pThis->pszFilePrefix != NULL)
		free(pThis->pszFilePrefix);

	if(pszPrefix == NULL) /* just unset the prefix! */
		ABORT_FINALIZE(RS_RET_OK);

	if((pThis->pszFilePrefix = malloc(sizeof(uchar) * iLenPrefix + 1)) == NULL)
		ABORT_FINALIZE(RS_RET_OUT_OF_MEMORY);
	memcpy(pThis->pszFilePrefix, pszPrefix, iLenPrefix + 1);
	pThis->lenFilePrefix = iLenPrefix;

finalize_it:
	return iRet;
}

/* set the queue's maximum file size
 * rgerhards, 2008-01-09
 */
rsRetVal
queueSetMaxFileSize(queue_t *pThis, size_t iMaxFileSize)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, queue);
	
	if(iMaxFileSize < 1024) {
		ABORT_FINALIZE(RS_RET_VALUE_TOO_LOW);
	}

	pThis->iMaxFileSize = iMaxFileSize;

finalize_it:
	return iRet;
}


/* enqueue a new user data element
 * Enqueues the new element and awakes worker thread.
 * TODO: this code still uses the "discard if queue full" approach from
 * the main queue. This needs to be reconsidered or, better, done via a
 * caller-selectable parameter mode. For the time being, I leave it in.
 * rgerhards, 2008-01-03
 */
rsRetVal
queueEnqObj(queue_t *pThis, void *pUsr)
{
	DEFiRet;
	int iCancelStateSave;
	int i;
	struct timespec t;
	int iSeverity = 8;
	rsRetVal iRetLocal;

	ISOBJ_TYPE_assert(pThis, queue);

	/* Please note that this function is not cancel-safe and consequently
	 * sets the calling thread's cancelibility state to PTHREAD_CANCEL_DISABLE
	 * during its execution. If that is not done, race conditions occur if the
	 * thread is canceled (most important use case is input module termination).
	 * rgerhards, 2008-01-08
	 */
	if(pThis->pWrkThrds != NULL) {
		pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &iCancelStateSave);
		pthread_mutex_lock(pThis->mut);
	}

	/* process any pending thread requests */
	queueChkWrkThrdChanges(pThis);

	/* first check if we can discard anything */
	if(pThis->iDiscardMrk > 0 && pThis->iQueueSize >= pThis->iDiscardMrk) {
		iRetLocal = objGetSeverity(pUsr, &iSeverity);
		if(iRetLocal == RS_RET_OK && iSeverity >= pThis->iDiscardSeverity) {
			dbgprintf("Queue 0x%lx: queue nearly full (%d entries), discarded severity %d message\n",
				  queueGetID(pThis), pThis->iQueueSize, iSeverity);
			objDestruct(pUsr);
			ABORT_FINALIZE(RS_RET_QUEUE_FULL);
		} else {
			dbgprintf("Queue 0x%lx: queue nearly full (%d entries), but could not drop msg "
				  "(iRet: %d, severity %d)\n", queueGetID(pThis), pThis->iQueueSize,
				  iRetLocal, iSeverity);
		}
	}

	/* then check if we need to add an assistance disk queue */
	if(pThis->bIsDA)
		CHKiRet(queueChkStrtDA(pThis));
	
	/* re-process any new pending thread requests and see if we need to start workers */
	queueChkAndStrtWrk(pThis);

	/* and finally (try to) enqueue what is left over */
	while(pThis->iMaxQueueSize > 0 && pThis->iQueueSize >= pThis->iMaxQueueSize) {
		dbgprintf("Queue 0x%lx: enqueueMsg: queue FULL - waiting to drain.\n", queueGetID(pThis));
		queueTimeoutComp(&t, pThis->toEnq);
		if(pthread_cond_timedwait (pThis->notFull, pThis->mut, &t) != 0) {
			dbgprintf("Queue 0x%lx: enqueueMsg: cond timeout, dropping message!\n", queueGetID(pThis));
			objDestruct(pUsr);
			ABORT_FINALIZE(RS_RET_QUEUE_FULL);
		}
	}
	CHKiRet(queueAdd(pThis, pUsr));
	queueChkPersist(pThis);

finalize_it:
	/* now activate the worker thread */
	if(pThis->pWrkThrds != NULL) {
		pthread_mutex_unlock(pThis->mut);
		i = pthread_cond_signal(pThis->notEmpty);
		dbgprintf("Queue 0x%lx: EnqueueMsg signaled condition (%d)\n", (unsigned long) pThis, i);
		pthread_setcancelstate(iCancelStateSave, NULL);
	}


	return iRet;
}

/* some simple object access methods */
DEFpropSetMeth(queue, iPersistUpdCnt, int);
DEFpropSetMeth(queue, toQShutdown, long);
DEFpropSetMeth(queue, toActShutdown, long);
DEFpropSetMeth(queue, toWrkShutdown, long);
DEFpropSetMeth(queue, toEnq, long);
DEFpropSetMeth(queue, iHighWtrMrk, int);
DEFpropSetMeth(queue, iLowWtrMrk, int);
DEFpropSetMeth(queue, iDiscardMrk, int);
DEFpropSetMeth(queue, iDiscardSeverity, int);
DEFpropSetMeth(queue, bIsDA, int);
DEFpropSetMeth(queue, iMinMsgsPerWrkr, int);


/* get the size of this queue. The important thing about this get method is that it
 * is synchronized via the queue mutex. So it provides the information back without
 * any chance of race. Obviously, this causes quite some overhead, so this
 * function should only be called in situations where a race needs to be avoided.
 * rgerhards, 2008-01-16
 */
static rsRetVal
queueGetQueueSize(queue_t *pThis, int *piQueueSize)
{
	DEFiRet;
	int iCancelStateSave;

	ISOBJ_TYPE_assert(pThis, queue);
	assert(piQueueSize != NULL);

	pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &iCancelStateSave);
	pthread_mutex_lock(pThis->mut);

	*piQueueSize = pThis->iQueueSize; /* tell the world there is one more worker */

	pthread_mutex_unlock(pThis->mut);
	pthread_setcancelstate(iCancelStateSave, NULL);

	return iRet;
}


/* This function can be used as a generic way to set properties. Only the subset
 * of properties required to read persisted property bags is supported. This
 * functions shall only be called by the property bag reader, thus it is static.
 * rgerhards, 2008-01-11
 */
#define isProp(name) !rsCStrSzStrCmp(pProp->pcsName, (uchar*) name, sizeof(name) - 1)
static rsRetVal queueSetProperty(queue_t *pThis, property_t *pProp)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, queue);
	assert(pProp != NULL);

 	if(isProp("iQueueSize")) {
		pThis->iQueueSize = pProp->val.vInt;
 	} else if(isProp("qType")) {
		if(pThis->qType != pProp->val.vLong)
			ABORT_FINALIZE(RS_RET_QTYPE_MISMATCH);
	}

finalize_it:
	return iRet;
}
#undef	isProp


/* Initialize the stream class. Must be called as the very first method
 * before anything else is called inside this class.
 * rgerhards, 2008-01-09
 */
BEGINObjClassInit(queue, 1)
	//OBJSetMethodHandler(objMethod_SERIALIZE, strmSerialize);
	OBJSetMethodHandler(objMethod_SETPROPERTY, queueSetProperty);
	//OBJSetMethodHandler(objMethod_CONSTRUCTION_FINALIZER, strmConstructFinalize);
ENDObjClassInit(queue)

/*
 * vi:set ai:
 */