summaryrefslogtreecommitdiffstats
path: root/runtime/stream.c
blob: d76f12e7e4175769f5c9df4b5cc6227eb5d64620 (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
/* The serial stream class.
 *
 * A serial stream provides serial data access. In theory, serial streams
 * can be implemented via a number of methods (e.g. files or in-memory
 * streams). In practice, there currently only exist the file type (aka
 * "driver").
 *
 * File begun on 2008-01-09 by RGerhards
 * Large modifications in 2009-06 to support using it with omfile, including zip writer.
 * Note that this file obtains the zlib wrapper object is needed, but it never frees it
 * again. While this sounds like a leak (and one may argue it actually is), there is no
 * harm associated with that. The reason is that strm is a core object, so it is terminated
 * only when rsyslogd exists. As we could only release on termination (or else bear more 
 * overhead for keeping track of how many users we have), not releasing zlibw is OK, because
 * it will be released when rsyslogd terminates. We may want to revisit this decision if
 * it turns out to be problematic. Then, we need to quasi-refcount the number of accesses
 * to the object.
 *
 * Copyright 2008, 2009 Rainer Gerhards and Adiscon GmbH.
 *
 * This file is part of the rsyslog runtime library.
 *
 * The rsyslog runtime library is free software: you can redistribute it and/or modify
 * it under the terms of the GNU Lesser General Public License as published by
 * the Free Software Foundation, either version 3 of the License, or
 * (at your option) any later version.
 *
 * The rsyslog runtime library 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 Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public License
 * along with the rsyslog runtime library.  If not, see <http://www.gnu.org/licenses/>.
 *
 * A copy of the GPL can be found in the file "COPYING" in this distribution.
 * A copy of the LGPL can be found in the file "COPYING.LESSER" in this distribution.
 */
#include "config.h"

#include <stdlib.h>
#include <string.h>
#include <assert.h>
#include <signal.h>
#include <pthread.h>
#include <fcntl.h>
#include <unistd.h>
#include <sys/stat.h>	 /* required for HP UX */
#include <errno.h>
#include <pthread.h>

#include "rsyslog.h"
#include "stringbuf.h"
#include "srUtils.h"
#include "obj.h"
#include "stream.h"
#include "unicode-helper.h"
#include "module-template.h"
#if HAVE_SYS_PRCTL_H
#  include <sys/prctl.h>
#endif

#define inline

/* static data */
DEFobjStaticHelpers
DEFobjCurrIf(zlibw)

/* forward definitions */
static rsRetVal strmFlush(strm_t *pThis);
static rsRetVal strmWrite(strm_t *pThis, uchar *pBuf, size_t lenBuf);
static rsRetVal strmCloseFile(strm_t *pThis);
static void *asyncWriterThread(void *pPtr);
static rsRetVal doZipWrite(strm_t *pThis, uchar *pBuf, size_t lenBuf);
static rsRetVal strmPhysWrite(strm_t *pThis, uchar *pBuf, size_t lenBuf);


/* methods */

/* Try to resolve a size limit situation. This is used to support custom-file size handlers
 * for omfile. It first runs the command, and then checks if we are still above the size
 * treshold. Note that this works only with single file names, NOT with circular names.
 * Note that pszCurrFName can NOT be taken from pThis, because the stream is closed when
 * we are called (and that destroys pszCurrFName, as there is NO CURRENT file name!). So
 * we need to receive the name as a parameter.
 * initially wirtten 2005-06-21, moved to this class & updates 2009-06-01, both rgerhards
 */
static rsRetVal
resolveFileSizeLimit(strm_t *pThis, uchar *pszCurrFName)
{
	uchar *pParams;
	uchar *pCmd;
	uchar *p;
	off_t actualFileSize;
	rsRetVal localRet;
	DEFiRet;
	ISOBJ_TYPE_assert(pThis, strm);
	assert(pszCurrFName != NULL);

	if(pThis->pszSizeLimitCmd == NULL) {
		ABORT_FINALIZE(RS_RET_NON_SIZELIMITCMD); /* nothing we can do in this case... */
	}
	
	/* we first check if we have command line parameters. We assume this, 
	 * when we have a space in the program name. If we find it, everything after
	 * the space is treated as a single argument.
	 */
	CHKmalloc(pCmd = ustrdup(pThis->pszSizeLimitCmd));

	for(p = pCmd ; *p && *p != ' ' ; ++p) {
		/* JUST SKIP */
	}

	if(*p == ' ') {
		*p = '\0'; /* pretend string-end */
		pParams = p+1;
	} else
		pParams = NULL;

	/* the execProg() below is probably not great, but at least is is
	 * fairly secure now. Once we change the way file size limits are
	 * handled, we should also revisit how this command is run (and
	 * with which parameters).   rgerhards, 2007-07-20
	 */
	execProg(pCmd, 1, pParams);

	free(pCmd);

	localRet = getFileSize(pszCurrFName, &actualFileSize);

	if(localRet == RS_RET_OK && actualFileSize >= pThis->iSizeLimit) {
		ABORT_FINALIZE(RS_RET_SIZELIMITCMD_DIDNT_RESOLVE); /* OK, it didn't work out... */
	} else if(localRet != RS_RET_FILE_NOT_FOUND) {
		/* file not found is OK, the command may have moved away the file */
		ABORT_FINALIZE(localRet);
	}

finalize_it:
	if(iRet != RS_RET_OK) {
		if(iRet == RS_RET_SIZELIMITCMD_DIDNT_RESOLVE) {
			DBGPRINTF("file size limit cmd for file '%s' did no resolve situation\n", pszCurrFName);
		} else {
			DBGPRINTF("file size limit cmd for file '%s' failed with code %d.\n", pszCurrFName, iRet);
		}
		pThis->bDisabled = 1;
	}

	RETiRet;
}


/* Check if the file has grown beyond the configured omfile iSizeLimit
 * and, if so, initiate processing.
 */
static rsRetVal
doSizeLimitProcessing(strm_t *pThis)
{
	uchar *pszCurrFName = NULL;
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, strm);
	ASSERT(pThis->iSizeLimit != 0);
	ASSERT(pThis->fd != -1);

	if(pThis->iCurrOffs >= pThis->iSizeLimit) {
		/* strmClosefile() destroys the current file name, so we
		 * need to preserve it.
		 */
		CHKmalloc(pszCurrFName = ustrdup(pThis->pszCurrFName));
		CHKiRet(strmCloseFile(pThis));
		CHKiRet(resolveFileSizeLimit(pThis, pszCurrFName));
	}

finalize_it:
	free(pszCurrFName);
	RETiRet;
}


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

/* do the physical open() call on a file.
 */
static rsRetVal
doPhysOpen(strm_t *pThis)
{
	int iFlags = 0;
	DEFiRet;
	ISOBJ_TYPE_assert(pThis, strm);

	/* compute which flags we need to provide to open */
	switch(pThis->tOperationsMode) {
		case STREAMMODE_READ:
			iFlags = O_CLOEXEC | O_NOCTTY | O_RDONLY;
			break;
		case STREAMMODE_WRITE:	/* legacy mode used inside queue engine */
			iFlags = O_CLOEXEC | O_NOCTTY | O_WRONLY | O_CREAT;
			break;
		case STREAMMODE_WRITE_TRUNC:
			iFlags = O_CLOEXEC | O_NOCTTY | O_WRONLY | O_CREAT | O_TRUNC;
			break;
		case STREAMMODE_WRITE_APPEND:
			iFlags = O_CLOEXEC | O_NOCTTY | O_WRONLY | O_CREAT | O_APPEND;
			break;
		default:assert(0);
			break;
	}
	if(pThis->sType == STREAMTYPE_NAMED_PIPE) {
		DBGPRINTF("Note: stream '%s' is a named pipe, open with O_NONBLOCK\n", pThis->pszCurrFName);
		iFlags |= O_NONBLOCK;
	}

	pThis->fd = open((char*)pThis->pszCurrFName, iFlags, pThis->tOpenMode);
	DBGPRINTF("file '%s' opened as #%d with mode %d\n", pThis->pszCurrFName, pThis->fd, pThis->tOpenMode);
	if(pThis->fd == -1) {
		char errStr[1024];
		int err = errno;
		rs_strerror_r(err, errStr, sizeof(errStr));
		dbgoprint((obj_t*) pThis, "open error %d, file '%s': %s\n", errno, pThis->pszCurrFName, errStr);
		if(err == ENOENT)
			ABORT_FINALIZE(RS_RET_FILE_NOT_FOUND);
		else
			ABORT_FINALIZE(RS_RET_IO_ERROR);
	} else {
		if(!ustrcmp(pThis->pszCurrFName, UCHAR_CONSTANT(_PATH_CONSOLE)) || isatty(pThis->fd)) {
			DBGPRINTF("file %d is a tty-type file\n", pThis->fd);
			pThis->bIsTTY = 1;
		} else {
			pThis->bIsTTY = 0;
		}
	}

finalize_it:
	RETiRet;
}


/* open a strm file
 * It is OK to call this function when the stream is already open. In that
 * case, it returns immediately with RS_RET_OK
 */
static rsRetVal strmOpenFile(strm_t *pThis)
{
	DEFiRet;

	ASSERT(pThis != NULL);

	if(pThis->fd != -1)
		ABORT_FINALIZE(RS_RET_OK);

	if(pThis->pszFName == NULL)
		ABORT_FINALIZE(RS_RET_FILE_PREFIX_MISSING);

	if(pThis->sType == STREAMTYPE_FILE_CIRCULAR) {
		CHKiRet(genFileName(&pThis->pszCurrFName, pThis->pszDir, pThis->lenDir,
				    pThis->pszFName, pThis->lenFName, pThis->iCurrFNum, pThis->iFileNumDigits));
	} else {
		if(pThis->pszDir == NULL) {
			if((pThis->pszCurrFName = (uchar*) strdup((char*) pThis->pszFName)) == NULL)
				ABORT_FINALIZE(RS_RET_OUT_OF_MEMORY);
		} else {
			CHKiRet(genFileName(&pThis->pszCurrFName, pThis->pszDir, pThis->lenDir,
					    pThis->pszFName, pThis->lenFName, -1, 0));
		}
	}

	CHKiRet(doPhysOpen(pThis));

	pThis->iCurrOffs = 0;
	if(pThis->tOperationsMode == STREAMMODE_WRITE_APPEND) {
		/* we need to obtain the current offset */
		off_t offset;
		CHKiRet(getFileSize(pThis->pszCurrFName, &offset));
		pThis->iCurrOffs = offset;
	}

	dbgoprint((obj_t*) pThis, "opened file '%s' for %s as %d\n", pThis->pszCurrFName,
		  (pThis->tOperationsMode == STREAMMODE_READ) ? "READ" : "WRITE", pThis->fd);

finalize_it:
	RETiRet;
}


/* wait for the output writer thread to be done. This must be called before actions
 * that require data to be persisted. May be called in non-async mode and is a null
 * operation than. Must be called with the mutex locked.
 */
static inline void
strmWaitAsyncWriterDone(strm_t *pThis)
{
	BEGINfunc
	if(pThis->bAsyncWrite) {
		/* awake writer thread and make it write out everything */
		pthread_cond_signal(&pThis->notEmpty);
		d_pthread_cond_wait(&pThis->isEmpty, &pThis->mut);
	}
	ENDfunc
}


/* close a strm file
 * Note that the bDeleteOnClose flag is honored. If it is set, the file will be
 * deleted after close. This is in support for the qRead thread.
 * Note: it is valid to call this function when the physical file is closed. If so,
 * strmCloseFile() will still check if there is any unwritten data inside buffers
 * (this may be the case) and, if so, will open the file, write the data, and then
 * close it again (this is done via strmFlush and friends).
 */
static rsRetVal strmCloseFile(strm_t *pThis)
{
	DEFiRet;

	ASSERT(pThis != NULL);
	dbgoprint((obj_t*) pThis, "file %d closing\n", pThis->fd);

	if(pThis->tOperationsMode != STREAMMODE_READ) {
		strmFlush(pThis);
		if(pThis->bAsyncWrite) {
			strmWaitAsyncWriterDone(pThis);
		}
	}

	close(pThis->fd);
	pThis->fd = -1;

	if(pThis->fdDir != -1) {
		/* close associated directory handle, if it is open */
		close(pThis->fdDir);
		pThis->fdDir = -1;
	}

	if(pThis->bDeleteOnClose) {
		if(unlink((char*) pThis->pszCurrFName) == -1) {
			char errStr[1024];
			int err = errno;
			rs_strerror_r(err, errStr, sizeof(errStr));
			DBGPRINTF("error %d unlinking '%s' - ignored: %s\n",
				   errno, pThis->pszCurrFName, errStr);
		}
	}

	pThis->iCurrOffs = 0;	/* we are back at begin of file */
	if(pThis->pszCurrFName != NULL) {
		free(pThis->pszCurrFName);	/* no longer needed in any case (just for open) */
		pThis->pszCurrFName = NULL;
	}

	RETiRet;
}


/* switch to next strm file
 * This method must only be called if we are in a multi-file mode!
 */
static rsRetVal
strmNextFile(strm_t *pThis)
{
	DEFiRet;

	ASSERT(pThis != NULL);
	ASSERT(pThis->iMaxFiles != 0);
	ASSERT(pThis->fd != -1);

	CHKiRet(strmCloseFile(pThis));

	/* we do modulo operation to ensure we obey the iMaxFile property. This will always
	 * result in a file number lower than iMaxFile, so it if wraps, the name is back to
	 * 0, which results in the first file being overwritten. Not desired for queues, so
	 * make sure their iMaxFiles is large enough. But it is well-desired for other
	 * use cases, e.g. a circular output log file. -- rgerhards, 2008-01-10
	 */
	pThis->iCurrFNum = (pThis->iCurrFNum + 1) % pThis->iMaxFiles;

finalize_it:
	RETiRet;
}


/* handle the eof case for monitored files.
 * If we are monitoring a file, someone may have rotated it. In this case, we
 * also need to close it and reopen it under the same name.
 * rgerhards, 2008-02-13
 */
static rsRetVal
strmHandleEOFMonitor(strm_t *pThis)
{
	DEFiRet;
	struct stat statOpen;
	struct stat statName;

	ISOBJ_TYPE_assert(pThis, strm);
	/* find inodes of both current descriptor as well as file now in file
	 * system. If they are different, the file has been rotated (or
	 * otherwise rewritten). We also check the size, because the inode
	 * does not change if the file is truncated (this, BTW, is also a case
	 * where we actually loose log lines, because we can not do anything
	 * against truncation...). We do NOT rely on the time of last
	 * modificaton because that may not be available under all
	 * circumstances. -- rgerhards, 2008-02-13
	 */
	if(fstat(pThis->fd, &statOpen) == -1)
		ABORT_FINALIZE(RS_RET_IO_ERROR);
	if(stat((char*) pThis->pszCurrFName, &statName) == -1)
		ABORT_FINALIZE(RS_RET_IO_ERROR);
	if(statOpen.st_ino == statName.st_ino && pThis->iCurrOffs == statName.st_size) {
		ABORT_FINALIZE(RS_RET_EOF);
	} else {
		/* we had a file change! */
		CHKiRet(strmCloseFile(pThis));
		CHKiRet(strmOpenFile(pThis));
	}

finalize_it:
	RETiRet;
}


/* handle the EOF case of a stream
 * The EOF case is somewhat complicated, as the proper action depends on the
 * mode the stream is in. If there are multiple files (circular logs, most
 * important use case is queue files!), we need to close the current file and
 * try to open the next one.
 * rgerhards, 2008-02-13
 */
static rsRetVal
strmHandleEOF(strm_t *pThis)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, strm);
	switch(pThis->sType) {
		case STREAMTYPE_FILE_SINGLE:
		case STREAMTYPE_NAMED_PIPE:
			ABORT_FINALIZE(RS_RET_EOF);
			break;
		case STREAMTYPE_FILE_CIRCULAR:
			/* we have multiple files and need to switch to the next one */
			/* TODO: think about emulating EOF in this case (not yet needed) */
			dbgoprint((obj_t*) pThis, "file %d EOF\n", pThis->fd);
			CHKiRet(strmNextFile(pThis));
			break;
		case STREAMTYPE_FILE_MONITOR:
			CHKiRet(strmHandleEOFMonitor(pThis));
			break;
	}

finalize_it:
	RETiRet;
}

/* read the next buffer from disk
 * rgerhards, 2008-02-13
 */
static rsRetVal
strmReadBuf(strm_t *pThis)
{
	DEFiRet;
	int bRun;
	long iLenRead;

	ISOBJ_TYPE_assert(pThis, strm);
	/* We need to try read at least twice because we may run into EOF and need to switch files. */
	bRun = 1;
	while(bRun) {
		/* first check if we need to (re)open the file. We may have switched to a new one in
		 * circular mode or it may have been rewritten (rotated) if we monitor a file
		 * rgerhards, 2008-02-13
		 */
		CHKiRet(strmOpenFile(pThis));
		iLenRead = read(pThis->fd, pThis->pIOBuf, pThis->sIOBufSize);
		dbgoprint((obj_t*) pThis, "file %d read %ld bytes\n", pThis->fd, iLenRead);
		if(iLenRead == 0) {
			CHKiRet(strmHandleEOF(pThis));
		} else if(iLenRead < 0)
			ABORT_FINALIZE(RS_RET_IO_ERROR);
		else { /* good read */
			pThis->iBufPtrMax = iLenRead;
			bRun = 0;	/* exit loop */
		}
	}
	/* if we reach this point, we had a good read */
	pThis->iBufPtr = 0;

finalize_it:
	RETiRet;
}


/* logically "read" a character from a file. What actually happens is that
 * data is taken from the buffer. Only if the buffer is full, data is read 
 * directly from file. In that case, a read is performed blockwise.
 * rgerhards, 2008-01-07
 * NOTE: needs to be enhanced to support sticking with a strm entry (if not
 * deleted).
 */
static rsRetVal strmReadChar(strm_t *pThis, uchar *pC)
{
	DEFiRet;
	
	ASSERT(pThis != NULL);
	ASSERT(pC != NULL);

	/* DEV debug only: dbgoprint((obj_t*) pThis, "strmRead index %d, max %d\n", pThis->iBufPtr, pThis->iBufPtrMax); */
	if(pThis->iUngetC != -1) {	/* do we have an "unread" char that we need to provide? */
		*pC = pThis->iUngetC;
		++pThis->iCurrOffs; /* one more octet read */
		pThis->iUngetC = -1;
		ABORT_FINALIZE(RS_RET_OK);
	}
	
	/* do we need to obtain a new buffer? */
	if(pThis->iBufPtr >= pThis->iBufPtrMax) {
		CHKiRet(strmReadBuf(pThis));
	}

	/* if we reach this point, we have data available in the buffer */

	*pC = pThis->pIOBuf[pThis->iBufPtr++];
	++pThis->iCurrOffs; /* one more octet read */

finalize_it:
	RETiRet;
}


/* unget a single character just like ungetc(). As with that call, there is only a single
 * character buffering capability.
 * rgerhards, 2008-01-07
 */
static rsRetVal strmUnreadChar(strm_t *pThis, uchar c)
{
	ASSERT(pThis != NULL);
	ASSERT(pThis->iUngetC == -1);
	pThis->iUngetC = c;
	--pThis->iCurrOffs; /* one less octet read - NOTE: this can cause problems if we got a file change
	and immediately do an unread and the file is on a buffer boundary and the stream is then persisted.
	With the queue, this can not happen as an Unread is only done on record begin, which is never split
	accross files. For other cases we accept the very remote risk. -- rgerhards, 2008-01-12 */

	return RS_RET_OK;
}


/* read a line from a strm file. A line is terminated by LF. The LF is read, but it
 * is not returned in the buffer (it is discared). The caller is responsible for
 * destruction of the returned CStr object! -- rgerhards, 2008-01-07
 * rgerhards, 2008-03-27: I now use the ppCStr directly, without any interim
 * string pointer. The reason is that this function my be called by inputs, which
 * are pthread_killed() upon termination. So if we use their native pointer, they
 * can cleanup (but only then).
 */
static rsRetVal
strmReadLine(strm_t *pThis, cstr_t **ppCStr)
{
	DEFiRet;
	uchar c;

	ASSERT(pThis != NULL);
	ASSERT(ppCStr != NULL);

	CHKiRet(cstrConstruct(ppCStr));

	/* now read the line */
	CHKiRet(strmReadChar(pThis, &c));
	while(c != '\n') {
		CHKiRet(cstrAppendChar(*ppCStr, c));
		CHKiRet(strmReadChar(pThis, &c));
	}
	CHKiRet(cstrFinalize(*ppCStr));

finalize_it:
	if(iRet != RS_RET_OK && *ppCStr != NULL)
		cstrDestruct(ppCStr);

	RETiRet;
}


/* Standard-Constructor for the strm object
 */
BEGINobjConstruct(strm) /* be sure to specify the object type also in END macro! */
	pThis->iCurrFNum = 1;
	pThis->fd = -1;
	pThis->fdDir = -1;
	pThis->iUngetC = -1;
	pThis->sType = STREAMTYPE_FILE_SINGLE;
	pThis->sIOBufSize = glblGetIOBufSize();
	pThis->tOpenMode = 0600;
ENDobjConstruct(strm)


/* ConstructionFinalizer
 * rgerhards, 2008-01-09
 */
static rsRetVal strmConstructFinalize(strm_t *pThis)
{
	rsRetVal localRet;
	int i;
	DEFiRet;

	ASSERT(pThis != NULL);

	pThis->iBufPtrMax = 0; /* results in immediate read request */
	if(pThis->iZipLevel) { /* do we need a zip buf? */
		localRet = objUse(zlibw, LM_ZLIBW_FILENAME);
		if(localRet != RS_RET_OK) {
			pThis->iZipLevel = 0;
			DBGPRINTF("stream was requested with zip mode, but zlibw module unavailable (%d) - using "
				  "without zip\n", localRet);
		} else {
			/* we use the same size as the original buf, as we would like
			 * to make sure we can write out everything with a SINGLE api call!
			 * We add another 128 bytes to take care of the gzip header and "all eventualities".
			 */
			CHKmalloc(pThis->pZipBuf = (Bytef*) malloc(sizeof(uchar) * (pThis->sIOBufSize + 128)));
		}
	}

	/* if we are set to sync, we must obtain a file handle to the directory for fsync() purposes */
	if(pThis->bSync && !pThis->bIsTTY) {
		pThis->fdDir = open((char*)pThis->pszDir, O_RDONLY | O_CLOEXEC | O_NOCTTY);
		if(pThis->fdDir == -1) {
			char errStr[1024];
			int err = errno;
			rs_strerror_r(err, errStr, sizeof(errStr));
			DBGPRINTF("error %d opening directory file for fsync() use - fsync for directory disabled: %s\n",
				   errno, errStr);
		}
	}

	/* if we have a flush interval, we need to do async writes in any case */
	if(pThis->iFlushInterval != 0) {
		pThis->bAsyncWrite = 1;
	}

	/* if we work asynchronously, we need a couple of synchronization objects */
	if(pThis->bAsyncWrite) {
		pthread_mutex_init(&pThis->mut, 0);
		pthread_cond_init(&pThis->notFull, 0);
		pthread_cond_init(&pThis->notEmpty, 0);
		pthread_cond_init(&pThis->isEmpty, 0);
		pThis->iCnt = pThis->iEnq = pThis->iDeq = 0;
		for(i = 0 ; i < STREAM_ASYNC_NUMBUFS ; ++i) {
			CHKmalloc(pThis->asyncBuf[i].pBuf = (uchar*) malloc(sizeof(uchar) * pThis->sIOBufSize));
		}
		pThis->pIOBuf = pThis->asyncBuf[0].pBuf;
		pThis->bStopWriter = 0;
		if(pthread_create(&pThis->writerThreadID, NULL, asyncWriterThread, pThis) != 0)
			DBGPRINTF("ERROR: stream %p cold not create writer thread\n", pThis);
	} else {
		/* we work synchronously, so we need to alloc a fixed pIOBuf */
		CHKmalloc(pThis->pIOBuf = (uchar*) malloc(sizeof(uchar) * pThis->sIOBufSize));
	}

finalize_it:
	RETiRet;
}


/* stop the writer thread (we MUST be runnnig asynchronously when this method
 * is called!). Note that the mutex must be locked! -- rgerhards, 2009-07-06
 */
static inline void
stopWriter(strm_t *pThis)
{
	BEGINfunc
	pThis->bStopWriter = 1;
	pthread_cond_signal(&pThis->notEmpty);
	d_pthread_mutex_unlock(&pThis->mut);
	pthread_join(pThis->writerThreadID, NULL);
	ENDfunc
}


/* destructor for the strm object */
BEGINobjDestruct(strm) /* be sure to specify the object type also in END and CODESTART macros! */
	int i;
CODESTARTobjDestruct(strm)
	if(pThis->bAsyncWrite)
		/* Note: mutex will be unlocked in stopWriter! */
		d_pthread_mutex_lock(&pThis->mut);

	/* strmClose() will handle read-only files as well as need to open
	 * files that have unwritten buffers. -- rgerhards, 2010-03-09
	 */
	strmCloseFile(pThis);

	if(pThis->bAsyncWrite) {
		stopWriter(pThis);
		pthread_mutex_destroy(&pThis->mut);
		pthread_cond_destroy(&pThis->notFull);
		pthread_cond_destroy(&pThis->notEmpty);
		pthread_cond_destroy(&pThis->isEmpty);
		for(i = 0 ; i < STREAM_ASYNC_NUMBUFS ; ++i) {
			free(pThis->asyncBuf[i].pBuf);
		}
	} else {
		free(pThis->pIOBuf);
	}

	/* Finally, we can free the resources.
	 * IMPORTANT: we MUST free this only AFTER the ansyncWriter has been stopped, else
	 * we get random errors...
	 */
	free(pThis->pszDir);
	free(pThis->pZipBuf);
	free(pThis->pszCurrFName);
	free(pThis->pszFName);

ENDobjDestruct(strm)


/* check if we need to open a new file (in output mode only).
 * The decision is based on file size AND record delimition state.
 * This method may also be called on a closed file, in which case
 * it immediately returns.
 */
static rsRetVal strmCheckNextOutputFile(strm_t *pThis)
{
	DEFiRet;

	if(pThis->fd == -1)
		FINALIZE;

	/* wait for output to be empty, so that our counts are correct */
	strmWaitAsyncWriterDone(pThis);

	if(pThis->iCurrOffs >= pThis->iMaxFileSize) {
		dbgoprint((obj_t*) pThis, "max file size %ld reached for %d, now %ld - starting new file\n",
			  (long) pThis->iMaxFileSize, pThis->fd, (long) pThis->iCurrOffs);
		CHKiRet(strmNextFile(pThis));
	}

finalize_it:
	RETiRet;
}


/* try to recover a tty after a write error. This may have happend
 * due to vhangup(), and, if so, we can simply re-open it.
 */
#ifdef linux
#	define ERR_TTYHUP EIO
#else
#	define ERR_TTYHUP EBADF
#endif
static rsRetVal
tryTTYRecover(strm_t *pThis, int err)
{
	DEFiRet;
	ISOBJ_TYPE_assert(pThis, strm);
	if(err == ERR_TTYHUP) {
		close(pThis->fd);
		CHKiRet(doPhysOpen(pThis));
	}

finalize_it:
	RETiRet;
}
#undef ER_TTYHUP


/* issue write() api calls until either the buffer is completely
 * written or an error occured (it may happen that multiple writes
 * are required, what is perfectly legal. On exit, *pLenBuf contains
 * the number of bytes actually written.
 * rgerhards, 2009-06-08
 */
static rsRetVal
doWriteCall(strm_t *pThis, uchar *pBuf, size_t *pLenBuf)
{
	ssize_t lenBuf;
	ssize_t iTotalWritten;
	ssize_t iWritten;
	char *pWriteBuf;
	DEFiRet;
	ISOBJ_TYPE_assert(pThis, strm);

	lenBuf = *pLenBuf;
	pWriteBuf = (char*) pBuf;
	iTotalWritten = 0;
	do {
		iWritten = write(pThis->fd, pWriteBuf, lenBuf);
		if(iWritten < 0) {
			char errStr[1024];
			int err = errno;
			iWritten = 0; /* we have written NO bytes! */
			rs_strerror_r(err, errStr, sizeof(errStr));
			DBGPRINTF("log file (%d) write error %d: %s\n", pThis->fd, err, errStr);
			if(err == EINTR) {
				/*NO ERROR, just continue */;
			} else {
				if(pThis->bIsTTY) {
					CHKiRet(tryTTYRecover(pThis, err));
				} else {
					ABORT_FINALIZE(RS_RET_IO_ERROR);
					/* Would it make sense to cover more error cases? So far, I 
					 * do not see good reason to do so.
					 */
				}
			}
	 	} 
		/* advance buffer to next write position */
		iTotalWritten += iWritten;
		lenBuf -= iWritten;
		pWriteBuf += iWritten;
	} while(lenBuf > 0);	/* Warning: do..while()! */

	dbgoprint((obj_t*) pThis, "file %d write wrote %d bytes\n", pThis->fd, (int) iWritten);

finalize_it:
	*pLenBuf = iTotalWritten;
	RETiRet;
}



/* write memory buffer to a stream object.
 */
static inline rsRetVal
doWriteInternal(strm_t *pThis, uchar *pBuf, size_t lenBuf)
{
	DEFiRet;

	ASSERT(pThis != NULL);

	if(pThis->iZipLevel) {
		CHKiRet(doZipWrite(pThis, pBuf, lenBuf));
	} else {
		/* write without zipping */
		CHKiRet(strmPhysWrite(pThis, pBuf, lenBuf));
	}

finalize_it:
	RETiRet;
}


/* This function is called to "do" an async write call, what primarily means that 
 * the data is handed over to the writer thread (which will then do the actual write
 * in parallel). Note that the stream mutex has already been locked by the
 * strmWrite...() calls. Also note that we always have only a single producer,
 * so we can simply serially assign the next free buffer to it and be sure that
 * the very some producer comes back in sequence to submit the then-filled buffers.
 * This also enables us to timout on partially written buffers. -- rgerhards, 2009-07-06
 */
static inline rsRetVal
doAsyncWriteInternal(strm_t *pThis, size_t lenBuf)
{
	DEFiRet;
	ISOBJ_TYPE_assert(pThis, strm);

	while(pThis->iCnt >= STREAM_ASYNC_NUMBUFS)
		d_pthread_cond_wait(&pThis->notFull, &pThis->mut);

	pThis->asyncBuf[pThis->iEnq % STREAM_ASYNC_NUMBUFS].lenBuf = lenBuf;
	pThis->pIOBuf = pThis->asyncBuf[++pThis->iEnq % STREAM_ASYNC_NUMBUFS].pBuf;

	pThis->bDoTimedWait = 0; /* everything written, no need to timeout partial buffer writes */
	if(++pThis->iCnt == 1)
		pthread_cond_signal(&pThis->notEmpty);

	RETiRet;
}


/* schedule writing to the stream. Depending on our concurrency settings,
 * this either directly writes to the stream or schedules writing via
 * the background thread. -- rgerhards, 2009-07-07
 */
static rsRetVal
strmSchedWrite(strm_t *pThis, uchar *pBuf, size_t lenBuf)
{
	DEFiRet;

	ASSERT(pThis != NULL);

	/* we need to reset the buffer pointer BEFORE calling the actual write
	 * function. Otherwise, in circular mode, the write function will 
	 * potentially close the file, then close will flush and as the 
	 * buffer pointer is nonzero, will re-call into this code here. In
	 * the end result, we than have a problem (and things are screwed
	 * up). So we reset the buffer pointer first, and all this can
	 * not happen. It is safe to do so, because that pointer is NOT
	 * used inside the write functions. -- rgerhads, 2010-03-10
	 */
	pThis->iBufPtr = 0; /* we are at the begin of a new buffer */
	if(pThis->bAsyncWrite) {
		CHKiRet(doAsyncWriteInternal(pThis, lenBuf));
	} else {
		CHKiRet(doWriteInternal(pThis, pBuf, lenBuf));
	}


finalize_it:
	RETiRet;
}



/* This is the writer thread for asynchronous mode.
 * -- rgerhards, 2009-07-06
 */
static void*
asyncWriterThread(void *pPtr)
{
	int iDeq;
	struct timespec t;
	bool bTimedOut = 0;
	strm_t *pThis = (strm_t*) pPtr;
	ISOBJ_TYPE_assert(pThis, strm);

	BEGINfunc
#	if HAVE_PRCTL && defined PR_SET_NAME
	if(prctl(PR_SET_NAME, "rs:asyn strmwr", 0, 0, 0) != 0) {
		DBGPRINTF("prctl failed, not setting thread name for '%s'\n", "stream writer");
	}
#	endif

	while(1) { /* loop broken inside */
		d_pthread_mutex_lock(&pThis->mut);
		while(pThis->iCnt == 0) {
			if(pThis->bStopWriter) {
				pthread_cond_broadcast(&pThis->isEmpty);
				d_pthread_mutex_unlock(&pThis->mut);
				goto finalize_it; /* break main loop */
			}
			if(bTimedOut && pThis->iBufPtr > 0) {
				/* if we timed out, we need to flush pending data */
				strmFlush(pThis);
				bTimedOut = 0;
				continue; /* now we should have data */
			}
			bTimedOut = 0;
			timeoutComp(&t, pThis->iFlushInterval * 2000); /* *1000 millisconds */
			if(pThis->bDoTimedWait) {
				if(pthread_cond_timedwait(&pThis->notEmpty, &pThis->mut, &t) != 0) {
					int err = errno;
					if(err == ETIMEDOUT) {
						bTimedOut = 1;
					} else {
						bTimedOut = 1;
						char errStr[1024];
						rs_strerror_r(err, errStr, sizeof(errStr));
						DBGPRINTF("stream async writer timeout with error (%d): %s - ignoring\n",
							   err, errStr);
					}
				}
			} else {
				d_pthread_cond_wait(&pThis->notEmpty, &pThis->mut);
			}
		}

		bTimedOut = 0; /* we may have timed out, but there *is* work to do... */

		iDeq = pThis->iDeq++ % STREAM_ASYNC_NUMBUFS;
		doWriteInternal(pThis, pThis->asyncBuf[iDeq].pBuf, pThis->asyncBuf[iDeq].lenBuf);
		// TODO: error check????? 2009-07-06

		--pThis->iCnt;
		if(pThis->iCnt < STREAM_ASYNC_NUMBUFS) {
			pthread_cond_signal(&pThis->notFull);
			if(pThis->iCnt == 0)
				pthread_cond_broadcast(&pThis->isEmpty);
		}
		d_pthread_mutex_unlock(&pThis->mut);
	}

finalize_it:
	ENDfunc
	return NULL; /* to keep pthreads happy */
}


/* sync the file to disk, so that any unwritten data is persisted. This
 * also syncs the directory and thus makes sure that the file survives
 * fatal failure. Note that we do NOT return an error status if the
 * sync fails. Doing so would probably cause more trouble than it
 * is worth (read: data loss may occur where we otherwise might not
 * have it). -- rgerhards, 2009-06-08
 */
#undef SYNCCALL
#if HAVE_FDATASYNC
#	define SYNCCALL(x) fdatasync(x)
#else
#	define SYNCCALL(x) fsync(x)
#endif
static rsRetVal
syncFile(strm_t *pThis)
{
	int ret;
	DEFiRet;

	if(pThis->bIsTTY)
		FINALIZE; /* TTYs can not be synced */

	DBGPRINTF("syncing file %d\n", pThis->fd);
	ret = SYNCCALL(pThis->fd);
	if(ret != 0) {
		char errStr[1024];
		int err = errno;
		rs_strerror_r(err, errStr, sizeof(errStr));
		DBGPRINTF("sync failed for file %d with error (%d): %s - ignoring\n",
			   pThis->fd, err, errStr);
	}
	
	if(pThis->fdDir != -1) {
		ret = fsync(pThis->fdDir);
	}

finalize_it:
	RETiRet;
}
#undef SYNCCALL

/* physically write to the output file. the provided data is ready for
 * writing (e.g. zipped if we are requested to do that).
 * Note that if the write() API fails, we do not reset any pointers, but return
 * an error code. That means we may redo work in the next iteration.
 * rgerhards, 2009-06-04
 */
static rsRetVal
strmPhysWrite(strm_t *pThis, uchar *pBuf, size_t lenBuf)
{
	size_t iWritten;
	DEFiRet;
	ISOBJ_TYPE_assert(pThis, strm);

	if(pThis->fd == -1)
		CHKiRet(strmOpenFile(pThis));

	iWritten = lenBuf;
	CHKiRet(doWriteCall(pThis, pBuf, &iWritten));

	pThis->iCurrOffs += iWritten;
	/* update user counter, if provided */
	if(pThis->pUsrWCntr != NULL)
		*pThis->pUsrWCntr += iWritten;

	if(pThis->bSync) {
		CHKiRet(syncFile(pThis));
	}

	if(pThis->sType == STREAMTYPE_FILE_CIRCULAR) {
		CHKiRet(strmCheckNextOutputFile(pThis));
	} else if(pThis->iSizeLimit != 0) {
		CHKiRet(doSizeLimitProcessing(pThis));
	}

finalize_it:
	RETiRet;
}


/* write the output buffer in zip mode
 * This means we compress it first and then do a physical write.
 * Note that we always do a full deflateInit ... deflate ... deflateEnd
 * sequence. While this is not optimal, we need to do it because we need
 * to ensure that the file is readable even when we are aborted. Doing the
 * full sequence brings us as far towards this goal as possible (and not
 * doing it would be a total failure). It may be worth considering to
 * add a config switch so that the user can decide the risk he is ready
 * to take, but so far this is not yet implemented (not even requested ;)).
 * rgerhards, 2009-06-04
 */
static rsRetVal
doZipWrite(strm_t *pThis, uchar *pBuf, size_t lenBuf)
{
	z_stream zstrm;
	int zRet;	/* zlib return state */
	bool bzInitDone = FALSE;
	DEFiRet;
	assert(pThis != NULL);
	assert(pBuf != NULL);

	/* allocate deflate state */
	zstrm.zalloc = Z_NULL;
	zstrm.zfree = Z_NULL;
	zstrm.opaque = Z_NULL;
	zstrm.next_in = (Bytef*) pBuf;	/* as of zlib doc, this must be set BEFORE DeflateInit2 */
	/* see note in file header for the params we use with deflateInit2() */
	zRet = zlibw.DeflateInit2(&zstrm, pThis->iZipLevel, Z_DEFLATED, 31, 9, Z_DEFAULT_STRATEGY);
	if(zRet != Z_OK) {
		DBGPRINTF("error %d returned from zlib/deflateInit2()\n", zRet);
		ABORT_FINALIZE(RS_RET_ZLIB_ERR);
	}
	bzInitDone = TRUE;

	/* now doing the compression */
	zstrm.next_in = (Bytef*) pBuf;	/* as of zlib doc, this must be set BEFORE DeflateInit2 */
	zstrm.avail_in = lenBuf;
	/* run deflate() on buffer until everything has been compressed */
	do {
		DBGPRINTF("in deflate() loop, avail_in %d, total_in %ld\n", zstrm.avail_in, zstrm.total_in);
		zstrm.avail_out = pThis->sIOBufSize;
		zstrm.next_out = pThis->pZipBuf;
		zRet = zlibw.Deflate(&zstrm, Z_FINISH);    /* no bad return value */
		DBGPRINTF("after deflate, ret %d, avail_out %d\n", zRet, zstrm.avail_out);
		assert(zRet != Z_STREAM_ERROR);  /* state not clobbered */
		if(zstrm.avail_out == pThis->sIOBufSize)
			break; /* this is valid, indicates end of compression --> see zlib howto */
		CHKiRet(strmPhysWrite(pThis, (uchar*)pThis->pZipBuf, pThis->sIOBufSize - zstrm.avail_out));
	} while (zstrm.avail_out == 0);
	assert(zstrm.avail_in == 0);     /* all input will be used */

finalize_it:
	if(bzInitDone) {
		zRet = zlibw.DeflateEnd(&zstrm);
		if(zRet != Z_OK) {
			DBGPRINTF("error %d returned from zlib/deflateEnd()\n", zRet);
		}
	}

	RETiRet;
}


/* flush stream output buffer to persistent storage. This can be called at any time
 * and is automatically called when the output buffer is full.
 * rgerhards, 2008-01-10
 */
static rsRetVal
strmFlush(strm_t *pThis)
{
	DEFiRet;

	ASSERT(pThis != NULL);
	dbgoprint((obj_t*) pThis, "file %d flush, buflen %ld\n", pThis->fd, (long) pThis->iBufPtr);

	if(pThis->tOperationsMode != STREAMMODE_READ && pThis->iBufPtr > 0) {
		iRet = strmSchedWrite(pThis, pThis->pIOBuf, pThis->iBufPtr);
	}

	RETiRet;
}


/* seek a stream to a specific location. Pending writes are flushed, read data
 * is invalidated.
 * rgerhards, 2008-01-12
 */
static rsRetVal strmSeek(strm_t *pThis, off_t offs)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, strm);

	if(pThis->fd == -1)
		strmOpenFile(pThis);
	else
		strmFlush(pThis);
	int i;
	dbgoprint((obj_t*) pThis, "file %d seek, pos %ld\n", pThis->fd, (long) offs);
	i = lseek(pThis->fd, offs, SEEK_SET); // TODO: check error!
	pThis->iCurrOffs = offs; /* we are now at *this* offset */
	pThis->iBufPtr = 0; /* buffer invalidated */

	RETiRet;
}


/* seek to current offset. This is primarily a helper to readjust the OS file
 * pointer after a strm object has been deserialized.
 */
static rsRetVal strmSeekCurrOffs(strm_t *pThis)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, strm);

	iRet = strmSeek(pThis, pThis->iCurrOffs);
	RETiRet;
}


/* write a *single* character to a stream object -- rgerhards, 2008-01-10
 */
static rsRetVal strmWriteChar(strm_t *pThis, uchar c)
{
	DEFiRet;

	ASSERT(pThis != NULL);

	if(pThis->bAsyncWrite)
		d_pthread_mutex_lock(&pThis->mut);

	if(pThis->bDisabled)
		ABORT_FINALIZE(RS_RET_STREAM_DISABLED);

	/* if the buffer is full, we need to flush before we can write */
	if(pThis->iBufPtr == pThis->sIOBufSize) {
		CHKiRet(strmFlush(pThis));
	}
	/* we now always have space for one character, so we simply copy it */
	*(pThis->pIOBuf + pThis->iBufPtr) = c;
	pThis->iBufPtr++;

finalize_it:
	if(pThis->bAsyncWrite)
		d_pthread_mutex_unlock(&pThis->mut);

	RETiRet;
}


/* write an integer value (actually a long) to a stream object
 * Note that we do not need to lock the mutex here, because we call
 * strmWrite(), which does the lock (aka: we must not lock it, else we
 * would run into a recursive lock, resulting in a deadlock!)
 */
static rsRetVal strmWriteLong(strm_t *pThis, long i)
{
	DEFiRet;
	uchar szBuf[32];

	ASSERT(pThis != NULL);

	CHKiRet(srUtilItoA((char*)szBuf, sizeof(szBuf), i));
	CHKiRet(strmWrite(pThis, szBuf, strlen((char*)szBuf)));

finalize_it:
	RETiRet;
}


/* write memory buffer to a stream object.
 * process the data in chunks and copy it over to our buffer. The caller-provided data
 * may theoritically be larger than our buffer. In that case, we do multiple copies. One
 * may argue if it were more efficient to write out the caller-provided buffer in that case
 * and earlier versions of rsyslog did this. However, this introduces a lot of complexity
 * inside the buffered writer and potential performance bottlenecks when trying to solve
 * it. Now keep in mind that we actually do (almost?) never have a case where the
 * caller-provided buffer is larger than our one. So instead of optimizing a case
 * which normally does not exist, we expect some degradation in its case but make us
 * perform better in the regular cases. -- rgerhards, 2009-07-07
 * Note: the pThis->iBufPtr == pThis->sIOBufSize logic below looks a bit like an
 * on-off error. In fact, it is not, because iBufPtr always points to the next
 * *free* byte in the buffer. So if it is sIOBufSize - 1, there actually is one
 * free byte left. This came up during a code walkthrough and was considered
 * worth nothing. -- rgerhards, 2010-03-10
 */
static rsRetVal
strmWrite(strm_t *pThis, uchar *pBuf, size_t lenBuf)
{
	DEFiRet;
	size_t iWrite;
	size_t iOffset;

	ASSERT(pThis != NULL);
	ASSERT(pBuf != NULL);

//DBGPRINTF("strmWrite(%p, '%65.65s', %ld);, disabled %d, sizelim %ld, size %lld\n", pThis, pBuf,lenBuf, pThis->bDisabled, pThis->iSizeLimit, pThis->iCurrOffs);
	if(pThis->bAsyncWrite)
		d_pthread_mutex_lock(&pThis->mut);

	if(pThis->bDisabled)
		ABORT_FINALIZE(RS_RET_STREAM_DISABLED);

	iOffset = 0;
	do {
		if(pThis->iBufPtr == pThis->sIOBufSize) {
			CHKiRet(strmFlush(pThis)); /* get a new buffer for rest of data */
		}
		iWrite = pThis->sIOBufSize - pThis->iBufPtr; /* this fits in current buf */
		if(iWrite > lenBuf)
			iWrite = lenBuf;
		memcpy(pThis->pIOBuf + pThis->iBufPtr, pBuf + iOffset, iWrite);
		pThis->iBufPtr += iWrite;
		iOffset += iWrite;
		lenBuf -= iWrite;
	} while(lenBuf > 0);

	/* now check if the buffer right at the end of the write is full and, if so,
	 * write it. This seems more natural than waiting (hours?) for the next message...
	 */
	if(pThis->iBufPtr == pThis->sIOBufSize) {
		CHKiRet(strmFlush(pThis)); /* get a new buffer for rest of data */
	}

finalize_it:
	if(pThis->bAsyncWrite) {
		if(pThis->bDoTimedWait == 0) {
			/* we potentially have a partial buffer, so re-activate the
			 * writer thread that it can set and pick up timeouts.
			 */
			pThis->bDoTimedWait = 1;
			pthread_cond_signal(&pThis->notEmpty);
		}
		d_pthread_mutex_unlock(&pThis->mut);
	}

	RETiRet;
}


/* property set methods */
/* simple ones first */
DEFpropSetMeth(strm, bDeleteOnClose, int)
DEFpropSetMeth(strm, iMaxFileSize, int)
DEFpropSetMeth(strm, iFileNumDigits, int)
DEFpropSetMeth(strm, tOperationsMode, int)
DEFpropSetMeth(strm, tOpenMode, mode_t)
DEFpropSetMeth(strm, sType, strmType_t)
DEFpropSetMeth(strm, iZipLevel, int)
DEFpropSetMeth(strm, bSync, int)
DEFpropSetMeth(strm, sIOBufSize, size_t)
DEFpropSetMeth(strm, iSizeLimit, off_t)
DEFpropSetMeth(strm, iFlushInterval, int)
DEFpropSetMeth(strm, pszSizeLimitCmd, uchar*)

static rsRetVal strmSetiMaxFiles(strm_t *pThis, int iNewVal)
{
	pThis->iMaxFiles = iNewVal;
	pThis->iFileNumDigits = getNumberDigits(iNewVal);
	return RS_RET_OK;
}


/* set the stream'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
 */
static rsRetVal
strmSetFName(strm_t *pThis, uchar *pszName, size_t iLenName)
{
	DEFiRet;

	ASSERT(pThis != NULL);
	ASSERT(pszName != NULL);
	
	if(iLenName < 1)
		ABORT_FINALIZE(RS_RET_FILE_PREFIX_MISSING);

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

	if((pThis->pszFName = malloc(sizeof(uchar) * (iLenName + 1))) == NULL)
		ABORT_FINALIZE(RS_RET_OUT_OF_MEMORY);

	memcpy(pThis->pszFName, pszName, iLenName + 1); /* always think about the \0! */
	pThis->lenFName = iLenName;

finalize_it:
	RETiRet;
}


/* set the stream's directory
 * The passed-in string is duplicated. So if the caller does not need
 * it any longer, it must free it.
 * rgerhards, 2008-01-09
 */
static rsRetVal
strmSetDir(strm_t *pThis, uchar *pszDir, size_t iLenDir)
{
	DEFiRet;

	ASSERT(pThis != NULL);
	ASSERT(pszDir != NULL);
	
	if(iLenDir < 1)
		ABORT_FINALIZE(RS_RET_FILE_PREFIX_MISSING);

	CHKmalloc(pThis->pszDir = malloc(sizeof(uchar) * iLenDir + 1));

	memcpy(pThis->pszDir, pszDir, iLenDir + 1); /* always think about the \0! */
	pThis->lenDir = iLenDir;

finalize_it:
	RETiRet;
}


/* support for data records
 * The stream class is able to write to multiple files. However, there are
 * situation (actually quite common), where a single data record should not
 * be split across files. This may be problematic if multiple stream write
 * calls are used to create the record. To support that, we provide the
 * bInRecord status variable. If it is set, no file spliting occurs. Once
 * it is set to 0, a check is done if a split is necessary and it then
 * happens. For a record-oriented caller, the proper sequence is:
 *
 * strmRecordBegin()
 * strmWrite...()
 * strmRecordEnd()
 *
 * Please note that records do not affect the writing of output buffers. They
 * are always written when full. The only thing affected is circular files
 * creation. So it is safe to write large records.
 *
 * IMPORTANT: RecordBegin() can not be nested! It is a programming error
 * if RecordBegin() is called while already in a record!
 *
 * rgerhards, 2008-01-10
 */
static rsRetVal strmRecordBegin(strm_t *pThis)
{
	ASSERT(pThis != NULL);
	ASSERT(pThis->bInRecord == 0);
	pThis->bInRecord = 1;
	return RS_RET_OK;
}

static rsRetVal strmRecordEnd(strm_t *pThis)
{
	DEFiRet;
	ASSERT(pThis != NULL);
	ASSERT(pThis->bInRecord == 1);

	pThis->bInRecord = 0;
	iRet = strmCheckNextOutputFile(pThis); /* check if we need to switch files */

	RETiRet;
}
/* end stream record support functions */


/* This method serializes a stream object. That means the whole
 * object is modified into text form. That text form is suitable for
 * later reconstruction of the object.
 * The most common use case for this method is the creation of an
 * on-disk representation of the message object.
 * We do not serialize the dynamic properties. 
 * rgerhards, 2008-01-10
 */
static rsRetVal strmSerialize(strm_t *pThis, strm_t *pStrm)
{
	DEFiRet;
	int i;
	long l;

	ISOBJ_TYPE_assert(pThis, strm);
	ISOBJ_TYPE_assert(pStrm, strm);

	strmFlush(pThis);
	CHKiRet(obj.BeginSerialize(pStrm, (obj_t*) pThis));

	objSerializeSCALAR(pStrm, iCurrFNum, INT);
	objSerializePTR(pStrm, pszFName, PSZ);
	objSerializeSCALAR(pStrm, iMaxFiles, INT);
	objSerializeSCALAR(pStrm, bDeleteOnClose, INT);

	i = pThis->sType;
	objSerializeSCALAR_VAR(pStrm, sType, INT, i);

	i = pThis->tOperationsMode;
	objSerializeSCALAR_VAR(pStrm, tOperationsMode, INT, i);

	i = pThis->tOpenMode;
	objSerializeSCALAR_VAR(pStrm, tOpenMode, INT, i);

	l = (long) pThis->iCurrOffs;
	objSerializeSCALAR_VAR(pStrm, iCurrOffs, LONG, l);

	CHKiRet(obj.EndSerialize(pStrm));

finalize_it:
	RETiRet;
}



/* set a user write-counter. This counter is initialized to zero and
 * receives the number of bytes written. It is accurate only after a
 * flush(). This hook is provided as a means to control disk size usage.
 * The pointer must be valid at all times (so if it is on the stack, be sure
 * to remove it when you exit the function). Pointers are removed by
 * calling strmSetWCntr() with a NULL param. Only one pointer is settable,
 * any new set overwrites the previous one.
 * rgerhards, 2008-02-27
 */
static rsRetVal
strmSetWCntr(strm_t *pThis, number_t *pWCnt)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, strm);

	if(pWCnt != NULL)
		*pWCnt = 0;
	pThis->pUsrWCntr = pWCnt;

	RETiRet;
}


#include "stringbuf.h"

/* This function can be used as a generic way to set properties.
 * rgerhards, 2008-01-11
 */
#define isProp(name) !rsCStrSzStrCmp(pProp->pcsName, UCHAR_CONSTANT(name), sizeof(name) - 1)
static rsRetVal strmSetProperty(strm_t *pThis, var_t *pProp)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, strm);
	ASSERT(pProp != NULL);

 	if(isProp("sType")) {
		CHKiRet(strmSetsType(pThis, (strmType_t) pProp->val.num));
 	} else if(isProp("iCurrFNum")) {
		pThis->iCurrFNum = pProp->val.num;
 	} else if(isProp("pszFName")) {
		CHKiRet(strmSetFName(pThis, rsCStrGetSzStrNoNULL(pProp->val.pStr), rsCStrLen(pProp->val.pStr)));
 	} else if(isProp("tOperationsMode")) {
		CHKiRet(strmSettOperationsMode(pThis, pProp->val.num));
 	} else if(isProp("tOpenMode")) {
		CHKiRet(strmSettOpenMode(pThis, pProp->val.num));
 	} else if(isProp("iCurrOffs")) {
		pThis->iCurrOffs = pProp->val.num;
 	} else if(isProp("iMaxFileSize")) {
		CHKiRet(strmSetiMaxFileSize(pThis, pProp->val.num));
 	} else if(isProp("iMaxFiles")) {
		CHKiRet(strmSetiMaxFiles(pThis, pProp->val.num));
 	} else if(isProp("iFileNumDigits")) {
		CHKiRet(strmSetiFileNumDigits(pThis, pProp->val.num));
 	} else if(isProp("bDeleteOnClose")) {
		CHKiRet(strmSetbDeleteOnClose(pThis, pProp->val.num));
	}

finalize_it:
	RETiRet;
}
#undef	isProp


/* return the current offset inside the stream. Note that on two consequtive calls, the offset
 * reported on the second call may actually be lower than on the first call. This is due to
 * file circulation. A caller must deal with that. -- rgerhards, 2008-01-30
 */
static rsRetVal
strmGetCurrOffset(strm_t *pThis, int64 *pOffs)
{
	DEFiRet;

	ISOBJ_TYPE_assert(pThis, strm);
	ASSERT(pOffs != NULL);

	*pOffs = pThis->iCurrOffs;

	RETiRet;
}


/* queryInterface function
 * rgerhards, 2008-02-29
 */
BEGINobjQueryInterface(strm)
CODESTARTobjQueryInterface(strm)
	if(pIf->ifVersion != strmCURR_IF_VERSION) { /* check for current version, increment on each change */
		ABORT_FINALIZE(RS_RET_INTERFACE_NOT_SUPPORTED);
	}

	/* ok, we have the right interface, so let's fill it
	 * Please note that we may also do some backwards-compatibility
	 * work here (if we can support an older interface version - that,
	 * of course, also affects the "if" above).
	 */
	pIf->Construct = strmConstruct;
	pIf->ConstructFinalize = strmConstructFinalize;
	pIf->Destruct = strmDestruct;
	pIf->ReadChar = strmReadChar;
	pIf->UnreadChar = strmUnreadChar;
	pIf->ReadLine = strmReadLine;
	pIf->SeekCurrOffs = strmSeekCurrOffs;
	pIf->Write = strmWrite;
	pIf->WriteChar = strmWriteChar;
	pIf->WriteLong = strmWriteLong;
	pIf->SetFName = strmSetFName;
	pIf->SetDir = strmSetDir;
	pIf->Flush = strmFlush;
	pIf->RecordBegin = strmRecordBegin;
	pIf->RecordEnd = strmRecordEnd;
	pIf->Serialize = strmSerialize;
	pIf->GetCurrOffset = strmGetCurrOffset;
	pIf->SetWCntr = strmSetWCntr;
	/* set methods */
	pIf->SetbDeleteOnClose = strmSetbDeleteOnClose;
	pIf->SetiMaxFileSize = strmSetiMaxFileSize;
	pIf->SetiMaxFiles = strmSetiMaxFiles;
	pIf->SetiFileNumDigits = strmSetiFileNumDigits;
	pIf->SettOperationsMode = strmSettOperationsMode;
	pIf->SettOpenMode = strmSettOpenMode;
	pIf->SetsType = strmSetsType;
	pIf->SetiZipLevel = strmSetiZipLevel;
	pIf->SetbSync = strmSetbSync;
	pIf->SetsIOBufSize = strmSetsIOBufSize;
	pIf->SetiSizeLimit = strmSetiSizeLimit;
	pIf->SetiFlushInterval = strmSetiFlushInterval;
	pIf->SetpszSizeLimitCmd = strmSetpszSizeLimitCmd;
finalize_it:
ENDobjQueryInterface(strm)


/* 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(strm, 1, OBJ_IS_CORE_MODULE)
	/* request objects we use */

	OBJSetMethodHandler(objMethod_SERIALIZE, strmSerialize);
	OBJSetMethodHandler(objMethod_SETPROPERTY, strmSetProperty);
	OBJSetMethodHandler(objMethod_CONSTRUCTION_FINALIZER, strmConstructFinalize);
ENDObjClassInit(strm)

/* vi:set ai:
 */