summaryrefslogtreecommitdiffstats
path: root/queue.c
blob: b31173d9cbaba62b79ae2eb151314fbde60e0a0b (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
// 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

/* methods */

/* first, 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->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;

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;

	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  -------------------- */

/* 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;

	assert(pThis != NULL);

	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));

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);
}

/* -------------------- 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",
			  (unsigned long) 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", (unsigned long) 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",
		  (unsigned long) pThis, iRet, pThis->iQueueSize);

	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;
	qWrkCmd_t tShutdownCmd;

	assert(pThis != NULL);

	/* select shutdown mode */
	tShutdownCmd = (pThis->bImmediateShutdown) ? eWRKTHRDCMD_SHUTDOWN_IMMEDIATE : eWRKTHRDCMD_SHUTDOWN;

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

	/* tell all workers to terminate */
	for(i = 0 ; i < pThis->iNumWorkerThreads ; ++i)
		pThis->pWrkThrds[i].tCurrCmd = tShutdownCmd;

	/* awake them... */
	pthread_cond_broadcast(pThis->notEmpty);
	
	/* and wait for their termination */
	for(i = 0 ; i < pThis->iNumWorkerThreads ; ++i) {
		pthread_join(pThis->pWrkThrds[i].thrdID, NULL);
	}

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

	return iRet;
}


/* Each queue has 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.
 *
 * There are two classes of worker threads, all implemented via the function
 * below. The queue may start multiple workers. The first one carries out normal
 * processing BUT also manages the other workers (the first one and all other
 * ones are the two different classes). This is so that the queue can dynamically
 * start and stop worker threads. So far, this dynamic mode is not yet supported,
 * but we will need it at least for disk-assisted queue types. There, multiple
 * workers are supported as long as the queue is running in memory, but only
 * a single worker is supported if running in disk mode. To start and stop
 * workers, we need to have one thread that is capable to wait. We could start
 * up a specific management thread. However, this means additional overhead. So
 * we have decided to use worker #0, which is always present, to carry out this
 * management as an additional chore. -- rgerhards, 2008-01-10
 */
static void *
queueWorker(void *arg)
{
	DEFiRet;
	queue_t *pThis = (queue_t*) arg;
	void *pUsr;
	sigset_t sigSet;
	int iMyThrdIndx;	/* index for this thread in queue thread table */

	assert(pThis != NULL);

	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", (unsigned long) pThis, iMyThrdIndx);

	/* now we have our identity, on to real processing */
	while(pThis->pWrkThrds[iMyThrdIndx].tCurrCmd == eWRKTHRDCMD_RUN
	      || (pThis->pWrkThrds[iMyThrdIndx].tCurrCmd == eWRKTHRDCMD_SHUTDOWN && pThis->iQueueSize > 0)) {
		pthread_mutex_lock(pThis->mut);
		while (pThis->iQueueSize == 0 && pThis->pWrkThrds[iMyThrdIndx].tCurrCmd == eWRKTHRDCMD_RUN) {
			dbgprintf("Queue 0x%lx/w%d: queue EMPTY, waiting for next message.\n",
				  (unsigned long) pThis, iMyThrdIndx);
			pthread_cond_wait (pThis->notEmpty, pThis->mut);
		}
		if(pThis->iQueueSize > 0) {
			/* dequeue element (still protected from mutex) */
			iRet = queueDel(pThis, &pUsr);
			pthread_mutex_unlock(pThis->mut);
			pthread_cond_signal (pThis->notFull);
			/* 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. Howerver, this is really not a good thing.
			 * rgerhards, 2008-01-03
			 */
			if(iRet == RS_RET_OK) {
				rsRetVal iRetLocal;
				dbgprintf("Queue 0x%lx/w%d: worker executes consumer...\n",
				           (unsigned long) pThis, iMyThrdIndx);
				iRetLocal = pThis->pConsumer(pUsr);
				dbgprintf("Queue 0x%lx/w%d: worker: consumer returnd %d\n",
				           (unsigned long) pThis, iMyThrdIndx, iRetLocal);
				if(iRetLocal != RS_RET_OK)
					dbgprintf("Queue 0x%lx/w%d: Consumer returned iRet %d\n",
					          (unsigned long) pThis, iMyThrdIndx, iRetLocal);
			} else {
				dbgprintf("Queue 0x%lx/w%d: error %d dequeueing element - ignoring, but strange things "
				          "may happen\n", (unsigned long) pThis, iMyThrdIndx, iRet);
			}
		} else { /* the mutex must be unlocked in any case (important for termination) */
			pthread_mutex_unlock(pThis->mut);
		}

		/* 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();
		
		if(Debug && (pThis->pWrkThrds[iMyThrdIndx].tCurrCmd == eWRKTHRDCMD_SHUTDOWN) && pThis->iQueueSize > 0)
			dbgprintf("Queue 0x%lx/w%d: worker does not yet terminate because it still has "
				  " %d messages to process.\n", (unsigned long) pThis, iMyThrdIndx, pThis->iQueueSize);
	}

	dbgprintf("Queue 0x%lx/w%d: worker thread terminates with %d entries left in queue.\n",
		  (unsigned long) pThis, iMyThrdIndx, pThis->iQueueSize);
	pThis->pWrkThrds[iMyThrdIndx].tCurrCmd = eWRKTHRDCMD_TERMINATED; /* indicate termination */
	pthread_exit(0);
}


/* This method checks if there is any persistent information on the
 * queue and, if so, tries to load it. This method can only legally be
 * called from the destructor (I moved it out from there to keep the
 * Constructor code somewhat smaller). -- 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_OK);
		} 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(objDeserializeObjAsPropBag(pThis->tVars.disk.pWrite, psQIF));
	CHKiRet(objDeserializeObjAsPropBag(pThis->tVars.disk.pRead, psQIF));

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;
}


/* 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;
	}

	/* call type-specific constructor */
	CHKiRet(pThis->qConstruct(pThis));

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;
	int iState;
	int i;

	assert(pThis != NULL);

	/* and now check if there is some persistent information that needs to be read in */
	CHKiRet(queueTryLoadPersistedInfo(pThis));

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

	if(pThis->qType != QUEUETYPE_DIRECT) {
		if((pThis->pWrkThrds = calloc(pThis->iNumWorkerThreads, sizeof(qWrkThrd_t))) == NULL)
			ABORT_FINALIZE(RS_RET_OUT_OF_MEMORY);

		/* fire up the worker threads */
		for(i = 0 ; i < pThis->iNumWorkerThreads ; ++i) {
			pThis->pWrkThrds[i].tCurrCmd = eWRKTHRDCMD_RUN;
			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);
		}
	}

finalize_it:
	return iRet;
}


#if 0
/* persist disk status on disk. This is necessary if we run either
 * a disk queue or in a disk assisted mode.
 */
static rsRetVal queuePersistDskFilInfo(queue_t *pThis)
{
	DEFiRet;

	assert(pThis != NULL);


finalize_it:
	return iRet;
}
#endif



/* 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; /* Queue Info File */
	uchar pszQIFNam[MAXFNAME];
	size_t lenQIFNam;
	int i;

	assert(pThis != NULL);
	if(pThis->iQueueSize == 0)
		FINALIZE; /* nothing left to do, so be happy */

	if(pThis->qType != QUEUETYPE_DISK)
		ABORT_FINALIZE(RS_RET_NOT_IMPLEMENTED); /* TODO: later... */

	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);
	CHKiRet(strmConstruct(&psQIF));
	CHKiRet(strmSetDir(psQIF, glblGetWorkDir(), strlen((char*)glblGetWorkDir())));
	CHKiRet(strmSettOperationsMode(psQIF, STREAMMODE_WRITE));
	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));
	i = pThis->qType;
	objSerializeSCALAR_VAR(psQIF, qType, INT, i);
	objSerializeSCALAR(psQIF, iQueueSize, INT);
	CHKiRet(objEndSerialize(psQIF));
dbgprintf("queue serial 1\n");

	/* 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));
dbgprintf("queue serial 2\n");
	
	/* persist queue object itself */

	/* ready with the queue, now call driver to persist queue data */
	//iRet = ;

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

finalize_it:
	strmDestruct(psQIF);
	return iRet;
}


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

	assert(pThis != NULL);

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

	/* now check if we need to persist the queue */
	if(pThis->bImmediateShutdown) {
		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((pThis->pszFilePrefix = malloc(sizeof(uchar) * iLenPrefix + 1)) == NULL)
		ABORT_FINALIZE(RS_RET_OUT_OF_MEMORY);
	memcpy(pThis->pszFilePrefix, pszPrefix, iLenPrefix + 1);
	pThis->lenFilePrefix = iLenPrefix;
	
	if(pThis->qType == QUEUETYPE_DISK) {
		CHKiRet(strmSetFName(pThis->tVars.disk.pWrite, pszPrefix, iLenPrefix));
		CHKiRet(strmSetFName(pThis->tVars.disk.pRead,  pszPrefix, 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;

	assert(pThis != NULL);
	
	if(iMaxFileSize < 1024) {
		ABORT_FINALIZE(RS_RET_VALUE_TOO_LOW);
	}

	pThis->iMaxFileSize = iMaxFileSize;
	if(pThis->qType == QUEUETYPE_DISK)
		CHKiRet(strmSetiMaxFileSize(pThis->tVars.disk.pWrite, 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;

	assert(pThis != NULL);

	/* 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
	 */
	pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &iCancelStateSave);
	if(pThis->pWrkThrds != NULL)
		pthread_mutex_lock(pThis->mut);
		
	while(pThis->iQueueSize >= pThis->iMaxQueueSize) {
		dbgprintf("Queue 0x%lx: enqueueMsg: queue FULL - waiting to drain.\n", (unsigned long) pThis);
		
		clock_gettime (CLOCK_REALTIME, &t);
		t.tv_sec += 2; /* TODO: configurable! */
		
		if(pthread_cond_timedwait (pThis->notFull,
					   pThis->mut, &t) != 0) {
			dbgprintf("Queue 0x%lx: enqueueMsg: cond timeout, dropping message!\n", (unsigned long) pThis);
			objDestruct(pUsr);
			ABORT_FINALIZE(RS_RET_QUEUE_FULL);
		}
	}
	CHKiRet(queueAdd(pThis, pUsr));

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, bImmediateShutdown, int);


/* 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:
 */