aboutsummaryrefslogtreecommitdiff
path: root/src/im-sandbox.c
blob: a1fab4973e06175f1b45d9db07a5ed2d27d29e9f (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
/*
 * im-sandbox.c
 *
 * Sandbox for indexing
 *
 * Copyright © 2012-2021 Deutsches Elektronen-Synchrotron DESY,
 *                       a research centre of the Helmholtz Association.
 * Copyright © 2012 Richard Kirian
 * Copyright © 2012 Lorenzo Galli
 *
 * Authors:
 *   2010-2020 Thomas White <taw@physics.org>
 *   2014      Valerio Mariani
 *   2011      Richard Kirian
 *   2012      Lorenzo Galli
 *   2012      Chunhong Yoon
 *   2017      Stijn de Graaf
 *
 * This file is part of CrystFEL.
 *
 * CrystFEL 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.
 *
 * CrystFEL 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 CrystFEL.  If not, see <http://www.gnu.org/licenses/>.
 *
 */

#ifdef HAVE_CONFIG_H
#include <config.h>
#endif

#include <stdarg.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <unistd.h>
#include <getopt.h>
#include <pthread.h>
#include <sys/wait.h>
#include <fcntl.h>
#include <signal.h>
#include <sys/stat.h>
#include <assert.h>
#include <sys/mman.h>
#include <semaphore.h>

#ifdef HAVE_CLOCK_GETTIME
#include <time.h>
#else
#include <sys/time.h>
#endif

#include "im-sandbox.h"
#include "process_image.h"
#include "time-accounts.h"
#include "im-zmq.h"


struct sandbox
{
	int n_processed_last_stats;
	time_t t_last_stats;

	/* Processing timeout in seconds.  After this long without responding
	 * to a ping, the worker will be killed.  After 3 times this long
	 * working on one image, even with ping responses, a warning will be
	 * shown to the user. */
	int timeout;

	struct index_args *iargs;

	/* Worker processes */
	int n_proc;
	pid_t *pids;
	int *running;
	time_t *last_response;
	int last_ping[MAX_NUM_WORKERS];
	int profile;  /* Whether to do wall-clock time profiling */

	/* Streams to read from (NB not the same indices as the above) */
	int n_read;
	FILE **fhs;
	int *fds;

	int serial;

	struct sb_shm *shared;
	sem_t *queue_sem;

	const char *tmpdir;

	/* ZMQ mode */
	int zmq;
	const char *zmq_address;
	char **zmq_subscriptions;
	int n_zmq_subscriptions;
	const char *zmq_request;

	/* Final output */
	Stream *stream;
};

struct get_pattern_ctx
{
	FILE *fh;
	int use_basename;
	const DataTemplate *dtempl;
	const char *prefix;
	char *filename;
	char **events;
	int n_events;
	int event_index;
};


#ifdef HAVE_CLOCK_GETTIME

static time_t get_monotonic_seconds()
{
	struct timespec tp;
	clock_gettime(CLOCK_MONOTONIC, &tp);
	return tp.tv_sec;
}

#else

/* Fallback version of the above.  The time according to gettimeofday() is not
 * monotonic, so measuring intervals based on it will screw up if there's a
 * timezone change (e.g. daylight savings) while the program is running. */
static time_t get_monotonic_seconds()
{
	struct timeval tp;
	gettimeofday(&tp, NULL);
	return tp.tv_sec;
}

#endif


static void stamp_response(struct sandbox *sb, int n)
{
	sb->last_response[n] = get_monotonic_seconds();
	sb->last_ping[n] = sb->shared->pings[n];
}


static void check_hung_workers(struct sandbox *sb)
{
	int i;
	time_t tnow = get_monotonic_seconds();
	for ( i=0; i<sb->n_proc; i++ ) {

		if ( !sb->running[i] ) continue;

		if ( sb->shared->pings[i] != sb->last_ping[i] ) {
			stamp_response(sb, i);
		}

		if ( tnow - sb->last_response[i] > sb->timeout ) {
			STATUS("Worker %i did not respond for %i seconds - "
			       "sending it SIGKILL.\n", i, sb->timeout);
			kill(sb->pids[i], SIGKILL);
			stamp_response(sb, i);
		}

		if ( tnow - sb->shared->time_last_start[i] > sb->timeout*3 ) {
			if ( !sb->shared->warned_long_running[i] ) {
				STATUS("Worker %i has been working on one "
				       "frame for more than %i seconds (just "
				       "for info).\n", i, sb->timeout);
				STATUS("Event ID is: %s\n",
				       sb->shared->last_ev[i]);
				STATUS("Task ID is: %s\n",
				       sb->shared->last_task[i]);
				sb->shared->warned_long_running[i] = 1;
			}
		}

	}
}


static char *read_prefixed_filename(struct get_pattern_ctx *gpctx,
                                    char **event)
{
	char* line;

	*event = NULL;

	line = malloc(1024);
	if ( line == NULL ) return NULL;

	do {
		if ( fgets(line, 1023, gpctx->fh) == NULL )
		{
			if ( !feof(gpctx->fh) ) {
				ERROR("Input file read error.\n");
			}
			free(line);
			return NULL;
		}
		chomp(line);

	} while ( line[0] == '\0' );

	/* Chop off event ID */
	size_t n = strlen(line);
	while ( line[n] != ' ' && n > 2 ) n--;
	if ( n != 2 ) {
		/* Event descriptor must contain "//".
		 * If it doesn't, assume the filename just contains a
		 * space. */
		if ( strstr(&line[n], "//") != NULL ) {
			line[n] = '\0';
			*event = strdup(&line[n+1]);
		}
	} /* else no spaces at all */

	if ( gpctx->use_basename ) {
		char *tmp;
		tmp = safe_basename(line);
		free(line);
		line = tmp;
	}

	/* Add prefix */
	if ( gpctx->prefix != NULL ) {
		char *tmp;
		size_t len = strlen(line) + strlen(gpctx->prefix) + 1;
		tmp = malloc(len);
		if ( tmp == NULL ) {
			ERROR("Couldn't allocate memory for filename\n");
			return NULL;
		}
		strcpy(tmp, gpctx->prefix);
		strcat(tmp, line);
		free(line);
		line = tmp;
	}

	return line;
}


/* Return 0 for "no more" */
static int get_pattern(struct get_pattern_ctx *gpctx,
                       char **pfilename, char **pevent)
{
	char *filename;
	char *evstr;

	/* Is an event available already? */
	if ( (gpctx->events != NULL)
	  && (gpctx->event_index < gpctx->n_events) )
	{
		*pfilename = gpctx->filename;
		*pevent = gpctx->events[gpctx->event_index++];
		return 1;
	}

	do {

		/* No events in list.  Time to top it up */
		filename = read_prefixed_filename(gpctx, &evstr);

		/* Nothing left in file -> we're done */
		if ( filename == NULL ) return 0;

		/* Does the line from the input file contain an event ID?
		 * If so, just send it straight back. */
		if ( evstr != NULL ) {
			*pfilename = filename;
			*pevent = evstr;
			return 1;
		}

		/* We got a filename, but no event.  Attempt to expand... */
		free(gpctx->events);  /* Free the old list.
		                       * NB The actual strings were freed
		                       * by fill_queue */
		gpctx->events = image_expand_frames(gpctx->dtempl, filename,
		                                    &gpctx->n_events);
		if ( gpctx->events == NULL ) {
			ERROR("Failed to get event list from %s.\n",
			      filename);
		}

	} while ( gpctx->events == NULL );

	/* Save filename for next time */
	free(gpctx->filename);
	gpctx->filename = filename;

	gpctx->event_index = 0;
	*pfilename = gpctx->filename;
	*pevent = gpctx->events[gpctx->event_index++];
	return 1;
}


static void shuffle_events(struct sb_shm *sb_shared)
{
	int i;

	for ( i=1; i<sb_shared->n_events; i++ ) {
		memcpy(sb_shared->queue[i-1], sb_shared->queue[i], MAX_EV_LEN);
	}
	sb_shared->n_events--;
}


void set_last_task(char *lt, const char *task)
{
	if ( lt == NULL ) return;
	assert(strlen(task) < MAX_TASK_LEN-1);
	strcpy(lt, task);
}


static int run_work(const struct index_args *iargs, Stream *st,
                    int cookie, const char *tmpdir, struct sandbox *sb)
{
	int allDone = 0;
	TimeAccounts *taccs;
	struct im_zmq *zmqstuff = NULL;

	/* Connect via ZMQ */
	if ( sb->zmq ) {
		zmqstuff = im_zmq_connect(sb->zmq_address,
		                          sb->zmq_subscriptions,
		                          sb->n_zmq_subscriptions,
		                          sb->zmq_request);
		if ( zmqstuff == NULL ) {
			ERROR("ZMQ setup failed.\n");
			return 1;
		}
	}

	taccs = time_accounts_init();

	while ( !allDone ) {

		struct pattern_args pargs;
		int ser;
		char *line;
		size_t len;
		int i;
		char *event_str = NULL;
		char *ser_str = NULL;
		int ok = 1;

		/* Wait until an event is ready */
		time_accounts_set(taccs, TACC_EVENTWAIT);
		set_last_task(sb->shared->last_task[cookie], "wait_event");
		if ( sem_wait(sb->queue_sem) != 0 ) {
			ERROR("Failed to wait on queue semaphore: %s\n",
			      strerror(errno));
		}

		/* Get the event from the queue */
		set_last_task(sb->shared->last_task[cookie], "read_queue");
		pthread_mutex_lock(&sb->shared->queue_lock);
		if ( ((sb->shared->n_events==0) && (sb->shared->no_more))
		   || (sb->shared->should_shutdown) )
		{
			/* Queue is empty and no more are coming,
			 * or another process has initiated a shutdown.
			 * Either way, it's time to get out of here. */
			pthread_mutex_unlock(&sb->shared->queue_lock);
			allDone = 1;
			continue;
		}
		if ( sb->shared->n_events == 0 ) {
			ERROR("Got the semaphore, but no events in queue!\n");
			ERROR("no_more = %i\n", sb->shared->no_more);
			pthread_mutex_unlock(&sb->shared->queue_lock);
			allDone = 1;
			continue;
		}

		line = strdup(sb->shared->queue[0]);

		len = strlen(line);
		assert(len > 1);
		for ( i=len-1; i>0; i-- ) {
			if ( line[i] == ' ' ) {
				line[i] = '\0';
				ser_str = &line[i+1];
				break;
			}
		}
		len = strlen(line);
		assert(len > 1);
		for ( i=len-1; i>0; i-- ) {
			if ( line[i] == ' ' ) {
				line[i] = '\0';
				event_str = &line[i+1];
				break;
			}
		}
		if ( (ser_str != NULL) && (event_str != NULL) ) {
			if ( sscanf(ser_str, "%i", &ser) != 1 ) {
				STATUS("Invalid serial number '%s'\n",
				       ser_str);
				ok = 0;
			}
		}
		if ( !ok ) {
			STATUS("Invalid event string '%s'\n",
			       sb->shared->queue[0]);
			ok = 0;
		}
		memcpy(sb->shared->last_ev[cookie], sb->shared->queue[0],
		       MAX_EV_LEN);
		shuffle_events(sb->shared);
		pthread_mutex_unlock(&sb->shared->queue_lock);

		if ( !ok ) continue;

		pargs.filename = strdup(line);
		pargs.event = strdup(event_str);

		free(line);

		if ( !sb->zmq ) {

			pargs.zmq_data = NULL;
			pargs.zmq_data_size = 0;

		} else {

			do {
				pargs.zmq_data = im_zmq_fetch(zmqstuff,
				                              &pargs.zmq_data_size);
			} while ( pargs.zmq_data_size < 15 );

			/* The filename/event, which will be 'fake' values in
			 * this case, still came via the event queue.  More
			 * importantly, the event queue gave us a unique
			 * serial number for this image. */

		}

		sb->shared->time_last_start[cookie] = get_monotonic_seconds();
		process_image(iargs, &pargs, st, cookie, tmpdir, ser,
		              sb->shared, taccs, sb->shared->last_task[cookie]);

		/* pargs.zmq_data will be copied into the image structure, so
		 * that it can be queried for "header" values etc.  It will
		 * eventually be freed by image_free() under process_image() */
	}

	im_zmq_shutdown(zmqstuff);

	time_accounts_set(taccs, TACC_FINALCLEANUP);
	cleanup_indexing(iargs->ipriv);
	cell_free(iargs->cell);
	if ( sb->profile ) time_accounts_print(taccs);
	time_accounts_free(taccs);
	return 0;
}


static ssize_t lwrite(int fd, const char *a)
{
	size_t l = strlen(a);
	return write(fd, a, l);
}


static int pump_chunk(FILE *fh, int ofd)
{
	int chunk_started = 0;

	do {

		char line[1024];
		char *rval;

		rval = fgets(line, 1024, fh);
		if ( rval == NULL ) {

			if ( feof(fh) ) {
				/* Whoops, connection lost */
				if ( chunk_started ) {
					ERROR("EOF during chunk!\n");
					lwrite(ofd, "Unfinished chunk!\n");
					lwrite(ofd, STREAM_CHUNK_END_MARKER"\n");
				} /* else normal end of output */
				return 1;
			}

			ERROR("fgets() failed: %s\n", strerror(errno));
			if ( errno != EINTR ) return 1;

		}

		if ( strcmp(line, "FLUSH\n") == 0 ) break;
		lwrite(ofd, line);

		if ( strcmp(line, STREAM_CHUNK_END_MARKER"\n") == 0 ) break;

	} while ( 1 );
	return 0;
}


/* Add an fd to the list of pipes to be read from */
static void add_pipe(struct sandbox *sb, int fd)
{
	int *fds_new;
	FILE **fhs_new;
	int slot;

	fds_new = realloc(sb->fds, (sb->n_read+1)*sizeof(int));
	if ( fds_new == NULL ) {
		ERROR("Failed to allocate memory for new pipe.\n");
		return;
	}

	fhs_new = realloc(sb->fhs, (sb->n_read+1)*sizeof(FILE *));
	if ( fhs_new == NULL ) {
		ERROR("Failed to allocate memory for new FH.\n");
		free(fds_new);
		return;
	}

	sb->fds = fds_new;
	sb->fhs = fhs_new;
	slot = sb->n_read;

	sb->fds[slot] = fd;

	sb->fhs[slot] = fdopen(fd, "r");
	if ( sb->fhs[slot] == NULL ) {
		ERROR("Couldn't fdopen() stream!\n");
		return;
	}

	sb->n_read++;
}


static void remove_pipe(struct sandbox *sb, int d)
{
	int i;

	fclose(sb->fhs[d]);

	for ( i=d; i<sb->n_read; i++ ) {
		if ( i < sb->n_read-1 ) {
			sb->fds[i] = sb->fds[i+1];
			sb->fhs[i] = sb->fhs[i+1];
		} /* else don't bother */
	}

	sb->n_read--;

	/* We don't bother shrinking the arrays */
}


static void try_read(struct sandbox *sb, TimeAccounts *taccs)
{
	int r, i;
	struct timeval tv;
	fd_set fds;
	int fdmax;
	const int ofd = stream_get_fd(sb->stream);

	time_accounts_set(taccs, TACC_SELECT);

	tv.tv_sec = 0;
	tv.tv_usec = 500000;

	FD_ZERO(&fds);
	fdmax = 0;
	for ( i=0; i<sb->n_read; i++ ) {

		int fd;

		fd = sb->fds[i];

		FD_SET(fd, &fds);
		if ( fd > fdmax ) fdmax = fd;

	}

	r = select(fdmax+1, &fds, NULL, NULL, &tv);

	if ( r == -1 ) {
		if ( errno != EINTR ) {
			ERROR("select() failed: %s\n", strerror(errno));
		} /* Otherwise no big deal */
		return;
	}

	for ( i=0; i<sb->n_read; i++ ) {

		if ( !FD_ISSET(sb->fds[i], &fds) ) {
			continue;
		}

		/* If the chunk cannot be read, assume the connection
		 * is broken and that the process will die soon. */
		time_accounts_set(taccs, TACC_STREAMREAD);
		if ( pump_chunk(sb->fhs[i], ofd) ) {
			remove_pipe(sb, i);
		}

	}
}


static void start_worker_process(struct sandbox *sb, int slot)
{
	pid_t p;
	int stream_pipe[2];

	if ( pipe(stream_pipe) == - 1 ) {
		ERROR("pipe() failed!\n");
		return;
	}

	sb->shared->pings[slot] = 0;
	sb->last_ping[slot] = 0;
	sb->shared->time_last_start[slot] = get_monotonic_seconds();
	sb->shared->warned_long_running[slot] = 0;

	p = fork();
	if ( p == -1 ) {
		ERROR("fork() failed!\n");
		return;
	}

	if ( p == 0 ) {

		Stream *st;
		struct sigaction sa;
		int r;
		char *tmp;
		struct stat s;
		size_t ll;
		int i;

	        /* First, disconnect the signal handlers */
	        sa.sa_flags = 0;
	        sigemptyset(&sa.sa_mask);
	        sa.sa_handler = SIG_DFL;
	        r = sigaction(SIGCHLD, &sa, NULL);
	        if ( r == -1 ) {
			ERROR("Failed to set signal handler!\n");
			exit(1);
	        }
	        r = sigaction(SIGINT, &sa, NULL);
	        if ( r == -1 ) {
			ERROR("Failed to set signal handler!\n");
			exit(1);
	        }
	        r = sigaction(SIGQUIT, &sa, NULL);
	        if ( r == -1 ) {
			ERROR("Failed to set signal handler!\n");
			exit(1);
	        }

		ll = 64 + strlen(sb->tmpdir);
		tmp = malloc(ll);
		if ( tmp == NULL ) {
			ERROR("Failed to allocate temporary dir\n");
			exit(1);
		}

		snprintf(tmp, 63, "%s/worker.%i", sb->tmpdir, slot);

		if ( stat(tmp, &s) == -1 ) {
			if ( errno != ENOENT ) {
				ERROR("Failed to stat temporary folder.\n");
				exit(1);
			}

			r = mkdir(tmp, S_IRWXU);
			if ( r ) {
				ERROR("Failed to create temporary folder: %s\n",
				strerror(errno));
				exit(1);
			}
		}

		/* Free resources which will not be needed by worker */
		free(sb->pids);
		for ( i=0; i<sb->n_read; i++ ) {
			fclose(sb->fhs[i]);
		}
		free(sb->fhs);
		free(sb->fds);
		free(sb->running);
		/* Not freed because it's not worth passing them down just for
		 * this purpose: event list file handle,
		 *               main output stream handle
		 *               original temp dir name (without indexamajig.XX)
		 *               prefix
		 */

		st = stream_open_fd_for_write(stream_pipe[1], sb->iargs->dtempl);
		r = run_work(sb->iargs, st, slot, tmp, sb);
		stream_close(st);

		free(tmp);

		munmap(sb->shared, sizeof(struct sb_shm));

		free(sb);

		exit(r);

	}

	/* Parent process gets the 'write' end of the filename pipe
	 * and the 'read' end of the result pipe. */
	sb->pids[slot] = p;
	sb->running[slot] = 1;
	stamp_response(sb, slot);
	add_pipe(sb, stream_pipe[0]);
	close(stream_pipe[1]);
}


static void handle_zombie(struct sandbox *sb, int respawn)
{
	int i;

	for ( i=0; i<sb->n_proc; i++ ) {

		int status, p;

		if ( !sb->running[i] ) continue;

		p = waitpid(sb->pids[i], &status, WNOHANG);

		if ( p == -1 ) {
			ERROR("waitpid() failed.\n");
			continue;
		}

		if ( p == sb->pids[i] ) {

			sb->running[i] = 0;

			if ( WIFEXITED(status) ) {
				continue;
			}

			if ( WIFSIGNALED(status) ) {

				if ( (WTERMSIG(status) == SIGINT)
				  || (WTERMSIG(status) == SIGQUIT) ) continue;

				STATUS("Worker %i was killed by signal %i\n",
				       i, WTERMSIG(status));
				STATUS("Event ID was: %s\n",
				       sb->shared->last_ev[i]);
				STATUS("Task ID was: %s\n",
				       sb->shared->last_task[i]);
				if ( respawn ) start_worker_process(sb, i);
			}

		}

	}
}


static int setup_shm(struct sandbox *sb)
{
	pthread_mutexattr_t attr;

	sb->shared = mmap(NULL, sizeof(struct sb_shm), PROT_READ | PROT_WRITE,
	                  MAP_SHARED | MAP_ANON, -1, 0);

	if ( sb->shared == MAP_FAILED ) {
		ERROR("SHM setup failed: %s\n", strerror(errno));
		return 1;
	}

	if ( pthread_mutexattr_init(&attr) ) {
		ERROR("Failed to initialise mutex attr.\n");
		return 1;
	}

	if ( pthread_mutexattr_setpshared(&attr, PTHREAD_PROCESS_SHARED) ) {
		ERROR("Failed to set process shared attribute.\n");
		return 1;
	}

	if ( pthread_mutex_init(&sb->shared->term_lock, &attr) ) {
		ERROR("Terminal lock setup failed.\n");
		return 1;
	}

	if ( pthread_mutex_init(&sb->shared->queue_lock, &attr) ) {
		ERROR("Queue lock setup failed.\n");
		return 1;
	}

	if ( pthread_mutex_init(&sb->shared->totals_lock, &attr) ) {
		ERROR("Totals lock setup failed.\n");
		return 1;
	}

	pthread_mutexattr_destroy(&attr);

	return 0;
}


/* Assumes the caller is already holding queue_lock! */
static int fill_queue(struct get_pattern_ctx *gpctx, struct sandbox *sb)
{
	while ( sb->shared->n_events < QUEUE_SIZE ) {

		char *filename;
		char *evstr;

		if ( sb->zmq ) {
			/* These values will be passed down to process_image,
			 * but ignored.  The 'real' filename, which is still a
			 * 'fake' filename - only for accounting purposes - will
			 * be generated by image_read_data_block(). */
			filename = "ZMQdata";
			evstr = strdup("//");
		} else {
			if ( !get_pattern(gpctx, &filename, &evstr) ) return 1;
		}

		memset(sb->shared->queue[sb->shared->n_events], 0, MAX_EV_LEN);
		snprintf(sb->shared->queue[sb->shared->n_events++], MAX_EV_LEN,
		         "%s %s %i", filename, evstr, sb->serial++);
		sem_post(sb->queue_sem);
		free(evstr);

	}
	return 0;
}

volatile sig_atomic_t at_zombies = 0;
volatile sig_atomic_t at_interrupt = 0;

static void sigchld_handler(int sig, siginfo_t *si, void *uc_v)
{
	at_zombies = 1;
}


static void sigint_handler(int sig, siginfo_t *si, void *uc_v)
{
	at_interrupt = 1;
}


static void check_signals(struct sandbox *sb, const char *semname_q,
                          int respawn)
{
	if ( at_zombies ) {
		at_zombies = 0;
		handle_zombie(sb, respawn);
	}

	if ( at_interrupt ) {
		sem_unlink(semname_q);
		exit(0);
	}
}


static void try_status(struct sandbox *sb, int final)
{
	int r;
	int n_proc_this;
	time_t tNow;
	time_t time_this;
	const char *finalstr;
	char persec[64];

	tNow = get_monotonic_seconds();
	time_this = tNow - sb->t_last_stats;
	if ( !final && (time_this < 5) ) return;

	n_proc_this = sb->shared->n_processed - sb->n_processed_last_stats;

	r = pthread_mutex_trylock(&sb->shared->term_lock);
	if ( r ) return; /* No lock -> don't bother */

	if ( final ) {
		finalstr = "Final: ";
		persec[0] = '\0';
	} else {
		finalstr = "";
		snprintf(persec, 64, ", %.1f images/sec",
		         (double)n_proc_this/time_this);
	}
	STATUS("%s%i images processed, %i hits (%.1f%%), "
	       "%i indexable (%.1f%% of hits, %.1f%% overall), "
	       "%i crystals%s.\n",
	       finalstr, sb->shared->n_processed,
	       sb->shared->n_hits,
	       100.0 * sb->shared->n_hits / sb->shared->n_processed,
	       sb->shared->n_hadcrystals,
	       100.0 * sb->shared->n_hadcrystals / sb->shared->n_hits,
	       100.0 * sb->shared->n_hadcrystals / sb->shared->n_processed,
	       sb->shared->n_crystals, persec);

	sb->n_processed_last_stats = sb->shared->n_processed;
	sb->t_last_stats = tNow;

	pthread_mutex_unlock(&sb->shared->term_lock);
}


static void delete_temporary_folder(const char *tmpdir, int n_proc)
{
	int slot;
	size_t len;
	char *workerdir;
	char *path;

	/* List of files which it's safe to delete */
	char *files[] = {"gmon.out", "mosflm.lp", "SUMMARY", "XDS.INP",
	                 "xfel_001.img", "xfel_001.spt", "xfel.drx",
	                 "xfel.felix", "xfel.gve", "xfel.ini", "xfel.log",
	                 "IDXREF.LP", "SPOT.XDS", "xfel.newmat", "XPARM.XDS"};

	/* Number of items in the above list */
	int n_files = 15;

	if ( n_proc > 99999 ) return;  /* Paranoia */

	len = strlen(tmpdir);
	workerdir = calloc(len+32, 1);
	path = calloc(len+64, 1);
	if ( (workerdir == NULL) || (path == NULL) ) return;

	snprintf(path, 127, "%s/mosflm.lp", tmpdir);
	unlink(path);
	snprintf(path, 127, "%s/SUMMARY", tmpdir);
	unlink(path);

	for ( slot=0; slot<n_proc; slot++ ) {

		struct stat s;
		int i;

		snprintf(workerdir, 63, "%s/worker.%i", tmpdir, slot);
		if ( stat(workerdir, &s) == -1 ) continue;

		for ( i=0; i<n_files; i++ ) {
			snprintf(path, 127, "%s/%s", workerdir, files[i]);
			unlink(path);
		}

		if ( rmdir(workerdir) ) {
			ERROR("Failed to delete worker temporary folder: %s\n",
			      strerror(errno));
		}

	}

	if ( rmdir(tmpdir) ) {
		ERROR("Failed to delete temporary folder: %s\n", strerror(errno));
	}

	free(workerdir);
	free(path);
}


char *create_tempdir(const char *temp_location)
{
	char *tmpdir;
	size_t ll;
	struct stat s;

	if ( temp_location == NULL ) {
		temp_location = "";
	}

	ll = 64+strlen(temp_location);
	tmpdir = malloc(ll);
	if ( tmpdir == NULL ) {
		ERROR("Failed to allocate temporary directory name\n");
		return NULL;
	}
	snprintf(tmpdir, ll, "%s/indexamajig.%i", temp_location, getpid());

	if ( stat(tmpdir, &s) == -1 ) {

		int r;

		if ( errno != ENOENT ) {
			ERROR("Failed to stat temporary folder.\n");
			return NULL;
		}

		r = mkdir(tmpdir, S_IRWXU);
		if ( r ) {
			ERROR("Failed to create temporary folder: %s\n",
			      strerror(errno));
			return NULL;
		}

	}

	return tmpdir;
}


/* Returns the number of frames processed (not necessarily indexed).
 * If the return value is zero, something is probably wrong. */
int create_sandbox(struct index_args *iargs, int n_proc, char *prefix,
                   int config_basename, FILE *fh,
                   Stream *stream, const char *tmpdir, int serial_start,
                   const char *zmq_address, char **zmq_subscriptions,
                   int n_zmq_subscriptions, const char *zmq_request,
                   int timeout, int profile)
{
	int i;
	struct sandbox *sb;
	char semname_q[64];
	struct sigaction sa;
	int r;
	int allDone = 0;
	TimeAccounts *taccs;
	struct get_pattern_ctx gpctx;

	if ( n_proc > MAX_NUM_WORKERS ) {
		ERROR("Number of workers (%i) is too large.  Using %i\n",
		      n_proc, MAX_NUM_WORKERS);
		n_proc = MAX_NUM_WORKERS;
	}

	sb = calloc(1, sizeof(struct sandbox));
	if ( sb == NULL ) {
		ERROR("Couldn't allocate memory for sandbox.\n");
		return 0;
	}

	sb->n_processed_last_stats = 0;
	sb->t_last_stats = get_monotonic_seconds();
	sb->n_proc = n_proc;
	sb->iargs = iargs;
	sb->serial = serial_start;
	sb->tmpdir = tmpdir;
	sb->profile = profile;
	sb->timeout = timeout;
	if ( zmq_address != NULL ) {
		sb->zmq = 1;
		sb->zmq_address = zmq_address;
		sb->zmq_subscriptions = zmq_subscriptions;
		sb->n_zmq_subscriptions = n_zmq_subscriptions;
		sb->zmq_request = zmq_request;
	} else {
		sb->zmq = 0;
	}

	sb->fds = NULL;
	sb->fhs = NULL;
	sb->stream = stream;

	gpctx.fh = fh;
	gpctx.use_basename = config_basename;
	gpctx.dtempl = iargs->dtempl;
	gpctx.prefix = prefix;
	gpctx.filename = NULL;
	gpctx.events = NULL;
	gpctx.event_index = 0;

	if ( setup_shm(sb) ) {
		ERROR("Failed to set up SHM.\n");
		free(sb);
		return 0;
	}

	sb->shared->n_processed = 0;
	sb->shared->n_hits = 0;
	sb->shared->n_hadcrystals = 0;
	sb->shared->n_crystals = 0;
	sb->shared->should_shutdown = 0;

	/* Set up semaphore to control work queue */
	snprintf(semname_q, 64, "indexamajig-q%i", getpid());
	sb->queue_sem = sem_open(semname_q, O_CREAT | O_EXCL,
	                         S_IRUSR | S_IWUSR, 0);
	if ( sb->queue_sem == SEM_FAILED ) {
		ERROR("Failed to create semaphore: %s\n", strerror(errno));
		return 0;
	}

	sb->pids = calloc(n_proc, sizeof(pid_t));
	sb->running = calloc(n_proc, sizeof(int));
	sb->last_response = calloc(n_proc, sizeof(time_t));
	if ( (sb->pids == NULL) || (sb->running == NULL)
	  || (sb->last_response == NULL) )
	{
		ERROR("Couldn't allocate memory for PIDs.\n");
		return 0;
	}

	/* Fill the queue */
	pthread_mutex_lock(&sb->shared->queue_lock);
	sb->shared->n_events = 0;
	sb->shared->no_more = fill_queue(&gpctx, sb);
	pthread_mutex_unlock(&sb->shared->queue_lock);

	/* Fork the right number of times */
	for ( i=0; i<n_proc; i++ ) {
		start_worker_process(sb, i);
	}

	/* Set up signal handler to take action if any children die */
	sa.sa_flags = SA_SIGINFO | SA_NOCLDSTOP | SA_RESTART;
	sigemptyset(&sa.sa_mask);
	sa.sa_sigaction = sigchld_handler;
	r = sigaction(SIGCHLD, &sa, NULL);
	if ( r == -1 ) {
	        ERROR("Failed to set signal handler!\n");
	        return 0;
	}

	/* Set up signal handler to clean up semaphore on exit */
	sa.sa_flags = SA_SIGINFO | SA_NOCLDSTOP | SA_RESTART;
	sigemptyset(&sa.sa_mask);
	sa.sa_sigaction = sigint_handler;
	r = sigaction(SIGINT, &sa, NULL);
	if ( r == -1 ) {
	        ERROR("Failed to set signal handler!\n");
	        return 0;
	}
	r = sigaction(SIGQUIT, &sa, NULL);
	if ( r == -1 ) {
	        ERROR("Failed to set signal handler!\n");
	        return 0;
	}

	taccs = time_accounts_init();

	do {

		/* Check for stream output from workers */
		try_read(sb, taccs);

		/* Check for interrupt or zombies */
		time_accounts_set(taccs, TACC_SIGNALS);
		check_signals(sb, semname_q, 1);

		/* Check for hung workers */
		check_hung_workers(sb);

		/* Top up the queue if necessary */
		time_accounts_set(taccs, TACC_QUEUETOPUP);
		pthread_mutex_lock(&sb->shared->queue_lock);
		if ( !sb->shared->no_more && (sb->shared->n_events < QUEUE_SIZE/2) ) {
			if ( fill_queue(&gpctx, sb) ) sb->shared->no_more = 1;
		}
		pthread_mutex_unlock(&sb->shared->queue_lock);

		/* Update progress */
		time_accounts_set(taccs, TACC_STATUS);
		try_status(sb, 0);

		/* Have all the events been swallowed? */
		time_accounts_set(taccs, TACC_ENDCHECK);
		pthread_mutex_lock(&sb->shared->queue_lock);
		if ( sb->shared->no_more && (sb->shared->n_events == 0) ) allDone = 1;
		if ( sb->shared->should_shutdown ) {
			/* Worker process requested immediate shutdown */
			allDone = 1;
			sb->shared->n_events = 0;
			sb->shared->no_more = 1;
		}
		pthread_mutex_unlock(&sb->shared->queue_lock);

	} while ( !allDone );

	fclose(fh);

	/* Indicate to the workers that we are finished, and wake them up one
	 * last time */
	time_accounts_set(taccs, TACC_WAKEUP);
	STATUS("Waiting for the last patterns to be processed...\n");
	pthread_mutex_lock(&sb->shared->queue_lock);
	sb->shared->no_more = 1;
	pthread_mutex_unlock(&sb->shared->queue_lock);
	for ( i=0; i<n_proc; i++ ) {
		sem_post(sb->queue_sem);
	}
	for ( i=0; i<n_proc; i++ ) {
		int status;
		time_accounts_set(taccs, TACC_WAITPID);
		while ( waitpid(sb->pids[i], &status, WNOHANG) == 0 ) {

			time_accounts_set(taccs, TACC_STREAMREAD);
			try_read(sb, taccs);

			time_accounts_set(taccs, TACC_SIGNALS);
			check_signals(sb, semname_q, 0);

			check_hung_workers(sb);

			time_accounts_set(taccs, TACC_STATUS);
			try_status(sb, 0);

			time_accounts_set(taccs, TACC_WAITPID);
		}
		/* If this worker died and got waited by the zombie handler,
		 * waitpid() returns -1 and the loop still exits. */
	}

	if ( profile ) time_accounts_print(taccs);
	time_accounts_free(taccs);

	sem_unlink(semname_q);

	for ( i=0; i<sb->n_read; i++ ) {
		fclose(sb->fhs[i]);
	}
	free(sb->fhs);
	free(sb->fds);
	free(sb->running);
	free(sb->last_response);
	free(sb->pids);

	try_status(sb, 1);
	r = (sb->shared->n_processed == 0);
	if ( sb->shared->should_shutdown ) r = 1;

	delete_temporary_folder(sb->tmpdir, n_proc);

	munmap(sb->shared, sizeof(struct sb_shm));
	free(sb);

	return r;
}