-
Notifications
You must be signed in to change notification settings - Fork 998
/
aio-stress.c
1375 lines (1102 loc) · 32.9 KB
/
aio-stress.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
// SPDX-License-Identifier: GPL-2.0-only
/*
* Copyright (c) 2004 SuSE, Inc. All Rights Reserved.
* Written by: Chris Mason <mason@suse.com>
* Copyright (C) 2022 SUSE LLC Andrea Cervesato <andrea.cervesato@suse.com>
*/
/*\
* [Description]
*
* Test creates a series of files and start AIO operations on them.
* AIO is done in a rotating loop: first file1.bin gets 8 requests, then
* file2.bin, then file3.bin etc. As each file finishes writing, test switches
* to reads. IO buffers are aligned in case we want to do direct IO.
*/
#define _FILE_OFFSET_BITS 64
#define _GNU_SOURCE
#include "tst_test.h"
#ifdef HAVE_LIBAIO
#include <stdio.h>
#include <errno.h>
#include <assert.h>
#include <stdlib.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <sys/time.h>
#include <sys/ipc.h>
#include <sys/shm.h>
#include <sys/mman.h>
#include <string.h>
#include <pthread.h>
#include <libaio.h>
#include "tst_safe_pthread.h"
#include "tst_safe_sysv_ipc.h"
#define IO_FREE 0
#define IO_PENDING 1
enum {
WRITE,
READ,
RWRITE,
RREAD,
LAST_STAGE,
};
#define USE_MALLOC 0
#define USE_SHM 1
#define USE_SHMFS 2
static char *str_num_files;
static char *str_max_io_submit;
static char *str_num_contexts;
static char *str_context_offset;
static char *str_file_size;
static char *str_rec_len;
static char *str_depth;
static char *str_io_iter;
static char *str_iterations;
static char *str_o_flag;
static char *str_stages;
static char *str_use_shm;
static char *str_num_threads;
static int num_files = 1;
static long long file_size = 1024 * 1024 * 1024;
static long stages;
static unsigned long page_size_mask;
static int o_flag;
static char *latency_stats;
static char *completion_latency_stats;
static int io_iter = 8;
static int iterations = 500;
static int max_io_submit;
static long long rec_len = 64 * 1024;
static int depth = 64;
static int num_threads = 1;
static int num_contexts = 1;
static long long context_offset = 2 * 1024 * 1024;
static char *no_fsync_stages;
static int use_shm;
static int shm_id;
static char *unaligned_buffer;
static char *aligned_buffer;
static int padded_reclen;
static char *verify;
static char *verify_buf;
static char *unlink_files;
/*
* latencies during io_submit are measured, these are the
* granularities for deviations
*/
#define DEVIATIONS 6
static int deviations[DEVIATIONS] = { 100, 250, 500, 1000, 5000, 10000 };
struct io_latency {
double max;
double min;
double total_io;
double total_lat;
double deviations[DEVIATIONS];
};
/* container for a series of operations to a file */
struct io_oper {
/* already open file descriptor, valid for whatever operation you want
*/
int fd;
/* starting byte of the operation */
off_t start;
/* ending byte of the operation */
off_t end;
/* size of the read/write buffer */
int reclen;
/* max number of pending requests before a wait is triggered */
int depth;
/* current number of pending requests */
int num_pending;
/* last error, zero if there were none */
int last_err;
/* total number of errors hit. */
int num_err;
/* read,write, random, etc */
int rw;
/* number of I/O that will get sent to aio */
int total_ios;
/* number of I/O we've already sent */
int started_ios;
/* last offset used in an io operation */
off_t last_offset;
/* stonewalled = 1 when we got cut off before submitting all our I/O */
int stonewalled;
/* list management */
struct io_oper *next;
struct io_oper *prev;
struct timeval start_time;
char *file_name;
};
/* a single io, and all the tracking needed for it */
struct io_unit {
/* note, iocb must go first! */
struct iocb iocb;
/* pointer to parent io operation struct */
struct io_oper *io_oper;
/* aligned buffer */
char *buf;
/* size of the aligned buffer (record size) */
int buf_size;
/* state of this io unit (free, pending, done) */
int busy;
/* result of last operation */
long res;
struct io_unit *next;
struct timeval io_start_time; /* time of io_submit */
};
struct thread_info {
io_context_t io_ctx;
pthread_t tid;
/* allocated array of io_unit structs */
struct io_unit *ios;
/* list of io units available for io */
struct io_unit *free_ious;
/* number of io units in the I/O array */
int num_global_ios;
/* number of io units in flight */
int num_global_pending;
/* preallocated array of iocb pointers, only used in run_active */
struct iocb **iocbs;
/* preallocated array of events */
struct io_event *events;
/* size of the events array */
int num_global_events;
/* latency stats for io_submit */
struct io_latency io_submit_latency;
/* list of operations still in progress, and of those finished */
struct io_oper *active_opers;
struct io_oper *finished_opers;
/* number of files this thread is doing io on */
int num_files;
/* how much io this thread did in the last stage */
double stage_mb_trans;
/* latency completion stats i/o time from io_submit until io_getevents */
struct io_latency io_completion_latency;
};
/* pthread mutexes and other globals for keeping the threads in sync */
static pthread_barrier_t worker_barrier;
static struct timeval global_stage_start_time;
static struct thread_info *global_thread_info;
/*
* return seconds between start_tv and stop_tv in double precision
*/
static double time_since(struct timeval *start_tv, struct timeval *stop_tv)
{
double sec, usec;
double ret;
sec = stop_tv->tv_sec - start_tv->tv_sec;
usec = stop_tv->tv_usec - start_tv->tv_usec;
if (sec > 0 && usec < 0) {
sec--;
usec += 1000000;
}
ret = sec + usec / (double)1000000;
if (ret < 0)
ret = 0;
return ret;
}
/*
* return seconds between start_tv and now in double precision
*/
static double time_since_now(struct timeval *start_tv)
{
struct timeval stop_time;
gettimeofday(&stop_time, NULL);
return time_since(start_tv, &stop_time);
}
/*
* Add latency info to latency struct
*/
static void calc_latency(struct timeval *start_tv, struct timeval *stop_tv,
struct io_latency *lat)
{
double delta;
int i;
delta = time_since(start_tv, stop_tv);
delta = delta * 1000;
if (delta > lat->max)
lat->max = delta;
if (!lat->min || delta < lat->min)
lat->min = delta;
lat->total_io++;
lat->total_lat += delta;
for (i = 0; i < DEVIATIONS; i++) {
if (delta < deviations[i]) {
lat->deviations[i]++;
break;
}
}
}
static void oper_list_add(struct io_oper *oper, struct io_oper **list)
{
if (!*list) {
*list = oper;
oper->prev = oper->next = oper;
return;
}
oper->prev = (*list)->prev;
oper->next = *list;
(*list)->prev->next = oper;
(*list)->prev = oper;
}
static void oper_list_del(struct io_oper *oper, struct io_oper **list)
{
if ((*list)->next == (*list)->prev && *list == (*list)->next) {
*list = NULL;
return;
}
oper->prev->next = oper->next;
oper->next->prev = oper->prev;
if (*list == oper)
*list = oper->next;
}
/* worker func to check error fields in the io unit */
static int check_finished_io(struct io_unit *io)
{
int i;
if (io->res != io->buf_size) {
struct stat s;
SAFE_FSTAT(io->io_oper->fd, &s);
/*
* If file size is large enough for the read, then this short
* read is an error.
*/
if ((io->io_oper->rw == READ || io->io_oper->rw == RREAD) &&
s.st_size > (io->iocb.u.c.offset + io->res)) {
tst_res(TINFO, "io err %lu (%s) op %d, off %llu size %d",
io->res, tst_strerrno(-io->res), io->iocb.aio_lio_opcode,
io->iocb.u.c.offset, io->buf_size);
io->io_oper->last_err = io->res;
io->io_oper->num_err++;
return -1;
}
}
if (verify && io->io_oper->rw == READ) {
if (memcmp(io->buf, verify_buf, io->io_oper->reclen)) {
tst_res(TINFO, "verify error, file %s offset %llu contents (offset:bad:good):",
io->io_oper->file_name, io->iocb.u.c.offset);
for (i = 0; i < io->io_oper->reclen; i++) {
if (io->buf[i] != verify_buf[i]) {
tst_res(TINFO, "%d:%c:%c ", i,
io->buf[i], verify_buf[i]);
}
}
}
}
return 0;
}
/* worker func to check the busy bits and get an io unit ready for use */
static int grab_iou(struct io_unit *io, struct io_oper *oper)
{
if (io->busy == IO_PENDING)
return -1;
io->busy = IO_PENDING;
io->res = 0;
io->io_oper = oper;
return 0;
}
static char *stage_name(int rw)
{
switch (rw) {
case WRITE:
return "write";
case READ:
return "read";
case RWRITE:
return "random write";
case RREAD:
return "random read";
}
return "unknown";
}
static inline double oper_mb_trans(struct io_oper *oper)
{
return ((double)oper->started_ios * (double)oper->reclen) / (double)(1024 * 1024);
}
static void print_time(struct io_oper *oper)
{
double runtime;
double tput;
double mb;
runtime = time_since_now(&oper->start_time);
mb = oper_mb_trans(oper);
tput = mb / runtime;
tst_res(TINFO, "%s on %s (%.2f MB/s) %.2f MB in %.2fs",
stage_name(oper->rw), oper->file_name, tput, mb, runtime);
}
static void print_lat(char *str, struct io_latency *lat)
{
char out[4 * 1024];
char *ptr = out;
double avg = lat->total_lat / lat->total_io;
int i;
double total_counted = 0;
tst_res(TINFO, "%s min %.2f avg %.2f max %.2f", str, lat->min, avg, lat->max);
for (i = 0; i < DEVIATIONS; i++) {
ptr += sprintf(ptr, "%.0f < %d", lat->deviations[i], deviations[i]);
total_counted += lat->deviations[i];
}
if (total_counted && lat->total_io - total_counted)
ptr += sprintf(ptr, " < %.0f", lat->total_io - total_counted);
tst_res(TINFO, "%s", out);
memset(lat, 0, sizeof(*lat));
}
static void print_latency(struct thread_info *t)
{
struct io_latency *lat = &t->io_submit_latency;
print_lat("latency", lat);
}
static void print_completion_latency(struct thread_info *t)
{
struct io_latency *lat = &t->io_completion_latency;
print_lat("completion latency", lat);
}
/*
* updates the fields in the io operation struct that belongs to this
* io unit, and make the io unit reusable again
*/
static void finish_io(struct thread_info *t, struct io_unit *io, long result,
struct timeval *tv_now)
{
struct io_oper *oper = io->io_oper;
calc_latency(&io->io_start_time, tv_now, &t->io_completion_latency);
io->res = result;
io->busy = IO_FREE;
io->next = t->free_ious;
t->free_ious = io;
oper->num_pending--;
t->num_global_pending--;
check_finished_io(io);
if (oper->num_pending == 0 &&
(oper->started_ios == oper->total_ios || oper->stonewalled)) {
print_time(oper);
}
}
static int read_some_events(struct thread_info *t)
{
struct io_unit *event_io;
struct io_event *event;
int nr;
int i;
int min_nr = io_iter;
struct timeval stop_time;
if (t->num_global_pending < io_iter)
min_nr = t->num_global_pending;
nr = io_getevents(t->io_ctx, min_nr, t->num_global_events, t->events, NULL);
if (nr <= 0)
return nr;
gettimeofday(&stop_time, NULL);
for (i = 0; i < nr; i++) {
event = t->events + i;
event_io = (struct io_unit *)((unsigned long)event->obj);
finish_io(t, event_io, event->res, &stop_time);
}
return nr;
}
/*
* finds a free io unit, waiting for pending requests if required. returns
* null if none could be found
*/
static struct io_unit *find_iou(struct thread_info *t, struct io_oper *oper)
{
struct io_unit *event_io;
int nr;
retry:
if (t->free_ious) {
event_io = t->free_ious;
t->free_ious = t->free_ious->next;
if (grab_iou(event_io, oper))
tst_brk(TBROK, "io unit on free list but not free");
return event_io;
}
nr = read_some_events(t);
if (nr > 0)
goto retry;
else
tst_res(TINFO, "no free ious after read_some_events");
return NULL;
}
/*
* wait for all pending requests for this io operation to finish
*/
static int io_oper_wait(struct thread_info *t, struct io_oper *oper)
{
struct io_event event;
struct io_unit *event_io;
if (!oper)
return 0;
if (oper->num_pending == 0)
goto done;
/* this func is not speed sensitive, no need to go wild reading
* more than one event at a time
*/
while (io_getevents(t->io_ctx, 1, 1, &event, NULL) > 0) {
struct timeval tv_now;
event_io = (struct io_unit *)((unsigned long)event.obj);
gettimeofday(&tv_now, NULL);
finish_io(t, event_io, event.res, &tv_now);
if (oper->num_pending == 0)
break;
}
done:
if (oper->num_err)
tst_res(TINFO, "%u errors on oper, last %u", oper->num_err, oper->last_err);
return 0;
}
static off_t random_byte_offset(struct io_oper *oper)
{
off_t num;
off_t rand_byte = oper->start;
off_t range;
off_t offset = 1;
range = (oper->end - oper->start) / (1024 * 1024);
if ((page_size_mask + 1) > (1024 * 1024))
offset = (page_size_mask + 1) / (1024 * 1024);
if (range < offset)
range = 0;
else
range -= offset;
/* find a random mb offset */
num = 1 + (int)((double)range * rand() / (RAND_MAX + 1.0));
rand_byte += num * 1024 * 1024;
/* find a random byte offset */
num = 1 + (int)((double)(1024 * 1024) * rand() / (RAND_MAX + 1.0));
/* page align */
num = (num + page_size_mask) & ~page_size_mask;
rand_byte += num;
if (rand_byte + oper->reclen > oper->end)
rand_byte -= oper->reclen;
return rand_byte;
}
/*
* build an aio iocb for an operation, based on oper->rw and the
* last offset used. This finds the struct io_unit that will be attached
* to the iocb, and things are ready for submission to aio after this
* is called.
*
* returns null on error
*/
static struct io_unit *build_iocb(struct thread_info *t, struct io_oper *oper)
{
struct io_unit *io;
off_t rand_byte;
io = find_iou(t, oper);
if (!io)
tst_brk(TBROK, "unable to find io unit");
switch (oper->rw) {
case WRITE:
io_prep_pwrite(&io->iocb, oper->fd, io->buf, oper->reclen, oper->last_offset);
oper->last_offset += oper->reclen;
break;
case READ:
io_prep_pread(&io->iocb, oper->fd, io->buf, oper->reclen, oper->last_offset);
oper->last_offset += oper->reclen;
break;
case RREAD:
rand_byte = random_byte_offset(oper);
oper->last_offset = rand_byte;
io_prep_pread(&io->iocb, oper->fd, io->buf, oper->reclen, rand_byte);
break;
case RWRITE:
rand_byte = random_byte_offset(oper);
oper->last_offset = rand_byte;
io_prep_pwrite(&io->iocb, oper->fd, io->buf, oper->reclen, rand_byte);
break;
}
return io;
}
/*
* wait for any pending requests, and then free all ram associated with
* an operation. returns the last error the operation hit (zero means none)
*/
static int finish_oper(struct thread_info *t, struct io_oper *oper)
{
unsigned long last_err;
io_oper_wait(t, oper);
last_err = oper->last_err;
if (oper->num_pending > 0)
tst_res(TINFO, "oper num_pending is %d", oper->num_pending);
SAFE_CLOSE(oper->fd);
free(oper);
return last_err;
}
/*
* allocates an io operation and fills in all the fields. returns
* null on error
*/
static struct io_oper *create_oper(int fd, int rw, off_t start, off_t end,
int reclen, int depth, char *file_name)
{
struct io_oper *oper;
oper = SAFE_MALLOC(sizeof(*oper));
memset(oper, 0, sizeof(*oper));
oper->depth = depth;
oper->start = start;
oper->end = end;
oper->last_offset = oper->start;
oper->fd = fd;
oper->reclen = reclen;
oper->rw = rw;
oper->total_ios = (oper->end - oper->start) / oper->reclen;
oper->file_name = file_name;
return oper;
}
/*
* does setup on num_ios worth of iocbs, but does not actually
* start any io
*/
static int build_oper(struct thread_info *t, struct io_oper *oper, int num_ios,
struct iocb **my_iocbs)
{
int i;
struct io_unit *io;
if (oper->started_ios == 0)
gettimeofday(&oper->start_time, NULL);
if (num_ios == 0)
num_ios = oper->total_ios;
if ((oper->started_ios + num_ios) > oper->total_ios)
num_ios = oper->total_ios - oper->started_ios;
for (i = 0; i < num_ios; i++) {
io = build_iocb(t, oper);
if (!io)
return -1;
my_iocbs[i] = &io->iocb;
}
return num_ios;
}
/*
* runs through the iocbs in the array provided and updates
* counters in the associated oper struct
*/
static void update_iou_counters(struct iocb **my_iocbs, int nr, struct timeval *tv_now)
{
struct io_unit *io;
int i;
for (i = 0; i < nr; i++) {
io = (struct io_unit *)(my_iocbs[i]);
io->io_oper->num_pending++;
io->io_oper->started_ios++;
io->io_start_time = *tv_now; /* set time of io_submit */
}
}
/* starts some io for a given file, returns zero if all went well */
static int run_built(struct thread_info *t, int num_ios, struct iocb **my_iocbs)
{
int ret;
struct timeval start_time;
struct timeval stop_time;
resubmit:
gettimeofday(&start_time, NULL);
ret = io_submit(t->io_ctx, num_ios, my_iocbs);
gettimeofday(&stop_time, NULL);
calc_latency(&start_time, &stop_time, &t->io_submit_latency);
if (ret != num_ios) {
/* some I/O got through */
if (ret > 0) {
update_iou_counters(my_iocbs, ret, &stop_time);
my_iocbs += ret;
t->num_global_pending += ret;
num_ios -= ret;
}
/*
* we've used all the requests allocated in aio_init, wait and
* retry
*/
if (ret > 0 || ret == -EAGAIN) {
int old_ret = ret;
ret = read_some_events(t);
if (ret <= 0)
tst_brk(TBROK, "ret was %d and now is %d", ret, old_ret);
goto resubmit;
}
tst_res(TINFO, "ret %d (%s) on io_submit", ret, tst_strerrno(-ret));
return -1;
}
update_iou_counters(my_iocbs, ret, &stop_time);
t->num_global_pending += ret;
return 0;
}
/*
* changes oper->rw to the next in a command sequence, or returns zero
* to say this operation is really, completely done for
*/
static int restart_oper(struct io_oper *oper)
{
int new_rw = 0;
if (oper->last_err)
return 0;
if (oper->rw == WRITE && (stages & (1 << READ)))
new_rw = READ;
if (oper->rw == READ && (!new_rw && stages & (1 << RWRITE)))
new_rw = RWRITE;
if (oper->rw == RWRITE && (!new_rw && stages & (1 << RREAD)))
new_rw = RREAD;
if (new_rw) {
oper->started_ios = 0;
oper->last_offset = oper->start;
oper->stonewalled = 0;
/*
* we're restarting an operation with pending requests, so the
* timing info won't be printed by finish_io. Printing it here
*/
if (oper->num_pending)
print_time(oper);
oper->rw = new_rw;
return 1;
}
return 0;
}
static int oper_runnable(struct io_oper *oper)
{
struct stat buf;
/* first context is always runnable, if started_ios > 0, no need to
* redo the calculations
*/
if (oper->started_ios || oper->start == 0)
return 1;
/* only the sequential phases force delays in starting */
if (oper->rw >= RWRITE)
return 1;
SAFE_FSTAT(oper->fd, &buf);
if (S_ISREG(buf.st_mode) && buf.st_size < oper->start)
return 0;
return 1;
}
/*
* runs through all the io operations on the active list, and starts
* a chunk of io on each. If any io operations are completely finished,
* it either switches them to the next stage or puts them on the
* finished list.
*
* this function stops after max_io_submit iocbs are sent down the
* pipe, even if it has not yet touched all the operations on the
* active list. Any operations that have finished are moved onto
* the finished_opers list.
*/
static int run_active_list(struct thread_info *t, int io_iter, int max_io_submit)
{
struct io_oper *oper;
struct io_oper *built_opers = NULL;
struct iocb **my_iocbs = t->iocbs;
int ret = 0;
int num_built = 0;
oper = t->active_opers;
while (oper) {
if (!oper_runnable(oper)) {
oper = oper->next;
if (oper == t->active_opers)
break;
continue;
}
ret = build_oper(t, oper, io_iter, my_iocbs);
if (ret >= 0) {
my_iocbs += ret;
num_built += ret;
oper_list_del(oper, &t->active_opers);
oper_list_add(oper, &built_opers);
oper = t->active_opers;
if (num_built + io_iter > max_io_submit)
break;
} else
break;
}
if (num_built) {
ret = run_built(t, num_built, t->iocbs);
if (ret < 0)
tst_brk(TBROK, "error %d on run_built", ret);
while (built_opers) {
oper = built_opers;
oper_list_del(oper, &built_opers);
oper_list_add(oper, &t->active_opers);
if (oper->started_ios == oper->total_ios) {
oper_list_del(oper, &t->active_opers);
oper_list_add(oper, &t->finished_opers);
}
}
}
return 0;
}
static void aio_setup(io_context_t *io_ctx, int n)
{
int res = io_queue_init(n, io_ctx);
if (res != 0)
tst_brk(TBROK, "io_queue_setup(%d) returned %d (%s)", n, res, tst_strerrno(-res));
}
/*
* allocate io operation and event arrays for a given thread
*/
static void setup_ious(struct thread_info *t, int num_files, int depth, int reclen, int max_io_submit)
{
int i;
size_t bytes = num_files * depth * sizeof(*t->ios);
t->ios = SAFE_MALLOC(bytes);
memset(t->ios, 0, bytes);
for (i = 0; i < depth * num_files; i++) {
t->ios[i].buf = aligned_buffer;
aligned_buffer += padded_reclen;
t->ios[i].buf_size = reclen;
if (verify)
memset(t->ios[i].buf, 'b', reclen);
else
memset(t->ios[i].buf, 0, reclen);
t->ios[i].next = t->free_ious;
t->free_ious = t->ios + i;
}
if (verify) {
verify_buf = aligned_buffer;
memset(verify_buf, 'b', reclen);
}
t->iocbs = SAFE_MALLOC(sizeof(struct iocb *) * max_io_submit);
memset(t->iocbs, 0, max_io_submit * sizeof(struct iocb *));
t->events = SAFE_MALLOC(sizeof(struct io_event) * depth * num_files);
memset(t->events, 0, num_files * sizeof(struct io_event) * depth);
t->num_global_ios = num_files * depth;
t->num_global_events = t->num_global_ios;
}
/*
* The buffers used for file data are allocated as a single big
* malloc, and then each thread and operation takes a piece and uses
* that for file data. This lets us do a large shm or bigpages alloc
* and without trying to find a special place in each thread to map the
* buffers to
*/
static int setup_shared_mem(int num_threads, int num_files, int depth, int reclen)
{
char *p = NULL;
size_t total_ram;
padded_reclen = (reclen + page_size_mask) / (page_size_mask + 1);
padded_reclen = padded_reclen * (page_size_mask + 1);
total_ram = num_files * depth * padded_reclen + num_threads;
if (verify)
total_ram += padded_reclen;
/* for aligning buffer after the allocation */
total_ram += page_size_mask;
if (use_shm == USE_MALLOC) {
p = SAFE_MALLOC(total_ram);
} else if (use_shm == USE_SHM) {
SAFE_SHMGET(IPC_PRIVATE, total_ram, IPC_CREAT | 0700);
p = SAFE_SHMAT(shm_id, (char *)0x50000000, 0);
} else if (use_shm == USE_SHMFS) {
char mmap_name[16]; /* /dev/shm/ + null + XXXXXX */
int fd;
strcpy(mmap_name, "/dev/shm/XXXXXX");
fd = mkstemp(mmap_name);
if (fd < 0)
tst_brk(TBROK, "mkstemp error");
SAFE_UNLINK(mmap_name);
SAFE_FTRUNCATE(fd, total_ram);
shm_id = fd;
p = SAFE_MMAP((char *)0x50000000, total_ram,
PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
}
unaligned_buffer = p;
p = (char *)((intptr_t)(p + page_size_mask) & ~page_size_mask);
aligned_buffer = p;
return 0;