-
Notifications
You must be signed in to change notification settings - Fork 1.3k
/
sg.c
1465 lines (1280 loc) · 35.6 KB
/
sg.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
/*
* sg engine
*
* IO engine that uses the Linux SG v3 interface to talk to SCSI devices
*
* This ioengine can operate in two modes:
* sync with block devices (/dev/sdX) or
* with character devices (/dev/sgY) with direct=1 or sync=1
* async with character devices with direct=0 and sync=0
*
* What value does queue() return for the different cases?
* queue() return value
* In sync mode:
* /dev/sdX RWT FIO_Q_COMPLETED
* /dev/sgY RWT FIO_Q_COMPLETED
* with direct=1 or sync=1
*
* In async mode:
* /dev/sgY RWT FIO_Q_QUEUED
* direct=0 and sync=0
*
* Because FIO_SYNCIO is set for this ioengine td_io_queue() will fill in
* issue_time *before* each IO is sent to queue()
*
* Where are the IO counting functions called for the different cases?
*
* In sync mode:
* /dev/sdX (commit==NULL)
* RWT
* io_u_mark_depth() called in td_io_queue()
* io_u_mark_submit/complete() called in td_io_queue()
* issue_time set in td_io_queue()
*
* /dev/sgY with direct=1 or sync=1 (commit does nothing)
* RWT
* io_u_mark_depth() called in td_io_queue()
* io_u_mark_submit/complete() called in queue()
* issue_time set in td_io_queue()
*
* In async mode:
* /dev/sgY with direct=0 and sync=0
* RW: read and write operations are submitted in queue()
* io_u_mark_depth() called in td_io_commit()
* io_u_mark_submit() called in queue()
* issue_time set in td_io_queue()
* T: trim operations are queued in queue() and submitted in commit()
* io_u_mark_depth() called in td_io_commit()
* io_u_mark_submit() called in commit()
* issue_time set in commit()
*
*/
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <errno.h>
#include <poll.h>
#include "../fio.h"
#include "../optgroup.h"
#ifdef FIO_HAVE_SGIO
#ifndef SGV4_FLAG_HIPRI
#define SGV4_FLAG_HIPRI 0x800
#endif
enum {
FIO_SG_WRITE = 1,
FIO_SG_WRITE_VERIFY,
FIO_SG_WRITE_SAME,
FIO_SG_WRITE_SAME_NDOB,
FIO_SG_WRITE_STREAM,
FIO_SG_VERIFY_BYTCHK_00,
FIO_SG_VERIFY_BYTCHK_01,
FIO_SG_VERIFY_BYTCHK_11,
};
struct sg_options {
void *pad;
unsigned int hipri;
unsigned int readfua;
unsigned int writefua;
unsigned int write_mode;
uint16_t stream_id;
};
static struct fio_option options[] = {
{
.name = "hipri",
.lname = "High Priority",
.type = FIO_OPT_STR_SET,
.off1 = offsetof(struct sg_options, hipri),
.help = "Use polled IO completions",
.category = FIO_OPT_C_ENGINE,
.group = FIO_OPT_G_SG,
},
{
.name = "readfua",
.lname = "sg engine read fua flag support",
.type = FIO_OPT_BOOL,
.off1 = offsetof(struct sg_options, readfua),
.help = "Set FUA flag (force unit access) for all Read operations",
.def = "0",
.category = FIO_OPT_C_ENGINE,
.group = FIO_OPT_G_SG,
},
{
.name = "writefua",
.lname = "sg engine write fua flag support",
.type = FIO_OPT_BOOL,
.off1 = offsetof(struct sg_options, writefua),
.help = "Set FUA flag (force unit access) for all Write operations",
.def = "0",
.category = FIO_OPT_C_ENGINE,
.group = FIO_OPT_G_SG,
},
{
.name = "sg_write_mode",
.lname = "specify sg write mode",
.type = FIO_OPT_STR,
.off1 = offsetof(struct sg_options, write_mode),
.help = "Specify SCSI WRITE mode",
.def = "write",
.posval = {
{ .ival = "write",
.oval = FIO_SG_WRITE,
.help = "Issue standard SCSI WRITE commands",
},
{ .ival = "write_and_verify",
.oval = FIO_SG_WRITE_VERIFY,
.help = "Issue SCSI WRITE AND VERIFY commands",
},
{ .ival = "verify",
.oval = FIO_SG_WRITE_VERIFY,
.help = "Issue SCSI WRITE AND VERIFY commands. This "
"option is deprecated. Use write_and_verify instead.",
},
{ .ival = "write_same",
.oval = FIO_SG_WRITE_SAME,
.help = "Issue SCSI WRITE SAME commands",
},
{ .ival = "same",
.oval = FIO_SG_WRITE_SAME,
.help = "Issue SCSI WRITE SAME commands. This "
"option is deprecated. Use write_same instead.",
},
{ .ival = "write_same_ndob",
.oval = FIO_SG_WRITE_SAME_NDOB,
.help = "Issue SCSI WRITE SAME(16) commands with NDOB flag set",
},
{ .ival = "verify_bytchk_00",
.oval = FIO_SG_VERIFY_BYTCHK_00,
.help = "Issue SCSI VERIFY commands with BYTCHK set to 00",
},
{ .ival = "verify_bytchk_01",
.oval = FIO_SG_VERIFY_BYTCHK_01,
.help = "Issue SCSI VERIFY commands with BYTCHK set to 01",
},
{ .ival = "verify_bytchk_11",
.oval = FIO_SG_VERIFY_BYTCHK_11,
.help = "Issue SCSI VERIFY commands with BYTCHK set to 11",
},
{ .ival = "write_stream",
.oval = FIO_SG_WRITE_STREAM,
.help = "Issue SCSI WRITE STREAM(16) commands",
},
},
.category = FIO_OPT_C_ENGINE,
.group = FIO_OPT_G_SG,
},
{
.name = "stream_id",
.lname = "stream id for WRITE STREAM(16) commands",
.type = FIO_OPT_INT,
.off1 = offsetof(struct sg_options, stream_id),
.help = "Stream ID for WRITE STREAM(16) commands",
.def = "0",
.category = FIO_OPT_C_ENGINE,
.group = FIO_OPT_G_SG,
},
{
.name = NULL,
},
};
#define MAX_10B_LBA 0xFFFFFFFFULL
#define SCSI_TIMEOUT_MS 30000 // 30 second timeout; currently no method to override
#define MAX_SB 64 // sense block maximum return size
/*
#define FIO_SGIO_DEBUG
*/
struct sgio_cmd {
unsigned char cdb[16]; // enhanced from 10 to support 16 byte commands
unsigned char sb[MAX_SB]; // add sense block to commands
int nr;
};
struct sgio_trim {
uint8_t *unmap_param;
unsigned int unmap_range_count;
struct io_u **trim_io_us;
};
struct sgio_data {
struct sgio_cmd *cmds;
struct io_u **events;
struct pollfd *pfds;
int *fd_flags;
void *sgbuf;
unsigned int bs;
int type_checked;
struct sgio_trim **trim_queues;
int current_queue;
#ifdef FIO_SGIO_DEBUG
unsigned int *trim_queue_map;
#endif
};
static inline uint16_t sgio_get_be16(uint8_t *buf)
{
return be16_to_cpu(*((uint16_t *) buf));
}
static inline uint32_t sgio_get_be32(uint8_t *buf)
{
return be32_to_cpu(*((uint32_t *) buf));
}
static inline uint64_t sgio_get_be64(uint8_t *buf)
{
return be64_to_cpu(*((uint64_t *) buf));
}
static inline void sgio_set_be16(uint16_t val, uint8_t *buf)
{
uint16_t t = cpu_to_be16(val);
memcpy(buf, &t, sizeof(uint16_t));
}
static inline void sgio_set_be32(uint32_t val, uint8_t *buf)
{
uint32_t t = cpu_to_be32(val);
memcpy(buf, &t, sizeof(uint32_t));
}
static inline void sgio_set_be64(uint64_t val, uint8_t *buf)
{
uint64_t t = cpu_to_be64(val);
memcpy(buf, &t, sizeof(uint64_t));
}
static inline bool sgio_unbuffered(struct thread_data *td)
{
return (td->o.odirect || td->o.sync_io);
}
static void sgio_hdr_init(struct sgio_data *sd, struct sg_io_hdr *hdr,
struct io_u *io_u, int fs)
{
struct sgio_cmd *sc = &sd->cmds[io_u->index];
memset(hdr, 0, sizeof(*hdr));
memset(sc->cdb, 0, sizeof(sc->cdb));
hdr->interface_id = 'S';
hdr->cmdp = sc->cdb;
hdr->cmd_len = sizeof(sc->cdb);
hdr->sbp = sc->sb;
hdr->mx_sb_len = sizeof(sc->sb);
hdr->pack_id = io_u->index;
hdr->usr_ptr = io_u;
hdr->timeout = SCSI_TIMEOUT_MS;
if (fs) {
hdr->dxferp = io_u->xfer_buf;
hdr->dxfer_len = io_u->xfer_buflen;
}
}
static int pollin_events(struct pollfd *pfds, int fds)
{
int i;
for (i = 0; i < fds; i++)
if (pfds[i].revents & POLLIN)
return 1;
return 0;
}
static int sg_fd_read(int fd, void *data, size_t size)
{
int err = 0;
while (size) {
ssize_t ret;
ret = read(fd, data, size);
if (ret < 0) {
if (errno == EAGAIN || errno == EINTR)
continue;
err = errno;
break;
} else if (!ret)
break;
else {
data += ret;
size -= ret;
}
}
if (err)
return err;
if (size)
return EAGAIN;
return 0;
}
static int fio_sgio_getevents(struct thread_data *td, unsigned int min,
unsigned int max,
const struct timespec fio_unused *t)
{
struct sgio_data *sd = td->io_ops_data;
int left = max, eventNum, ret, r = 0, trims = 0;
void *buf = sd->sgbuf;
unsigned int i, j, events;
struct fio_file *f;
struct io_u *io_u;
/*
* Fill in the file descriptors
*/
for_each_file(td, f, i) {
/*
* don't block for min events == 0
*/
if (!min)
sd->fd_flags[i] = fio_set_fd_nonblocking(f->fd, "sg");
else
sd->fd_flags[i] = -1;
sd->pfds[i].fd = f->fd;
sd->pfds[i].events = POLLIN;
}
/*
** There are two counters here:
** - number of SCSI commands completed
** - number of io_us completed
**
** These are the same with reads and writes, but
** could differ with trim/unmap commands because
** a single unmap can include multiple io_us
*/
while (left > 0) {
char *p;
dprint(FD_IO, "sgio_getevents: sd %p: min=%d, max=%d, left=%d\n", sd, min, max, left);
do {
if (!min)
break;
ret = poll(sd->pfds, td->o.nr_files, -1);
if (ret < 0) {
if (!r)
r = -errno;
td_verror(td, errno, "poll");
break;
} else if (!ret)
continue;
if (pollin_events(sd->pfds, td->o.nr_files))
break;
} while (1);
if (r < 0)
break;
re_read:
p = buf;
events = 0;
for_each_file(td, f, i) {
for (eventNum = 0; eventNum < left; eventNum++) {
ret = sg_fd_read(f->fd, p, sizeof(struct sg_io_hdr));
dprint(FD_IO, "sgio_getevents: sg_fd_read ret: %d\n", ret);
if (ret) {
r = -ret;
td_verror(td, r, "sg_read");
break;
}
io_u = ((struct sg_io_hdr *)p)->usr_ptr;
if (io_u->ddir == DDIR_TRIM) {
events += sd->trim_queues[io_u->index]->unmap_range_count;
eventNum += sd->trim_queues[io_u->index]->unmap_range_count - 1;
} else
events++;
p += sizeof(struct sg_io_hdr);
dprint(FD_IO, "sgio_getevents: events: %d, eventNum: %d, left: %d\n", events, eventNum, left);
}
}
if (r < 0 && !events)
break;
if (!events) {
usleep(1000);
goto re_read;
}
left -= events;
r += events;
for (i = 0; i < events; i++) {
struct sg_io_hdr *hdr = (struct sg_io_hdr *) buf + i;
sd->events[i + trims] = hdr->usr_ptr;
io_u = (struct io_u *)(hdr->usr_ptr);
if (hdr->info & SG_INFO_CHECK) {
/* record if an io error occurred, ignore resid */
memcpy(&io_u->hdr, hdr, sizeof(struct sg_io_hdr));
sd->events[i + trims]->error = EIO;
}
if (io_u->ddir == DDIR_TRIM) {
struct sgio_trim *st = sd->trim_queues[io_u->index];
#ifdef FIO_SGIO_DEBUG
assert(st->trim_io_us[0] == io_u);
assert(sd->trim_queue_map[io_u->index] == io_u->index);
dprint(FD_IO, "sgio_getevents: reaping %d io_us from trim queue %d\n", st->unmap_range_count, io_u->index);
dprint(FD_IO, "sgio_getevents: reaped io_u %d and stored in events[%d]\n", io_u->index, i+trims);
#endif
for (j = 1; j < st->unmap_range_count; j++) {
++trims;
sd->events[i + trims] = st->trim_io_us[j];
#ifdef FIO_SGIO_DEBUG
dprint(FD_IO, "sgio_getevents: reaped io_u %d and stored in events[%d]\n", st->trim_io_us[j]->index, i+trims);
assert(sd->trim_queue_map[st->trim_io_us[j]->index] == io_u->index);
#endif
if (hdr->info & SG_INFO_CHECK) {
/* record if an io error occurred, ignore resid */
memcpy(&st->trim_io_us[j]->hdr, hdr, sizeof(struct sg_io_hdr));
sd->events[i + trims]->error = EIO;
}
}
events -= st->unmap_range_count - 1;
st->unmap_range_count = 0;
}
}
}
if (!min) {
for_each_file(td, f, i) {
if (sd->fd_flags[i] == -1)
continue;
if (fcntl(f->fd, F_SETFL, sd->fd_flags[i]) < 0)
log_err("fio: sg failed to restore fcntl flags: %s\n", strerror(errno));
}
}
return r;
}
static enum fio_q_status fio_sgio_ioctl_doio(struct thread_data *td,
struct fio_file *f,
struct io_u *io_u)
{
struct sgio_data *sd = td->io_ops_data;
struct sg_io_hdr *hdr = &io_u->hdr;
int ret;
sd->events[0] = io_u;
ret = ioctl(f->fd, SG_IO, hdr);
if (ret < 0)
return ret;
/* record if an io error occurred */
if (hdr->info & SG_INFO_CHECK)
io_u->error = EIO;
return FIO_Q_COMPLETED;
}
static enum fio_q_status fio_sgio_rw_doio(struct thread_data *td,
struct fio_file *f,
struct io_u *io_u, int do_sync)
{
struct sg_io_hdr *hdr = &io_u->hdr;
int ret;
ret = write(f->fd, hdr, sizeof(*hdr));
if (ret < 0)
return ret;
if (do_sync) {
/*
* We can't just read back the first command that completes
* and assume it's the one we need, it could be any command
* that is inflight.
*/
do {
struct io_u *__io_u;
ret = read(f->fd, hdr, sizeof(*hdr));
if (ret < 0)
return ret;
__io_u = hdr->usr_ptr;
/* record if an io error occurred */
if (hdr->info & SG_INFO_CHECK)
__io_u->error = EIO;
if (__io_u == io_u)
break;
if (io_u_sync_complete(td, __io_u))
break;
} while (1);
return FIO_Q_COMPLETED;
}
return FIO_Q_QUEUED;
}
static enum fio_q_status fio_sgio_doio(struct thread_data *td,
struct io_u *io_u, int do_sync)
{
struct fio_file *f = io_u->file;
enum fio_q_status ret;
if (f->filetype == FIO_TYPE_BLOCK) {
ret = fio_sgio_ioctl_doio(td, f, io_u);
if (io_u->error)
td_verror(td, io_u->error, __func__);
} else {
ret = fio_sgio_rw_doio(td, f, io_u, do_sync);
if (io_u->error && do_sync)
td_verror(td, io_u->error, __func__);
}
return ret;
}
static void fio_sgio_rw_lba(struct sg_io_hdr *hdr, unsigned long long lba,
unsigned long long nr_blocks, bool override16)
{
if (lba < MAX_10B_LBA && !override16) {
sgio_set_be32((uint32_t) lba, &hdr->cmdp[2]);
sgio_set_be16((uint16_t) nr_blocks, &hdr->cmdp[7]);
} else {
sgio_set_be64(lba, &hdr->cmdp[2]);
sgio_set_be32((uint32_t) nr_blocks, &hdr->cmdp[10]);
}
return;
}
static int fio_sgio_prep(struct thread_data *td, struct io_u *io_u)
{
struct sg_io_hdr *hdr = &io_u->hdr;
struct sg_options *o = td->eo;
struct sgio_data *sd = td->io_ops_data;
unsigned long long nr_blocks, lba;
int offset;
if (io_u->xfer_buflen & (sd->bs - 1)) {
log_err("read/write not sector aligned\n");
return EINVAL;
}
nr_blocks = io_u->xfer_buflen / sd->bs;
lba = io_u->offset / sd->bs;
if (io_u->ddir == DDIR_READ) {
sgio_hdr_init(sd, hdr, io_u, 1);
hdr->dxfer_direction = SG_DXFER_FROM_DEV;
if (lba < MAX_10B_LBA)
hdr->cmdp[0] = 0x28; // read(10)
else
hdr->cmdp[0] = 0x88; // read(16)
if (o->hipri)
hdr->flags |= SGV4_FLAG_HIPRI;
if (o->readfua)
hdr->cmdp[1] |= 0x08;
fio_sgio_rw_lba(hdr, lba, nr_blocks, false);
} else if (io_u->ddir == DDIR_WRITE) {
sgio_hdr_init(sd, hdr, io_u, 1);
hdr->dxfer_direction = SG_DXFER_TO_DEV;
switch(o->write_mode) {
case FIO_SG_WRITE:
if (lba < MAX_10B_LBA)
hdr->cmdp[0] = 0x2a; // write(10)
else
hdr->cmdp[0] = 0x8a; // write(16)
if (o->hipri)
hdr->flags |= SGV4_FLAG_HIPRI;
if (o->writefua)
hdr->cmdp[1] |= 0x08;
break;
case FIO_SG_WRITE_VERIFY:
if (lba < MAX_10B_LBA)
hdr->cmdp[0] = 0x2e; // write and verify(10)
else
hdr->cmdp[0] = 0x8e; // write and verify(16)
break;
// BYTCHK is disabled by virtue of the memset in sgio_hdr_init
case FIO_SG_WRITE_SAME:
hdr->dxfer_len = sd->bs;
if (lba < MAX_10B_LBA)
hdr->cmdp[0] = 0x41; // write same(10)
else
hdr->cmdp[0] = 0x93; // write same(16)
break;
case FIO_SG_WRITE_SAME_NDOB:
hdr->cmdp[0] = 0x93; // write same(16)
hdr->cmdp[1] |= 0x1; // no data output buffer
hdr->dxfer_len = 0;
break;
case FIO_SG_WRITE_STREAM:
hdr->cmdp[0] = 0x9a; // write stream (16)
if (o->writefua)
hdr->cmdp[1] |= 0x08;
sgio_set_be64(lba, &hdr->cmdp[2]);
sgio_set_be16((uint16_t) io_u->file->engine_pos, &hdr->cmdp[10]);
sgio_set_be16((uint16_t) nr_blocks, &hdr->cmdp[12]);
break;
case FIO_SG_VERIFY_BYTCHK_00:
if (lba < MAX_10B_LBA)
hdr->cmdp[0] = 0x2f; // VERIFY(10)
else
hdr->cmdp[0] = 0x8f; // VERIFY(16)
hdr->dxfer_len = 0;
break;
case FIO_SG_VERIFY_BYTCHK_01:
if (lba < MAX_10B_LBA)
hdr->cmdp[0] = 0x2f; // VERIFY(10)
else
hdr->cmdp[0] = 0x8f; // VERIFY(16)
hdr->cmdp[1] |= 0x02; // BYTCHK = 01b
break;
case FIO_SG_VERIFY_BYTCHK_11:
if (lba < MAX_10B_LBA)
hdr->cmdp[0] = 0x2f; // VERIFY(10)
else
hdr->cmdp[0] = 0x8f; // VERIFY(16)
hdr->cmdp[1] |= 0x06; // BYTCHK = 11b
hdr->dxfer_len = sd->bs;
break;
};
if (o->write_mode != FIO_SG_WRITE_STREAM)
fio_sgio_rw_lba(hdr, lba, nr_blocks,
o->write_mode == FIO_SG_WRITE_SAME_NDOB);
} else if (io_u->ddir == DDIR_TRIM) {
struct sgio_trim *st;
if (sd->current_queue == -1) {
sgio_hdr_init(sd, hdr, io_u, 0);
hdr->cmd_len = 10;
hdr->dxfer_direction = SG_DXFER_TO_DEV;
hdr->cmdp[0] = 0x42; // unmap
sd->current_queue = io_u->index;
st = sd->trim_queues[sd->current_queue];
hdr->dxferp = st->unmap_param;
#ifdef FIO_SGIO_DEBUG
assert(sd->trim_queues[io_u->index]->unmap_range_count == 0);
dprint(FD_IO, "sg: creating new queue based on io_u %d\n", io_u->index);
#endif
}
else
st = sd->trim_queues[sd->current_queue];
dprint(FD_IO, "sg: adding io_u %d to trim queue %d\n", io_u->index, sd->current_queue);
st->trim_io_us[st->unmap_range_count] = io_u;
#ifdef FIO_SGIO_DEBUG
sd->trim_queue_map[io_u->index] = sd->current_queue;
#endif
offset = 8 + 16 * st->unmap_range_count;
sgio_set_be64(lba, &st->unmap_param[offset]);
sgio_set_be32((uint32_t) nr_blocks, &st->unmap_param[offset + 8]);
st->unmap_range_count++;
} else if (ddir_sync(io_u->ddir)) {
sgio_hdr_init(sd, hdr, io_u, 0);
hdr->dxfer_direction = SG_DXFER_NONE;
if (lba < MAX_10B_LBA)
hdr->cmdp[0] = 0x35; // synccache(10)
else
hdr->cmdp[0] = 0x91; // synccache(16)
} else
assert(0);
return 0;
}
static void fio_sgio_unmap_setup(struct sg_io_hdr *hdr, struct sgio_trim *st)
{
uint16_t cnt = st->unmap_range_count * 16;
hdr->dxfer_len = cnt + 8;
sgio_set_be16(cnt + 8, &hdr->cmdp[7]);
sgio_set_be16(cnt + 6, st->unmap_param);
sgio_set_be16(cnt, &st->unmap_param[2]);
return;
}
static enum fio_q_status fio_sgio_queue(struct thread_data *td,
struct io_u *io_u)
{
struct sg_io_hdr *hdr = &io_u->hdr;
struct sgio_data *sd = td->io_ops_data;
int ret, do_sync = 0;
fio_ro_check(td, io_u);
if (sgio_unbuffered(td) || ddir_sync(io_u->ddir))
do_sync = 1;
if (io_u->ddir == DDIR_TRIM) {
if (do_sync || io_u->file->filetype == FIO_TYPE_BLOCK) {
struct sgio_trim *st = sd->trim_queues[sd->current_queue];
/* finish cdb setup for unmap because we are
** doing unmap commands synchronously */
#ifdef FIO_SGIO_DEBUG
assert(st->unmap_range_count == 1);
assert(io_u == st->trim_io_us[0]);
#endif
hdr = &io_u->hdr;
fio_sgio_unmap_setup(hdr, st);
st->unmap_range_count = 0;
sd->current_queue = -1;
} else
/* queue up trim ranges and submit in commit() */
return FIO_Q_QUEUED;
}
ret = fio_sgio_doio(td, io_u, do_sync);
if (ret < 0)
io_u->error = errno;
else if (hdr->status) {
io_u->resid = hdr->resid;
io_u->error = EIO;
} else if (td->io_ops->commit != NULL) {
if (do_sync && !ddir_sync(io_u->ddir)) {
io_u_mark_submit(td, 1);
io_u_mark_complete(td, 1);
} else if (io_u->ddir == DDIR_READ || io_u->ddir == DDIR_WRITE) {
io_u_mark_submit(td, 1);
io_u_queued(td, io_u);
}
}
if (io_u->error) {
td_verror(td, io_u->error, "xfer");
return FIO_Q_COMPLETED;
}
return ret;
}
static int fio_sgio_commit(struct thread_data *td)
{
struct sgio_data *sd = td->io_ops_data;
struct sgio_trim *st;
struct io_u *io_u;
struct sg_io_hdr *hdr;
struct timespec now;
unsigned int i;
int ret;
if (sd->current_queue == -1)
return 0;
st = sd->trim_queues[sd->current_queue];
io_u = st->trim_io_us[0];
hdr = &io_u->hdr;
fio_sgio_unmap_setup(hdr, st);
sd->current_queue = -1;
ret = fio_sgio_rw_doio(td, io_u->file, io_u, 0);
if (ret < 0 || hdr->status) {
int error;
if (ret < 0)
error = errno;
else {
error = EIO;
ret = -EIO;
}
for (i = 0; i < st->unmap_range_count; i++) {
st->trim_io_us[i]->error = error;
clear_io_u(td, st->trim_io_us[i]);
if (hdr->status)
st->trim_io_us[i]->resid = hdr->resid;
}
td_verror(td, error, "xfer");
return ret;
}
if (fio_fill_issue_time(td)) {
fio_gettime(&now, NULL);
for (i = 0; i < st->unmap_range_count; i++) {
memcpy(&st->trim_io_us[i]->issue_time, &now, sizeof(now));
io_u_queued(td, io_u);
}
}
io_u_mark_submit(td, st->unmap_range_count);
return 0;
}
static struct io_u *fio_sgio_event(struct thread_data *td, int event)
{
struct sgio_data *sd = td->io_ops_data;
return sd->events[event];
}
static int fio_sgio_read_capacity(struct thread_data *td, unsigned int *bs,
unsigned long long *max_lba)
{
/*
* need to do read capacity operation w/o benefit of sd or
* io_u structures, which are not initialized until later.
*/
struct sg_io_hdr hdr;
unsigned long long hlba;
unsigned int blksz = 0;
unsigned char cmd[16];
unsigned char sb[64];
unsigned char buf[32]; // read capacity return
int ret;
int fd = -1;
struct fio_file *f = td->files[0];
/* open file independent of rest of application */
fd = open(f->file_name, O_RDONLY);
if (fd < 0)
return -errno;
memset(&hdr, 0, sizeof(hdr));
memset(cmd, 0, sizeof(cmd));
memset(sb, 0, sizeof(sb));
memset(buf, 0, sizeof(buf));
/* First let's try a 10 byte read capacity. */
hdr.interface_id = 'S';
hdr.cmdp = cmd;
hdr.cmd_len = 10;
hdr.sbp = sb;
hdr.mx_sb_len = sizeof(sb);
hdr.timeout = SCSI_TIMEOUT_MS;
hdr.cmdp[0] = 0x25; // Read Capacity(10)
hdr.dxfer_direction = SG_DXFER_FROM_DEV;
hdr.dxferp = buf;
hdr.dxfer_len = sizeof(buf);
ret = ioctl(fd, SG_IO, &hdr);
if (ret < 0) {
close(fd);
return ret;
}
if (hdr.info & SG_INFO_CHECK) {
/* RCAP(10) might be unsupported by device. Force RCAP(16) */
hlba = MAX_10B_LBA;
} else {
blksz = sgio_get_be32(&buf[4]);
hlba = sgio_get_be32(buf);
}
/*
* If max lba masked by MAX_10B_LBA equals MAX_10B_LBA,
* then need to retry with 16 byte Read Capacity command.
*/
if (hlba == MAX_10B_LBA) {
hdr.cmd_len = 16;
hdr.cmdp[0] = 0x9e; // service action
hdr.cmdp[1] = 0x10; // Read Capacity(16)
sgio_set_be32(sizeof(buf), &hdr.cmdp[10]);
hdr.dxfer_direction = SG_DXFER_FROM_DEV;
hdr.dxferp = buf;
hdr.dxfer_len = sizeof(buf);
ret = ioctl(fd, SG_IO, &hdr);
if (ret < 0) {
close(fd);
return ret;
}
/* record if an io error occurred */
if (hdr.info & SG_INFO_CHECK)
td_verror(td, EIO, "fio_sgio_read_capacity");
blksz = sgio_get_be32(&buf[8]);
hlba = sgio_get_be64(buf);
}
if (blksz) {
*bs = blksz;
*max_lba = hlba;
ret = 0;
} else {
ret = EIO;
}
close(fd);
return ret;
}
static void fio_sgio_cleanup(struct thread_data *td)
{
struct sgio_data *sd = td->io_ops_data;
int i;
if (sd) {
free(sd->events);
free(sd->cmds);
free(sd->fd_flags);
free(sd->pfds);
free(sd->sgbuf);
#ifdef FIO_SGIO_DEBUG
free(sd->trim_queue_map);
#endif
for (i = 0; i < td->o.iodepth; i++) {
free(sd->trim_queues[i]->unmap_param);
free(sd->trim_queues[i]->trim_io_us);
free(sd->trim_queues[i]);
}
free(sd->trim_queues);
free(sd);
}
}
static int fio_sgio_init(struct thread_data *td)
{
struct sgio_data *sd;
struct sgio_trim *st;
struct sg_io_hdr *h3p;
int i;
sd = calloc(1, sizeof(*sd));
sd->cmds = calloc(td->o.iodepth, sizeof(struct sgio_cmd));
sd->sgbuf = calloc(td->o.iodepth, sizeof(struct sg_io_hdr));
sd->events = calloc(td->o.iodepth, sizeof(struct io_u *));
sd->pfds = calloc(td->o.nr_files, sizeof(struct pollfd));
sd->fd_flags = calloc(td->o.nr_files, sizeof(int));
sd->type_checked = 0;
sd->trim_queues = calloc(td->o.iodepth, sizeof(struct sgio_trim *));
sd->current_queue = -1;
#ifdef FIO_SGIO_DEBUG
sd->trim_queue_map = calloc(td->o.iodepth, sizeof(int));
#endif
for (i = 0, h3p = sd->sgbuf; i < td->o.iodepth; i++, ++h3p) {
sd->trim_queues[i] = calloc(1, sizeof(struct sgio_trim));
st = sd->trim_queues[i];
st->unmap_param = calloc(td->o.iodepth + 1, sizeof(char[16]));
st->unmap_range_count = 0;
st->trim_io_us = calloc(td->o.iodepth, sizeof(struct io_u *));
h3p->interface_id = 'S';
}
td->io_ops_data = sd;
/*