projects / fio.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
Add job number specific dumping and fix atexit() error
[fio.git] / io_u.c
... / ...
CommitLineData
1#include <unistd.h>
2#include <fcntl.h>
3#include <string.h>
4#include <signal.h>
5#include <time.h>
6#include <assert.h>
7
8#include "fio.h"
9#include "hash.h"
10
11/*
12 * Change this define to play with the timeout handling
13 */
14#undef FIO_USE_TIMEOUT
15
16struct io_completion_data {
17 int nr; /* input */
18
19 int error; /* output */
20 unsigned long bytes_done[2]; /* output */
21 struct timeval time; /* output */
22};
23
24/*
25 * The ->file_map[] contains a map of blocks we have or have not done io
26 * to yet. Used to make sure we cover the entire range in a fair fashion.
27 */
28static int random_map_free(struct fio_file *f, const unsigned long long block)
29{
30 unsigned int idx = RAND_MAP_IDX(f, block);
31 unsigned int bit = RAND_MAP_BIT(f, block);
32
33 dprint(FD_RANDOM, "free: b=%llu, idx=%u, bit=%u\n", block, idx, bit);
34
35 return (f->file_map[idx] & (1UL << bit)) == 0;
36}
37
38/*
39 * Mark a given offset as used in the map.
40 */
41static void mark_random_map(struct thread_data *td, struct io_u *io_u)
42{
43 unsigned int min_bs = td->o.rw_min_bs;
44 struct fio_file *f = io_u->file;
45 unsigned long long block;
46 unsigned int blocks;
47 unsigned int nr_blocks;
48
49 block = (io_u->offset - f->file_offset) / (unsigned long long) min_bs;
50 blocks = 0;
51 nr_blocks = (io_u->buflen + min_bs - 1) / min_bs;
52
53 while (blocks < nr_blocks) {
54 unsigned int idx, bit;
55
56 /*
57 * If we have a mixed random workload, we may
58 * encounter blocks we already did IO to.
59 */
60 if ((td->o.ddir_nr == 1) && !random_map_free(f, block))
61 break;
62
63 idx = RAND_MAP_IDX(f, block);
64 bit = RAND_MAP_BIT(f, block);
65
66 fio_assert(td, idx < f->num_maps);
67
68 f->file_map[idx] |= (1UL << bit);
69 block++;
70 blocks++;
71 }
72
73 if ((blocks * min_bs) < io_u->buflen)
74 io_u->buflen = blocks * min_bs;
75}
76
77static inline unsigned long long last_block(struct thread_data *td,
78 struct fio_file *f,
79 enum fio_ddir ddir)
80{
81 unsigned long long max_blocks;
82
83 max_blocks = f->io_size / (unsigned long long) td->o.min_bs[ddir];
84 if (!max_blocks)
85 return 0;
86
87 return max_blocks;
88}
89
90/*
91 * Return the next free block in the map.
92 */
93static int get_next_free_block(struct thread_data *td, struct fio_file *f,
94 enum fio_ddir ddir, unsigned long long *b)
95{
96 unsigned long long min_bs = td->o.rw_min_bs;
97 int i;
98
99 i = f->last_free_lookup;
100 *b = (i * BLOCKS_PER_MAP);
101 while ((*b) * min_bs < f->real_file_size) {
102 if (f->file_map[i] != -1UL) {
103 *b += fio_ffz(f->file_map[i]);
104 if (*b > last_block(td, f, ddir))
105 break;
106 f->last_free_lookup = i;
107 return 0;
108 }
109
110 *b += BLOCKS_PER_MAP;
111 i++;
112 }
113
114 dprint(FD_IO, "failed finding a free block\n");
115 return 1;
116}
117
118static int get_next_rand_offset(struct thread_data *td, struct fio_file *f,
119 enum fio_ddir ddir, unsigned long long *b)
120{
121 unsigned long long r;
122 int loops = 5;
123
124 do {
125 r = os_random_long(&td->random_state);
126 dprint(FD_RANDOM, "off rand %llu\n", r);
127 *b = (last_block(td, f, ddir) - 1)
128 * (r / ((unsigned long long) RAND_MAX + 1.0));
129
130 /*
131 * if we are not maintaining a random map, we are done.
132 */
133 if (!file_randommap(td, f))
134 return 0;
135
136 /*
137 * calculate map offset and check if it's free
138 */
139 if (random_map_free(f, *b))
140 return 0;
141
142 dprint(FD_RANDOM, "get_next_rand_offset: offset %llu busy\n",
143 *b);
144 } while (--loops);
145
146 /*
147 * we get here, if we didn't suceed in looking up a block. generate
148 * a random start offset into the filemap, and find the first free
149 * block from there.
150 */
151 loops = 10;
152 do {
153 f->last_free_lookup = (f->num_maps - 1) * (r / (RAND_MAX+1.0));
154 if (!get_next_free_block(td, f, ddir, b))
155 return 0;
156
157 r = os_random_long(&td->random_state);
158 } while (--loops);
159
160 /*
161 * that didn't work either, try exhaustive search from the start
162 */
163 f->last_free_lookup = 0;
164 return get_next_free_block(td, f, ddir, b);
165}
166
167/*
168 * For random io, generate a random new block and see if it's used. Repeat
169 * until we find a free one. For sequential io, just return the end of
170 * the last io issued.
171 */
172static int get_next_offset(struct thread_data *td, struct io_u *io_u)
173{
174 struct fio_file *f = io_u->file;
175 unsigned long long b;
176 enum fio_ddir ddir = io_u->ddir;
177
178 if (td_random(td) && (td->o.ddir_nr && !--td->ddir_nr)) {
179 td->ddir_nr = td->o.ddir_nr;
180
181 if (get_next_rand_offset(td, f, ddir, &b))
182 return 1;
183 } else {
184 if (f->last_pos >= f->real_file_size) {
185 if (!td_random(td) ||
186 get_next_rand_offset(td, f, ddir, &b))
187 return 1;
188 } else
189 b = (f->last_pos - f->file_offset) / td->o.min_bs[ddir];
190 }
191
192 io_u->offset = b * td->o.min_bs[ddir];
193 if (io_u->offset >= f->io_size) {
194 dprint(FD_IO, "get_next_offset: offset %llu >= io_size %llu\n",
195 io_u->offset, f->io_size);
196 return 1;
197 }
198
199 io_u->offset += f->file_offset;
200 if (io_u->offset >= f->real_file_size) {
201 dprint(FD_IO, "get_next_offset: offset %llu >= size %llu\n",
202 io_u->offset, f->real_file_size);
203 return 1;
204 }
205
206 return 0;
207}
208
209static unsigned int get_next_buflen(struct thread_data *td, struct io_u *io_u)
210{
211 const int ddir = io_u->ddir;
212 unsigned int buflen;
213 long r;
214
215 if (td->o.min_bs[ddir] == td->o.max_bs[ddir])
216 buflen = td->o.min_bs[ddir];
217 else {
218 r = os_random_long(&td->bsrange_state);
219 if (!td->o.bssplit_nr) {
220 buflen = (unsigned int)
221 (1 + (double) (td->o.max_bs[ddir] - 1)
222 * r / (RAND_MAX + 1.0));
223 } else {
224 long perc = 0;
225 unsigned int i;
226
227 for (i = 0; i < td->o.bssplit_nr; i++) {
228 struct bssplit *bsp = &td->o.bssplit[i];
229
230 buflen = bsp->bs;
231 perc += bsp->perc;
232 if (r <= ((LONG_MAX / 100L) * perc))
233 break;
234 }
235 }
236 if (!td->o.bs_unaligned) {
237 buflen = (buflen + td->o.min_bs[ddir] - 1)
238 & ~(td->o.min_bs[ddir] - 1);
239 }
240 }
241
242 if (io_u->offset + buflen > io_u->file->real_file_size) {
243 dprint(FD_IO, "lower buflen %u -> %u (ddir=%d)\n", buflen,
244 td->o.min_bs[ddir], ddir);
245 buflen = td->o.min_bs[ddir];
246 }
247
248 return buflen;
249}
250
251static void set_rwmix_bytes(struct thread_data *td)
252{
253 unsigned long issues;
254 unsigned int diff;
255
256 /*
257 * we do time or byte based switch. this is needed because
258 * buffered writes may issue a lot quicker than they complete,
259 * whereas reads do not.
260 */
261 issues = td->io_issues[td->rwmix_ddir] - td->rwmix_issues;
262 diff = td->o.rwmix[td->rwmix_ddir ^ 1];
263
264 td->rwmix_issues = td->io_issues[td->rwmix_ddir]
265 + (issues * ((100 - diff)) / diff);
266}
267
268static inline enum fio_ddir get_rand_ddir(struct thread_data *td)
269{
270 unsigned int v;
271 long r;
272
273 r = os_random_long(&td->rwmix_state);
274 v = 1 + (int) (100.0 * (r / (RAND_MAX + 1.0)));
275 if (v < td->o.rwmix[DDIR_READ])
276 return DDIR_READ;
277
278 return DDIR_WRITE;
279}
280
281/*
282 * Return the data direction for the next io_u. If the job is a
283 * mixed read/write workload, check the rwmix cycle and switch if
284 * necessary.
285 */
286static enum fio_ddir get_rw_ddir(struct thread_data *td)
287{
288 if (td_rw(td)) {
289 /*
290 * Check if it's time to seed a new data direction.
291 */
292 if (td->io_issues[td->rwmix_ddir] >= td->rwmix_issues) {
293 unsigned long long max_bytes;
294 enum fio_ddir ddir;
295
296 /*
297 * Put a top limit on how many bytes we do for
298 * one data direction, to avoid overflowing the
299 * ranges too much
300 */
301 ddir = get_rand_ddir(td);
302 max_bytes = td->this_io_bytes[ddir];
303 if (max_bytes >=
304 (td->o.size * td->o.rwmix[ddir] / 100)) {
305 if (!td->rw_end_set[ddir])
306 td->rw_end_set[ddir] = 1;
307
308 ddir ^= 1;
309 }
310
311 if (ddir != td->rwmix_ddir)
312 set_rwmix_bytes(td);
313
314 td->rwmix_ddir = ddir;
315 }
316 return td->rwmix_ddir;
317 } else if (td_read(td))
318 return DDIR_READ;
319 else
320 return DDIR_WRITE;
321}
322
323static void put_file_log(struct thread_data *td, struct fio_file *f)
324{
325 int ret = put_file(td, f);
326
327 if (ret)
328 td_verror(td, ret, "file close");
329}
330
331void put_io_u(struct thread_data *td, struct io_u *io_u)
332{
333 assert((io_u->flags & IO_U_F_FREE) == 0);
334 io_u->flags |= IO_U_F_FREE;
335
336 if (io_u->file)
337 put_file_log(td, io_u->file);
338
339 io_u->file = NULL;
340 list_del(&io_u->list);
341 list_add(&io_u->list, &td->io_u_freelist);
342 td->cur_depth--;
343}
344
345void requeue_io_u(struct thread_data *td, struct io_u **io_u)
346{
347 struct io_u *__io_u = *io_u;
348
349 __io_u->flags |= IO_U_F_FREE;
350 if ((__io_u->flags & IO_U_F_FLIGHT) && (__io_u->ddir != DDIR_SYNC))
351 td->io_issues[__io_u->ddir]--;
352
353 __io_u->flags &= ~IO_U_F_FLIGHT;
354
355 list_del(&__io_u->list);
356 list_add_tail(&__io_u->list, &td->io_u_requeues);
357 td->cur_depth--;
358 *io_u = NULL;
359}
360
361static int fill_io_u(struct thread_data *td, struct io_u *io_u)
362{
363 if (td->io_ops->flags & FIO_NOIO)
364 goto out;
365
366 /*
367 * see if it's time to sync
368 */
369 if (td->o.fsync_blocks &&
370 !(td->io_issues[DDIR_WRITE] % td->o.fsync_blocks) &&
371 td->io_issues[DDIR_WRITE] && should_fsync(td)) {
372 io_u->ddir = DDIR_SYNC;
373 goto out;
374 }
375
376 io_u->ddir = get_rw_ddir(td);
377
378 /*
379 * See if it's time to switch to a new zone
380 */
381 if (td->zone_bytes >= td->o.zone_size) {
382 td->zone_bytes = 0;
383 io_u->file->last_pos += td->o.zone_skip;
384 td->io_skip_bytes += td->o.zone_skip;
385 }
386
387 /*
388 * No log, let the seq/rand engine retrieve the next buflen and
389 * position.
390 */
391 if (get_next_offset(td, io_u)) {
392 dprint(FD_IO, "io_u %p, failed getting offset\n", io_u);
393 return 1;
394 }
395
396 io_u->buflen = get_next_buflen(td, io_u);
397 if (!io_u->buflen) {
398 dprint(FD_IO, "io_u %p, failed getting buflen\n", io_u);
399 return 1;
400 }
401
402 if (io_u->offset + io_u->buflen > io_u->file->real_file_size) {
403 dprint(FD_IO, "io_u %p, offset too large\n", io_u);
404 dprint(FD_IO, " off=%llu/%lu > %llu\n", io_u->offset,
405 io_u->buflen, io_u->file->real_file_size);
406 return 1;
407 }
408
409 /*
410 * mark entry before potentially trimming io_u
411 */
412 if (td_random(td) && file_randommap(td, io_u->file))
413 mark_random_map(td, io_u);
414
415 /*
416 * If using a write iolog, store this entry.
417 */
418out:
419 dprint_io_u(io_u, "fill_io_u");
420 td->zone_bytes += io_u->buflen;
421 log_io_u(td, io_u);
422 return 0;
423}
424
425static void __io_u_mark_map(unsigned int *map, unsigned int nr)
426{
427 int index = 0;
428
429 switch (nr) {
430 default:
431 index = 6;
432 break;
433 case 33 ... 64:
434 index = 5;
435 break;
436 case 17 ... 32:
437 index = 4;
438 break;
439 case 9 ... 16:
440 index = 3;
441 break;
442 case 5 ... 8:
443 index = 2;
444 break;
445 case 1 ... 4:
446 index = 1;
447 case 0:
448 break;
449 }
450
451 map[index]++;
452}
453
454void io_u_mark_submit(struct thread_data *td, unsigned int nr)
455{
456 __io_u_mark_map(td->ts.io_u_submit, nr);
457 td->ts.total_submit++;
458}
459
460void io_u_mark_complete(struct thread_data *td, unsigned int nr)
461{
462 __io_u_mark_map(td->ts.io_u_complete, nr);
463 td->ts.total_complete++;
464}
465
466void io_u_mark_depth(struct thread_data *td, unsigned int nr)
467{
468 int index = 0;
469
470 switch (td->cur_depth) {
471 default:
472 index = 6;
473 break;
474 case 32 ... 63:
475 index = 5;
476 break;
477 case 16 ... 31:
478 index = 4;
479 break;
480 case 8 ... 15:
481 index = 3;
482 break;
483 case 4 ... 7:
484 index = 2;
485 break;
486 case 2 ... 3:
487 index = 1;
488 case 1:
489 break;
490 }
491
492 td->ts.io_u_map[index] += nr;
493}
494
495static void io_u_mark_lat_usec(struct thread_data *td, unsigned long usec)
496{
497 int index = 0;
498
499 assert(usec < 1000);
500
501 switch (usec) {
502 case 750 ... 999:
503 index = 9;
504 break;
505 case 500 ... 749:
506 index = 8;
507 break;
508 case 250 ... 499:
509 index = 7;
510 break;
511 case 100 ... 249:
512 index = 6;
513 break;
514 case 50 ... 99:
515 index = 5;
516 break;
517 case 20 ... 49:
518 index = 4;
519 break;
520 case 10 ... 19:
521 index = 3;
522 break;
523 case 4 ... 9:
524 index = 2;
525 break;
526 case 2 ... 3:
527 index = 1;
528 case 0 ... 1:
529 break;
530 }
531
532 assert(index < FIO_IO_U_LAT_U_NR);
533 td->ts.io_u_lat_u[index]++;
534}
535
536static void io_u_mark_lat_msec(struct thread_data *td, unsigned long msec)
537{
538 int index = 0;
539
540 switch (msec) {
541 default:
542 index = 11;
543 break;
544 case 1000 ... 1999:
545 index = 10;
546 break;
547 case 750 ... 999:
548 index = 9;
549 break;
550 case 500 ... 749:
551 index = 8;
552 break;
553 case 250 ... 499:
554 index = 7;
555 break;
556 case 100 ... 249:
557 index = 6;
558 break;
559 case 50 ... 99:
560 index = 5;
561 break;
562 case 20 ... 49:
563 index = 4;
564 break;
565 case 10 ... 19:
566 index = 3;
567 break;
568 case 4 ... 9:
569 index = 2;
570 break;
571 case 2 ... 3:
572 index = 1;
573 case 0 ... 1:
574 break;
575 }
576
577 assert(index < FIO_IO_U_LAT_M_NR);
578 td->ts.io_u_lat_m[index]++;
579}
580
581static void io_u_mark_latency(struct thread_data *td, unsigned long usec)
582{
583 if (usec < 1000)
584 io_u_mark_lat_usec(td, usec);
585 else
586 io_u_mark_lat_msec(td, usec / 1000);
587}
588
589/*
590 * Get next file to service by choosing one at random
591 */
592static struct fio_file *get_next_file_rand(struct thread_data *td, int goodf,
593 int badf)
594{
595 struct fio_file *f;
596 int fno;
597
598 do {
599 long r = os_random_long(&td->next_file_state);
600
601 fno = (unsigned int) ((double) td->o.nr_files
602 * (r / (RAND_MAX + 1.0)));
603 f = td->files[fno];
604 if (f->flags & FIO_FILE_DONE)
605 continue;
606
607 if ((!goodf || (f->flags & goodf)) && !(f->flags & badf)) {
608 dprint(FD_FILE, "get_next_file_rand: %p\n", f);
609 return f;
610 }
611 } while (1);
612}
613
614/*
615 * Get next file to service by doing round robin between all available ones
616 */
617static struct fio_file *get_next_file_rr(struct thread_data *td, int goodf,
618 int badf)
619{
620 unsigned int old_next_file = td->next_file;
621 struct fio_file *f;
622
623 do {
624 f = td->files[td->next_file];
625
626 td->next_file++;
627 if (td->next_file >= td->o.nr_files)
628 td->next_file = 0;
629
630 if (f->flags & FIO_FILE_DONE) {
631 f = NULL;
632 continue;
633 }
634
635 if ((!goodf || (f->flags & goodf)) && !(f->flags & badf))
636 break;
637
638 f = NULL;
639 } while (td->next_file != old_next_file);
640
641 dprint(FD_FILE, "get_next_file_rr: %p\n", f);
642 return f;
643}
644
645static struct fio_file *get_next_file(struct thread_data *td)
646{
647 struct fio_file *f;
648
649 assert(td->o.nr_files <= td->files_index);
650
651 if (!td->nr_open_files || td->nr_done_files >= td->o.nr_files) {
652 dprint(FD_FILE, "get_next_file: nr_open=%d, nr_done=%d,"
653 " nr_files=%d\n", td->nr_open_files,
654 td->nr_done_files,
655 td->o.nr_files);
656 return NULL;
657 }
658
659 f = td->file_service_file;
660 if (f && (f->flags & FIO_FILE_OPEN) && td->file_service_left--)
661 goto out;
662
663 if (td->o.file_service_type == FIO_FSERVICE_RR)
664 f = get_next_file_rr(td, FIO_FILE_OPEN, FIO_FILE_CLOSING);
665 else
666 f = get_next_file_rand(td, FIO_FILE_OPEN, FIO_FILE_CLOSING);
667
668 td->file_service_file = f;
669 td->file_service_left = td->file_service_nr - 1;
670out:
671 dprint(FD_FILE, "get_next_file: %p\n", f);
672 return f;
673}
674
675static struct fio_file *find_next_new_file(struct thread_data *td)
676{
677 struct fio_file *f;
678
679 if (!td->nr_open_files || td->nr_done_files >= td->o.nr_files)
680 return NULL;
681
682 if (td->o.file_service_type == FIO_FSERVICE_RR)
683 f = get_next_file_rr(td, 0, FIO_FILE_OPEN);
684 else
685 f = get_next_file_rand(td, 0, FIO_FILE_OPEN);
686
687 return f;
688}
689
690static int set_io_u_file(struct thread_data *td, struct io_u *io_u)
691{
692 struct fio_file *f;
693
694 do {
695 f = get_next_file(td);
696 if (!f)
697 return 1;
698
699set_file:
700 io_u->file = f;
701 get_file(f);
702
703 if (!fill_io_u(td, io_u))
704 break;
705
706 /*
707 * optimization to prevent close/open of the same file. This
708 * way we preserve queueing etc.
709 */
710 if (td->o.nr_files == 1 && td->o.time_based) {
711 put_file_log(td, f);
712 fio_file_reset(f);
713 goto set_file;
714 }
715
716 /*
717 * td_io_close() does a put_file() as well, so no need to
718 * do that here.
719 */
720 io_u->file = NULL;
721 td_io_close_file(td, f);
722 f->flags |= FIO_FILE_DONE;
723 td->nr_done_files++;
724
725 /*
726 * probably not the right place to do this, but see
727 * if we need to open a new file
728 */
729 if (td->nr_open_files < td->o.open_files &&
730 td->o.open_files != td->o.nr_files) {
731 f = find_next_new_file(td);
732
733 if (!f || td_io_open_file(td, f))
734 return 1;
735
736 goto set_file;
737 }
738 } while (1);
739
740 return 0;
741}
742
743
744struct io_u *__get_io_u(struct thread_data *td)
745{
746 struct io_u *io_u = NULL;
747
748 if (!list_empty(&td->io_u_requeues))
749 io_u = list_entry(td->io_u_requeues.next, struct io_u, list);
750 else if (!queue_full(td)) {
751 io_u = list_entry(td->io_u_freelist.next, struct io_u, list);
752
753 io_u->buflen = 0;
754 io_u->resid = 0;
755 io_u->file = NULL;
756 io_u->end_io = NULL;
757 }
758
759 if (io_u) {
760 assert(io_u->flags & IO_U_F_FREE);
761 io_u->flags &= ~IO_U_F_FREE;
762
763 io_u->error = 0;
764 list_del(&io_u->list);
765 list_add(&io_u->list, &td->io_u_busylist);
766 td->cur_depth++;
767 }
768
769 return io_u;
770}
771
772/*
773 * Return an io_u to be processed. Gets a buflen and offset, sets direction,
774 * etc. The returned io_u is fully ready to be prepped and submitted.
775 */
776struct io_u *get_io_u(struct thread_data *td)
777{
778 struct fio_file *f;
779 struct io_u *io_u;
780
781 io_u = __get_io_u(td);
782 if (!io_u) {
783 dprint(FD_IO, "__get_io_u failed\n");
784 return NULL;
785 }
786
787 /*
788 * from a requeue, io_u already setup
789 */
790 if (io_u->file)
791 goto out;
792
793 /*
794 * If using an iolog, grab next piece if any available.
795 */
796 if (td->o.read_iolog_file) {
797 if (read_iolog_get(td, io_u))
798 goto err_put;
799 } else if (set_io_u_file(td, io_u)) {
800 dprint(FD_IO, "io_u %p, setting file failed\n", io_u);
801 goto err_put;
802 }
803
804 f = io_u->file;
805 assert(f->flags & FIO_FILE_OPEN);
806
807 if (io_u->ddir != DDIR_SYNC) {
808 if (!io_u->buflen && !(td->io_ops->flags & FIO_NOIO)) {
809 dprint(FD_IO, "get_io_u: zero buflen on %p\n", io_u);
810 goto err_put;
811 }
812
813 f->last_pos = io_u->offset + io_u->buflen;
814
815 if (td->o.verify != VERIFY_NONE)
816 populate_verify_io_u(td, io_u);
817 }
818
819 /*
820 * Set io data pointers.
821 */
822 io_u->endpos = io_u->offset + io_u->buflen;
823 io_u->xfer_buf = io_u->buf;
824 io_u->xfer_buflen = io_u->buflen;
825
826 if (td->o.refill_buffers && io_u->ddir == DDIR_WRITE)
827 io_u_fill_buffer(td, io_u, io_u->xfer_buflen);
828out:
829 if (!td_io_prep(td, io_u)) {
830 fio_gettime(&io_u->start_time, NULL);
831 return io_u;
832 }
833err_put:
834 dprint(FD_IO, "get_io_u failed\n");
835 put_io_u(td, io_u);
836 return NULL;
837}
838
839void io_u_log_error(struct thread_data *td, struct io_u *io_u)
840{
841 const char *msg[] = { "read", "write", "sync" };
842
843 log_err("fio: io_u error");
844
845 if (io_u->file)
846 log_err(" on file %s", io_u->file->file_name);
847
848 log_err(": %s\n", strerror(io_u->error));
849
850 log_err(" %s offset=%llu, buflen=%lu\n", msg[io_u->ddir],
851 io_u->offset, io_u->xfer_buflen);
852
853 if (!td->error)
854 td_verror(td, io_u->error, "io_u error");
855}
856
857static void io_completed(struct thread_data *td, struct io_u *io_u,
858 struct io_completion_data *icd)
859{
860 unsigned long usec;
861
862 dprint_io_u(io_u, "io complete");
863
864 assert(io_u->flags & IO_U_F_FLIGHT);
865 io_u->flags &= ~IO_U_F_FLIGHT;
866
867 if (io_u->ddir == DDIR_SYNC) {
868 td->last_was_sync = 1;
869 return;
870 }
871
872 td->last_was_sync = 0;
873
874 if (!io_u->error) {
875 unsigned int bytes = io_u->buflen - io_u->resid;
876 const enum fio_ddir idx = io_u->ddir;
877 int ret;
878
879 td->io_blocks[idx]++;
880 td->io_bytes[idx] += bytes;
881 td->this_io_bytes[idx] += bytes;
882
883 usec = utime_since(&io_u->issue_time, &icd->time);
884
885 add_clat_sample(td, idx, usec);
886 add_bw_sample(td, idx, &icd->time);
887 io_u_mark_latency(td, usec);
888
889 if (td_write(td) && idx == DDIR_WRITE &&
890 td->o.do_verify &&
891 td->o.verify != VERIFY_NONE)
892 log_io_piece(td, io_u);
893
894 icd->bytes_done[idx] += bytes;
895
896 if (io_u->end_io) {
897 ret = io_u->end_io(td, io_u);
898 if (ret && !icd->error)
899 icd->error = ret;
900 }
901 } else {
902 icd->error = io_u->error;
903 io_u_log_error(td, io_u);
904 }
905}
906
907static void init_icd(struct io_completion_data *icd, int nr)
908{
909 fio_gettime(&icd->time, NULL);
910
911 icd->nr = nr;
912
913 icd->error = 0;
914 icd->bytes_done[0] = icd->bytes_done[1] = 0;
915}
916
917static void ios_completed(struct thread_data *td,
918 struct io_completion_data *icd)
919{
920 struct io_u *io_u;
921 int i;
922
923 for (i = 0; i < icd->nr; i++) {
924 io_u = td->io_ops->event(td, i);
925
926 io_completed(td, io_u, icd);
927 put_io_u(td, io_u);
928 }
929}
930
931/*
932 * Complete a single io_u for the sync engines.
933 */
934long io_u_sync_complete(struct thread_data *td, struct io_u *io_u)
935{
936 struct io_completion_data icd;
937
938 init_icd(&icd, 1);
939 io_completed(td, io_u, &icd);
940 put_io_u(td, io_u);
941
942 if (!icd.error)
943 return icd.bytes_done[0] + icd.bytes_done[1];
944
945 td_verror(td, icd.error, "io_u_sync_complete");
946 return -1;
947}
948
949/*
950 * Called to complete min_events number of io for the async engines.
951 */
952long io_u_queued_complete(struct thread_data *td, int min_events)
953{
954 struct io_completion_data icd;
955 struct timespec *tvp = NULL;
956 int ret;
957 struct timespec ts = { .tv_sec = 0, .tv_nsec = 0, };
958
959 dprint(FD_IO, "io_u_queued_completed: min=%d\n", min_events);
960
961 if (!min_events)
962 tvp = &ts;
963
964 ret = td_io_getevents(td, min_events, td->cur_depth, tvp);
965 if (ret < 0) {
966 td_verror(td, -ret, "td_io_getevents");
967 return ret;
968 } else if (!ret)
969 return ret;
970
971 init_icd(&icd, ret);
972 ios_completed(td, &icd);
973 if (!icd.error)
974 return icd.bytes_done[0] + icd.bytes_done[1];
975
976 td_verror(td, icd.error, "io_u_queued_complete");
977 return -1;
978}
979
980/*
981 * Call when io_u is really queued, to update the submission latency.
982 */
983void io_u_queued(struct thread_data *td, struct io_u *io_u)
984{
985 unsigned long slat_time;
986
987 slat_time = utime_since(&io_u->start_time, &io_u->issue_time);
988 add_slat_sample(td, io_u->ddir, slat_time);
989}
990
991/*
992 * "randomly" fill the buffer contents
993 */
994void io_u_fill_buffer(struct thread_data *td, struct io_u *io_u,
995 unsigned int max_bs)
996{
997 long *ptr = io_u->buf;
998
999 if (!td->o.zero_buffers) {
1000 while ((void *) ptr - io_u->buf < max_bs) {
1001 *ptr = rand() * GOLDEN_RATIO_PRIME;
1002 ptr++;
1003 }
1004 } else
1005 memset(ptr, 0, max_bs);
1006}
1007
1008#ifdef FIO_USE_TIMEOUT
1009void io_u_set_timeout(struct thread_data *td)
1010{
1011 assert(td->cur_depth);
1012
1013 td->timer.it_interval.tv_sec = 0;
1014 td->timer.it_interval.tv_usec = 0;
1015 td->timer.it_value.tv_sec = IO_U_TIMEOUT + IO_U_TIMEOUT_INC;
1016 td->timer.it_value.tv_usec = 0;
1017 setitimer(ITIMER_REAL, &td->timer, NULL);
1018 fio_gettime(&td->timeout_end, NULL);
1019}
1020
1021static void io_u_dump(struct io_u *io_u)
1022{
1023 unsigned long t_start = mtime_since_now(&io_u->start_time);
1024 unsigned long t_issue = mtime_since_now(&io_u->issue_time);
1025
1026 log_err("io_u=%p, t_start=%lu, t_issue=%lu\n", io_u, t_start, t_issue);
1027 log_err(" buf=%p/%p, len=%lu/%lu, offset=%llu\n", io_u->buf,
1028 io_u->xfer_buf, io_u->buflen,
1029 io_u->xfer_buflen,
1030 io_u->offset);
1031 log_err(" ddir=%d, fname=%s\n", io_u->ddir, io_u->file->file_name);
1032}
1033#else
1034void io_u_set_timeout(struct thread_data fio_unused *td)
1035{
1036}
1037#endif
1038
1039#ifdef FIO_USE_TIMEOUT
1040static void io_u_timeout_handler(int fio_unused sig)
1041{
1042 struct thread_data *td, *__td;
1043 pid_t pid = getpid();
1044 struct list_head *entry;
1045 struct io_u *io_u;
1046 int i;
1047
1048 log_err("fio: io_u timeout\n");
1049
1050 /*
1051 * TLS would be nice...
1052 */
1053 td = NULL;
1054 for_each_td(__td, i) {
1055 if (__td->pid == pid) {
1056 td = __td;
1057 break;
1058 }
1059 }
1060
1061 if (!td) {
1062 log_err("fio: io_u timeout, can't find job\n");
1063 exit(1);
1064 }
1065
1066 if (!td->cur_depth) {
1067 log_err("fio: timeout without pending work?\n");
1068 return;
1069 }
1070
1071 log_err("fio: io_u timeout: job=%s, pid=%d\n", td->o.name, td->pid);
1072
1073 list_for_each(entry, &td->io_u_busylist) {
1074 io_u = list_entry(entry, struct io_u, list);
1075
1076 io_u_dump(io_u);
1077 }
1078
1079 td_verror(td, ETIMEDOUT, "io_u timeout");
1080 exit(1);
1081}
1082#endif
1083
1084void io_u_init_timeout(void)
1085{
1086#ifdef FIO_USE_TIMEOUT
1087 signal(SIGALRM, io_u_timeout_handler);
1088#endif
1089}
fio - Flexible IO Tester