-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathreactor.c
1561 lines (1276 loc) · 39.2 KB
/
reactor.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
/*-
* BSD LICENSE
*
* Copyright (c) Intel Corporation.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in
* the documentation and/or other materials provided with the
* distribution.
* * Neither the name of Intel Corporation nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "spdk/stdinc.h"
#include "spdk/likely.h"
#include "spdk_internal/event.h"
#include "spdk/log.h"
#include "spdk/thread.h"
#include "spdk/env.h"
#include "spdk/util.h"
#include "spdk/string.h"
#include "spdk/fd_group.h"
#ifdef __linux__
#include <sys/prctl.h>
#include <sys/eventfd.h>
#endif
#ifdef __FreeBSD__
#include <pthread_np.h>
#endif
#define SPDK_EVENT_BATCH_SIZE 8
static struct spdk_reactor *g_reactors;
static uint32_t g_reactor_count;
static struct spdk_cpuset g_reactor_core_mask;
static enum spdk_reactor_state g_reactor_state = SPDK_REACTOR_STATE_UNINITIALIZED;
static bool g_framework_context_switch_monitor_enabled = true;
static struct spdk_mempool *g_spdk_event_mempool = NULL;
TAILQ_HEAD(, spdk_scheduler) g_scheduler_list
= TAILQ_HEAD_INITIALIZER(g_scheduler_list);
static struct spdk_scheduler *g_scheduler;
static struct spdk_scheduler *g_new_scheduler;
static struct spdk_reactor *g_scheduling_reactor;
static uint64_t g_scheduler_period;
static uint32_t g_scheduler_core_number;
static struct spdk_scheduler_core_info *g_core_infos = NULL;
TAILQ_HEAD(, spdk_governor) g_governor_list
= TAILQ_HEAD_INITIALIZER(g_governor_list);
static int _governor_get_capabilities(uint32_t lcore_id,
struct spdk_governor_capabilities *capabilities);
static struct spdk_governor g_governor = {
.name = "default",
.get_core_capabilities = _governor_get_capabilities,
};
static int reactor_interrupt_init(struct spdk_reactor *reactor);
static void reactor_interrupt_fini(struct spdk_reactor *reactor);
static struct spdk_scheduler *
_scheduler_find(char *name)
{
struct spdk_scheduler *tmp;
TAILQ_FOREACH(tmp, &g_scheduler_list, link) {
if (strcmp(name, tmp->name) == 0) {
return tmp;
}
}
return NULL;
}
int
_spdk_scheduler_set(char *name)
{
struct spdk_scheduler *scheduler;
scheduler = _scheduler_find(name);
if (scheduler == NULL) {
SPDK_ERRLOG("Requested scheduler is missing\n");
return -ENOENT;
}
if (g_scheduling_reactor->flags.is_scheduling) {
if (g_scheduler != g_new_scheduler) {
/* Scheduler already changed, cannot defer multiple deinits */
return -EBUSY;
}
} else {
if (g_scheduler != NULL && g_scheduler->deinit != NULL) {
g_scheduler->deinit(&g_governor);
}
g_scheduler = scheduler;
}
g_new_scheduler = scheduler;
if (scheduler->init != NULL) {
scheduler->init(&g_governor);
}
return 0;
}
struct spdk_scheduler *
_spdk_scheduler_get(void)
{
return g_scheduler;
}
uint64_t
_spdk_scheduler_period_get(void)
{
/* Convert from ticks to microseconds */
return (g_scheduler_period * SPDK_SEC_TO_USEC / spdk_get_ticks_hz());
}
void
_spdk_scheduler_period_set(uint64_t period)
{
/* Convert microseconds to ticks */
g_scheduler_period = period * spdk_get_ticks_hz() / SPDK_SEC_TO_USEC;
}
void
_spdk_scheduler_disable(void)
{
g_scheduler_period = 0;
}
void
_spdk_scheduler_list_add(struct spdk_scheduler *scheduler)
{
if (_scheduler_find(scheduler->name)) {
SPDK_ERRLOG("scheduler named '%s' already registered.\n", scheduler->name);
assert(false);
return;
}
TAILQ_INSERT_TAIL(&g_scheduler_list, scheduler, link);
}
static void
reactor_construct(struct spdk_reactor *reactor, uint32_t lcore)
{
reactor->lcore = lcore;
reactor->flags.is_valid = true;
TAILQ_INIT(&reactor->threads);
reactor->thread_count = 0;
spdk_cpuset_zero(&reactor->notify_cpuset);
reactor->events = spdk_ring_create(SPDK_RING_TYPE_MP_SC, 65536, SPDK_ENV_SOCKET_ID_ANY);
if (reactor->events == NULL) {
SPDK_ERRLOG("Failed to allocate events ring\n");
assert(false);
}
/* Always initialize interrupt facilities for reactor */
if (reactor_interrupt_init(reactor) != 0) {
/* Reactor interrupt facilities are necessary if seting app to interrupt mode. */
if (spdk_interrupt_mode_is_enabled()) {
SPDK_ERRLOG("Failed to prepare intr facilities\n");
assert(false);
}
return;
}
/* If application runs with full interrupt ability,
* all reactors are going to run in interrupt mode.
*/
if (spdk_interrupt_mode_is_enabled()) {
uint32_t i;
SPDK_ENV_FOREACH_CORE(i) {
spdk_cpuset_set_cpu(&reactor->notify_cpuset, i, true);
}
reactor->in_interrupt = true;
}
}
struct spdk_reactor *
spdk_reactor_get(uint32_t lcore)
{
struct spdk_reactor *reactor;
if (g_reactors == NULL) {
SPDK_WARNLOG("Called spdk_reactor_get() while the g_reactors array was NULL!\n");
return NULL;
}
if (lcore >= g_reactor_count) {
return NULL;
}
reactor = &g_reactors[lcore];
if (reactor->flags.is_valid == false) {
return NULL;
}
return reactor;
}
struct spdk_reactor *
_spdk_get_scheduling_reactor(void)
{
return g_scheduling_reactor;
}
static int reactor_thread_op(struct spdk_thread *thread, enum spdk_thread_op op);
static bool reactor_thread_op_supported(enum spdk_thread_op op);
int
spdk_reactors_init(void)
{
struct spdk_reactor *reactor;
int rc;
uint32_t i, current_core;
char mempool_name[32];
snprintf(mempool_name, sizeof(mempool_name), "evtpool_%d", getpid());
g_spdk_event_mempool = spdk_mempool_create(mempool_name,
262144 - 1, /* Power of 2 minus 1 is optimal for memory consumption */
sizeof(struct spdk_event),
SPDK_MEMPOOL_DEFAULT_CACHE_SIZE,
SPDK_ENV_SOCKET_ID_ANY);
if (g_spdk_event_mempool == NULL) {
SPDK_ERRLOG("spdk_event_mempool creation failed\n");
return -1;
}
/* struct spdk_reactor must be aligned on 64 byte boundary */
g_reactor_count = spdk_env_get_last_core() + 1;
rc = posix_memalign((void **)&g_reactors, 64,
g_reactor_count * sizeof(struct spdk_reactor));
if (rc != 0) {
SPDK_ERRLOG("Could not allocate array size=%u for g_reactors\n",
g_reactor_count);
spdk_mempool_free(g_spdk_event_mempool);
return -1;
}
g_core_infos = calloc(g_reactor_count, sizeof(*g_core_infos));
if (g_core_infos == NULL) {
SPDK_ERRLOG("Could not allocate memory for g_core_infos\n");
spdk_mempool_free(g_spdk_event_mempool);
free(g_reactors);
return -ENOMEM;
}
memset(g_reactors, 0, (g_reactor_count) * sizeof(struct spdk_reactor));
spdk_thread_lib_init_ext(reactor_thread_op, reactor_thread_op_supported,
sizeof(struct spdk_lw_thread));
SPDK_ENV_FOREACH_CORE(i) {
reactor_construct(&g_reactors[i], i);
}
current_core = spdk_env_get_current_core();
reactor = spdk_reactor_get(current_core);
assert(reactor != NULL);
g_scheduling_reactor = reactor;
/* set default scheduling period to one second */
g_scheduler_period = spdk_get_ticks_hz();
rc = _spdk_scheduler_set("static");
assert(rc == 0);
g_reactor_state = SPDK_REACTOR_STATE_INITIALIZED;
return 0;
}
void
spdk_reactors_fini(void)
{
uint32_t i;
struct spdk_reactor *reactor;
if (g_reactor_state == SPDK_REACTOR_STATE_UNINITIALIZED) {
return;
}
if (g_scheduler->deinit != NULL) {
g_scheduler->deinit(&g_governor);
}
spdk_thread_lib_fini();
SPDK_ENV_FOREACH_CORE(i) {
reactor = spdk_reactor_get(i);
assert(reactor != NULL);
assert(reactor->thread_count == 0);
if (reactor->events != NULL) {
spdk_ring_free(reactor->events);
}
reactor_interrupt_fini(reactor);
if (g_core_infos != NULL) {
free(g_core_infos[i].threads);
}
}
spdk_mempool_free(g_spdk_event_mempool);
free(g_reactors);
g_reactors = NULL;
free(g_core_infos);
g_core_infos = NULL;
}
static void _reactor_set_interrupt_mode(void *arg1, void *arg2);
static void
_reactor_set_notify_cpuset(void *arg1, void *arg2)
{
struct spdk_reactor *target = arg1;
struct spdk_reactor *reactor = spdk_reactor_get(spdk_env_get_current_core());
assert(reactor != NULL);
spdk_cpuset_set_cpu(&reactor->notify_cpuset, target->lcore, target->new_in_interrupt);
}
static void
_reactor_set_notify_cpuset_cpl(void *arg1, void *arg2)
{
struct spdk_reactor *target = arg1;
if (target->new_in_interrupt == false) {
target->set_interrupt_mode_in_progress = false;
spdk_thread_send_msg(_spdk_get_app_thread(), target->set_interrupt_mode_cb_fn,
target->set_interrupt_mode_cb_arg);
} else {
struct spdk_event *ev;
ev = spdk_event_allocate(target->lcore, _reactor_set_interrupt_mode, target, NULL);
assert(ev);
spdk_event_call(ev);
}
}
static void
_reactor_set_thread_interrupt_mode(void *ctx)
{
struct spdk_reactor *reactor = ctx;
spdk_thread_set_interrupt_mode(reactor->in_interrupt);
}
static void
_reactor_set_interrupt_mode(void *arg1, void *arg2)
{
struct spdk_reactor *target = arg1;
struct spdk_thread *thread;
struct spdk_lw_thread *lw_thread, *tmp;
assert(target == spdk_reactor_get(spdk_env_get_current_core()));
assert(target != NULL);
assert(target->in_interrupt != target->new_in_interrupt);
SPDK_DEBUGLOG(reactor, "Do reactor set on core %u from %s to state %s\n",
target->lcore, target->in_interrupt ? "intr" : "poll", target->new_in_interrupt ? "intr" : "poll");
target->in_interrupt = target->new_in_interrupt;
/* Align spdk_thread with reactor to interrupt mode or poll mode */
TAILQ_FOREACH_SAFE(lw_thread, &target->threads, link, tmp) {
thread = spdk_thread_get_from_ctx(lw_thread);
spdk_thread_send_msg(thread, _reactor_set_thread_interrupt_mode, target);
}
if (target->new_in_interrupt == false) {
spdk_for_each_reactor(_reactor_set_notify_cpuset, target, NULL, _reactor_set_notify_cpuset_cpl);
} else {
uint64_t notify = 1;
int rc = 0;
/* Always trigger spdk_event and resched event in case of race condition */
rc = write(target->events_fd, ¬ify, sizeof(notify));
if (rc < 0) {
SPDK_ERRLOG("failed to notify event queue: %s.\n", spdk_strerror(errno));
}
rc = write(target->resched_fd, ¬ify, sizeof(notify));
if (rc < 0) {
SPDK_ERRLOG("failed to notify reschedule: %s.\n", spdk_strerror(errno));
}
target->set_interrupt_mode_in_progress = false;
spdk_thread_send_msg(_spdk_get_app_thread(), target->set_interrupt_mode_cb_fn,
target->set_interrupt_mode_cb_arg);
}
}
int
spdk_reactor_set_interrupt_mode(uint32_t lcore, bool new_in_interrupt,
spdk_reactor_set_interrupt_mode_cb cb_fn, void *cb_arg)
{
struct spdk_reactor *target;
target = spdk_reactor_get(lcore);
if (target == NULL) {
return -EINVAL;
}
if (spdk_get_thread() != _spdk_get_app_thread()) {
SPDK_ERRLOG("It is only permitted within spdk application thread.\n");
return -EPERM;
}
if (target->in_interrupt == new_in_interrupt) {
cb_fn(cb_arg);
return 0;
}
if (target->set_interrupt_mode_in_progress) {
SPDK_NOTICELOG("Reactor(%u) is already in progress to set interrupt mode\n", lcore);
return -EBUSY;
}
target->set_interrupt_mode_in_progress = true;
target->new_in_interrupt = new_in_interrupt;
target->set_interrupt_mode_cb_fn = cb_fn;
target->set_interrupt_mode_cb_arg = cb_arg;
SPDK_DEBUGLOG(reactor, "Starting reactor event from %d to %d\n",
spdk_env_get_current_core(), lcore);
if (new_in_interrupt == false) {
/* For potential race cases, when setting the reactor to poll mode,
* first change the mode of the reactor and then clear the corresponding
* bit of the notify_cpuset of each reactor.
*/
struct spdk_event *ev;
ev = spdk_event_allocate(lcore, _reactor_set_interrupt_mode, target, NULL);
assert(ev);
spdk_event_call(ev);
} else {
/* For race caces, when setting the reactor to interrupt mode, first set the
* corresponding bit of the notify_cpuset of each reactor and then change the mode.
*/
spdk_for_each_reactor(_reactor_set_notify_cpuset, target, NULL, _reactor_set_notify_cpuset_cpl);
}
return 0;
}
struct spdk_event *
spdk_event_allocate(uint32_t lcore, spdk_event_fn fn, void *arg1, void *arg2)
{
struct spdk_event *event = NULL;
struct spdk_reactor *reactor = spdk_reactor_get(lcore);
if (!reactor) {
assert(false);
return NULL;
}
event = spdk_mempool_get(g_spdk_event_mempool);
if (event == NULL) {
assert(false);
return NULL;
}
event->lcore = lcore;
event->fn = fn;
event->arg1 = arg1;
event->arg2 = arg2;
return event;
}
void
spdk_event_call(struct spdk_event *event)
{
int rc;
struct spdk_reactor *reactor;
struct spdk_reactor *local_reactor = NULL;
uint32_t current_core = spdk_env_get_current_core();
reactor = spdk_reactor_get(event->lcore);
assert(reactor != NULL);
assert(reactor->events != NULL);
rc = spdk_ring_enqueue(reactor->events, (void **)&event, 1, NULL);
if (rc != 1) {
assert(false);
}
if (current_core != SPDK_ENV_LCORE_ID_ANY) {
local_reactor = spdk_reactor_get(current_core);
}
/* If spdk_event_call isn't called on a reactor, always send a notification.
* If it is called on a reactor, send a notification if the destination reactor
* is indicated in interrupt mode state.
*/
if (spdk_unlikely(local_reactor == NULL) ||
spdk_unlikely(spdk_cpuset_get_cpu(&local_reactor->notify_cpuset, event->lcore))) {
uint64_t notify = 1;
rc = write(reactor->events_fd, ¬ify, sizeof(notify));
if (rc < 0) {
SPDK_ERRLOG("failed to notify event queue: %s.\n", spdk_strerror(errno));
}
}
}
static inline uint32_t
event_queue_run_batch(struct spdk_reactor *reactor)
{
unsigned count, i;
void *events[SPDK_EVENT_BATCH_SIZE];
struct spdk_thread *thread;
struct spdk_lw_thread *lw_thread;
#ifdef DEBUG
/*
* spdk_ring_dequeue() fills events and returns how many entries it wrote,
* so we will never actually read uninitialized data from events, but just to be sure
* (and to silence a static analyzer false positive), initialize the array to NULL pointers.
*/
memset(events, 0, sizeof(events));
#endif
/* Operate event notification if this reactor currently runs in interrupt state */
if (spdk_unlikely(reactor->in_interrupt)) {
uint64_t notify = 1;
int rc;
/* There may be race between event_acknowledge and another producer's event_notify,
* so event_acknowledge should be applied ahead. And then check for self's event_notify.
* This can avoid event notification missing.
*/
rc = read(reactor->events_fd, ¬ify, sizeof(notify));
if (rc < 0) {
SPDK_ERRLOG("failed to acknowledge event queue: %s.\n", spdk_strerror(errno));
return -errno;
}
count = spdk_ring_dequeue(reactor->events, events, SPDK_EVENT_BATCH_SIZE);
if (spdk_ring_count(reactor->events) != 0) {
/* Trigger new notification if there are still events in event-queue waiting for processing. */
rc = write(reactor->events_fd, ¬ify, sizeof(notify));
if (rc < 0) {
SPDK_ERRLOG("failed to notify event queue: %s.\n", spdk_strerror(errno));
return -errno;
}
}
} else {
count = spdk_ring_dequeue(reactor->events, events, SPDK_EVENT_BATCH_SIZE);
}
if (count == 0) {
return 0;
}
/* Execute the events. There are still some remaining events
* that must occur on an SPDK thread. To accomodate those, try to
* run them on the first thread in the list, if it exists. */
lw_thread = TAILQ_FIRST(&reactor->threads);
if (lw_thread) {
thread = spdk_thread_get_from_ctx(lw_thread);
} else {
thread = NULL;
}
spdk_set_thread(thread);
for (i = 0; i < count; i++) {
struct spdk_event *event = events[i];
assert(event != NULL);
event->fn(event->arg1, event->arg2);
}
spdk_set_thread(NULL);
spdk_mempool_put_bulk(g_spdk_event_mempool, events, count);
return count;
}
/* 1s */
#define CONTEXT_SWITCH_MONITOR_PERIOD 1000000
static int
get_rusage(struct spdk_reactor *reactor)
{
struct rusage rusage;
if (getrusage(RUSAGE_THREAD, &rusage) != 0) {
return -1;
}
if (rusage.ru_nvcsw != reactor->rusage.ru_nvcsw || rusage.ru_nivcsw != reactor->rusage.ru_nivcsw) {
SPDK_INFOLOG(reactor,
"Reactor %d: %ld voluntary context switches and %ld involuntary context switches in the last second.\n",
reactor->lcore, rusage.ru_nvcsw - reactor->rusage.ru_nvcsw,
rusage.ru_nivcsw - reactor->rusage.ru_nivcsw);
}
reactor->rusage = rusage;
return -1;
}
void
spdk_framework_enable_context_switch_monitor(bool enable)
{
/* This global is being read by multiple threads, so this isn't
* strictly thread safe. However, we're toggling between true and
* false here, and if a thread sees the value update later than it
* should, it's no big deal. */
g_framework_context_switch_monitor_enabled = enable;
}
bool
spdk_framework_context_switch_monitor_enabled(void)
{
return g_framework_context_switch_monitor_enabled;
}
static void
_set_thread_name(const char *thread_name)
{
#if defined(__linux__)
prctl(PR_SET_NAME, thread_name, 0, 0, 0);
#elif defined(__FreeBSD__)
pthread_set_name_np(pthread_self(), thread_name);
#else
pthread_setname_np(pthread_self(), thread_name);
#endif
}
static void
_init_thread_stats(struct spdk_reactor *reactor, struct spdk_lw_thread *lw_thread)
{
struct spdk_thread *thread = spdk_thread_get_from_ctx(lw_thread);
lw_thread->lcore = reactor->lcore;
spdk_set_thread(thread);
spdk_thread_get_stats(&lw_thread->current_stats);
}
static void
_threads_reschedule(struct spdk_scheduler_core_info *cores_info)
{
struct spdk_scheduler_core_info *core;
struct spdk_lw_thread *lw_thread;
uint32_t i, j;
SPDK_ENV_FOREACH_CORE(i) {
core = &cores_info[i];
for (j = 0; j < core->threads_count; j++) {
lw_thread = core->threads[j];
if (lw_thread->lcore != lw_thread->new_lcore) {
_spdk_lw_thread_set_core(lw_thread, lw_thread->new_lcore);
}
}
}
}
static void
_reactors_scheduler_fini(void)
{
struct spdk_reactor *reactor;
uint32_t i;
/* Reschedule based on the balancing output */
_threads_reschedule(g_core_infos);
SPDK_ENV_FOREACH_CORE(i) {
reactor = spdk_reactor_get(i);
assert(reactor != NULL);
reactor->flags.is_scheduling = false;
}
}
static void
_reactors_scheduler_update_core_mode(void *ctx)
{
struct spdk_reactor *reactor;
int rc = 0;
if (g_scheduler_core_number == SPDK_ENV_LCORE_ID_ANY) {
g_scheduler_core_number = spdk_env_get_first_core();
} else {
g_scheduler_core_number = spdk_env_get_next_core(g_scheduler_core_number);
}
if (g_scheduler_core_number == SPDK_ENV_LCORE_ID_ANY) {
_reactors_scheduler_fini();
return;
}
reactor = spdk_reactor_get(g_scheduler_core_number);
assert(reactor != NULL);
if (reactor->in_interrupt != g_core_infos[g_scheduler_core_number].interrupt_mode) {
/* Switch next found reactor to new state */
rc = spdk_reactor_set_interrupt_mode(g_scheduler_core_number,
g_core_infos[g_scheduler_core_number].interrupt_mode, _reactors_scheduler_update_core_mode, NULL);
if (rc == 0) {
return;
}
}
_reactors_scheduler_update_core_mode(NULL);
}
static void
_reactors_scheduler_balance(void *arg1, void *arg2)
{
if (g_reactor_state == SPDK_REACTOR_STATE_RUNNING) {
g_scheduler->balance(g_core_infos, g_reactor_count, &g_governor);
g_scheduler_core_number = SPDK_ENV_LCORE_ID_ANY;
_reactors_scheduler_update_core_mode(NULL);
}
}
static void
_reactors_scheduler_cancel(void *arg1, void *arg2)
{
struct spdk_reactor *reactor;
uint32_t i;
SPDK_ENV_FOREACH_CORE(i) {
reactor = spdk_reactor_get(i);
assert(reactor != NULL);
reactor->flags.is_scheduling = false;
}
}
/* Phase 1 of thread scheduling is to gather metrics on the existing threads */
static void
_reactors_scheduler_gather_metrics(void *arg1, void *arg2)
{
struct spdk_scheduler_core_info *core_info;
struct spdk_lw_thread *lw_thread;
struct spdk_reactor *reactor;
struct spdk_event *evt;
uint32_t next_core;
uint32_t i;
reactor = spdk_reactor_get(spdk_env_get_current_core());
assert(reactor != NULL);
reactor->flags.is_scheduling = true;
core_info = &g_core_infos[reactor->lcore];
core_info->lcore = reactor->lcore;
core_info->core_idle_tsc = reactor->idle_tsc;
core_info->core_busy_tsc = reactor->busy_tsc;
core_info->interrupt_mode = reactor->in_interrupt;
SPDK_DEBUGLOG(reactor, "Gathering metrics on %u\n", reactor->lcore);
free(core_info->threads);
core_info->threads = NULL;
i = 0;
TAILQ_FOREACH(lw_thread, &reactor->threads, link) {
_init_thread_stats(reactor, lw_thread);
i++;
}
core_info->threads_count = i;
if (core_info->threads_count > 0) {
core_info->threads = calloc(core_info->threads_count, sizeof(struct spdk_lw_thread *));
if (core_info->threads == NULL) {
SPDK_ERRLOG("Failed to allocate memory when gathering metrics on %u\n", reactor->lcore);
/* Cancel this round of schedule work */
evt = spdk_event_allocate(g_scheduling_reactor->lcore, _reactors_scheduler_cancel, NULL, NULL);
spdk_event_call(evt);
return;
}
i = 0;
TAILQ_FOREACH(lw_thread, &reactor->threads, link) {
core_info->threads[i] = lw_thread;
_spdk_lw_thread_get_current_stats(lw_thread, &lw_thread->snapshot_stats);
i++;
}
}
next_core = spdk_env_get_next_core(reactor->lcore);
if (next_core == UINT32_MAX) {
next_core = spdk_env_get_first_core();
}
/* If we've looped back around to the scheduler thread, move to the next phase */
if (next_core == g_scheduling_reactor->lcore) {
/* Phase 2 of scheduling is rebalancing - deciding which threads to move where */
evt = spdk_event_allocate(next_core, _reactors_scheduler_balance, NULL, NULL);
spdk_event_call(evt);
return;
}
evt = spdk_event_allocate(next_core, _reactors_scheduler_gather_metrics, NULL, NULL);
spdk_event_call(evt);
}
static int _reactor_schedule_thread(struct spdk_thread *thread);
static uint64_t g_rusage_period;
static void
_reactor_remove_lw_thread(struct spdk_reactor *reactor, struct spdk_lw_thread *lw_thread)
{
struct spdk_thread *thread = spdk_thread_get_from_ctx(lw_thread);
int efd;
TAILQ_REMOVE(&reactor->threads, lw_thread, link);
assert(reactor->thread_count > 0);
reactor->thread_count--;
/* Operate thread intr if running with full interrupt ability */
if (spdk_interrupt_mode_is_enabled()) {
efd = spdk_thread_get_interrupt_fd(thread);
spdk_fd_group_remove(reactor->fgrp, efd);
}
}
static bool
reactor_post_process_lw_thread(struct spdk_reactor *reactor, struct spdk_lw_thread *lw_thread)
{
struct spdk_thread *thread = spdk_thread_get_from_ctx(lw_thread);
if (spdk_unlikely(lw_thread->resched)) {
lw_thread->resched = false;
_reactor_remove_lw_thread(reactor, lw_thread);
_reactor_schedule_thread(thread);
return true;
}
if (spdk_unlikely(spdk_thread_is_exited(thread) &&
spdk_thread_is_idle(thread))) {
if (reactor->flags.is_scheduling == false) {
_reactor_remove_lw_thread(reactor, lw_thread);
spdk_thread_destroy(thread);
return true;
}
}
return false;
}
static void
reactor_interrupt_run(struct spdk_reactor *reactor)
{
int block_timeout = -1; /* _EPOLL_WAIT_FOREVER */
spdk_fd_group_wait(reactor->fgrp, block_timeout);
}
static void
_reactor_run(struct spdk_reactor *reactor)
{
struct spdk_thread *thread;
struct spdk_lw_thread *lw_thread, *tmp;
uint64_t now;
int rc;
event_queue_run_batch(reactor);
TAILQ_FOREACH_SAFE(lw_thread, &reactor->threads, link, tmp) {
thread = spdk_thread_get_from_ctx(lw_thread);
rc = spdk_thread_poll(thread, 0, reactor->tsc_last);
now = spdk_thread_get_last_tsc(thread);
if (rc == 0) {
reactor->idle_tsc += now - reactor->tsc_last;
} else if (rc > 0) {
reactor->busy_tsc += now - reactor->tsc_last;
}
reactor->tsc_last = now;
reactor_post_process_lw_thread(reactor, lw_thread);
}
}
static int
reactor_run(void *arg)
{
struct spdk_reactor *reactor = arg;
struct spdk_thread *thread;
struct spdk_lw_thread *lw_thread, *tmp;
char thread_name[32];
uint64_t last_sched = 0;
SPDK_NOTICELOG("Reactor started on core %u\n", reactor->lcore);
/* Rename the POSIX thread because the reactor is tied to the POSIX
* thread in the SPDK event library.
*/
snprintf(thread_name, sizeof(thread_name), "reactor_%u", reactor->lcore);
_set_thread_name(thread_name);
reactor->tsc_last = spdk_get_ticks();
while (1) {
/* Execute interrupt process fn if this reactor currently runs in interrupt state */
if (spdk_unlikely(reactor->in_interrupt)) {
reactor_interrupt_run(reactor);
} else {
_reactor_run(reactor);
}
if (g_framework_context_switch_monitor_enabled) {
if ((reactor->last_rusage + g_rusage_period) < reactor->tsc_last) {
get_rusage(reactor);
reactor->last_rusage = reactor->tsc_last;
}
}
if (spdk_unlikely(g_scheduler_period > 0 &&
(reactor->tsc_last - last_sched) > g_scheduler_period &&
reactor == g_scheduling_reactor &&
!reactor->flags.is_scheduling)) {
if (spdk_unlikely(g_scheduler != g_new_scheduler)) {
if (g_scheduler->deinit != NULL) {
g_scheduler->deinit(&g_governor);
}
g_scheduler = g_new_scheduler;
}
if (spdk_unlikely(g_scheduler->balance != NULL)) {
last_sched = reactor->tsc_last;
_reactors_scheduler_gather_metrics(NULL, NULL);
}
}
if (g_reactor_state != SPDK_REACTOR_STATE_RUNNING) {
break;
}
}
TAILQ_FOREACH(lw_thread, &reactor->threads, link) {
thread = spdk_thread_get_from_ctx(lw_thread);
spdk_set_thread(thread);
spdk_thread_exit(thread);
}
while (!TAILQ_EMPTY(&reactor->threads)) {
TAILQ_FOREACH_SAFE(lw_thread, &reactor->threads, link, tmp) {
thread = spdk_thread_get_from_ctx(lw_thread);
spdk_set_thread(thread);
if (spdk_thread_is_exited(thread)) {
_reactor_remove_lw_thread(reactor, lw_thread);
spdk_thread_destroy(thread);
} else {
spdk_thread_poll(thread, 0, 0);
}
}
}
return 0;
}
int
spdk_app_parse_core_mask(const char *mask, struct spdk_cpuset *cpumask)
{