-
Notifications
You must be signed in to change notification settings - Fork 206
/
Copy pathmachine_routines_common.c
1440 lines (1276 loc) · 45.5 KB
/
machine_routines_common.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
/*
* Copyright (c) 2007-2021 Apple Inc. All rights reserved.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_START@
*
* This file contains Original Code and/or Modifications of Original Code
* as defined in and that are subject to the Apple Public Source License
* Version 2.0 (the 'License'). You may not use this file except in
* compliance with the License. The rights granted to you under the License
* may not be used to create, or enable the creation or redistribution of,
* unlawful or unlicensed copies of an Apple operating system, or to
* circumvent, violate, or enable the circumvention or violation of, any
* terms of an Apple operating system software license agreement.
*
* Please obtain a copy of the License at
* http://www.opensource.apple.com/apsl/ and read it before using this file.
*
* The Original Code and all software distributed under the License are
* distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
* EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
* INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
* Please see the License for the specific language governing rights and
* limitations under the License.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_END@
*/
#include <arm/machine_cpu.h>
#include <arm/cpu_internal.h>
#include <arm/cpuid.h>
#include <arm/cpuid_internal.h>
#include <arm/cpu_data.h>
#include <arm/cpu_data_internal.h>
#include <arm/misc_protos.h>
#include <arm/machdep_call.h>
#include <arm/machine_routines.h>
#include <arm/rtclock.h>
#include <kern/machine.h>
#include <kern/thread.h>
#include <kern/thread_group.h>
#include <kern/policy_internal.h>
#include <kern/sched_hygiene.h>
#include <kern/startup.h>
#include <machine/config.h>
#include <machine/atomic.h>
#include <pexpert/pexpert.h>
#include <pexpert/device_tree.h>
#if MONOTONIC
#include <kern/monotonic.h>
#include <machine/monotonic.h>
#endif /* MONOTONIC */
#include <mach/machine.h>
#include <mach/machine/sdt.h>
#if !HAS_CONTINUOUS_HWCLOCK
extern uint64_t mach_absolutetime_asleep;
#else
extern uint64_t wake_abstime;
static uint64_t wake_conttime = UINT64_MAX;
#endif
extern volatile uint32_t debug_enabled;
extern _Atomic unsigned int cluster_type_num_active_cpus[MAX_CPU_TYPES];
const char *cluster_type_names[MAX_CPU_TYPES] = {
[CLUSTER_TYPE_SMP] = "Standard",
[CLUSTER_TYPE_P] = "Performance",
[CLUSTER_TYPE_E] = "Efficiency",
};
static int max_cpus_initialized = 0;
#define MAX_CPUS_SET 0x1
#define MAX_CPUS_WAIT 0x2
LCK_GRP_DECLARE(max_cpus_grp, "max_cpus");
LCK_MTX_DECLARE(max_cpus_lock, &max_cpus_grp);
uint32_t lockdown_done = 0;
boolean_t is_clock_configured = FALSE;
static void
sched_perfcontrol_oncore_default(perfcontrol_state_t new_thread_state __unused, going_on_core_t on __unused)
{
}
static void
sched_perfcontrol_switch_default(perfcontrol_state_t old_thread_state __unused, perfcontrol_state_t new_thread_state __unused)
{
}
static void
sched_perfcontrol_offcore_default(perfcontrol_state_t old_thread_state __unused, going_off_core_t off __unused, boolean_t thread_terminating __unused)
{
}
static void
sched_perfcontrol_thread_group_default(thread_group_data_t data __unused)
{
}
static void
sched_perfcontrol_max_runnable_latency_default(perfcontrol_max_runnable_latency_t latencies __unused)
{
}
static void
sched_perfcontrol_work_interval_notify_default(perfcontrol_state_t thread_state __unused,
perfcontrol_work_interval_t work_interval __unused)
{
}
static void
sched_perfcontrol_work_interval_ctl_default(perfcontrol_state_t thread_state __unused,
perfcontrol_work_interval_instance_t instance __unused)
{
}
static void
sched_perfcontrol_deadline_passed_default(__unused uint64_t deadline)
{
}
static void
sched_perfcontrol_csw_default(
__unused perfcontrol_event event, __unused uint32_t cpu_id, __unused uint64_t timestamp,
__unused uint32_t flags, __unused struct perfcontrol_thread_data *offcore,
__unused struct perfcontrol_thread_data *oncore,
__unused struct perfcontrol_cpu_counters *cpu_counters, __unused void *unused)
{
}
static void
sched_perfcontrol_state_update_default(
__unused perfcontrol_event event, __unused uint32_t cpu_id, __unused uint64_t timestamp,
__unused uint32_t flags, __unused struct perfcontrol_thread_data *thr_data,
__unused void *unused)
{
}
static void
sched_perfcontrol_thread_group_blocked_default(
__unused thread_group_data_t blocked_tg, __unused thread_group_data_t blocking_tg,
__unused uint32_t flags, __unused perfcontrol_state_t blocked_thr_state)
{
}
static void
sched_perfcontrol_thread_group_unblocked_default(
__unused thread_group_data_t unblocked_tg, __unused thread_group_data_t unblocking_tg,
__unused uint32_t flags, __unused perfcontrol_state_t unblocked_thr_state)
{
}
sched_perfcontrol_offcore_t sched_perfcontrol_offcore = sched_perfcontrol_offcore_default;
sched_perfcontrol_context_switch_t sched_perfcontrol_switch = sched_perfcontrol_switch_default;
sched_perfcontrol_oncore_t sched_perfcontrol_oncore = sched_perfcontrol_oncore_default;
sched_perfcontrol_thread_group_init_t sched_perfcontrol_thread_group_init = sched_perfcontrol_thread_group_default;
sched_perfcontrol_thread_group_deinit_t sched_perfcontrol_thread_group_deinit = sched_perfcontrol_thread_group_default;
sched_perfcontrol_thread_group_flags_update_t sched_perfcontrol_thread_group_flags_update = sched_perfcontrol_thread_group_default;
sched_perfcontrol_max_runnable_latency_t sched_perfcontrol_max_runnable_latency = sched_perfcontrol_max_runnable_latency_default;
sched_perfcontrol_work_interval_notify_t sched_perfcontrol_work_interval_notify = sched_perfcontrol_work_interval_notify_default;
sched_perfcontrol_work_interval_ctl_t sched_perfcontrol_work_interval_ctl = sched_perfcontrol_work_interval_ctl_default;
sched_perfcontrol_deadline_passed_t sched_perfcontrol_deadline_passed = sched_perfcontrol_deadline_passed_default;
sched_perfcontrol_csw_t sched_perfcontrol_csw = sched_perfcontrol_csw_default;
sched_perfcontrol_state_update_t sched_perfcontrol_state_update = sched_perfcontrol_state_update_default;
sched_perfcontrol_thread_group_blocked_t sched_perfcontrol_thread_group_blocked = sched_perfcontrol_thread_group_blocked_default;
sched_perfcontrol_thread_group_unblocked_t sched_perfcontrol_thread_group_unblocked = sched_perfcontrol_thread_group_unblocked_default;
boolean_t sched_perfcontrol_thread_shared_rsrc_flags_enabled = false;
void
sched_perfcontrol_register_callbacks(sched_perfcontrol_callbacks_t callbacks, unsigned long size_of_state)
{
assert(callbacks == NULL || callbacks->version >= SCHED_PERFCONTROL_CALLBACKS_VERSION_2);
if (size_of_state > sizeof(struct perfcontrol_state)) {
panic("%s: Invalid required state size %lu", __FUNCTION__, size_of_state);
}
if (callbacks) {
#if CONFIG_THREAD_GROUPS
if (callbacks->version >= SCHED_PERFCONTROL_CALLBACKS_VERSION_3) {
if (callbacks->thread_group_init != NULL) {
sched_perfcontrol_thread_group_init = callbacks->thread_group_init;
} else {
sched_perfcontrol_thread_group_init = sched_perfcontrol_thread_group_default;
}
if (callbacks->thread_group_deinit != NULL) {
sched_perfcontrol_thread_group_deinit = callbacks->thread_group_deinit;
} else {
sched_perfcontrol_thread_group_deinit = sched_perfcontrol_thread_group_default;
}
// tell CLPC about existing thread groups
thread_group_resync(TRUE);
}
if (callbacks->version >= SCHED_PERFCONTROL_CALLBACKS_VERSION_6) {
if (callbacks->thread_group_flags_update != NULL) {
sched_perfcontrol_thread_group_flags_update = callbacks->thread_group_flags_update;
} else {
sched_perfcontrol_thread_group_flags_update = sched_perfcontrol_thread_group_default;
}
}
if (callbacks->version >= SCHED_PERFCONTROL_CALLBACKS_VERSION_8) {
if (callbacks->thread_group_blocked != NULL) {
sched_perfcontrol_thread_group_blocked = callbacks->thread_group_blocked;
} else {
sched_perfcontrol_thread_group_blocked = sched_perfcontrol_thread_group_blocked_default;
}
if (callbacks->thread_group_unblocked != NULL) {
sched_perfcontrol_thread_group_unblocked = callbacks->thread_group_unblocked;
} else {
sched_perfcontrol_thread_group_unblocked = sched_perfcontrol_thread_group_unblocked_default;
}
}
#endif
if (callbacks->version >= SCHED_PERFCONTROL_CALLBACKS_VERSION_9) {
sched_perfcontrol_thread_shared_rsrc_flags_enabled = true;
}
if (callbacks->version >= SCHED_PERFCONTROL_CALLBACKS_VERSION_7) {
if (callbacks->work_interval_ctl != NULL) {
sched_perfcontrol_work_interval_ctl = callbacks->work_interval_ctl;
} else {
sched_perfcontrol_work_interval_ctl = sched_perfcontrol_work_interval_ctl_default;
}
}
if (callbacks->version >= SCHED_PERFCONTROL_CALLBACKS_VERSION_5) {
if (callbacks->csw != NULL) {
sched_perfcontrol_csw = callbacks->csw;
} else {
sched_perfcontrol_csw = sched_perfcontrol_csw_default;
}
if (callbacks->state_update != NULL) {
sched_perfcontrol_state_update = callbacks->state_update;
} else {
sched_perfcontrol_state_update = sched_perfcontrol_state_update_default;
}
}
if (callbacks->version >= SCHED_PERFCONTROL_CALLBACKS_VERSION_4) {
if (callbacks->deadline_passed != NULL) {
sched_perfcontrol_deadline_passed = callbacks->deadline_passed;
} else {
sched_perfcontrol_deadline_passed = sched_perfcontrol_deadline_passed_default;
}
}
if (callbacks->offcore != NULL) {
sched_perfcontrol_offcore = callbacks->offcore;
} else {
sched_perfcontrol_offcore = sched_perfcontrol_offcore_default;
}
if (callbacks->context_switch != NULL) {
sched_perfcontrol_switch = callbacks->context_switch;
} else {
sched_perfcontrol_switch = sched_perfcontrol_switch_default;
}
if (callbacks->oncore != NULL) {
sched_perfcontrol_oncore = callbacks->oncore;
} else {
sched_perfcontrol_oncore = sched_perfcontrol_oncore_default;
}
if (callbacks->max_runnable_latency != NULL) {
sched_perfcontrol_max_runnable_latency = callbacks->max_runnable_latency;
} else {
sched_perfcontrol_max_runnable_latency = sched_perfcontrol_max_runnable_latency_default;
}
if (callbacks->work_interval_notify != NULL) {
sched_perfcontrol_work_interval_notify = callbacks->work_interval_notify;
} else {
sched_perfcontrol_work_interval_notify = sched_perfcontrol_work_interval_notify_default;
}
} else {
/* reset to defaults */
#if CONFIG_THREAD_GROUPS
thread_group_resync(FALSE);
#endif
sched_perfcontrol_offcore = sched_perfcontrol_offcore_default;
sched_perfcontrol_switch = sched_perfcontrol_switch_default;
sched_perfcontrol_oncore = sched_perfcontrol_oncore_default;
sched_perfcontrol_thread_group_init = sched_perfcontrol_thread_group_default;
sched_perfcontrol_thread_group_deinit = sched_perfcontrol_thread_group_default;
sched_perfcontrol_thread_group_flags_update = sched_perfcontrol_thread_group_default;
sched_perfcontrol_max_runnable_latency = sched_perfcontrol_max_runnable_latency_default;
sched_perfcontrol_work_interval_notify = sched_perfcontrol_work_interval_notify_default;
sched_perfcontrol_work_interval_ctl = sched_perfcontrol_work_interval_ctl_default;
sched_perfcontrol_csw = sched_perfcontrol_csw_default;
sched_perfcontrol_state_update = sched_perfcontrol_state_update_default;
sched_perfcontrol_thread_group_blocked = sched_perfcontrol_thread_group_blocked_default;
sched_perfcontrol_thread_group_unblocked = sched_perfcontrol_thread_group_unblocked_default;
}
}
static void
machine_switch_populate_perfcontrol_thread_data(struct perfcontrol_thread_data *data,
thread_t thread,
uint64_t same_pri_latency)
{
bzero(data, sizeof(struct perfcontrol_thread_data));
data->perfctl_class = thread_get_perfcontrol_class(thread);
data->energy_estimate_nj = 0;
data->thread_id = thread->thread_id;
#if CONFIG_THREAD_GROUPS
struct thread_group *tg = thread_group_get(thread);
data->thread_group_id = thread_group_get_id(tg);
data->thread_group_data = thread_group_get_machine_data(tg);
#endif
data->scheduling_latency_at_same_basepri = same_pri_latency;
data->perfctl_state = FIND_PERFCONTROL_STATE(thread);
}
static void
machine_switch_populate_perfcontrol_cpu_counters(struct perfcontrol_cpu_counters *cpu_counters)
{
#if MONOTONIC
mt_perfcontrol(&cpu_counters->instructions, &cpu_counters->cycles);
#else /* MONOTONIC */
cpu_counters->instructions = 0;
cpu_counters->cycles = 0;
#endif /* !MONOTONIC */
}
int perfcontrol_callout_stats_enabled = 0;
static _Atomic uint64_t perfcontrol_callout_stats[PERFCONTROL_CALLOUT_MAX][PERFCONTROL_STAT_MAX];
static _Atomic uint64_t perfcontrol_callout_count[PERFCONTROL_CALLOUT_MAX];
#if MONOTONIC
static inline
bool
perfcontrol_callout_counters_begin(uint64_t *counters)
{
if (!perfcontrol_callout_stats_enabled) {
return false;
}
mt_fixed_counts(counters);
return true;
}
static inline
void
perfcontrol_callout_counters_end(uint64_t *start_counters,
perfcontrol_callout_type_t type)
{
uint64_t end_counters[MT_CORE_NFIXED];
mt_fixed_counts(end_counters);
os_atomic_add(&perfcontrol_callout_stats[type][PERFCONTROL_STAT_CYCLES],
end_counters[MT_CORE_CYCLES] - start_counters[MT_CORE_CYCLES], relaxed);
os_atomic_add(&perfcontrol_callout_stats[type][PERFCONTROL_STAT_INSTRS],
end_counters[MT_CORE_INSTRS] - start_counters[MT_CORE_INSTRS], relaxed);
os_atomic_inc(&perfcontrol_callout_count[type], relaxed);
}
#endif /* MONOTONIC */
uint64_t
perfcontrol_callout_stat_avg(perfcontrol_callout_type_t type,
perfcontrol_callout_stat_t stat)
{
if (!perfcontrol_callout_stats_enabled) {
return 0;
}
return os_atomic_load_wide(&perfcontrol_callout_stats[type][stat], relaxed) /
os_atomic_load_wide(&perfcontrol_callout_count[type], relaxed);
}
#if CONFIG_SCHED_EDGE
/*
* The Edge scheduler allows the performance controller to update properties about the
* threads as part of the callouts. These properties typically include shared cluster
* resource usage. This allows the scheduler to manage specific threads within the
* workload more optimally.
*/
static void
sched_perfcontrol_thread_flags_update(thread_t thread,
struct perfcontrol_thread_data *thread_data,
shared_rsrc_policy_agent_t agent)
{
kern_return_t kr = KERN_SUCCESS;
if (thread_data->thread_flags_mask & PERFCTL_THREAD_FLAGS_MASK_CLUSTER_SHARED_RSRC_RR) {
if (thread_data->thread_flags & PERFCTL_THREAD_FLAGS_MASK_CLUSTER_SHARED_RSRC_RR) {
kr = thread_shared_rsrc_policy_set(thread, 0, CLUSTER_SHARED_RSRC_TYPE_RR, agent);
} else {
kr = thread_shared_rsrc_policy_clear(thread, CLUSTER_SHARED_RSRC_TYPE_RR, agent);
}
}
if (thread_data->thread_flags_mask & PERFCTL_THREAD_FLAGS_MASK_CLUSTER_SHARED_RSRC_NATIVE_FIRST) {
if (thread_data->thread_flags & PERFCTL_THREAD_FLAGS_MASK_CLUSTER_SHARED_RSRC_NATIVE_FIRST) {
kr = thread_shared_rsrc_policy_set(thread, 0, CLUSTER_SHARED_RSRC_TYPE_NATIVE_FIRST, agent);
} else {
kr = thread_shared_rsrc_policy_clear(thread, CLUSTER_SHARED_RSRC_TYPE_NATIVE_FIRST, agent);
}
}
/*
* The thread_shared_rsrc_policy_* routines only fail if the performance controller is
* attempting to double set/clear a policy on the thread.
*/
assert(kr == KERN_SUCCESS);
}
#endif /* CONFIG_SCHED_EDGE */
void
machine_switch_perfcontrol_context(perfcontrol_event event,
uint64_t timestamp,
uint32_t flags,
uint64_t new_thread_same_pri_latency,
thread_t old,
thread_t new)
{
if (sched_perfcontrol_switch != sched_perfcontrol_switch_default) {
perfcontrol_state_t old_perfcontrol_state = FIND_PERFCONTROL_STATE(old);
perfcontrol_state_t new_perfcontrol_state = FIND_PERFCONTROL_STATE(new);
sched_perfcontrol_switch(old_perfcontrol_state, new_perfcontrol_state);
}
if (sched_perfcontrol_csw != sched_perfcontrol_csw_default) {
uint32_t cpu_id = (uint32_t)cpu_number();
struct perfcontrol_cpu_counters cpu_counters;
struct perfcontrol_thread_data offcore, oncore;
machine_switch_populate_perfcontrol_thread_data(&offcore, old, 0);
machine_switch_populate_perfcontrol_thread_data(&oncore, new,
new_thread_same_pri_latency);
machine_switch_populate_perfcontrol_cpu_counters(&cpu_counters);
#if MONOTONIC
uint64_t counters[MT_CORE_NFIXED];
bool ctrs_enabled = perfcontrol_callout_counters_begin(counters);
#endif /* MONOTONIC */
sched_perfcontrol_csw(event, cpu_id, timestamp, flags,
&offcore, &oncore, &cpu_counters, NULL);
#if MONOTONIC
if (ctrs_enabled) {
perfcontrol_callout_counters_end(counters, PERFCONTROL_CALLOUT_CONTEXT);
}
#endif /* MONOTONIC */
recount_add_energy(old, get_threadtask(old),
offcore.energy_estimate_nj);
#if CONFIG_SCHED_EDGE
if (sched_perfcontrol_thread_shared_rsrc_flags_enabled) {
sched_perfcontrol_thread_flags_update(old, &offcore, SHARED_RSRC_POLICY_AGENT_PERFCTL_CSW);
}
#endif /* CONFIG_SCHED_EDGE */
}
}
void
machine_switch_perfcontrol_state_update(perfcontrol_event event,
uint64_t timestamp,
uint32_t flags,
thread_t thread)
{
if (sched_perfcontrol_state_update == sched_perfcontrol_state_update_default) {
return;
}
uint32_t cpu_id = (uint32_t)cpu_number();
struct perfcontrol_thread_data data;
machine_switch_populate_perfcontrol_thread_data(&data, thread, 0);
#if MONOTONIC
uint64_t counters[MT_CORE_NFIXED];
bool ctrs_enabled = perfcontrol_callout_counters_begin(counters);
#endif /* MONOTONIC */
sched_perfcontrol_state_update(event, cpu_id, timestamp, flags,
&data, NULL);
#if MONOTONIC
if (ctrs_enabled) {
perfcontrol_callout_counters_end(counters, PERFCONTROL_CALLOUT_STATE_UPDATE);
}
#endif /* MONOTONIC */
#if CONFIG_PERVASIVE_ENERGY
recount_add_energy(thread, get_threadtask(thread), data.energy_estimate_nj);
#endif /* CONFIG_PERVASIVE_ENERGY */
#if CONFIG_SCHED_EDGE
if (sched_perfcontrol_thread_shared_rsrc_flags_enabled && (event == QUANTUM_EXPIRY)) {
sched_perfcontrol_thread_flags_update(thread, &data, SHARED_RSRC_POLICY_AGENT_PERFCTL_QUANTUM);
} else {
assert(data.thread_flags_mask == 0);
}
#endif /* CONFIG_SCHED_EDGE */
}
void
machine_thread_going_on_core(thread_t new_thread,
thread_urgency_t urgency,
uint64_t sched_latency,
uint64_t same_pri_latency,
uint64_t timestamp)
{
if (sched_perfcontrol_oncore == sched_perfcontrol_oncore_default) {
return;
}
struct going_on_core on_core;
perfcontrol_state_t state = FIND_PERFCONTROL_STATE(new_thread);
on_core.thread_id = new_thread->thread_id;
on_core.energy_estimate_nj = 0;
on_core.qos_class = (uint16_t)proc_get_effective_thread_policy(new_thread, TASK_POLICY_QOS);
on_core.urgency = (uint16_t)urgency;
on_core.is_32_bit = thread_is_64bit_data(new_thread) ? FALSE : TRUE;
on_core.is_kernel_thread = get_threadtask(new_thread) == kernel_task;
#if CONFIG_THREAD_GROUPS
struct thread_group *tg = thread_group_get(new_thread);
on_core.thread_group_id = thread_group_get_id(tg);
on_core.thread_group_data = thread_group_get_machine_data(tg);
#endif
on_core.scheduling_latency = sched_latency;
on_core.start_time = timestamp;
on_core.scheduling_latency_at_same_basepri = same_pri_latency;
#if MONOTONIC
uint64_t counters[MT_CORE_NFIXED];
bool ctrs_enabled = perfcontrol_callout_counters_begin(counters);
#endif /* MONOTONIC */
sched_perfcontrol_oncore(state, &on_core);
#if MONOTONIC
if (ctrs_enabled) {
perfcontrol_callout_counters_end(counters, PERFCONTROL_CALLOUT_ON_CORE);
}
#endif /* MONOTONIC */
}
void
machine_thread_going_off_core(thread_t old_thread, boolean_t thread_terminating,
uint64_t last_dispatch, __unused boolean_t thread_runnable)
{
if (sched_perfcontrol_offcore == sched_perfcontrol_offcore_default) {
return;
}
struct going_off_core off_core;
perfcontrol_state_t state = FIND_PERFCONTROL_STATE(old_thread);
off_core.thread_id = old_thread->thread_id;
off_core.energy_estimate_nj = 0;
off_core.end_time = last_dispatch;
#if CONFIG_THREAD_GROUPS
struct thread_group *tg = thread_group_get(old_thread);
off_core.thread_group_id = thread_group_get_id(tg);
off_core.thread_group_data = thread_group_get_machine_data(tg);
#endif
#if MONOTONIC
uint64_t counters[MT_CORE_NFIXED];
bool ctrs_enabled = perfcontrol_callout_counters_begin(counters);
#endif /* MONOTONIC */
sched_perfcontrol_offcore(state, &off_core, thread_terminating);
#if MONOTONIC
if (ctrs_enabled) {
perfcontrol_callout_counters_end(counters, PERFCONTROL_CALLOUT_OFF_CORE);
}
#endif /* MONOTONIC */
}
#if CONFIG_THREAD_GROUPS
void
machine_thread_group_init(struct thread_group *tg)
{
if (sched_perfcontrol_thread_group_init == sched_perfcontrol_thread_group_default) {
return;
}
struct thread_group_data data;
data.thread_group_id = thread_group_get_id(tg);
data.thread_group_data = thread_group_get_machine_data(tg);
data.thread_group_size = thread_group_machine_data_size();
data.thread_group_flags = thread_group_get_flags(tg);
sched_perfcontrol_thread_group_init(&data);
}
void
machine_thread_group_deinit(struct thread_group *tg)
{
if (sched_perfcontrol_thread_group_deinit == sched_perfcontrol_thread_group_default) {
return;
}
struct thread_group_data data;
data.thread_group_id = thread_group_get_id(tg);
data.thread_group_data = thread_group_get_machine_data(tg);
data.thread_group_size = thread_group_machine_data_size();
data.thread_group_flags = thread_group_get_flags(tg);
sched_perfcontrol_thread_group_deinit(&data);
}
void
machine_thread_group_flags_update(struct thread_group *tg, uint32_t flags)
{
if (sched_perfcontrol_thread_group_flags_update == sched_perfcontrol_thread_group_default) {
return;
}
struct thread_group_data data;
data.thread_group_id = thread_group_get_id(tg);
data.thread_group_data = thread_group_get_machine_data(tg);
data.thread_group_size = thread_group_machine_data_size();
data.thread_group_flags = flags;
sched_perfcontrol_thread_group_flags_update(&data);
}
void
machine_thread_group_blocked(struct thread_group *blocked_tg,
struct thread_group *blocking_tg,
uint32_t flags,
thread_t blocked_thread)
{
if (sched_perfcontrol_thread_group_blocked == sched_perfcontrol_thread_group_blocked_default) {
return;
}
spl_t s = splsched();
perfcontrol_state_t state = FIND_PERFCONTROL_STATE(blocked_thread);
struct thread_group_data blocked_data;
assert(blocked_tg != NULL);
blocked_data.thread_group_id = thread_group_get_id(blocked_tg);
blocked_data.thread_group_data = thread_group_get_machine_data(blocked_tg);
blocked_data.thread_group_size = thread_group_machine_data_size();
if (blocking_tg == NULL) {
/*
* For special cases such as the render server, the blocking TG is a
* well known TG. Only in that case, the blocking_tg should be NULL.
*/
assert(flags & PERFCONTROL_CALLOUT_BLOCKING_TG_RENDER_SERVER);
sched_perfcontrol_thread_group_blocked(&blocked_data, NULL, flags, state);
} else {
struct thread_group_data blocking_data;
blocking_data.thread_group_id = thread_group_get_id(blocking_tg);
blocking_data.thread_group_data = thread_group_get_machine_data(blocking_tg);
blocking_data.thread_group_size = thread_group_machine_data_size();
sched_perfcontrol_thread_group_blocked(&blocked_data, &blocking_data, flags, state);
}
KDBG(MACHDBG_CODE(DBG_MACH_THREAD_GROUP, MACH_THREAD_GROUP_BLOCK) | DBG_FUNC_START,
thread_tid(blocked_thread), thread_group_get_id(blocked_tg),
blocking_tg ? thread_group_get_id(blocking_tg) : THREAD_GROUP_INVALID,
flags);
splx(s);
}
void
machine_thread_group_unblocked(struct thread_group *unblocked_tg,
struct thread_group *unblocking_tg,
uint32_t flags,
thread_t unblocked_thread)
{
if (sched_perfcontrol_thread_group_unblocked == sched_perfcontrol_thread_group_unblocked_default) {
return;
}
spl_t s = splsched();
perfcontrol_state_t state = FIND_PERFCONTROL_STATE(unblocked_thread);
struct thread_group_data unblocked_data;
assert(unblocked_tg != NULL);
unblocked_data.thread_group_id = thread_group_get_id(unblocked_tg);
unblocked_data.thread_group_data = thread_group_get_machine_data(unblocked_tg);
unblocked_data.thread_group_size = thread_group_machine_data_size();
if (unblocking_tg == NULL) {
/*
* For special cases such as the render server, the unblocking TG is a
* well known TG. Only in that case, the unblocking_tg should be NULL.
*/
assert(flags & PERFCONTROL_CALLOUT_BLOCKING_TG_RENDER_SERVER);
sched_perfcontrol_thread_group_unblocked(&unblocked_data, NULL, flags, state);
} else {
struct thread_group_data unblocking_data;
unblocking_data.thread_group_id = thread_group_get_id(unblocking_tg);
unblocking_data.thread_group_data = thread_group_get_machine_data(unblocking_tg);
unblocking_data.thread_group_size = thread_group_machine_data_size();
sched_perfcontrol_thread_group_unblocked(&unblocked_data, &unblocking_data, flags, state);
}
KDBG(MACHDBG_CODE(DBG_MACH_THREAD_GROUP, MACH_THREAD_GROUP_BLOCK) | DBG_FUNC_END,
thread_tid(unblocked_thread), thread_group_get_id(unblocked_tg),
unblocking_tg ? thread_group_get_id(unblocking_tg) : THREAD_GROUP_INVALID,
flags);
splx(s);
}
#endif /* CONFIG_THREAD_GROUPS */
void
machine_max_runnable_latency(uint64_t bg_max_latency,
uint64_t default_max_latency,
uint64_t realtime_max_latency)
{
if (sched_perfcontrol_max_runnable_latency == sched_perfcontrol_max_runnable_latency_default) {
return;
}
struct perfcontrol_max_runnable_latency latencies = {
.max_scheduling_latencies = {
[THREAD_URGENCY_NONE] = 0,
[THREAD_URGENCY_BACKGROUND] = bg_max_latency,
[THREAD_URGENCY_NORMAL] = default_max_latency,
[THREAD_URGENCY_REAL_TIME] = realtime_max_latency
}
};
sched_perfcontrol_max_runnable_latency(&latencies);
}
void
machine_work_interval_notify(thread_t thread,
struct kern_work_interval_args* kwi_args)
{
if (sched_perfcontrol_work_interval_notify == sched_perfcontrol_work_interval_notify_default) {
return;
}
perfcontrol_state_t state = FIND_PERFCONTROL_STATE(thread);
struct perfcontrol_work_interval work_interval = {
.thread_id = thread->thread_id,
.qos_class = (uint16_t)proc_get_effective_thread_policy(thread, TASK_POLICY_QOS),
.urgency = kwi_args->urgency,
.flags = kwi_args->notify_flags,
.work_interval_id = kwi_args->work_interval_id,
.start = kwi_args->start,
.finish = kwi_args->finish,
.deadline = kwi_args->deadline,
.next_start = kwi_args->next_start,
.create_flags = kwi_args->create_flags,
};
#if CONFIG_THREAD_GROUPS
struct thread_group *tg;
tg = thread_group_get(thread);
work_interval.thread_group_id = thread_group_get_id(tg);
work_interval.thread_group_data = thread_group_get_machine_data(tg);
#endif
sched_perfcontrol_work_interval_notify(state, &work_interval);
}
void
machine_perfcontrol_deadline_passed(uint64_t deadline)
{
if (sched_perfcontrol_deadline_passed != sched_perfcontrol_deadline_passed_default) {
sched_perfcontrol_deadline_passed(deadline);
}
}
#if SCHED_HYGIENE_DEBUG
__options_decl(int_mask_hygiene_flags_t, uint8_t, {
INT_MASK_BASE = 0x00,
INT_MASK_FROM_HANDLER = 0x01,
INT_MASK_IS_STACKSHOT = 0x02,
});
/*
* ml_spin_debug_reset()
* Reset the timestamp on a thread that has been unscheduled
* to avoid false alarms. Alarm will go off if interrupts are held
* disabled for too long, starting from now.
*
* Call ml_get_timebase() directly to prevent extra overhead on newer
* platforms that's enabled in DEVELOPMENT kernel configurations.
*/
void
ml_spin_debug_reset(thread_t thread)
{
if (thread->machine.intmask_timestamp) {
thread->machine.intmask_timestamp = ml_get_sched_hygiene_timebase();
INTERRUPT_MASKED_DEBUG_CAPTURE_PMC(thread);
}
}
/*
* ml_spin_debug_clear()
* Clear the timestamp and cycle/instruction counts on a thread that
* has been unscheduled to avoid false alarms
*/
void
ml_spin_debug_clear(thread_t thread)
{
thread->machine.intmask_timestamp = 0;
thread->machine.intmask_cycles = 0;
thread->machine.intmask_instr = 0;
}
/*
* ml_spin_debug_clear_self()
* Clear the timestamp on the current thread to prevent
* false alarms
*/
void
ml_spin_debug_clear_self(void)
{
ml_spin_debug_clear(current_thread());
}
#ifndef KASAN
/*
* Get a character representing the provided thread's kind of CPU.
*/
#if !MONOTONIC
__unused
#endif // !MONOTONIC
static char
__ml_interrupts_disabled_cpu_kind(thread_t thread)
{
#if __AMP__
processor_t processor = thread->last_processor;
if (!processor) {
return '!';
}
switch (processor->processor_set->pset_cluster_type) {
case PSET_AMP_P:
return 'P';
case PSET_AMP_E:
return 'E';
default:
return '?';
}
#else // __AMP__
#pragma unused(thread)
return '-';
#endif // !__AMP__
}
#define EXTRA_INFO_STRING_SIZE 256
#define LOW_FREQ_THRESHOLD_MHZ 500
#define HIGH_CPI_THRESHOLD 3
static void
__ml_trigger_interrupts_disabled_handle(thread_t thread, uint64_t start, uint64_t now, uint64_t timeout, int_mask_hygiene_flags_t flags)
{
mach_timebase_info_data_t timebase;
clock_timebase_info(&timebase);
bool is_int_handler = flags & INT_MASK_FROM_HANDLER;
bool is_stackshot = flags & INT_MASK_IS_STACKSHOT;
const uint64_t time_elapsed = now - start;
const uint64_t time_elapsed_ns = (time_elapsed * timebase.numer) / timebase.denom;
uint64_t current_cycles = 0, current_instrs = 0;
#if MONOTONIC
if (sched_hygiene_debug_pmc) {
mt_cur_cpu_cycles_instrs_speculative(¤t_cycles, ¤t_instrs);
}
#endif // MONOTONIC
const uint64_t cycles_elapsed = current_cycles - thread->machine.intmask_cycles;
const uint64_t instrs_elapsed = current_instrs - thread->machine.intmask_instr;
if (interrupt_masked_debug_mode == SCHED_HYGIENE_MODE_PANIC) {
const uint64_t timeout_ns = ((timeout * debug_cpu_performance_degradation_factor) * timebase.numer) / timebase.denom;
char extra_info_string[EXTRA_INFO_STRING_SIZE] = { '\0' };
#if MONOTONIC
if (sched_hygiene_debug_pmc) {
const uint64_t time_elapsed_us = time_elapsed_ns / 1000;
const uint64_t average_freq_mhz = cycles_elapsed / time_elapsed_us;
const uint64_t average_cpi_whole = cycles_elapsed / instrs_elapsed;
const uint64_t average_cpi_fractional = ((cycles_elapsed * 100) / instrs_elapsed) % 100;
bool high_cpi = average_cpi_whole >= HIGH_CPI_THRESHOLD;
char core_kind = __ml_interrupts_disabled_cpu_kind(thread);
bool low_mhz = average_freq_mhz < LOW_FREQ_THRESHOLD_MHZ;
snprintf(extra_info_string, EXTRA_INFO_STRING_SIZE,
", %sfreq = %llu MHz, %sCPI = %llu.%llu, CPU kind = %c",
low_mhz ? "low " : "",
average_freq_mhz,
high_cpi ? "high " : "",
average_cpi_whole,
average_cpi_fractional,
core_kind);
}
#endif // MONOTONIC
if (is_int_handler) {
panic("Processing of an interrupt (type = %u, handler address = %p, vector = %p) "
"took %llu nanoseconds (start = %llu, now = %llu, timeout = %llu ns%s)",
thread->machine.int_type, (void *)thread->machine.int_handler_addr, (void *)thread->machine.int_vector,
time_elapsed_ns, start, now, timeout_ns, extra_info_string);
} else {
panic("%s for %llu nanoseconds (start = %llu, now = %llu, timeout = %llu ns%s)",
is_stackshot ? "Stackshot disabled interrupts" : "Interrupts held disabled",
time_elapsed_ns, start, now, timeout_ns, extra_info_string);
}
} else if (interrupt_masked_debug_mode == SCHED_HYGIENE_MODE_TRACE) {
if (is_int_handler) {
static const uint32_t interrupt_handled_dbgid =
MACHDBG_CODE(DBG_MACH_SCHED, MACH_INT_HANDLED_EXPIRED);
DTRACE_SCHED3(interrupt_handled_dbgid, uint64_t, time_elapsed,
uint64_t, cycles_elapsed, uint64_t, instrs_elapsed);
KDBG(interrupt_handled_dbgid, time_elapsed,
cycles_elapsed, instrs_elapsed);
} else {
static const uint32_t interrupt_masked_dbgid =
MACHDBG_CODE(DBG_MACH_SCHED, MACH_INT_MASKED_EXPIRED);
DTRACE_SCHED3(interrupt_masked_dbgid, uint64_t, time_elapsed,
uint64_t, cycles_elapsed, uint64_t, instrs_elapsed);
KDBG(interrupt_masked_dbgid, time_elapsed,
cycles_elapsed, instrs_elapsed);
}
}
}
#endif // !defined(KASAN)
static inline void
__ml_handle_interrupts_disabled_duration(thread_t thread, uint64_t timeout, bool is_int_handler)
{
if (timeout == 0) {
return; // 0 means timeout disabled.
}
uint64_t start = is_int_handler ? thread->machine.inthandler_timestamp : thread->machine.intmask_timestamp;
if (start != 0) {
uint64_t now = ml_get_sched_hygiene_timebase();
if (interrupt_masked_debug_mode &&
((now - start) > timeout * debug_cpu_performance_degradation_factor) &&
!thread->machine.inthandler_abandon) {
/*
* Disable the actual panic for KASAN due to the overhead of KASAN itself, leave the rest of the
* mechanism enabled so that KASAN can catch any bugs in the mechanism itself.
*/
#ifndef KASAN
__ml_trigger_interrupts_disabled_handle(thread, start, now, timeout, is_int_handler);
#endif
}
if (is_int_handler) {
uint64_t const duration = now - start;
/*
* No need for an atomic add, the only thread modifying
* this is ourselves. Other threads querying will just see
* either the old or the new value. (This will also just
* resolve to regular loads and stores on relevant
* platforms.)
*/
uint64_t const old_duration = os_atomic_load_wide(&thread->machine.int_time_mt, relaxed);
os_atomic_store_wide(&thread->machine.int_time_mt, old_duration + duration, relaxed);
}
}
}
void
ml_handle_interrupts_disabled_duration(thread_t thread)
{
__ml_handle_interrupts_disabled_duration(thread, os_atomic_load(&interrupt_masked_timeout, relaxed), INT_MASK_BASE);
}
void
ml_handle_stackshot_interrupt_disabled_duration(thread_t thread)
{
/* Use MAX() to let the user bump the timeout further if needed */
uint64_t stackshot_timeout = os_atomic_load(&stackshot_interrupt_masked_timeout, relaxed);
uint64_t normal_timeout = os_atomic_load(&interrupt_masked_timeout, relaxed);
uint64_t timeout = MAX(stackshot_timeout, normal_timeout);
__ml_handle_interrupts_disabled_duration(thread, timeout, INT_MASK_IS_STACKSHOT);
}
void
ml_handle_interrupt_handler_duration(thread_t thread)
{
__ml_handle_interrupts_disabled_duration(thread, os_atomic_load(&interrupt_masked_timeout, relaxed), INT_MASK_FROM_HANDLER);
}
void
ml_irq_debug_start(uintptr_t handler, uintptr_t vector)
{
INTERRUPT_MASKED_DEBUG_START(handler, DBG_INTR_TYPE_OTHER);
current_thread()->machine.int_vector = (uintptr_t)VM_KERNEL_STRIP_PTR(vector);
}
void
ml_irq_debug_end()
{
INTERRUPT_MASKED_DEBUG_END();
}
/*
* Abandon a potential timeout when handling an interrupt. It is important to
* continue to keep track of the interrupt time so the time-stamp can't be
* reset. (Interrupt time is subtracted from preemption time to maintain
* accurate preemption time measurement).
* When `inthandler_abandon` is true, a timeout will be ignored when the
* interrupt handler finishes.
*/
void