summaryrefslogtreecommitdiff
path: root/src/cpu/kvm/base.cc
blob: ddf80dbaaa1756063e6a8aabf8b099945e44951f (plain)
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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
/*
 * Copyright (c) 2012, 2015 ARM Limited
 * All rights reserved
 *
 * The license below extends only to copyright in the software and shall
 * not be construed as granting a license to any other intellectual
 * property including but not limited to intellectual property relating
 * to a hardware implementation of the functionality of the software
 * licensed hereunder.  You may use the software subject to the license
 * terms below provided that you ensure that this notice is replicated
 * unmodified and in its entirety in all distributions of the software,
 * modified or unmodified, in source code or in binary form.
 *
 * 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 the copyright holders 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.
 *
 * Authors: Andreas Sandberg
 */

#include <linux/kvm.h>
#include <sys/ioctl.h>
#include <sys/mman.h>
#include <unistd.h>

#include <cerrno>
#include <csignal>
#include <ostream>

#include "arch/mmapped_ipr.hh"
#include "arch/utility.hh"
#include "cpu/kvm/base.hh"
#include "debug/Checkpoint.hh"
#include "debug/Drain.hh"
#include "debug/Kvm.hh"
#include "debug/KvmIO.hh"
#include "debug/KvmRun.hh"
#include "params/BaseKvmCPU.hh"
#include "sim/process.hh"
#include "sim/system.hh"

#include <signal.h>

/* Used by some KVM macros */
#define PAGE_SIZE pageSize

BaseKvmCPU::BaseKvmCPU(BaseKvmCPUParams *params)
    : BaseCPU(params),
      vm(*params->kvmVM),
      _status(Idle),
      dataPort(name() + ".dcache_port", this),
      instPort(name() + ".icache_port", this),
      alwaysSyncTC(params->alwaysSyncTC),
      threadContextDirty(true),
      kvmStateDirty(false),
      vcpuID(vm.allocVCPUID()), vcpuFD(-1), vcpuMMapSize(0),
      _kvmRun(NULL), mmioRing(NULL),
      pageSize(sysconf(_SC_PAGE_SIZE)),
      tickEvent(*this),
      activeInstPeriod(0),
      perfControlledByTimer(params->usePerfOverflow),
      hostFactor(params->hostFactor),
      ctrInsts(0)
{
    if (pageSize == -1)
        panic("KVM: Failed to determine host page size (%i)\n",
              errno);

    if (FullSystem)
        thread = new SimpleThread(this, 0, params->system, params->itb, params->dtb,
                                  params->isa[0]);
    else
        thread = new SimpleThread(this, /* thread_num */ 0, params->system,
                                  params->workload[0], params->itb,
                                  params->dtb, params->isa[0]);

    thread->setStatus(ThreadContext::Halted);
    tc = thread->getTC();
    threadContexts.push_back(tc);
}

BaseKvmCPU::~BaseKvmCPU()
{
    if (_kvmRun)
        munmap(_kvmRun, vcpuMMapSize);
    close(vcpuFD);
}

void
BaseKvmCPU::init()
{
    BaseCPU::init();

    if (numThreads != 1)
        fatal("KVM: Multithreading not supported");

    tc->initMemProxies(tc);

    // initialize CPU, including PC
    if (FullSystem && !switchedOut())
        TheISA::initCPU(tc, tc->contextId());
}

void
BaseKvmCPU::startup()
{
    const BaseKvmCPUParams * const p(
        dynamic_cast<const BaseKvmCPUParams *>(params()));

    Kvm &kvm(*vm.kvm);

    BaseCPU::startup();

    assert(vcpuFD == -1);

    // Tell the VM that a CPU is about to start.
    vm.cpuStartup();

    // We can't initialize KVM CPUs in BaseKvmCPU::init() since we are
    // not guaranteed that the parent KVM VM has initialized at that
    // point. Initialize virtual CPUs here instead.
    vcpuFD = vm.createVCPU(vcpuID);

    // Map the KVM run structure */
    vcpuMMapSize = kvm.getVCPUMMapSize();
    _kvmRun = (struct kvm_run *)mmap(0, vcpuMMapSize,
                                     PROT_READ | PROT_WRITE, MAP_SHARED,
                                     vcpuFD, 0);
    if (_kvmRun == MAP_FAILED)
        panic("KVM: Failed to map run data structure\n");

    // Setup a pointer to the MMIO ring buffer if coalesced MMIO is
    // available. The offset into the KVM's communication page is
    // provided by the coalesced MMIO capability.
    int mmioOffset(kvm.capCoalescedMMIO());
    if (!p->useCoalescedMMIO) {
        inform("KVM: Coalesced MMIO disabled by config.\n");
    } else if (mmioOffset) {
        inform("KVM: Coalesced IO available\n");
        mmioRing = (struct kvm_coalesced_mmio_ring *)(
            (char *)_kvmRun + (mmioOffset * pageSize));
    } else {
        inform("KVM: Coalesced not supported by host OS\n");
    }

    thread->startup();

    Event *startupEvent(
        new EventWrapper<BaseKvmCPU,
                         &BaseKvmCPU::startupThread>(this, true));
    schedule(startupEvent, curTick());
}

BaseKvmCPU::Status
BaseKvmCPU::KVMCpuPort::nextIOState() const
{
    return (activeMMIOReqs || pendingMMIOPkts.size())
        ? RunningMMIOPending : RunningServiceCompletion;
}

Tick
BaseKvmCPU::KVMCpuPort::submitIO(PacketPtr pkt)
{
    if (cpu->system->isAtomicMode()) {
        Tick delay = sendAtomic(pkt);
        delete pkt->req;
        delete pkt;
        return delay;
    } else {
        if (pendingMMIOPkts.empty() && sendTimingReq(pkt)) {
            activeMMIOReqs++;
        } else {
            pendingMMIOPkts.push(pkt);
        }
        // Return value is irrelevant for timing-mode accesses.
        return 0;
    }
}

bool
BaseKvmCPU::KVMCpuPort::recvTimingResp(PacketPtr pkt)
{
    DPRINTF(KvmIO, "KVM: Finished timing request\n");

    delete pkt->req;
    delete pkt;
    activeMMIOReqs--;

    // We can switch back into KVM when all pending and in-flight MMIO
    // operations have completed.
    if (!(activeMMIOReqs || pendingMMIOPkts.size())) {
        DPRINTF(KvmIO, "KVM: Finished all outstanding timing requests\n");
        cpu->finishMMIOPending();
    }
    return true;
}

void
BaseKvmCPU::KVMCpuPort::recvReqRetry()
{
    DPRINTF(KvmIO, "KVM: Retry for timing request\n");

    assert(pendingMMIOPkts.size());

    // Assuming that we can issue infinite requests this cycle is a bit
    // unrealistic, but it's not worth modeling something more complex in
    // KVM.
    while (pendingMMIOPkts.size() && sendTimingReq(pendingMMIOPkts.front())) {
        pendingMMIOPkts.pop();
        activeMMIOReqs++;
    }
}

void
BaseKvmCPU::finishMMIOPending()
{
    assert(_status = RunningMMIOPending);
    assert(!tickEvent.scheduled());

    _status = RunningServiceCompletion;
    schedule(tickEvent, nextCycle());
}

void
BaseKvmCPU::startupThread()
{
    // Do thread-specific initialization. We need to setup signal
    // delivery for counters and timers from within the thread that
    // will execute the event queue to ensure that signals are
    // delivered to the right threads.
    const BaseKvmCPUParams * const p(
        dynamic_cast<const BaseKvmCPUParams *>(params()));

    vcpuThread = pthread_self();

    // Setup signal handlers. This has to be done after the vCPU is
    // created since it manipulates the vCPU signal mask.
    setupSignalHandler();

    setupCounters();

    if (p->usePerfOverflow)
        runTimer.reset(new PerfKvmTimer(hwCycles,
                                        KVM_KICK_SIGNAL,
                                        p->hostFactor,
                                        p->hostFreq));
    else
        runTimer.reset(new PosixKvmTimer(KVM_KICK_SIGNAL, CLOCK_MONOTONIC,
                                         p->hostFactor,
                                         p->hostFreq));

}

void
BaseKvmCPU::regStats()
{
    using namespace Stats;

    BaseCPU::regStats();

    numInsts
        .name(name() + ".committedInsts")
        .desc("Number of instructions committed")
        ;

    numVMExits
        .name(name() + ".numVMExits")
        .desc("total number of KVM exits")
        ;

    numVMHalfEntries
        .name(name() + ".numVMHalfEntries")
        .desc("number of KVM entries to finalize pending operations")
        ;

    numExitSignal
        .name(name() + ".numExitSignal")
        .desc("exits due to signal delivery")
        ;

    numMMIO
        .name(name() + ".numMMIO")
        .desc("number of VM exits due to memory mapped IO")
        ;

    numCoalescedMMIO
        .name(name() + ".numCoalescedMMIO")
        .desc("number of coalesced memory mapped IO requests")
        ;

    numIO
        .name(name() + ".numIO")
        .desc("number of VM exits due to legacy IO")
        ;

    numHalt
        .name(name() + ".numHalt")
        .desc("number of VM exits due to wait for interrupt instructions")
        ;

    numInterrupts
        .name(name() + ".numInterrupts")
        .desc("number of interrupts delivered")
        ;

    numHypercalls
        .name(name() + ".numHypercalls")
        .desc("number of hypercalls")
        ;
}

void
BaseKvmCPU::serializeThread(CheckpointOut &cp, ThreadID tid) const
{
    if (DTRACE(Checkpoint)) {
        DPRINTF(Checkpoint, "KVM: Serializing thread %i:\n", tid);
        dump();
    }

    assert(tid == 0);
    assert(_status == Idle);
    thread->serialize(cp);
}

void
BaseKvmCPU::unserializeThread(CheckpointIn &cp, ThreadID tid)
{
    DPRINTF(Checkpoint, "KVM: Unserialize thread %i:\n", tid);

    assert(tid == 0);
    assert(_status == Idle);
    thread->unserialize(cp);
    threadContextDirty = true;
}

DrainState
BaseKvmCPU::drain()
{
    if (switchedOut())
        return DrainState::Drained;

    DPRINTF(Drain, "BaseKvmCPU::drain\n");
    switch (_status) {
      case Running:
        // The base KVM code is normally ready when it is in the
        // Running state, but the architecture specific code might be
        // of a different opinion. This may happen when the CPU been
        // notified of an event that hasn't been accepted by the vCPU
        // yet.
        if (!archIsDrained())
            return DrainState::Draining;

        // The state of the CPU is consistent, so we don't need to do
        // anything special to drain it. We simply de-schedule the
        // tick event and enter the Idle state to prevent nasty things
        // like MMIOs from happening.
        if (tickEvent.scheduled())
            deschedule(tickEvent);
        _status = Idle;

        /** FALLTHROUGH */
      case Idle:
        // Idle, no need to drain
        assert(!tickEvent.scheduled());

        // Sync the thread context here since we'll need it when we
        // switch CPUs or checkpoint the CPU.
        syncThreadContext();

        return DrainState::Drained;

      case RunningServiceCompletion:
        // The CPU has just requested a service that was handled in
        // the RunningService state, but the results have still not
        // been reported to the CPU. Now, we /could/ probably just
        // update the register state ourselves instead of letting KVM
        // handle it, but that would be tricky. Instead, we enter KVM
        // and let it do its stuff.
        DPRINTF(Drain, "KVM CPU is waiting for service completion, "
                "requesting drain.\n");
        return DrainState::Draining;

      case RunningMMIOPending:
        // We need to drain since there are in-flight timing accesses
        DPRINTF(Drain, "KVM CPU is waiting for timing accesses to complete, "
                "requesting drain.\n");
        return DrainState::Draining;

      case RunningService:
        // We need to drain since the CPU is waiting for service (e.g., MMIOs)
        DPRINTF(Drain, "KVM CPU is waiting for service, requesting drain.\n");
        return DrainState::Draining;

      default:
        panic("KVM: Unhandled CPU state in drain()\n");
        return DrainState::Drained;
    }
}

void
BaseKvmCPU::drainResume()
{
    assert(!tickEvent.scheduled());

    // We might have been switched out. In that case, we don't need to
    // do anything.
    if (switchedOut())
        return;

    DPRINTF(Kvm, "drainResume\n");
    verifyMemoryMode();

    // The tick event is de-scheduled as a part of the draining
    // process. Re-schedule it if the thread context is active.
    if (tc->status() == ThreadContext::Active) {
        schedule(tickEvent, nextCycle());
        _status = Running;
    } else {
        _status = Idle;
    }
}

void
BaseKvmCPU::notifyFork()
{
    // We should have drained prior to forking, which means that the
    // tick event shouldn't be scheduled and the CPU is idle.
    assert(!tickEvent.scheduled());
    assert(_status == Idle);

    if (vcpuFD != -1) {
        if (close(vcpuFD) == -1)
            warn("kvm CPU: notifyFork failed to close vcpuFD\n");

        if (_kvmRun)
            munmap(_kvmRun, vcpuMMapSize);

        vcpuFD = -1;
        _kvmRun = NULL;

        hwInstructions.detach();
        hwCycles.detach();
    }
}

void
BaseKvmCPU::switchOut()
{
    DPRINTF(Kvm, "switchOut\n");

    BaseCPU::switchOut();

    // We should have drained prior to executing a switchOut, which
    // means that the tick event shouldn't be scheduled and the CPU is
    // idle.
    assert(!tickEvent.scheduled());
    assert(_status == Idle);
}

void
BaseKvmCPU::takeOverFrom(BaseCPU *cpu)
{
    DPRINTF(Kvm, "takeOverFrom\n");

    BaseCPU::takeOverFrom(cpu);

    // We should have drained prior to executing a switchOut, which
    // means that the tick event shouldn't be scheduled and the CPU is
    // idle.
    assert(!tickEvent.scheduled());
    assert(_status == Idle);
    assert(threadContexts.size() == 1);

    // Force an update of the KVM state here instead of flagging the
    // TC as dirty. This is not ideal from a performance point of
    // view, but it makes debugging easier as it allows meaningful KVM
    // state to be dumped before and after a takeover.
    updateKvmState();
    threadContextDirty = false;
}

void
BaseKvmCPU::verifyMemoryMode() const
{
    if (!(system->bypassCaches())) {
        fatal("The KVM-based CPUs requires the memory system to be in the "
              "'noncaching' mode.\n");
    }
}

void
BaseKvmCPU::wakeup(ThreadID tid)
{
    DPRINTF(Kvm, "wakeup()\n");
    // This method might have been called from another
    // context. Migrate to this SimObject's event queue when
    // delivering the wakeup signal.
    EventQueue::ScopedMigration migrate(eventQueue());

    // Kick the vCPU to get it to come out of KVM.
    kick();

    if (thread->status() != ThreadContext::Suspended)
        return;

    thread->activate();
}

void
BaseKvmCPU::activateContext(ThreadID thread_num)
{
    DPRINTF(Kvm, "ActivateContext %d\n", thread_num);

    assert(thread_num == 0);
    assert(thread);

    assert(_status == Idle);
    assert(!tickEvent.scheduled());

    numCycles += ticksToCycles(thread->lastActivate - thread->lastSuspend);

    schedule(tickEvent, clockEdge(Cycles(0)));
    _status = Running;
}


void
BaseKvmCPU::suspendContext(ThreadID thread_num)
{
    DPRINTF(Kvm, "SuspendContext %d\n", thread_num);

    assert(thread_num == 0);
    assert(thread);

    if (_status == Idle)
        return;

    assert(_status == Running || _status == RunningServiceCompletion);

    // The tick event may no be scheduled if the quest has requested
    // the monitor to wait for interrupts. The normal CPU models can
    // get their tick events descheduled by quiesce instructions, but
    // that can't happen here.
    if (tickEvent.scheduled())
        deschedule(tickEvent);

    _status = Idle;
}

void
BaseKvmCPU::deallocateContext(ThreadID thread_num)
{
    // for now, these are equivalent
    suspendContext(thread_num);
}

void
BaseKvmCPU::haltContext(ThreadID thread_num)
{
    // for now, these are equivalent
    suspendContext(thread_num);
}

ThreadContext *
BaseKvmCPU::getContext(int tn)
{
    assert(tn == 0);
    syncThreadContext();
    return tc;
}


Counter
BaseKvmCPU::totalInsts() const
{
    return ctrInsts;
}

Counter
BaseKvmCPU::totalOps() const
{
    hack_once("Pretending totalOps is equivalent to totalInsts()\n");
    return ctrInsts;
}

void
BaseKvmCPU::dump() const
{
    inform("State dumping not implemented.");
}

void
BaseKvmCPU::tick()
{
    Tick delay(0);
    assert(_status != Idle && _status != RunningMMIOPending);

    switch (_status) {
      case RunningService:
        // handleKvmExit() will determine the next state of the CPU
        delay = handleKvmExit();

        if (tryDrain())
            _status = Idle;
        break;

      case RunningServiceCompletion:
      case Running: {
          const uint64_t nextInstEvent(
              !comInstEventQueue[0]->empty() ?
              comInstEventQueue[0]->nextTick() : UINT64_MAX);
          // Enter into KVM and complete pending IO instructions if we
          // have an instruction event pending.
          const Tick ticksToExecute(
              nextInstEvent > ctrInsts ?
              curEventQueue()->nextTick() - curTick() : 0);

          if (alwaysSyncTC)
              threadContextDirty = true;

          // We might need to update the KVM state.
          syncKvmState();

          // Setup any pending instruction count breakpoints using
          // PerfEvent if we are going to execute more than just an IO
          // completion.
          if (ticksToExecute > 0)
              setupInstStop();

          DPRINTF(KvmRun, "Entering KVM...\n");
          if (drainState() == DrainState::Draining) {
              // Force an immediate exit from KVM after completing
              // pending operations. The architecture-specific code
              // takes care to run until it is in a state where it can
              // safely be drained.
              delay = kvmRunDrain();
          } else {
              delay = kvmRun(ticksToExecute);
          }

          // The CPU might have been suspended before entering into
          // KVM. Assume that the CPU was suspended /before/ entering
          // into KVM and skip the exit handling.
          if (_status == Idle)
              break;

          // Entering into KVM implies that we'll have to reload the thread
          // context from KVM if we want to access it. Flag the KVM state as
          // dirty with respect to the cached thread context.
          kvmStateDirty = true;

          if (alwaysSyncTC)
              syncThreadContext();

          // Enter into the RunningService state unless the
          // simulation was stopped by a timer.
          if (_kvmRun->exit_reason !=  KVM_EXIT_INTR) {
              _status = RunningService;
          } else {
              ++numExitSignal;
              _status = Running;
          }

          // Service any pending instruction events. The vCPU should
          // have exited in time for the event using the instruction
          // counter configured by setupInstStop().
          comInstEventQueue[0]->serviceEvents(ctrInsts);
          system->instEventQueue.serviceEvents(system->totalNumInsts);

          if (tryDrain())
              _status = Idle;
      } break;

      default:
        panic("BaseKvmCPU entered tick() in an illegal state (%i)\n",
              _status);
    }

    // Schedule a new tick if we are still running
    if (_status != Idle && _status != RunningMMIOPending)
        schedule(tickEvent, clockEdge(ticksToCycles(delay)));
}

Tick
BaseKvmCPU::kvmRunDrain()
{
    // By default, the only thing we need to drain is a pending IO
    // operation which assumes that we are in the
    // RunningServiceCompletion or RunningMMIOPending state.
    assert(_status == RunningServiceCompletion ||
           _status == RunningMMIOPending);

    // Deliver the data from the pending IO operation and immediately
    // exit.
    return kvmRun(0);
}

uint64_t
BaseKvmCPU::getHostCycles() const
{
    return hwCycles.read();
}

Tick
BaseKvmCPU::kvmRun(Tick ticks)
{
    Tick ticksExecuted;
    fatal_if(vcpuFD == -1,
             "Trying to run a KVM CPU in a forked child process. "
             "This is not supported.\n");
    DPRINTF(KvmRun, "KVM: Executing for %i ticks\n", ticks);

    if (ticks == 0) {
        // Settings ticks == 0 is a special case which causes an entry
        // into KVM that finishes pending operations (e.g., IO) and
        // then immediately exits.
        DPRINTF(KvmRun, "KVM: Delivering IO without full guest entry\n");

        ++numVMHalfEntries;

        // Send a KVM_KICK_SIGNAL to the vCPU thread (i.e., this
        // thread). The KVM control signal is masked while executing
        // in gem5 and gets unmasked temporarily as when entering
        // KVM. See setSignalMask() and setupSignalHandler().
        kick();

        // Start the vCPU. KVM will check for signals after completing
        // pending operations (IO). Since the KVM_KICK_SIGNAL is
        // pending, this forces an immediate exit to gem5 again. We
        // don't bother to setup timers since this shouldn't actually
        // execute any code (other than completing half-executed IO
        // instructions) in the guest.
        ioctlRun();

        // We always execute at least one cycle to prevent the
        // BaseKvmCPU::tick() to be rescheduled on the same tick
        // twice.
        ticksExecuted = clockPeriod();
    } else {
        // This method is executed as a result of a tick event. That
        // means that the event queue will be locked when entering the
        // method. We temporarily unlock the event queue to allow
        // other threads to steal control of this thread to inject
        // interrupts. They will typically lock the queue and then
        // force an exit from KVM by kicking the vCPU.
        EventQueue::ScopedRelease release(curEventQueue());

        if (ticks < runTimer->resolution()) {
            DPRINTF(KvmRun, "KVM: Adjusting tick count (%i -> %i)\n",
                    ticks, runTimer->resolution());
            ticks = runTimer->resolution();
        }

        // Get hardware statistics after synchronizing contexts. The KVM
        // state update might affect guest cycle counters.
        uint64_t baseCycles(getHostCycles());
        uint64_t baseInstrs(hwInstructions.read());

        // Arm the run timer and start the cycle timer if it isn't
        // controlled by the overflow timer. Starting/stopping the cycle
        // timer automatically starts the other perf timers as they are in
        // the same counter group.
        runTimer->arm(ticks);
        if (!perfControlledByTimer)
            hwCycles.start();

        ioctlRun();

        runTimer->disarm();
        if (!perfControlledByTimer)
            hwCycles.stop();

        // The control signal may have been delivered after we exited
        // from KVM. It will be pending in that case since it is
        // masked when we aren't executing in KVM. Discard it to make
        // sure we don't deliver it immediately next time we try to
        // enter into KVM.
        discardPendingSignal(KVM_KICK_SIGNAL);

        const uint64_t hostCyclesExecuted(getHostCycles() - baseCycles);
        const uint64_t simCyclesExecuted(hostCyclesExecuted * hostFactor);
        const uint64_t instsExecuted(hwInstructions.read() - baseInstrs);
        ticksExecuted = runTimer->ticksFromHostCycles(hostCyclesExecuted);

        /* Update statistics */
        numCycles += simCyclesExecuted;;
        numInsts += instsExecuted;
        ctrInsts += instsExecuted;
        system->totalNumInsts += instsExecuted;

        DPRINTF(KvmRun,
                "KVM: Executed %i instructions in %i cycles "
                "(%i ticks, sim cycles: %i).\n",
                instsExecuted, hostCyclesExecuted, ticksExecuted, simCyclesExecuted);
    }

    ++numVMExits;

    return ticksExecuted + flushCoalescedMMIO();
}

void
BaseKvmCPU::kvmNonMaskableInterrupt()
{
    ++numInterrupts;
    if (ioctl(KVM_NMI) == -1)
        panic("KVM: Failed to deliver NMI to virtual CPU\n");
}

void
BaseKvmCPU::kvmInterrupt(const struct kvm_interrupt &interrupt)
{
    ++numInterrupts;
    if (ioctl(KVM_INTERRUPT, (void *)&interrupt) == -1)
        panic("KVM: Failed to deliver interrupt to virtual CPU\n");
}

void
BaseKvmCPU::getRegisters(struct kvm_regs &regs) const
{
    if (ioctl(KVM_GET_REGS, &regs) == -1)
        panic("KVM: Failed to get guest registers\n");
}

void
BaseKvmCPU::setRegisters(const struct kvm_regs &regs)
{
    if (ioctl(KVM_SET_REGS, (void *)&regs) == -1)
        panic("KVM: Failed to set guest registers\n");
}

void
BaseKvmCPU::getSpecialRegisters(struct kvm_sregs &regs) const
{
    if (ioctl(KVM_GET_SREGS, &regs) == -1)
        panic("KVM: Failed to get guest special registers\n");
}

void
BaseKvmCPU::setSpecialRegisters(const struct kvm_sregs &regs)
{
    if (ioctl(KVM_SET_SREGS, (void *)&regs) == -1)
        panic("KVM: Failed to set guest special registers\n");
}

void
BaseKvmCPU::getFPUState(struct kvm_fpu &state) const
{
    if (ioctl(KVM_GET_FPU, &state) == -1)
        panic("KVM: Failed to get guest FPU state\n");
}

void
BaseKvmCPU::setFPUState(const struct kvm_fpu &state)
{
    if (ioctl(KVM_SET_FPU, (void *)&state) == -1)
        panic("KVM: Failed to set guest FPU state\n");
}


void
BaseKvmCPU::setOneReg(uint64_t id, const void *addr)
{
#ifdef KVM_SET_ONE_REG
    struct kvm_one_reg reg;
    reg.id = id;
    reg.addr = (uint64_t)addr;

    if (ioctl(KVM_SET_ONE_REG, &reg) == -1) {
        panic("KVM: Failed to set register (0x%x) value (errno: %i)\n",
              id, errno);
    }
#else
    panic("KVM_SET_ONE_REG is unsupported on this platform.\n");
#endif
}

void
BaseKvmCPU::getOneReg(uint64_t id, void *addr) const
{
#ifdef KVM_GET_ONE_REG
    struct kvm_one_reg reg;
    reg.id = id;
    reg.addr = (uint64_t)addr;

    if (ioctl(KVM_GET_ONE_REG, &reg) == -1) {
        panic("KVM: Failed to get register (0x%x) value (errno: %i)\n",
              id, errno);
    }
#else
    panic("KVM_GET_ONE_REG is unsupported on this platform.\n");
#endif
}

std::string
BaseKvmCPU::getAndFormatOneReg(uint64_t id) const
{
#ifdef KVM_GET_ONE_REG
    std::ostringstream ss;

    ss.setf(std::ios::hex, std::ios::basefield);
    ss.setf(std::ios::showbase);
#define HANDLE_INTTYPE(len)                      \
    case KVM_REG_SIZE_U ## len: {                \
        uint ## len ## _t value;                 \
        getOneReg(id, &value);                   \
        ss << value;                             \
    }  break

#define HANDLE_ARRAY(len)                               \
    case KVM_REG_SIZE_U ## len: {                       \
        uint8_t value[len / 8];                         \
        getOneReg(id, value);                           \
        ccprintf(ss, "[0x%x", value[0]);                \
        for (int i = 1; i < len  / 8; ++i)              \
            ccprintf(ss, ", 0x%x", value[i]);           \
        ccprintf(ss, "]");                              \
      } break

    switch (id & KVM_REG_SIZE_MASK) {
        HANDLE_INTTYPE(8);
        HANDLE_INTTYPE(16);
        HANDLE_INTTYPE(32);
        HANDLE_INTTYPE(64);
        HANDLE_ARRAY(128);
        HANDLE_ARRAY(256);
        HANDLE_ARRAY(512);
        HANDLE_ARRAY(1024);
      default:
        ss << "??";
    }

#undef HANDLE_INTTYPE
#undef HANDLE_ARRAY

    return ss.str();
#else
    panic("KVM_GET_ONE_REG is unsupported on this platform.\n");
#endif
}

void
BaseKvmCPU::syncThreadContext()
{
    if (!kvmStateDirty)
        return;

    assert(!threadContextDirty);

    updateThreadContext();
    kvmStateDirty = false;
}

void
BaseKvmCPU::syncKvmState()
{
    if (!threadContextDirty)
        return;

    assert(!kvmStateDirty);

    updateKvmState();
    threadContextDirty = false;
}

Tick
BaseKvmCPU::handleKvmExit()
{
    DPRINTF(KvmRun, "handleKvmExit (exit_reason: %i)\n", _kvmRun->exit_reason);
    assert(_status == RunningService);

    // Switch into the running state by default. Individual handlers
    // can override this.
    _status = Running;
    switch (_kvmRun->exit_reason) {
      case KVM_EXIT_UNKNOWN:
        return handleKvmExitUnknown();

      case KVM_EXIT_EXCEPTION:
        return handleKvmExitException();

      case KVM_EXIT_IO:
      {
        ++numIO;
        Tick ticks = handleKvmExitIO();
        _status = dataPort.nextIOState();
        return ticks;
      }

      case KVM_EXIT_HYPERCALL:
        ++numHypercalls;
        return handleKvmExitHypercall();

      case KVM_EXIT_HLT:
        /* The guest has halted and is waiting for interrupts */
        DPRINTF(Kvm, "handleKvmExitHalt\n");
        ++numHalt;

        // Suspend the thread until the next interrupt arrives
        thread->suspend();

        // This is actually ignored since the thread is suspended.
        return 0;

      case KVM_EXIT_MMIO:
      {
        /* Service memory mapped IO requests */
        DPRINTF(KvmIO, "KVM: Handling MMIO (w: %u, addr: 0x%x, len: %u)\n",
                _kvmRun->mmio.is_write,
                _kvmRun->mmio.phys_addr, _kvmRun->mmio.len);

        ++numMMIO;
        Tick ticks = doMMIOAccess(_kvmRun->mmio.phys_addr, _kvmRun->mmio.data,
                                  _kvmRun->mmio.len, _kvmRun->mmio.is_write);
        // doMMIOAccess could have triggered a suspend, in which case we don't
        // want to overwrite the _status.
        if (_status != Idle)
            _status = dataPort.nextIOState();
        return ticks;
      }

      case KVM_EXIT_IRQ_WINDOW_OPEN:
        return handleKvmExitIRQWindowOpen();

      case KVM_EXIT_FAIL_ENTRY:
        return handleKvmExitFailEntry();

      case KVM_EXIT_INTR:
        /* KVM was interrupted by a signal, restart it in the next
         * tick. */
        return 0;

      case KVM_EXIT_INTERNAL_ERROR:
        panic("KVM: Internal error (suberror: %u)\n",
              _kvmRun->internal.suberror);

      default:
        dump();
        panic("KVM: Unexpected exit (exit_reason: %u)\n", _kvmRun->exit_reason);
    }
}

Tick
BaseKvmCPU::handleKvmExitIO()
{
    panic("KVM: Unhandled guest IO (dir: %i, size: %i, port: 0x%x, count: %i)\n",
          _kvmRun->io.direction, _kvmRun->io.size,
          _kvmRun->io.port, _kvmRun->io.count);
}

Tick
BaseKvmCPU::handleKvmExitHypercall()
{
    panic("KVM: Unhandled hypercall\n");
}

Tick
BaseKvmCPU::handleKvmExitIRQWindowOpen()
{
    warn("KVM: Unhandled IRQ window.\n");
    return 0;
}


Tick
BaseKvmCPU::handleKvmExitUnknown()
{
    dump();
    panic("KVM: Unknown error when starting vCPU (hw reason: 0x%llx)\n",
          _kvmRun->hw.hardware_exit_reason);
}

Tick
BaseKvmCPU::handleKvmExitException()
{
    dump();
    panic("KVM: Got exception when starting vCPU "
          "(exception: %u, error_code: %u)\n",
          _kvmRun->ex.exception, _kvmRun->ex.error_code);
}

Tick
BaseKvmCPU::handleKvmExitFailEntry()
{
    dump();
    panic("KVM: Failed to enter virtualized mode (hw reason: 0x%llx)\n",
          _kvmRun->fail_entry.hardware_entry_failure_reason);
}

Tick
BaseKvmCPU::doMMIOAccess(Addr paddr, void *data, int size, bool write)
{
    ThreadContext *tc(thread->getTC());
    syncThreadContext();

    RequestPtr mmio_req = new Request(paddr, size, Request::UNCACHEABLE,
                                      dataMasterId());
    mmio_req->setContext(tc->contextId());
    // Some architectures do need to massage physical addresses a bit
    // before they are inserted into the memory system. This enables
    // APIC accesses on x86 and m5ops where supported through a MMIO
    // interface.
    BaseTLB::Mode tlb_mode(write ? BaseTLB::Write : BaseTLB::Read);
    Fault fault(tc->getDTBPtr()->finalizePhysical(mmio_req, tc, tlb_mode));
    if (fault != NoFault)
        warn("Finalization of MMIO address failed: %s\n", fault->name());


    const MemCmd cmd(write ? MemCmd::WriteReq : MemCmd::ReadReq);
    PacketPtr pkt = new Packet(mmio_req, cmd);
    pkt->dataStatic(data);

    if (mmio_req->isMmappedIpr()) {
        // We currently assume that there is no need to migrate to a
        // different event queue when doing IPRs. Currently, IPRs are
        // only used for m5ops, so it should be a valid assumption.
        const Cycles ipr_delay(write ?
                             TheISA::handleIprWrite(tc, pkt) :
                             TheISA::handleIprRead(tc, pkt));
        threadContextDirty = true;
        delete pkt->req;
        delete pkt;
        return clockPeriod() * ipr_delay;
    } else {
        // Temporarily lock and migrate to the event queue of the
        // VM. This queue is assumed to "own" all devices we need to
        // access if running in multi-core mode.
        EventQueue::ScopedMigration migrate(vm.eventQueue());

        return dataPort.submitIO(pkt);
    }
}

void
BaseKvmCPU::setSignalMask(const sigset_t *mask)
{
    std::unique_ptr<struct kvm_signal_mask> kvm_mask;

    if (mask) {
        kvm_mask.reset((struct kvm_signal_mask *)operator new(
                           sizeof(struct kvm_signal_mask) + sizeof(*mask)));
        // The kernel and the user-space headers have different ideas
        // about the size of sigset_t. This seems like a massive hack,
        // but is actually what qemu does.
        assert(sizeof(*mask) >= 8);
        kvm_mask->len = 8;
        memcpy(kvm_mask->sigset, mask, kvm_mask->len);
    }

    if (ioctl(KVM_SET_SIGNAL_MASK, (void *)kvm_mask.get()) == -1)
        panic("KVM: Failed to set vCPU signal mask (errno: %i)\n",
              errno);
}

int
BaseKvmCPU::ioctl(int request, long p1) const
{
    if (vcpuFD == -1)
        panic("KVM: CPU ioctl called before initialization\n");

    return ::ioctl(vcpuFD, request, p1);
}

Tick
BaseKvmCPU::flushCoalescedMMIO()
{
    if (!mmioRing)
        return 0;

    DPRINTF(KvmIO, "KVM: Flushing the coalesced MMIO ring buffer\n");

    // TODO: We might need to do synchronization when we start to
    // support multiple CPUs
    Tick ticks(0);
    while (mmioRing->first != mmioRing->last) {
        struct kvm_coalesced_mmio &ent(
            mmioRing->coalesced_mmio[mmioRing->first]);

        DPRINTF(KvmIO, "KVM: Handling coalesced MMIO (addr: 0x%x, len: %u)\n",
                ent.phys_addr, ent.len);

        ++numCoalescedMMIO;
        ticks += doMMIOAccess(ent.phys_addr, ent.data, ent.len, true);

        mmioRing->first = (mmioRing->first + 1) % KVM_COALESCED_MMIO_MAX;
    }

    return ticks;
}

/**
 * Dummy handler for KVM kick signals.
 *
 * @note This function is usually not called since the kernel doesn't
 * seem to deliver signals when the signal is only unmasked when
 * running in KVM. This doesn't matter though since we are only
 * interested in getting KVM to exit, which happens as expected. See
 * setupSignalHandler() and kvmRun() for details about KVM signal
 * handling.
 */
static void
onKickSignal(int signo, siginfo_t *si, void *data)
{
}

void
BaseKvmCPU::setupSignalHandler()
{
    struct sigaction sa;

    memset(&sa, 0, sizeof(sa));
    sa.sa_sigaction = onKickSignal;
    sa.sa_flags = SA_SIGINFO | SA_RESTART;
    if (sigaction(KVM_KICK_SIGNAL, &sa, NULL) == -1)
        panic("KVM: Failed to setup vCPU timer signal handler\n");

    sigset_t sigset;
    if (pthread_sigmask(SIG_BLOCK, NULL, &sigset) == -1)
        panic("KVM: Failed get signal mask\n");

    // Request KVM to setup the same signal mask as we're currently
    // running with except for the KVM control signal. We'll sometimes
    // need to raise the KVM_KICK_SIGNAL to cause immediate exits from
    // KVM after servicing IO requests. See kvmRun().
    sigdelset(&sigset, KVM_KICK_SIGNAL);
    setSignalMask(&sigset);

    // Mask our control signals so they aren't delivered unless we're
    // actually executing inside KVM.
    sigaddset(&sigset, KVM_KICK_SIGNAL);
    if (pthread_sigmask(SIG_SETMASK, &sigset, NULL) == -1)
        panic("KVM: Failed mask the KVM control signals\n");
}

bool
BaseKvmCPU::discardPendingSignal(int signum) const
{
    int discardedSignal;

    // Setting the timeout to zero causes sigtimedwait to return
    // immediately.
    struct timespec timeout;
    timeout.tv_sec = 0;
    timeout.tv_nsec = 0;

    sigset_t sigset;
    sigemptyset(&sigset);
    sigaddset(&sigset, signum);

    do {
        discardedSignal = sigtimedwait(&sigset, NULL, &timeout);
    } while (discardedSignal == -1 && errno == EINTR);

    if (discardedSignal == signum)
        return true;
    else if (discardedSignal == -1 && errno == EAGAIN)
        return false;
    else
        panic("Unexpected return value from sigtimedwait: %i (errno: %i)\n",
              discardedSignal, errno);
}

void
BaseKvmCPU::setupCounters()
{
    DPRINTF(Kvm, "Attaching cycle counter...\n");
    PerfKvmCounterConfig cfgCycles(PERF_TYPE_HARDWARE,
                                PERF_COUNT_HW_CPU_CYCLES);
    cfgCycles.disabled(true)
        .pinned(true);

    // Try to exclude the host. We set both exclude_hv and
    // exclude_host since different architectures use slightly
    // different APIs in the kernel.
    cfgCycles.exclude_hv(true)
        .exclude_host(true);

    if (perfControlledByTimer) {
        // We need to configure the cycles counter to send overflows
        // since we are going to use it to trigger timer signals that
        // trap back into m5 from KVM. In practice, this means that we
        // need to set some non-zero sample period that gets
        // overridden when the timer is armed.
        cfgCycles.wakeupEvents(1)
            .samplePeriod(42);
    }

    hwCycles.attach(cfgCycles,
                    0); // TID (0 => currentThread)

    setupInstCounter();
}

bool
BaseKvmCPU::tryDrain()
{
    if (drainState() != DrainState::Draining)
        return false;

    if (!archIsDrained()) {
        DPRINTF(Drain, "tryDrain: Architecture code is not ready.\n");
        return false;
    }

    if (_status == Idle || _status == Running) {
        DPRINTF(Drain,
                "tryDrain: CPU transitioned into the Idle state, drain done\n");
        signalDrainDone();
        return true;
    } else {
        DPRINTF(Drain, "tryDrain: CPU not ready.\n");
        return false;
    }
}

void
BaseKvmCPU::ioctlRun()
{
    if (ioctl(KVM_RUN) == -1) {
        if (errno != EINTR)
            panic("KVM: Failed to start virtual CPU (errno: %i)\n",
                  errno);
    }
}

void
BaseKvmCPU::setupInstStop()
{
    if (comInstEventQueue[0]->empty()) {
        setupInstCounter(0);
    } else {
        const uint64_t next(comInstEventQueue[0]->nextTick());

        assert(next > ctrInsts);
        setupInstCounter(next - ctrInsts);
    }
}

void
BaseKvmCPU::setupInstCounter(uint64_t period)
{
    // No need to do anything if we aren't attaching for the first
    // time or the period isn't changing.
    if (period == activeInstPeriod && hwInstructions.attached())
        return;

    PerfKvmCounterConfig cfgInstructions(PERF_TYPE_HARDWARE,
                                         PERF_COUNT_HW_INSTRUCTIONS);

    // Try to exclude the host. We set both exclude_hv and
    // exclude_host since different architectures use slightly
    // different APIs in the kernel.
    cfgInstructions.exclude_hv(true)
        .exclude_host(true);

    if (period) {
        // Setup a sampling counter if that has been requested.
        cfgInstructions.wakeupEvents(1)
            .samplePeriod(period);
    }

    // We need to detach and re-attach the counter to reliably change
    // sampling settings. See PerfKvmCounter::period() for details.
    if (hwInstructions.attached())
        hwInstructions.detach();
    assert(hwCycles.attached());
    hwInstructions.attach(cfgInstructions,
                          0, // TID (0 => currentThread)
                          hwCycles);

    if (period)
        hwInstructions.enableSignals(KVM_KICK_SIGNAL);

    activeInstPeriod = period;
}