summaryrefslogtreecommitdiffstats
path: root/runtime/thread.h
blob: ec276b59f57ad308474ab4316376abc8a63cf615 (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
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
/*
 * Copyright (C) 2011 The Android Open Source Project
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *      http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

#ifndef ART_RUNTIME_THREAD_H_
#define ART_RUNTIME_THREAD_H_

#include <atomic>
#include <bitset>
#include <deque>
#include <iosfwd>
#include <list>
#include <memory>
#include <string>

#include "base/atomic.h"
#include "base/enums.h"
#include "base/locks.h"
#include "base/macros.h"
#include "base/safe_map.h"
#include "base/value_object.h"
#include "entrypoints/jni/jni_entrypoints.h"
#include "entrypoints/quick/quick_entrypoints.h"
#include "handle_scope.h"
#include "interpreter/interpreter_cache.h"
#include "jvalue.h"
#include "managed_stack.h"
#include "offsets.h"
#include "read_barrier_config.h"
#include "runtime_globals.h"
#include "runtime_stats.h"
#include "suspend_reason.h"
#include "thread_state.h"

class BacktraceMap;

namespace art {

namespace gc {
namespace accounting {
template<class T> class AtomicStack;
}  // namespace accounting
namespace collector {
class SemiSpace;
}  // namespace collector
}  // namespace gc

namespace instrumentation {
struct InstrumentationStackFrame;
}  // namespace instrumentation

namespace mirror {
class Array;
class Class;
class ClassLoader;
class Object;
template<class T> class ObjectArray;
template<class T> class PrimitiveArray;
typedef PrimitiveArray<int32_t> IntArray;
class StackTraceElement;
class String;
class Throwable;
}  // namespace mirror

namespace verifier {
class MethodVerifier;
class VerifierDeps;
}  // namespace verifier

class ArtMethod;
class BaseMutex;
class ClassLinker;
class Closure;
class Context;
struct DebugInvokeReq;
class DeoptimizationContextRecord;
class DexFile;
class FrameIdToShadowFrame;
class JavaVMExt;
class JNIEnvExt;
class Monitor;
class RootVisitor;
class ScopedObjectAccessAlreadyRunnable;
class ShadowFrame;
class SingleStepControl;
class StackedShadowFrameRecord;
class Thread;
class ThreadList;
enum VisitRootFlags : uint8_t;

// A piece of data that can be held in the CustomTls. The destructor will be called during thread
// shutdown. The thread the destructor is called on is not necessarily the same thread it was stored
// on.
class TLSData {
 public:
  virtual ~TLSData() {}
};

// Thread priorities. These must match the Thread.MIN_PRIORITY,
// Thread.NORM_PRIORITY, and Thread.MAX_PRIORITY constants.
enum ThreadPriority {
  kMinThreadPriority = 1,
  kNormThreadPriority = 5,
  kMaxThreadPriority = 10,
};

enum ThreadFlag {
  kSuspendRequest   = 1,  // If set implies that suspend_count_ > 0 and the Thread should enter the
                          // safepoint handler.
  kCheckpointRequest = 2,  // Request that the thread do some checkpoint work and then continue.
  kEmptyCheckpointRequest = 4,  // Request that the thread do empty checkpoint and then continue.
  kActiveSuspendBarrier = 8,  // Register that at least 1 suspend barrier needs to be passed.
};

enum class StackedShadowFrameType {
  kShadowFrameUnderConstruction,
  kDeoptimizationShadowFrame,
};

// The type of method that triggers deoptimization. It contains info on whether
// the deoptimized method should advance dex_pc.
enum class DeoptimizationMethodType {
  kKeepDexPc,  // dex pc is required to be kept upon deoptimization.
  kDefault     // dex pc may or may not advance depending on other conditions.
};

// This should match RosAlloc::kNumThreadLocalSizeBrackets.
static constexpr size_t kNumRosAllocThreadLocalSizeBracketsInThread = 16;

// Thread's stack layout for implicit stack overflow checks:
//
//   +---------------------+  <- highest address of stack memory
//   |                     |
//   .                     .  <- SP
//   |                     |
//   |                     |
//   +---------------------+  <- stack_end
//   |                     |
//   |  Gap                |
//   |                     |
//   +---------------------+  <- stack_begin
//   |                     |
//   | Protected region    |
//   |                     |
//   +---------------------+  <- lowest address of stack memory
//
// The stack always grows down in memory.  At the lowest address is a region of memory
// that is set mprotect(PROT_NONE).  Any attempt to read/write to this region will
// result in a segmentation fault signal.  At any point, the thread's SP will be somewhere
// between the stack_end and the highest address in stack memory.  An implicit stack
// overflow check is a read of memory at a certain offset below the current SP (4K typically).
// If the thread's SP is below the stack_end address this will be a read into the protected
// region.  If the SP is above the stack_end address, the thread is guaranteed to have
// at least 4K of space.  Because stack overflow checks are only performed in generated code,
// if the thread makes a call out to a native function (through JNI), that native function
// might only have 4K of memory (if the SP is adjacent to stack_end).

class Thread {
 public:
  static const size_t kStackOverflowImplicitCheckSize;
  static constexpr bool kVerifyStack = kIsDebugBuild;

  // Creates a new native thread corresponding to the given managed peer.
  // Used to implement Thread.start.
  static void CreateNativeThread(JNIEnv* env, jobject peer, size_t stack_size, bool daemon);

  // Attaches the calling native thread to the runtime, returning the new native peer.
  // Used to implement JNI AttachCurrentThread and AttachCurrentThreadAsDaemon calls.
  static Thread* Attach(const char* thread_name, bool as_daemon, jobject thread_group,
                        bool create_peer);
  // Attaches the calling native thread to the runtime, returning the new native peer.
  static Thread* Attach(const char* thread_name, bool as_daemon, jobject thread_peer);

  // Reset internal state of child thread after fork.
  void InitAfterFork();

  // Get the currently executing thread, frequently referred to as 'self'. This call has reasonably
  // high cost and so we favor passing self around when possible.
  // TODO: mark as PURE so the compiler may coalesce and remove?
  static Thread* Current();

  // On a runnable thread, check for pending thread suspension request and handle if pending.
  void AllowThreadSuspension() REQUIRES_SHARED(Locks::mutator_lock_);

  // Process pending thread suspension request and handle if pending.
  void CheckSuspend() REQUIRES_SHARED(Locks::mutator_lock_);

  // Process a pending empty checkpoint if pending.
  void CheckEmptyCheckpointFromWeakRefAccess(BaseMutex* cond_var_mutex);
  void CheckEmptyCheckpointFromMutex();

  static Thread* FromManagedThread(const ScopedObjectAccessAlreadyRunnable& ts,
                                   ObjPtr<mirror::Object> thread_peer)
      REQUIRES(Locks::thread_list_lock_, !Locks::thread_suspend_count_lock_)
      REQUIRES_SHARED(Locks::mutator_lock_);
  static Thread* FromManagedThread(const ScopedObjectAccessAlreadyRunnable& ts, jobject thread)
      REQUIRES(Locks::thread_list_lock_, !Locks::thread_suspend_count_lock_)
      REQUIRES_SHARED(Locks::mutator_lock_);

  // Translates 172 to pAllocArrayFromCode and so on.
  template<PointerSize size_of_pointers>
  static void DumpThreadOffset(std::ostream& os, uint32_t offset);

  // Dumps a one-line summary of thread state (used for operator<<).
  void ShortDump(std::ostream& os) const;

  // Dumps the detailed thread state and the thread stack (used for SIGQUIT).
  void Dump(std::ostream& os,
            bool dump_native_stack = true,
            BacktraceMap* backtrace_map = nullptr,
            bool force_dump_stack = false) const
      REQUIRES(!Locks::thread_suspend_count_lock_)
      REQUIRES_SHARED(Locks::mutator_lock_);

  void DumpJavaStack(std::ostream& os,
                     bool check_suspended = true,
                     bool dump_locks = true) const
      REQUIRES(!Locks::thread_suspend_count_lock_)
      REQUIRES_SHARED(Locks::mutator_lock_);

  // Dumps the SIGQUIT per-thread header. 'thread' can be null for a non-attached thread, in which
  // case we use 'tid' to identify the thread, and we'll include as much information as we can.
  static void DumpState(std::ostream& os, const Thread* thread, pid_t tid)
      REQUIRES(!Locks::thread_suspend_count_lock_)
      REQUIRES_SHARED(Locks::mutator_lock_);

  ThreadState GetState() const {
    DCHECK_GE(tls32_.state_and_flags.as_struct.state, kTerminated);
    DCHECK_LE(tls32_.state_and_flags.as_struct.state, kSuspended);
    return static_cast<ThreadState>(tls32_.state_and_flags.as_struct.state);
  }

  ThreadState SetState(ThreadState new_state);

  int GetSuspendCount() const REQUIRES(Locks::thread_suspend_count_lock_) {
    return tls32_.suspend_count;
  }

  int GetUserCodeSuspendCount() const REQUIRES(Locks::thread_suspend_count_lock_,
                                               Locks::user_code_suspension_lock_) {
    return tls32_.user_code_suspend_count;
  }

  int GetDebugSuspendCount() const REQUIRES(Locks::thread_suspend_count_lock_) {
    return tls32_.debug_suspend_count;
  }

  bool IsSuspended() const {
    union StateAndFlags state_and_flags;
    state_and_flags.as_int = tls32_.state_and_flags.as_int;
    return state_and_flags.as_struct.state != kRunnable &&
        (state_and_flags.as_struct.flags & kSuspendRequest) != 0;
  }

  // If delta > 0 and (this != self or suspend_barrier is not null), this function may temporarily
  // release thread_suspend_count_lock_ internally.
  ALWAYS_INLINE
  bool ModifySuspendCount(Thread* self,
                          int delta,
                          AtomicInteger* suspend_barrier,
                          SuspendReason reason)
      WARN_UNUSED
      REQUIRES(Locks::thread_suspend_count_lock_);

  // Requests a checkpoint closure to run on another thread. The closure will be run when the thread
  // gets suspended. This will return true if the closure was added and will (eventually) be
  // executed. It returns false otherwise.
  //
  // Since multiple closures can be queued and some closures can delay other threads from running no
  // closure should attempt to suspend another thread while running.
  // TODO We should add some debug option that verifies this.
  bool RequestCheckpoint(Closure* function)
      REQUIRES(Locks::thread_suspend_count_lock_);

  // RequestSynchronousCheckpoint releases the thread_list_lock_ as a part of its execution. This is
  // due to the fact that Thread::Current() needs to go to sleep to allow the targeted thread to
  // execute the checkpoint for us if it is Runnable. The suspend_state is the state that the thread
  // will go into while it is awaiting the checkpoint to be run.
  // NB Passing ThreadState::kRunnable may cause the current thread to wait in a condition variable
  // while holding the mutator_lock_.  Callers should ensure that this will not cause any problems
  // for the closure or the rest of the system.
  // NB Since multiple closures can be queued and some closures can delay other threads from running
  // no closure should attempt to suspend another thread while running.
  bool RequestSynchronousCheckpoint(Closure* function,
                                    ThreadState suspend_state = ThreadState::kWaiting)
      REQUIRES_SHARED(Locks::mutator_lock_)
      RELEASE(Locks::thread_list_lock_)
      REQUIRES(!Locks::thread_suspend_count_lock_);

  bool RequestEmptyCheckpoint()
      REQUIRES(Locks::thread_suspend_count_lock_);

  void SetFlipFunction(Closure* function);
  Closure* GetFlipFunction();

  gc::accounting::AtomicStack<mirror::Object>* GetThreadLocalMarkStack() {
    CHECK(kUseReadBarrier);
    return tlsPtr_.thread_local_mark_stack;
  }
  void SetThreadLocalMarkStack(gc::accounting::AtomicStack<mirror::Object>* stack) {
    CHECK(kUseReadBarrier);
    tlsPtr_.thread_local_mark_stack = stack;
  }

  // Called when thread detected that the thread_suspend_count_ was non-zero. Gives up share of
  // mutator_lock_ and waits until it is resumed and thread_suspend_count_ is zero.
  void FullSuspendCheck()
      REQUIRES(!Locks::thread_suspend_count_lock_)
      REQUIRES_SHARED(Locks::mutator_lock_);

  // Transition from non-runnable to runnable state acquiring share on mutator_lock_.
  ALWAYS_INLINE ThreadState TransitionFromSuspendedToRunnable()
      REQUIRES(!Locks::thread_suspend_count_lock_)
      SHARED_LOCK_FUNCTION(Locks::mutator_lock_);

  // Transition from runnable into a state where mutator privileges are denied. Releases share of
  // mutator lock.
  ALWAYS_INLINE void TransitionFromRunnableToSuspended(ThreadState new_state)
      REQUIRES(!Locks::thread_suspend_count_lock_, !Roles::uninterruptible_)
      UNLOCK_FUNCTION(Locks::mutator_lock_);

  // Once called thread suspension will cause an assertion failure.
  const char* StartAssertNoThreadSuspension(const char* cause) ACQUIRE(Roles::uninterruptible_) {
    Roles::uninterruptible_.Acquire();  // No-op.
    if (kIsDebugBuild) {
      CHECK(cause != nullptr);
      const char* previous_cause = tlsPtr_.last_no_thread_suspension_cause;
      tls32_.no_thread_suspension++;
      tlsPtr_.last_no_thread_suspension_cause = cause;
      return previous_cause;
    } else {
      return nullptr;
    }
  }

  // End region where no thread suspension is expected.
  void EndAssertNoThreadSuspension(const char* old_cause) RELEASE(Roles::uninterruptible_) {
    if (kIsDebugBuild) {
      CHECK(old_cause != nullptr || tls32_.no_thread_suspension == 1);
      CHECK_GT(tls32_.no_thread_suspension, 0U);
      tls32_.no_thread_suspension--;
      tlsPtr_.last_no_thread_suspension_cause = old_cause;
    }
    Roles::uninterruptible_.Release();  // No-op.
  }

  void AssertThreadSuspensionIsAllowable(bool check_locks = true) const;

  // Return true if thread suspension is allowable.
  bool IsThreadSuspensionAllowable() const;

  bool IsDaemon() const {
    return tls32_.daemon;
  }

  size_t NumberOfHeldMutexes() const;

  bool HoldsLock(ObjPtr<mirror::Object> object) const REQUIRES_SHARED(Locks::mutator_lock_);

  /*
   * Changes the priority of this thread to match that of the java.lang.Thread object.
   *
   * We map a priority value from 1-10 to Linux "nice" values, where lower
   * numbers indicate higher priority.
   */
  void SetNativePriority(int newPriority);

  /*
   * Returns the thread priority for the current thread by querying the system.
   * This is useful when attaching a thread through JNI.
   *
   * Returns a value from 1 to 10 (compatible with java.lang.Thread values).
   */
  static int GetNativePriority();

  // Guaranteed to be non-zero.
  uint32_t GetThreadId() const {
    return tls32_.thin_lock_thread_id;
  }

  pid_t GetTid() const {
    return tls32_.tid;
  }

  // Returns the java.lang.Thread's name, or null if this Thread* doesn't have a peer.
  mirror::String* GetThreadName() const REQUIRES_SHARED(Locks::mutator_lock_);

  // Sets 'name' to the java.lang.Thread's name. This requires no transition to managed code,
  // allocation, or locking.
  void GetThreadName(std::string& name) const;

  // Sets the thread's name.
  void SetThreadName(const char* name) REQUIRES_SHARED(Locks::mutator_lock_);

  // Returns the thread-specific CPU-time clock in microseconds or -1 if unavailable.
  uint64_t GetCpuMicroTime() const;

  mirror::Object* GetPeer() const REQUIRES_SHARED(Locks::mutator_lock_) {
    DCHECK(Thread::Current() == this) << "Use GetPeerFromOtherThread instead";
    CHECK(tlsPtr_.jpeer == nullptr);
    return tlsPtr_.opeer;
  }
  // GetPeer is not safe if called on another thread in the middle of the CC thread flip and
  // the thread's stack may have not been flipped yet and peer may be a from-space (stale) ref.
  // This function will explicitly mark/forward it.
  mirror::Object* GetPeerFromOtherThread() const REQUIRES_SHARED(Locks::mutator_lock_);

  bool HasPeer() const {
    return tlsPtr_.jpeer != nullptr || tlsPtr_.opeer != nullptr;
  }

  RuntimeStats* GetStats() {
    return &tls64_.stats;
  }

  bool IsStillStarting() const;

  bool IsExceptionPending() const {
    return tlsPtr_.exception != nullptr;
  }

  bool IsAsyncExceptionPending() const {
    return tlsPtr_.async_exception != nullptr;
  }

  mirror::Throwable* GetException() const REQUIRES_SHARED(Locks::mutator_lock_) {
    return tlsPtr_.exception;
  }

  void AssertPendingException() const;
  void AssertPendingOOMException() const REQUIRES_SHARED(Locks::mutator_lock_);
  void AssertNoPendingException() const;
  void AssertNoPendingExceptionForNewException(const char* msg) const;

  void SetException(ObjPtr<mirror::Throwable> new_exception) REQUIRES_SHARED(Locks::mutator_lock_);

  // Set an exception that is asynchronously thrown from a different thread. This will be checked
  // periodically and might overwrite the current 'Exception'. This can only be called from a
  // checkpoint.
  //
  // The caller should also make sure that the thread has been deoptimized so that the exception
  // could be detected on back-edges.
  void SetAsyncException(ObjPtr<mirror::Throwable> new_exception)
      REQUIRES_SHARED(Locks::mutator_lock_);

  void ClearException() REQUIRES_SHARED(Locks::mutator_lock_) {
    tlsPtr_.exception = nullptr;
  }

  // Move the current async-exception to the main exception. This should be called when the current
  // thread is ready to deal with any async exceptions. Returns true if there is an async exception
  // that needs to be dealt with, false otherwise.
  bool ObserveAsyncException() REQUIRES_SHARED(Locks::mutator_lock_);

  // Find catch block and perform long jump to appropriate exception handle
  NO_RETURN void QuickDeliverException() REQUIRES_SHARED(Locks::mutator_lock_);

  Context* GetLongJumpContext();
  void ReleaseLongJumpContext(Context* context) {
    if (tlsPtr_.long_jump_context != nullptr) {
      ReleaseLongJumpContextInternal();
    }
    tlsPtr_.long_jump_context = context;
  }

  // Get the current method and dex pc. If there are errors in retrieving the dex pc, this will
  // abort the runtime iff abort_on_error is true.
  ArtMethod* GetCurrentMethod(uint32_t* dex_pc,
                              bool check_suspended = true,
                              bool abort_on_error = true) const
      REQUIRES_SHARED(Locks::mutator_lock_);

  // Returns whether the given exception was thrown by the current Java method being executed
  // (Note that this includes native Java methods).
  bool IsExceptionThrownByCurrentMethod(ObjPtr<mirror::Throwable> exception) const
      REQUIRES_SHARED(Locks::mutator_lock_);

  void SetTopOfStack(ArtMethod** top_method) {
    tlsPtr_.managed_stack.SetTopQuickFrame(top_method);
  }

  void SetTopOfStackTagged(ArtMethod** top_method) {
    tlsPtr_.managed_stack.SetTopQuickFrameTagged(top_method);
  }

  void SetTopOfShadowStack(ShadowFrame* top) {
    tlsPtr_.managed_stack.SetTopShadowFrame(top);
  }

  bool HasManagedStack() const {
    return tlsPtr_.managed_stack.HasTopQuickFrame() || tlsPtr_.managed_stack.HasTopShadowFrame();
  }

  // If 'msg' is null, no detail message is set.
  void ThrowNewException(const char* exception_class_descriptor, const char* msg)
      REQUIRES_SHARED(Locks::mutator_lock_) REQUIRES(!Roles::uninterruptible_);

  // If 'msg' is null, no detail message is set. An exception must be pending, and will be
  // used as the new exception's cause.
  void ThrowNewWrappedException(const char* exception_class_descriptor, const char* msg)
      REQUIRES_SHARED(Locks::mutator_lock_) REQUIRES(!Roles::uninterruptible_);

  void ThrowNewExceptionF(const char* exception_class_descriptor, const char* fmt, ...)
      __attribute__((format(printf, 3, 4)))
      REQUIRES_SHARED(Locks::mutator_lock_) REQUIRES(!Roles::uninterruptible_);

  void ThrowNewExceptionV(const char* exception_class_descriptor, const char* fmt, va_list ap)
      REQUIRES_SHARED(Locks::mutator_lock_) REQUIRES(!Roles::uninterruptible_);

  // OutOfMemoryError is special, because we need to pre-allocate an instance.
  // Only the GC should call this.
  void ThrowOutOfMemoryError(const char* msg) REQUIRES_SHARED(Locks::mutator_lock_)
      REQUIRES(!Roles::uninterruptible_);

  static void Startup();
  static void FinishStartup();
  static void Shutdown();

  // Notify this thread's thread-group that this thread has started.
  // Note: the given thread-group is used as a fast path and verified in debug build. If the value
  //       is null, the thread's thread-group is loaded from the peer.
  void NotifyThreadGroup(ScopedObjectAccessAlreadyRunnable& soa, jobject thread_group = nullptr)
      REQUIRES_SHARED(Locks::mutator_lock_);

  // JNI methods
  JNIEnvExt* GetJniEnv() const {
    return tlsPtr_.jni_env;
  }

  // Convert a jobject into a Object*
  ObjPtr<mirror::Object> DecodeJObject(jobject obj) const REQUIRES_SHARED(Locks::mutator_lock_);
  // Checks if the weak global ref has been cleared by the GC without decoding it.
  bool IsJWeakCleared(jweak obj) const REQUIRES_SHARED(Locks::mutator_lock_);

  mirror::Object* GetMonitorEnterObject() const REQUIRES_SHARED(Locks::mutator_lock_) {
    return tlsPtr_.monitor_enter_object;
  }

  void SetMonitorEnterObject(mirror::Object* obj) REQUIRES_SHARED(Locks::mutator_lock_) {
    tlsPtr_.monitor_enter_object = obj;
  }

  // Implements java.lang.Thread.interrupted.
  bool Interrupted();
  // Implements java.lang.Thread.isInterrupted.
  bool IsInterrupted();
  void Interrupt(Thread* self) REQUIRES(!wait_mutex_);
  void SetInterrupted(bool i) {
    tls32_.interrupted.store(i, std::memory_order_seq_cst);
  }
  void Notify() REQUIRES(!wait_mutex_);

  ALWAYS_INLINE void PoisonObjectPointers() {
    ++poison_object_cookie_;
  }

  ALWAYS_INLINE static void PoisonObjectPointersIfDebug();

  ALWAYS_INLINE uintptr_t GetPoisonObjectCookie() const {
    return poison_object_cookie_;
  }

  // Parking for 0ns of relative time means an untimed park, negative (though
  // should be handled in java code) returns immediately
  void Park(bool is_absolute, int64_t time) REQUIRES_SHARED(Locks::mutator_lock_);
  void Unpark();

 private:
  void NotifyLocked(Thread* self) REQUIRES(wait_mutex_);

 public:
  Mutex* GetWaitMutex() const LOCK_RETURNED(wait_mutex_) {
    return wait_mutex_;
  }

  ConditionVariable* GetWaitConditionVariable() const REQUIRES(wait_mutex_) {
    return wait_cond_;
  }

  Monitor* GetWaitMonitor() const REQUIRES(wait_mutex_) {
    return wait_monitor_;
  }

  void SetWaitMonitor(Monitor* mon) REQUIRES(wait_mutex_) {
    wait_monitor_ = mon;
  }

  // Waiter link-list support.
  Thread* GetWaitNext() const {
    return tlsPtr_.wait_next;
  }

  void SetWaitNext(Thread* next) {
    tlsPtr_.wait_next = next;
  }

  jobject GetClassLoaderOverride() {
    return tlsPtr_.class_loader_override;
  }

  void SetClassLoaderOverride(jobject class_loader_override);

  // Create the internal representation of a stack trace, that is more time
  // and space efficient to compute than the StackTraceElement[].
  template<bool kTransactionActive>
  jobject CreateInternalStackTrace(const ScopedObjectAccessAlreadyRunnable& soa) const
      REQUIRES_SHARED(Locks::mutator_lock_);

  // Convert an internal stack trace representation (returned by CreateInternalStackTrace) to a
  // StackTraceElement[]. If output_array is null, a new array is created, otherwise as many
  // frames as will fit are written into the given array. If stack_depth is non-null, it's updated
  // with the number of valid frames in the returned array.
  static jobjectArray InternalStackTraceToStackTraceElementArray(
      const ScopedObjectAccessAlreadyRunnable& soa, jobject internal,
      jobjectArray output_array = nullptr, int* stack_depth = nullptr)
      REQUIRES_SHARED(Locks::mutator_lock_);

  jobjectArray CreateAnnotatedStackTrace(const ScopedObjectAccessAlreadyRunnable& soa) const
      REQUIRES_SHARED(Locks::mutator_lock_);

  bool HasDebuggerShadowFrames() const {
    return tlsPtr_.frame_id_to_shadow_frame != nullptr;
  }

  void VisitRoots(RootVisitor* visitor, VisitRootFlags flags)
      REQUIRES_SHARED(Locks::mutator_lock_);

  void VerifyStack() REQUIRES_SHARED(Locks::mutator_lock_) {
    if (kVerifyStack) {
      VerifyStackImpl();
    }
  }

  //
  // Offsets of various members of native Thread class, used by compiled code.
  //

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> ThinLockIdOffset() {
    return ThreadOffset<pointer_size>(
        OFFSETOF_MEMBER(Thread, tls32_) +
        OFFSETOF_MEMBER(tls_32bit_sized_values, thin_lock_thread_id));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> InterruptedOffset() {
    return ThreadOffset<pointer_size>(
        OFFSETOF_MEMBER(Thread, tls32_) +
        OFFSETOF_MEMBER(tls_32bit_sized_values, interrupted));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> ThreadFlagsOffset() {
    return ThreadOffset<pointer_size>(
        OFFSETOF_MEMBER(Thread, tls32_) +
        OFFSETOF_MEMBER(tls_32bit_sized_values, state_and_flags));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> UseMterpOffset() {
    return ThreadOffset<pointer_size>(
        OFFSETOF_MEMBER(Thread, tls32_) +
        OFFSETOF_MEMBER(tls_32bit_sized_values, use_mterp));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> IsGcMarkingOffset() {
    return ThreadOffset<pointer_size>(
        OFFSETOF_MEMBER(Thread, tls32_) +
        OFFSETOF_MEMBER(tls_32bit_sized_values, is_gc_marking));
  }

  static constexpr size_t IsGcMarkingSize() {
    return sizeof(tls32_.is_gc_marking);
  }

  // Deoptimize the Java stack.
  void DeoptimizeWithDeoptimizationException(JValue* result) REQUIRES_SHARED(Locks::mutator_lock_);

 private:
  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> ThreadOffsetFromTlsPtr(size_t tls_ptr_offset) {
    size_t base = OFFSETOF_MEMBER(Thread, tlsPtr_);
    size_t scale = (pointer_size > kRuntimePointerSize) ?
      static_cast<size_t>(pointer_size) / static_cast<size_t>(kRuntimePointerSize) : 1;
    size_t shrink = (kRuntimePointerSize > pointer_size) ?
      static_cast<size_t>(kRuntimePointerSize) / static_cast<size_t>(pointer_size) : 1;
    return ThreadOffset<pointer_size>(base + ((tls_ptr_offset * scale) / shrink));
  }

 public:
  static uint32_t QuickEntryPointOffsetWithSize(size_t quick_entrypoint_offset,
                                                PointerSize pointer_size) {
    if (pointer_size == PointerSize::k32) {
      return QuickEntryPointOffset<PointerSize::k32>(quick_entrypoint_offset).
          Uint32Value();
    } else {
      return QuickEntryPointOffset<PointerSize::k64>(quick_entrypoint_offset).
          Uint32Value();
    }
  }

  template<PointerSize pointer_size>
  static ThreadOffset<pointer_size> QuickEntryPointOffset(size_t quick_entrypoint_offset) {
    return ThreadOffsetFromTlsPtr<pointer_size>(
        OFFSETOF_MEMBER(tls_ptr_sized_values, quick_entrypoints) + quick_entrypoint_offset);
  }

  template<PointerSize pointer_size>
  static ThreadOffset<pointer_size> JniEntryPointOffset(size_t jni_entrypoint_offset) {
    return ThreadOffsetFromTlsPtr<pointer_size>(
        OFFSETOF_MEMBER(tls_ptr_sized_values, jni_entrypoints) + jni_entrypoint_offset);
  }

  // Return the entry point offset integer value for ReadBarrierMarkRegX, where X is `reg`.
  template <PointerSize pointer_size>
  static int32_t ReadBarrierMarkEntryPointsOffset(size_t reg) {
    // The entry point list defines 30 ReadBarrierMarkRegX entry points.
    DCHECK_LT(reg, 30u);
    // The ReadBarrierMarkRegX entry points are ordered by increasing
    // register number in Thread::tls_Ptr_.quick_entrypoints.
    return QUICK_ENTRYPOINT_OFFSET(pointer_size, pReadBarrierMarkReg00).Int32Value()
        + static_cast<size_t>(pointer_size) * reg;
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> SelfOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(OFFSETOF_MEMBER(tls_ptr_sized_values, self));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> MterpCurrentIBaseOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(
        OFFSETOF_MEMBER(tls_ptr_sized_values, mterp_current_ibase));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> ExceptionOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(OFFSETOF_MEMBER(tls_ptr_sized_values, exception));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> PeerOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(OFFSETOF_MEMBER(tls_ptr_sized_values, opeer));
  }


  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> CardTableOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(OFFSETOF_MEMBER(tls_ptr_sized_values, card_table));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> ThreadSuspendTriggerOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(
        OFFSETOF_MEMBER(tls_ptr_sized_values, suspend_trigger));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> ThreadLocalPosOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(OFFSETOF_MEMBER(tls_ptr_sized_values,
                                                                thread_local_pos));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> ThreadLocalEndOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(OFFSETOF_MEMBER(tls_ptr_sized_values,
                                                                thread_local_end));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> ThreadLocalObjectsOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(OFFSETOF_MEMBER(tls_ptr_sized_values,
                                                                thread_local_objects));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> RosAllocRunsOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(OFFSETOF_MEMBER(tls_ptr_sized_values,
                                                                rosalloc_runs));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> ThreadLocalAllocStackTopOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(OFFSETOF_MEMBER(tls_ptr_sized_values,
                                                                thread_local_alloc_stack_top));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> ThreadLocalAllocStackEndOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(OFFSETOF_MEMBER(tls_ptr_sized_values,
                                                                thread_local_alloc_stack_end));
  }

  // Size of stack less any space reserved for stack overflow
  size_t GetStackSize() const {
    return tlsPtr_.stack_size - (tlsPtr_.stack_end - tlsPtr_.stack_begin);
  }

  ALWAYS_INLINE uint8_t* GetStackEndForInterpreter(bool implicit_overflow_check) const;

  uint8_t* GetStackEnd() const {
    return tlsPtr_.stack_end;
  }

  // Set the stack end to that to be used during a stack overflow
  void SetStackEndForStackOverflow() REQUIRES_SHARED(Locks::mutator_lock_);

  // Set the stack end to that to be used during regular execution
  ALWAYS_INLINE void ResetDefaultStackEnd();

  bool IsHandlingStackOverflow() const {
    return tlsPtr_.stack_end == tlsPtr_.stack_begin;
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> StackEndOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(
        OFFSETOF_MEMBER(tls_ptr_sized_values, stack_end));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> JniEnvOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(
        OFFSETOF_MEMBER(tls_ptr_sized_values, jni_env));
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> TopOfManagedStackOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(
        OFFSETOF_MEMBER(tls_ptr_sized_values, managed_stack) +
        ManagedStack::TaggedTopQuickFrameOffset());
  }

  const ManagedStack* GetManagedStack() const {
    return &tlsPtr_.managed_stack;
  }

  // Linked list recording fragments of managed stack.
  void PushManagedStackFragment(ManagedStack* fragment) {
    tlsPtr_.managed_stack.PushManagedStackFragment(fragment);
  }
  void PopManagedStackFragment(const ManagedStack& fragment) {
    tlsPtr_.managed_stack.PopManagedStackFragment(fragment);
  }

  ALWAYS_INLINE ShadowFrame* PushShadowFrame(ShadowFrame* new_top_frame);
  ALWAYS_INLINE ShadowFrame* PopShadowFrame();

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> TopShadowFrameOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(
        OFFSETOF_MEMBER(tls_ptr_sized_values, managed_stack) +
        ManagedStack::TopShadowFrameOffset());
  }

  // Is the given obj in this thread's stack indirect reference table?
  bool HandleScopeContains(jobject obj) const;

  void HandleScopeVisitRoots(RootVisitor* visitor, pid_t thread_id)
      REQUIRES_SHARED(Locks::mutator_lock_);

  BaseHandleScope* GetTopHandleScope() {
    return tlsPtr_.top_handle_scope;
  }

  void PushHandleScope(BaseHandleScope* handle_scope) {
    DCHECK_EQ(handle_scope->GetLink(), tlsPtr_.top_handle_scope);
    tlsPtr_.top_handle_scope = handle_scope;
  }

  BaseHandleScope* PopHandleScope() {
    BaseHandleScope* handle_scope = tlsPtr_.top_handle_scope;
    DCHECK(handle_scope != nullptr);
    tlsPtr_.top_handle_scope = tlsPtr_.top_handle_scope->GetLink();
    return handle_scope;
  }

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> TopHandleScopeOffset() {
    return ThreadOffsetFromTlsPtr<pointer_size>(OFFSETOF_MEMBER(tls_ptr_sized_values,
                                                                top_handle_scope));
  }

  DebugInvokeReq* GetInvokeReq() const {
    return tlsPtr_.debug_invoke_req;
  }

  SingleStepControl* GetSingleStepControl() const {
    return tlsPtr_.single_step_control;
  }

  // Indicates whether this thread is ready to invoke a method for debugging. This
  // is only true if the thread has been suspended by a debug event.
  bool IsReadyForDebugInvoke() const {
    return tls32_.ready_for_debug_invoke;
  }

  void SetReadyForDebugInvoke(bool ready) {
    tls32_.ready_for_debug_invoke = ready;
  }

  bool IsDebugMethodEntry() const {
    return tls32_.debug_method_entry_;
  }

  void SetDebugMethodEntry() {
    tls32_.debug_method_entry_ = true;
  }

  void ClearDebugMethodEntry() {
    tls32_.debug_method_entry_ = false;
  }

  bool GetIsGcMarking() const {
    CHECK(kUseReadBarrier);
    return tls32_.is_gc_marking;
  }

  void SetIsGcMarkingAndUpdateEntrypoints(bool is_marking);

  bool GetWeakRefAccessEnabled() const {
    CHECK(kUseReadBarrier);
    return tls32_.weak_ref_access_enabled;
  }

  void SetWeakRefAccessEnabled(bool enabled) {
    CHECK(kUseReadBarrier);
    tls32_.weak_ref_access_enabled = enabled;
  }

  uint32_t GetDisableThreadFlipCount() const {
    CHECK(kUseReadBarrier);
    return tls32_.disable_thread_flip_count;
  }

  void IncrementDisableThreadFlipCount() {
    CHECK(kUseReadBarrier);
    ++tls32_.disable_thread_flip_count;
  }

  void DecrementDisableThreadFlipCount() {
    CHECK(kUseReadBarrier);
    DCHECK_GT(tls32_.disable_thread_flip_count, 0U);
    --tls32_.disable_thread_flip_count;
  }

  // Returns true if the thread is a runtime thread (eg from a ThreadPool).
  bool IsRuntimeThread() const {
    return is_runtime_thread_;
  }

  void SetIsRuntimeThread(bool is_runtime_thread) {
    is_runtime_thread_ = is_runtime_thread;
  }

  // Returns true if the thread is allowed to load java classes.
  bool CanLoadClasses() const;

  // Activates single step control for debugging. The thread takes the
  // ownership of the given SingleStepControl*. It is deleted by a call
  // to DeactivateSingleStepControl or upon thread destruction.
  void ActivateSingleStepControl(SingleStepControl* ssc);

  // Deactivates single step control for debugging.
  void DeactivateSingleStepControl();

  // Sets debug invoke request for debugging. When the thread is resumed,
  // it executes the method described by this request then sends the reply
  // before suspending itself. The thread takes the ownership of the given
  // DebugInvokeReq*. It is deleted by a call to ClearDebugInvokeReq.
  void SetDebugInvokeReq(DebugInvokeReq* req);

  // Clears debug invoke request for debugging. When the thread completes
  // method invocation, it deletes its debug invoke request and suspends
  // itself.
  void ClearDebugInvokeReq();

  // Returns the fake exception used to activate deoptimization.
  static mirror::Throwable* GetDeoptimizationException() {
    // Note that the mirror::Throwable must be aligned to kObjectAlignment or else it cannot be
    // represented by ObjPtr.
    return reinterpret_cast<mirror::Throwable*>(0x100);
  }

  // Currently deoptimization invokes verifier which can trigger class loading
  // and execute Java code, so there might be nested deoptimizations happening.
  // We need to save the ongoing deoptimization shadow frames and return
  // values on stacks.
  // 'from_code' denotes whether the deoptimization was explicitly made from
  // compiled code.
  // 'method_type' contains info on whether deoptimization should advance
  // dex_pc.
  void PushDeoptimizationContext(const JValue& return_value,
                                 bool is_reference,
                                 ObjPtr<mirror::Throwable> exception,
                                 bool from_code,
                                 DeoptimizationMethodType method_type)
      REQUIRES_SHARED(Locks::mutator_lock_);
  void PopDeoptimizationContext(JValue* result,
                                ObjPtr<mirror::Throwable>* exception,
                                bool* from_code,
                                DeoptimizationMethodType* method_type)
      REQUIRES_SHARED(Locks::mutator_lock_);
  void AssertHasDeoptimizationContext()
      REQUIRES_SHARED(Locks::mutator_lock_);
  void PushStackedShadowFrame(ShadowFrame* sf, StackedShadowFrameType type);
  ShadowFrame* PopStackedShadowFrame(StackedShadowFrameType type, bool must_be_present = true);

  // For debugger, find the shadow frame that corresponds to a frame id.
  // Or return null if there is none.
  ShadowFrame* FindDebuggerShadowFrame(size_t frame_id)
      REQUIRES_SHARED(Locks::mutator_lock_);
  // For debugger, find the bool array that keeps track of the updated vreg set
  // for a frame id.
  bool* GetUpdatedVRegFlags(size_t frame_id) REQUIRES_SHARED(Locks::mutator_lock_);
  // For debugger, find the shadow frame that corresponds to a frame id. If
  // one doesn't exist yet, create one and track it in frame_id_to_shadow_frame.
  ShadowFrame* FindOrCreateDebuggerShadowFrame(size_t frame_id,
                                               uint32_t num_vregs,
                                               ArtMethod* method,
                                               uint32_t dex_pc)
      REQUIRES_SHARED(Locks::mutator_lock_);

  // Delete the entry that maps from frame_id to shadow_frame.
  void RemoveDebuggerShadowFrameMapping(size_t frame_id)
      REQUIRES_SHARED(Locks::mutator_lock_);

  std::deque<instrumentation::InstrumentationStackFrame>* GetInstrumentationStack() {
    return tlsPtr_.instrumentation_stack;
  }

  std::vector<ArtMethod*>* GetStackTraceSample() const {
    DCHECK(!IsAotCompiler());
    return tlsPtr_.deps_or_stack_trace_sample.stack_trace_sample;
  }

  void SetStackTraceSample(std::vector<ArtMethod*>* sample) {
    DCHECK(!IsAotCompiler());
    tlsPtr_.deps_or_stack_trace_sample.stack_trace_sample = sample;
  }

  verifier::VerifierDeps* GetVerifierDeps() const {
    DCHECK(IsAotCompiler());
    return tlsPtr_.deps_or_stack_trace_sample.verifier_deps;
  }

  // It is the responsability of the caller to make sure the verifier_deps
  // entry in the thread is cleared before destruction of the actual VerifierDeps
  // object, or the thread.
  void SetVerifierDeps(verifier::VerifierDeps* verifier_deps) {
    DCHECK(IsAotCompiler());
    DCHECK(verifier_deps == nullptr || tlsPtr_.deps_or_stack_trace_sample.verifier_deps == nullptr);
    tlsPtr_.deps_or_stack_trace_sample.verifier_deps = verifier_deps;
  }

  uint64_t GetTraceClockBase() const {
    return tls64_.trace_clock_base;
  }

  void SetTraceClockBase(uint64_t clock_base) {
    tls64_.trace_clock_base = clock_base;
  }

  BaseMutex* GetHeldMutex(LockLevel level) const {
    return tlsPtr_.held_mutexes[level];
  }

  void SetHeldMutex(LockLevel level, BaseMutex* mutex) {
    tlsPtr_.held_mutexes[level] = mutex;
  }

  void ClearSuspendBarrier(AtomicInteger* target)
      REQUIRES(Locks::thread_suspend_count_lock_);

  bool ReadFlag(ThreadFlag flag) const {
    return (tls32_.state_and_flags.as_struct.flags & flag) != 0;
  }

  bool TestAllFlags() const {
    return (tls32_.state_and_flags.as_struct.flags != 0);
  }

  void AtomicSetFlag(ThreadFlag flag) {
    tls32_.state_and_flags.as_atomic_int.fetch_or(flag, std::memory_order_seq_cst);
  }

  void AtomicClearFlag(ThreadFlag flag) {
    tls32_.state_and_flags.as_atomic_int.fetch_and(-1 ^ flag, std::memory_order_seq_cst);
  }

  bool UseMterp() const {
    return tls32_.use_mterp.load();
  }

  void ResetQuickAllocEntryPointsForThread(bool is_marking);

  // Returns the remaining space in the TLAB.
  size_t TlabSize() const {
    return tlsPtr_.thread_local_end - tlsPtr_.thread_local_pos;
  }

  // Returns the remaining space in the TLAB if we were to expand it to maximum capacity.
  size_t TlabRemainingCapacity() const {
    return tlsPtr_.thread_local_limit - tlsPtr_.thread_local_pos;
  }

  // Expand the TLAB by a fixed number of bytes. There must be enough capacity to do so.
  void ExpandTlab(size_t bytes) {
    tlsPtr_.thread_local_end += bytes;
    DCHECK_LE(tlsPtr_.thread_local_end, tlsPtr_.thread_local_limit);
  }

  // Doesn't check that there is room.
  mirror::Object* AllocTlab(size_t bytes);
  void SetTlab(uint8_t* start, uint8_t* end, uint8_t* limit);
  bool HasTlab() const;
  uint8_t* GetTlabStart() {
    return tlsPtr_.thread_local_start;
  }
  uint8_t* GetTlabPos() {
    return tlsPtr_.thread_local_pos;
  }

  // Remove the suspend trigger for this thread by making the suspend_trigger_ TLS value
  // equal to a valid pointer.
  // TODO: does this need to atomic?  I don't think so.
  void RemoveSuspendTrigger() {
    tlsPtr_.suspend_trigger = reinterpret_cast<uintptr_t*>(&tlsPtr_.suspend_trigger);
  }

  // Trigger a suspend check by making the suspend_trigger_ TLS value an invalid pointer.
  // The next time a suspend check is done, it will load from the value at this address
  // and trigger a SIGSEGV.
  void TriggerSuspend() {
    tlsPtr_.suspend_trigger = nullptr;
  }


  // Push an object onto the allocation stack.
  bool PushOnThreadLocalAllocationStack(mirror::Object* obj)
      REQUIRES_SHARED(Locks::mutator_lock_);

  // Set the thread local allocation pointers to the given pointers.
  void SetThreadLocalAllocationStack(StackReference<mirror::Object>* start,
                                     StackReference<mirror::Object>* end);

  // Resets the thread local allocation pointers.
  void RevokeThreadLocalAllocationStack();

  size_t GetThreadLocalBytesAllocated() const {
    return tlsPtr_.thread_local_end - tlsPtr_.thread_local_start;
  }

  size_t GetThreadLocalObjectsAllocated() const {
    return tlsPtr_.thread_local_objects;
  }

  void* GetRosAllocRun(size_t index) const {
    return tlsPtr_.rosalloc_runs[index];
  }

  void SetRosAllocRun(size_t index, void* run) {
    tlsPtr_.rosalloc_runs[index] = run;
  }

  bool ProtectStack(bool fatal_on_error = true);
  bool UnprotectStack();

  void SetMterpCurrentIBase(void* ibase) {
    tlsPtr_.mterp_current_ibase = ibase;
  }

  const void* GetMterpCurrentIBase() const {
    return tlsPtr_.mterp_current_ibase;
  }

  bool HandlingSignal() const {
    return tls32_.handling_signal_;
  }

  void SetHandlingSignal(bool handling_signal) {
    tls32_.handling_signal_ = handling_signal;
  }

  bool IsTransitioningToRunnable() const {
    return tls32_.is_transitioning_to_runnable;
  }

  void SetIsTransitioningToRunnable(bool value) {
    tls32_.is_transitioning_to_runnable = value;
  }

  void PushVerifier(verifier::MethodVerifier* verifier);
  void PopVerifier(verifier::MethodVerifier* verifier);

  void InitStringEntryPoints();

  void ModifyDebugDisallowReadBarrier(int8_t delta) {
    debug_disallow_read_barrier_ += delta;
  }

  uint8_t GetDebugDisallowReadBarrierCount() const {
    return debug_disallow_read_barrier_;
  }

  // Gets the current TLSData associated with the key or nullptr if there isn't any. Note that users
  // do not gain ownership of TLSData and must synchronize with SetCustomTls themselves to prevent
  // it from being deleted.
  TLSData* GetCustomTLS(const char* key) REQUIRES(!Locks::custom_tls_lock_);

  // Sets the tls entry at 'key' to data. The thread takes ownership of the TLSData. The destructor
  // will be run when the thread exits or when SetCustomTLS is called again with the same key.
  void SetCustomTLS(const char* key, TLSData* data) REQUIRES(!Locks::custom_tls_lock_);

  // Returns true if the current thread is the jit sensitive thread.
  bool IsJitSensitiveThread() const {
    return this == jit_sensitive_thread_;
  }

  bool IsSystemDaemon() const REQUIRES_SHARED(Locks::mutator_lock_);

  // Returns true if StrictMode events are traced for the current thread.
  static bool IsSensitiveThread() {
    if (is_sensitive_thread_hook_ != nullptr) {
      return (*is_sensitive_thread_hook_)();
    }
    return false;
  }

  // Set to the read barrier marking entrypoints to be non-null.
  void SetReadBarrierEntrypoints();

  static jobject CreateCompileTimePeer(JNIEnv* env,
                                       const char* name,
                                       bool as_daemon,
                                       jobject thread_group)
      REQUIRES_SHARED(Locks::mutator_lock_);

  ALWAYS_INLINE InterpreterCache* GetInterpreterCache() {
    return &interpreter_cache_;
  }

  // Clear all thread-local interpreter caches.
  //
  // Since the caches are keyed by memory pointer to dex instructions, this must be
  // called when any dex code is unloaded (before different code gets loaded at the
  // same memory location).
  //
  // If presence of cache entry implies some pre-conditions, this must also be
  // called if the pre-conditions might no longer hold true.
  static void ClearAllInterpreterCaches();

  template<PointerSize pointer_size>
  static constexpr ThreadOffset<pointer_size> InterpreterCacheOffset() {
    return ThreadOffset<pointer_size>(OFFSETOF_MEMBER(Thread, interpreter_cache_));
  }

  static constexpr int InterpreterCacheSizeLog2() {
    return WhichPowerOf2(InterpreterCache::kSize);
  }

 private:
  explicit Thread(bool daemon);
  ~Thread() REQUIRES(!Locks::mutator_lock_, !Locks::thread_suspend_count_lock_);
  void Destroy();

  void NotifyInTheadList()
      REQUIRES_SHARED(Locks::thread_list_lock_);

  // Attaches the calling native thread to the runtime, returning the new native peer.
  // Used to implement JNI AttachCurrentThread and AttachCurrentThreadAsDaemon calls.
  template <typename PeerAction>
  static Thread* Attach(const char* thread_name,
                        bool as_daemon,
                        PeerAction p);

  void CreatePeer(const char* name, bool as_daemon, jobject thread_group);

  template<bool kTransactionActive>
  static void InitPeer(ScopedObjectAccessAlreadyRunnable& soa,
                       ObjPtr<mirror::Object> peer,
                       jboolean thread_is_daemon,
                       jobject thread_group,
                       jobject thread_name,
                       jint thread_priority)
      REQUIRES_SHARED(Locks::mutator_lock_);

  // Avoid use, callers should use SetState. Used only by SignalCatcher::HandleSigQuit, ~Thread and
  // Dbg::ManageDeoptimization.
  ThreadState SetStateUnsafe(ThreadState new_state) {
    ThreadState old_state = GetState();
    if (old_state == kRunnable && new_state != kRunnable) {
      // Need to run pending checkpoint and suspend barriers. Run checkpoints in runnable state in
      // case they need to use a ScopedObjectAccess. If we are holding the mutator lock and a SOA
      // attempts to TransitionFromSuspendedToRunnable, it results in a deadlock.
      TransitionToSuspendedAndRunCheckpoints(new_state);
      // Since we transitioned to a suspended state, check the pass barrier requests.
      PassActiveSuspendBarriers();
    } else {
      tls32_.state_and_flags.as_struct.state = new_state;
    }
    return old_state;
  }

  void VerifyStackImpl() REQUIRES_SHARED(Locks::mutator_lock_);

  void DumpState(std::ostream& os) const REQUIRES_SHARED(Locks::mutator_lock_);
  void DumpStack(std::ostream& os,
                 bool dump_native_stack = true,
                 BacktraceMap* backtrace_map = nullptr,
                 bool force_dump_stack = false) const
      REQUIRES(!Locks::thread_suspend_count_lock_)
      REQUIRES_SHARED(Locks::mutator_lock_);

  // Out-of-line conveniences for debugging in gdb.
  static Thread* CurrentFromGdb();  // Like Thread::Current.
  // Like Thread::Dump(std::cerr).
  void DumpFromGdb() const REQUIRES_SHARED(Locks::mutator_lock_);

  static void* CreateCallback(void* arg);

  void HandleUncaughtExceptions(ScopedObjectAccessAlreadyRunnable& soa)
      REQUIRES_SHARED(Locks::mutator_lock_);
  void RemoveFromThreadGroup(ScopedObjectAccessAlreadyRunnable& soa)
      REQUIRES_SHARED(Locks::mutator_lock_);

  // Initialize a thread.
  //
  // The third parameter is not mandatory. If given, the thread will use this JNIEnvExt. In case
  // Init succeeds, this means the thread takes ownership of it. If Init fails, it is the caller's
  // responsibility to destroy the given JNIEnvExt. If the parameter is null, Init will try to
  // create a JNIEnvExt on its own (and potentially fail at that stage, indicated by a return value
  // of false).
  bool Init(ThreadList*, JavaVMExt*, JNIEnvExt* jni_env_ext = nullptr)
      REQUIRES(Locks::runtime_shutdown_lock_);
  void InitCardTable();
  void InitCpu();
  void CleanupCpu();
  void InitTlsEntryPoints();
  void InitTid();
  void InitPthreadKeySelf();
  bool InitStackHwm();

  void SetUpAlternateSignalStack();
  void TearDownAlternateSignalStack();

  ALWAYS_INLINE void TransitionToSuspendedAndRunCheckpoints(ThreadState new_state)
      REQUIRES(!Locks::thread_suspend_count_lock_, !Roles::uninterruptible_);

  ALWAYS_INLINE void PassActiveSuspendBarriers()
      REQUIRES(!Locks::thread_suspend_count_lock_, !Roles::uninterruptible_);

  // Registers the current thread as the jit sensitive thread. Should be called just once.
  static void SetJitSensitiveThread() {
    if (jit_sensitive_thread_ == nullptr) {
      jit_sensitive_thread_ = Thread::Current();
    } else {
      LOG(WARNING) << "Attempt to set the sensitive thread twice. Tid:"
          << Thread::Current()->GetTid();
    }
  }

  static void SetSensitiveThreadHook(bool (*is_sensitive_thread_hook)()) {
    is_sensitive_thread_hook_ = is_sensitive_thread_hook;
  }

  bool ModifySuspendCountInternal(Thread* self,
                                  int delta,
                                  AtomicInteger* suspend_barrier,
                                  SuspendReason reason)
      WARN_UNUSED
      REQUIRES(Locks::thread_suspend_count_lock_);

  // Runs a single checkpoint function. If there are no more pending checkpoint functions it will
  // clear the kCheckpointRequest flag. The caller is responsible for calling this in a loop until
  // the kCheckpointRequest flag is cleared.
  void RunCheckpointFunction();
  void RunEmptyCheckpoint();

  bool PassActiveSuspendBarriers(Thread* self)
      REQUIRES(!Locks::thread_suspend_count_lock_);

  // Install the protected region for implicit stack checks.
  void InstallImplicitProtection();

  template <bool kPrecise>
  void VisitRoots(RootVisitor* visitor) REQUIRES_SHARED(Locks::mutator_lock_);

  static bool IsAotCompiler();

  void ReleaseLongJumpContextInternal();

  // 32 bits of atomically changed state and flags. Keeping as 32 bits allows and atomic CAS to
  // change from being Suspended to Runnable without a suspend request occurring.
  union PACKED(4) StateAndFlags {
    StateAndFlags() {}
    struct PACKED(4) {
      // Bitfield of flag values. Must be changed atomically so that flag values aren't lost. See
      // ThreadFlags for bit field meanings.
      volatile uint16_t flags;
      // Holds the ThreadState. May be changed non-atomically between Suspended (ie not Runnable)
      // transitions. Changing to Runnable requires that the suspend_request be part of the atomic
      // operation. If a thread is suspended and a suspend_request is present, a thread may not
      // change to Runnable as a GC or other operation is in progress.
      volatile uint16_t state;
    } as_struct;
    AtomicInteger as_atomic_int;
    volatile int32_t as_int;

   private:
    // gcc does not handle struct with volatile member assignments correctly.
    // See http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47409
    DISALLOW_COPY_AND_ASSIGN(StateAndFlags);
  };
  static_assert(sizeof(StateAndFlags) == sizeof(int32_t), "Weird state_and_flags size");

  static void ThreadExitCallback(void* arg);

  // Maximum number of suspend barriers.
  static constexpr uint32_t kMaxSuspendBarriers = 3;

  // Has Thread::Startup been called?
  static bool is_started_;

  // TLS key used to retrieve the Thread*.
  static pthread_key_t pthread_key_self_;

  // Used to notify threads that they should attempt to resume, they will suspend again if
  // their suspend count is > 0.
  static ConditionVariable* resume_cond_ GUARDED_BY(Locks::thread_suspend_count_lock_);

  // Hook passed by framework which returns true
  // when StrictMode events are traced for the current thread.
  static bool (*is_sensitive_thread_hook_)();
  // Stores the jit sensitive thread (which for now is the UI thread).
  static Thread* jit_sensitive_thread_;

  /***********************************************************************************************/
  // Thread local storage. Fields are grouped by size to enable 32 <-> 64 searching to account for
  // pointer size differences. To encourage shorter encoding, more frequently used values appear
  // first if possible.
  /***********************************************************************************************/

  struct PACKED(4) tls_32bit_sized_values {
    // We have no control over the size of 'bool', but want our boolean fields
    // to be 4-byte quantities.
    typedef uint32_t bool32_t;

    explicit tls_32bit_sized_values(bool is_daemon) :
      suspend_count(0), debug_suspend_count(0), thin_lock_thread_id(0), tid(0),
      daemon(is_daemon), throwing_OutOfMemoryError(false), no_thread_suspension(0),
      thread_exit_check_count(0), handling_signal_(false),
      is_transitioning_to_runnable(false), ready_for_debug_invoke(false),
      debug_method_entry_(false), is_gc_marking(false), weak_ref_access_enabled(true),
      disable_thread_flip_count(0), user_code_suspend_count(0) {
    }

    union StateAndFlags state_and_flags;
    static_assert(sizeof(union StateAndFlags) == sizeof(int32_t),
                  "Size of state_and_flags and int32 are different");

    // A non-zero value is used to tell the current thread to enter a safe point
    // at the next poll.
    int suspend_count GUARDED_BY(Locks::thread_suspend_count_lock_);

    // How much of 'suspend_count_' is by request of the debugger, used to set things right
    // when the debugger detaches. Must be <= suspend_count_.
    int debug_suspend_count GUARDED_BY(Locks::thread_suspend_count_lock_);

    // Thin lock thread id. This is a small integer used by the thin lock implementation.
    // This is not to be confused with the native thread's tid, nor is it the value returned
    // by java.lang.Thread.getId --- this is a distinct value, used only for locking. One
    // important difference between this id and the ids visible to managed code is that these
    // ones get reused (to ensure that they fit in the number of bits available).
    uint32_t thin_lock_thread_id;

    // System thread id.
    uint32_t tid;

    // Is the thread a daemon?
    const bool32_t daemon;

    // A boolean telling us whether we're recursively throwing OOME.
    bool32_t throwing_OutOfMemoryError;

    // A positive value implies we're in a region where thread suspension isn't expected.
    uint32_t no_thread_suspension;

    // How many times has our pthread key's destructor been called?
    uint32_t thread_exit_check_count;

    // True if signal is being handled by this thread.
    bool32_t handling_signal_;

    // True if the thread is in TransitionFromSuspendedToRunnable(). This is used to distinguish the
    // non-runnable threads (eg. kNative, kWaiting) that are about to transition to runnable from
    // the rest of them.
    bool32_t is_transitioning_to_runnable;

    // True if the thread has been suspended by a debugger event. This is
    // used to invoke method from the debugger which is only allowed when
    // the thread is suspended by an event.
    bool32_t ready_for_debug_invoke;

    // True if the thread enters a method. This is used to detect method entry
    // event for the debugger.
    bool32_t debug_method_entry_;

    // True if the GC is in the marking phase. This is used for the CC collector only. This is
    // thread local so that we can simplify the logic to check for the fast path of read barriers of
    // GC roots.
    bool32_t is_gc_marking;

    // Thread "interrupted" status; stays raised until queried or thrown.
    Atomic<bool32_t> interrupted;

    AtomicInteger park_state_;

    // True if the thread is allowed to access a weak ref (Reference::GetReferent() and system
    // weaks) and to potentially mark an object alive/gray. This is used for concurrent reference
    // processing of the CC collector only. This is thread local so that we can enable/disable weak
    // ref access by using a checkpoint and avoid a race around the time weak ref access gets
    // disabled and concurrent reference processing begins (if weak ref access is disabled during a
    // pause, this is not an issue.) Other collectors use Runtime::DisallowNewSystemWeaks() and
    // ReferenceProcessor::EnableSlowPath().
    bool32_t weak_ref_access_enabled;

    // A thread local version of Heap::disable_thread_flip_count_. This keeps track of how many
    // levels of (nested) JNI critical sections the thread is in and is used to detect a nested JNI
    // critical section enter.
    uint32_t disable_thread_flip_count;

    // How much of 'suspend_count_' is by request of user code, used to distinguish threads
    // suspended by the runtime from those suspended by user code.
    // This should have GUARDED_BY(Locks::user_code_suspension_lock_) but auto analysis cannot be
    // told that AssertHeld should be good enough.
    int user_code_suspend_count GUARDED_BY(Locks::thread_suspend_count_lock_);

    // True if everything is in the ideal state for fast interpretation.
    // False if we need to switch to the C++ interpreter to handle special cases.
    std::atomic<bool32_t> use_mterp;
  } tls32_;

  struct PACKED(8) tls_64bit_sized_values {
    tls_64bit_sized_values() : trace_clock_base(0) {
    }

    // The clock base used for tracing.
    uint64_t trace_clock_base;

    RuntimeStats stats;
  } tls64_;

  struct PACKED(sizeof(void*)) tls_ptr_sized_values {
      tls_ptr_sized_values() : card_table(nullptr), exception(nullptr), stack_end(nullptr),
      managed_stack(), suspend_trigger(nullptr), jni_env(nullptr), tmp_jni_env(nullptr),
      self(nullptr), opeer(nullptr), jpeer(nullptr), stack_begin(nullptr), stack_size(0),
      deps_or_stack_trace_sample(), wait_next(nullptr), monitor_enter_object(nullptr),
      top_handle_scope(nullptr), class_loader_override(nullptr), long_jump_context(nullptr),
      instrumentation_stack(nullptr), debug_invoke_req(nullptr), single_step_control(nullptr),
      stacked_shadow_frame_record(nullptr), deoptimization_context_stack(nullptr),
      frame_id_to_shadow_frame(nullptr), name(nullptr), pthread_self(0),
      last_no_thread_suspension_cause(nullptr), checkpoint_function(nullptr),
      thread_local_start(nullptr), thread_local_pos(nullptr), thread_local_end(nullptr),
      thread_local_limit(nullptr),
      thread_local_objects(0), mterp_current_ibase(nullptr), thread_local_alloc_stack_top(nullptr),
      thread_local_alloc_stack_end(nullptr),
      flip_function(nullptr), method_verifier(nullptr), thread_local_mark_stack(nullptr),
      async_exception(nullptr) {
      std::fill(held_mutexes, held_mutexes + kLockLevelCount, nullptr);
    }

    // The biased card table, see CardTable for details.
    uint8_t* card_table;

    // The pending exception or null.
    mirror::Throwable* exception;

    // The end of this thread's stack. This is the lowest safely-addressable address on the stack.
    // We leave extra space so there's room for the code that throws StackOverflowError.
    uint8_t* stack_end;

    // The top of the managed stack often manipulated directly by compiler generated code.
    ManagedStack managed_stack;

    // In certain modes, setting this to 0 will trigger a SEGV and thus a suspend check.  It is
    // normally set to the address of itself.
    uintptr_t* suspend_trigger;

    // Every thread may have an associated JNI environment
    JNIEnvExt* jni_env;

    // Temporary storage to transfer a pre-allocated JNIEnvExt from the creating thread to the
    // created thread.
    JNIEnvExt* tmp_jni_env;

    // Initialized to "this". On certain architectures (such as x86) reading off of Thread::Current
    // is easy but getting the address of Thread::Current is hard. This field can be read off of
    // Thread::Current to give the address.
    Thread* self;

    // Our managed peer (an instance of java.lang.Thread). The jobject version is used during thread
    // start up, until the thread is registered and the local opeer_ is used.
    mirror::Object* opeer;
    jobject jpeer;

    // The "lowest addressable byte" of the stack.
    uint8_t* stack_begin;

    // Size of the stack.
    size_t stack_size;

    // Sampling profiler and AOT verification cannot happen on the same run, so we share
    // the same entry for the stack trace and the verifier deps.
    union DepsOrStackTraceSample {
      DepsOrStackTraceSample() {
        verifier_deps = nullptr;
        stack_trace_sample = nullptr;
      }
      // Pointer to previous stack trace captured by sampling profiler.
      std::vector<ArtMethod*>* stack_trace_sample;
      // When doing AOT verification, per-thread VerifierDeps.
      verifier::VerifierDeps* verifier_deps;
    } deps_or_stack_trace_sample;

    // The next thread in the wait set this thread is part of or null if not waiting.
    Thread* wait_next;

    // If we're blocked in MonitorEnter, this is the object we're trying to lock.
    mirror::Object* monitor_enter_object;

    // Top of linked list of handle scopes or null for none.
    BaseHandleScope* top_handle_scope;

    // Needed to get the right ClassLoader in JNI_OnLoad, but also
    // useful for testing.
    jobject class_loader_override;

    // Thread local, lazily allocated, long jump context. Used to deliver exceptions.
    Context* long_jump_context;

    // Additional stack used by method instrumentation to store method and return pc values.
    // Stored as a pointer since std::deque is not PACKED.
    std::deque<instrumentation::InstrumentationStackFrame>* instrumentation_stack;

    // JDWP invoke-during-breakpoint support.
    DebugInvokeReq* debug_invoke_req;

    // JDWP single-stepping support.
    SingleStepControl* single_step_control;

    // For gc purpose, a shadow frame record stack that keeps track of:
    // 1) shadow frames under construction.
    // 2) deoptimization shadow frames.
    StackedShadowFrameRecord* stacked_shadow_frame_record;

    // Deoptimization return value record stack.
    DeoptimizationContextRecord* deoptimization_context_stack;

    // For debugger, a linked list that keeps the mapping from frame_id to shadow frame.
    // Shadow frames may be created before deoptimization happens so that the debugger can
    // set local values there first.
    FrameIdToShadowFrame* frame_id_to_shadow_frame;

    // A cached copy of the java.lang.Thread's name.
    std::string* name;

    // A cached pthread_t for the pthread underlying this Thread*.
    pthread_t pthread_self;

    // If no_thread_suspension_ is > 0, what is causing that assertion.
    const char* last_no_thread_suspension_cause;

    // Pending checkpoint function or null if non-pending. If this checkpoint is set and someone\
    // requests another checkpoint, it goes to the checkpoint overflow list.
    Closure* checkpoint_function GUARDED_BY(Locks::thread_suspend_count_lock_);

    // Pending barriers that require passing or NULL if non-pending. Installation guarding by
    // Locks::thread_suspend_count_lock_.
    // They work effectively as art::Barrier, but implemented directly using AtomicInteger and futex
    // to avoid additional cost of a mutex and a condition variable, as used in art::Barrier.
    AtomicInteger* active_suspend_barriers[kMaxSuspendBarriers];

    // Thread-local allocation pointer. Moved here to force alignment for thread_local_pos on ARM.
    uint8_t* thread_local_start;

    // thread_local_pos and thread_local_end must be consecutive for ldrd and are 8 byte aligned for
    // potentially better performance.
    uint8_t* thread_local_pos;
    uint8_t* thread_local_end;

    // Thread local limit is how much we can expand the thread local buffer to, it is greater or
    // equal to thread_local_end.
    uint8_t* thread_local_limit;

    size_t thread_local_objects;

    // Entrypoint function pointers.
    // TODO: move this to more of a global offset table model to avoid per-thread duplication.
    JniEntryPoints jni_entrypoints;
    QuickEntryPoints quick_entrypoints;

    // Mterp jump table base.
    void* mterp_current_ibase;

    // There are RosAlloc::kNumThreadLocalSizeBrackets thread-local size brackets per thread.
    void* rosalloc_runs[kNumRosAllocThreadLocalSizeBracketsInThread];

    // Thread-local allocation stack data/routines.
    StackReference<mirror::Object>* thread_local_alloc_stack_top;
    StackReference<mirror::Object>* thread_local_alloc_stack_end;

    // Support for Mutex lock hierarchy bug detection.
    BaseMutex* held_mutexes[kLockLevelCount];

    // The function used for thread flip.
    Closure* flip_function;

    // Current method verifier, used for root marking.
    verifier::MethodVerifier* method_verifier;

    // Thread-local mark stack for the concurrent copying collector.
    gc::accounting::AtomicStack<mirror::Object>* thread_local_mark_stack;

    // The pending async-exception or null.
    mirror::Throwable* async_exception;
  } tlsPtr_;

  // Small thread-local cache to be used from the interpreter.
  // It is keyed by dex instruction pointer.
  // The value is opcode-depended (e.g. field offset).
  InterpreterCache interpreter_cache_;

  // All fields below this line should not be accessed by native code. This means these fields can
  // be modified, rearranged, added or removed without having to modify asm_support.h

  // Guards the 'wait_monitor_' members.
  Mutex* wait_mutex_ DEFAULT_MUTEX_ACQUIRED_AFTER;

  // Condition variable waited upon during a wait.
  ConditionVariable* wait_cond_ GUARDED_BY(wait_mutex_);
  // Pointer to the monitor lock we're currently waiting on or null if not waiting.
  Monitor* wait_monitor_ GUARDED_BY(wait_mutex_);

  // Debug disable read barrier count, only is checked for debug builds and only in the runtime.
  uint8_t debug_disallow_read_barrier_ = 0;

  // Note that it is not in the packed struct, may not be accessed for cross compilation.
  uintptr_t poison_object_cookie_ = 0;

  // Pending extra checkpoints if checkpoint_function_ is already used.
  std::list<Closure*> checkpoint_overflow_ GUARDED_BY(Locks::thread_suspend_count_lock_);

  // Custom TLS field that can be used by plugins or the runtime. Should not be accessed directly by
  // compiled code or entrypoints.
  SafeMap<std::string, std::unique_ptr<TLSData>> custom_tls_ GUARDED_BY(Locks::custom_tls_lock_);

  // True if the thread is some form of runtime thread (ex, GC or JIT).
  bool is_runtime_thread_;

  friend class Dbg;  // For SetStateUnsafe.
  friend class gc::collector::SemiSpace;  // For getting stack traces.
  friend class Runtime;  // For CreatePeer.
  friend class QuickExceptionHandler;  // For dumping the stack.
  friend class ScopedThreadStateChange;
  friend class StubTest;  // For accessing entrypoints.
  friend class ThreadList;  // For ~Thread and Destroy.

  friend class EntrypointsOrderTest;  // To test the order of tls entries.

  DISALLOW_COPY_AND_ASSIGN(Thread);
};

class SCOPED_CAPABILITY ScopedAssertNoThreadSuspension {
 public:
  ALWAYS_INLINE ScopedAssertNoThreadSuspension(const char* cause,
                                               bool enabled = true)
      ACQUIRE(Roles::uninterruptible_)
      : enabled_(enabled) {
    if (!enabled_) {
      return;
    }
    if (kIsDebugBuild) {
      self_ = Thread::Current();
      old_cause_ = self_->StartAssertNoThreadSuspension(cause);
    } else {
      Roles::uninterruptible_.Acquire();  // No-op.
    }
  }
  ALWAYS_INLINE ~ScopedAssertNoThreadSuspension() RELEASE(Roles::uninterruptible_) {
    if (!enabled_) {
      return;
    }
    if (kIsDebugBuild) {
      self_->EndAssertNoThreadSuspension(old_cause_);
    } else {
      Roles::uninterruptible_.Release();  // No-op.
    }
  }

 private:
  Thread* self_;
  const bool enabled_;
  const char* old_cause_;
};

class ScopedStackedShadowFramePusher {
 public:
  ScopedStackedShadowFramePusher(Thread* self, ShadowFrame* sf, StackedShadowFrameType type)
    : self_(self), type_(type) {
    self_->PushStackedShadowFrame(sf, type);
  }
  ~ScopedStackedShadowFramePusher() {
    self_->PopStackedShadowFrame(type_);
  }

 private:
  Thread* const self_;
  const StackedShadowFrameType type_;

  DISALLOW_COPY_AND_ASSIGN(ScopedStackedShadowFramePusher);
};

// Only works for debug builds.
class ScopedDebugDisallowReadBarriers {
 public:
  explicit ScopedDebugDisallowReadBarriers(Thread* self) : self_(self) {
    self_->ModifyDebugDisallowReadBarrier(1);
  }
  ~ScopedDebugDisallowReadBarriers() {
    self_->ModifyDebugDisallowReadBarrier(-1);
  }

 private:
  Thread* const self_;
};

class ScopedTransitioningToRunnable : public ValueObject {
 public:
  explicit ScopedTransitioningToRunnable(Thread* self)
      : self_(self) {
    DCHECK_EQ(self, Thread::Current());
    if (kUseReadBarrier) {
      self_->SetIsTransitioningToRunnable(true);
    }
  }

  ~ScopedTransitioningToRunnable() {
    if (kUseReadBarrier) {
      self_->SetIsTransitioningToRunnable(false);
    }
  }

 private:
  Thread* const self_;
};

class ThreadLifecycleCallback {
 public:
  virtual ~ThreadLifecycleCallback() {}

  virtual void ThreadStart(Thread* self) REQUIRES_SHARED(Locks::mutator_lock_) = 0;
  virtual void ThreadDeath(Thread* self) REQUIRES_SHARED(Locks::mutator_lock_) = 0;
};

std::ostream& operator<<(std::ostream& os, const Thread& thread);
std::ostream& operator<<(std::ostream& os, const StackedShadowFrameType& thread);

}  // namespace art

#endif  // ART_RUNTIME_THREAD_H_