summaryrefslogtreecommitdiffstats
path: root/Python/pystate.c
blob: cdd975f3672b9bd66bfc3415bab54ac848a92aba (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
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931

/* Thread and interpreter state structures and their interfaces */

#include "Python.h"
#include "pycore_ceval.h"
#include "pycore_code.h"          // stats
#include "pycore_dtoa.h"          // _dtoa_state_INIT()
#include "pycore_frame.h"
#include "pycore_initconfig.h"
#include "pycore_object.h"        // _PyType_InitCache()
#include "pycore_pyerrors.h"
#include "pycore_pylifecycle.h"
#include "pycore_pymem.h"         // _PyMem_SetDefaultAllocator()
#include "pycore_pystate.h"
#include "pycore_runtime_init.h"  // _PyRuntimeState_INIT
#include "pycore_sysmodule.h"     // _PySys_Audit()
#include "pycore_weakref.h"       // _PyWeakref_GET_REF()

/* --------------------------------------------------------------------------
CAUTION

Always use PyMem_RawMalloc() and PyMem_RawFree() directly in this file.  A
number of these functions are advertised as safe to call when the GIL isn't
held, and in a debug build Python redirects (e.g.) PyMem_NEW (etc) to Python's
debugging obmalloc functions.  Those aren't thread-safe (they rely on the GIL
to avoid the expense of doing their own locking).
-------------------------------------------------------------------------- */

#ifdef HAVE_DLOPEN
#ifdef HAVE_DLFCN_H
#include <dlfcn.h>
#endif
#if !HAVE_DECL_RTLD_LAZY
#define RTLD_LAZY 1
#endif
#endif

#ifdef __cplusplus
extern "C" {
#endif


/****************************************/
/* helpers for the current thread state */
/****************************************/

// API for the current thread state is further down.

/* "current" means one of:
   - bound to the current OS thread
   - holds the GIL
 */

//-------------------------------------------------
// a highly efficient lookup for the current thread
//-------------------------------------------------

/*
   The stored thread state is set by PyThreadState_Swap().

   For each of these functions, the GIL must be held by the current thread.
 */


#ifdef HAVE_THREAD_LOCAL
_Py_thread_local PyThreadState *_Py_tss_tstate = NULL;
#endif

static inline PyThreadState *
current_fast_get(_PyRuntimeState *Py_UNUSED(runtime))
{
#ifdef HAVE_THREAD_LOCAL
    return _Py_tss_tstate;
#else
    // XXX Fall back to the PyThread_tss_*() API.
#  error "no supported thread-local variable storage classifier"
#endif
}

static inline void
current_fast_set(_PyRuntimeState *Py_UNUSED(runtime), PyThreadState *tstate)
{
    assert(tstate != NULL);
#ifdef HAVE_THREAD_LOCAL
    _Py_tss_tstate = tstate;
#else
    // XXX Fall back to the PyThread_tss_*() API.
#  error "no supported thread-local variable storage classifier"
#endif
}

static inline void
current_fast_clear(_PyRuntimeState *Py_UNUSED(runtime))
{
#ifdef HAVE_THREAD_LOCAL
    _Py_tss_tstate = NULL;
#else
    // XXX Fall back to the PyThread_tss_*() API.
#  error "no supported thread-local variable storage classifier"
#endif
}

#define tstate_verify_not_active(tstate) \
    if (tstate == current_fast_get((tstate)->interp->runtime)) { \
        _Py_FatalErrorFormat(__func__, "tstate %p is still current", tstate); \
    }

PyThreadState *
_PyThreadState_GetCurrent(void)
{
    return current_fast_get(&_PyRuntime);
}


//------------------------------------------------
// the thread state bound to the current OS thread
//------------------------------------------------

static inline int
tstate_tss_initialized(Py_tss_t *key)
{
    return PyThread_tss_is_created(key);
}

static inline int
tstate_tss_init(Py_tss_t *key)
{
    assert(!tstate_tss_initialized(key));
    return PyThread_tss_create(key);
}

static inline void
tstate_tss_fini(Py_tss_t *key)
{
    assert(tstate_tss_initialized(key));
    PyThread_tss_delete(key);
}

static inline PyThreadState *
tstate_tss_get(Py_tss_t *key)
{
    assert(tstate_tss_initialized(key));
    return (PyThreadState *)PyThread_tss_get(key);
}

static inline int
tstate_tss_set(Py_tss_t *key, PyThreadState *tstate)
{
    assert(tstate != NULL);
    assert(tstate_tss_initialized(key));
    return PyThread_tss_set(key, (void *)tstate);
}

static inline int
tstate_tss_clear(Py_tss_t *key)
{
    assert(tstate_tss_initialized(key));
    return PyThread_tss_set(key, (void *)NULL);
}

#ifdef HAVE_FORK
/* Reset the TSS key - called by PyOS_AfterFork_Child().
 * This should not be necessary, but some - buggy - pthread implementations
 * don't reset TSS upon fork(), see issue #10517.
 */
static PyStatus
tstate_tss_reinit(Py_tss_t *key)
{
    if (!tstate_tss_initialized(key)) {
        return _PyStatus_OK();
    }
    PyThreadState *tstate = tstate_tss_get(key);

    tstate_tss_fini(key);
    if (tstate_tss_init(key) != 0) {
        return _PyStatus_NO_MEMORY();
    }

    /* If the thread had an associated auto thread state, reassociate it with
     * the new key. */
    if (tstate && tstate_tss_set(key, tstate) != 0) {
        return _PyStatus_ERR("failed to re-set autoTSSkey");
    }
    return _PyStatus_OK();
}
#endif


/*
   The stored thread state is set by bind_tstate() (AKA PyThreadState_Bind().

   The GIL does no need to be held for these.
  */

#define gilstate_tss_initialized(runtime) \
    tstate_tss_initialized(&(runtime)->autoTSSkey)
#define gilstate_tss_init(runtime) \
    tstate_tss_init(&(runtime)->autoTSSkey)
#define gilstate_tss_fini(runtime) \
    tstate_tss_fini(&(runtime)->autoTSSkey)
#define gilstate_tss_get(runtime) \
    tstate_tss_get(&(runtime)->autoTSSkey)
#define _gilstate_tss_set(runtime, tstate) \
    tstate_tss_set(&(runtime)->autoTSSkey, tstate)
#define _gilstate_tss_clear(runtime) \
    tstate_tss_clear(&(runtime)->autoTSSkey)
#define gilstate_tss_reinit(runtime) \
    tstate_tss_reinit(&(runtime)->autoTSSkey)

static inline void
gilstate_tss_set(_PyRuntimeState *runtime, PyThreadState *tstate)
{
    assert(tstate != NULL && tstate->interp->runtime == runtime);
    if (_gilstate_tss_set(runtime, tstate) != 0) {
        Py_FatalError("failed to set current tstate (TSS)");
    }
}

static inline void
gilstate_tss_clear(_PyRuntimeState *runtime)
{
    if (_gilstate_tss_clear(runtime) != 0) {
        Py_FatalError("failed to clear current tstate (TSS)");
    }
}


#ifndef NDEBUG
static inline int tstate_is_alive(PyThreadState *tstate);

static inline int
tstate_is_bound(PyThreadState *tstate)
{
    return tstate->_status.bound && !tstate->_status.unbound;
}
#endif  // !NDEBUG

static void bind_gilstate_tstate(PyThreadState *);
static void unbind_gilstate_tstate(PyThreadState *);

static void
bind_tstate(PyThreadState *tstate)
{
    assert(tstate != NULL);
    assert(tstate_is_alive(tstate) && !tstate->_status.bound);
    assert(!tstate->_status.unbound);  // just in case
    assert(!tstate->_status.bound_gilstate);
    assert(tstate != gilstate_tss_get(tstate->interp->runtime));
    assert(!tstate->_status.active);
    assert(tstate->thread_id == 0);
    assert(tstate->native_thread_id == 0);

    // Currently we don't necessarily store the thread state
    // in thread-local storage (e.g. per-interpreter).

    tstate->thread_id = PyThread_get_thread_ident();
#ifdef PY_HAVE_THREAD_NATIVE_ID
    tstate->native_thread_id = PyThread_get_thread_native_id();
#endif

    tstate->_status.bound = 1;
}

static void
unbind_tstate(PyThreadState *tstate)
{
    assert(tstate != NULL);
    // XXX assert(tstate_is_alive(tstate));
    assert(tstate_is_bound(tstate));
    // XXX assert(!tstate->_status.active);
    assert(tstate->thread_id > 0);
#ifdef PY_HAVE_THREAD_NATIVE_ID
    assert(tstate->native_thread_id > 0);
#endif

    // We leave thread_id and native_thread_id alone
    // since they can be useful for debugging.
    // Check the `_status` field to know if these values
    // are still valid.

    // We leave tstate->_status.bound set to 1
    // to indicate it was previously bound.
    tstate->_status.unbound = 1;
}


/* Stick the thread state for this thread in thread specific storage.

   When a thread state is created for a thread by some mechanism
   other than PyGILState_Ensure(), it's important that the GILState
   machinery knows about it so it doesn't try to create another
   thread state for the thread.
   (This is a better fix for SF bug #1010677 than the first one attempted.)

   The only situation where you can legitimately have more than one
   thread state for an OS level thread is when there are multiple
   interpreters.

   Before 3.12, the PyGILState_*() APIs didn't work with multiple
   interpreters (see bpo-10915 and bpo-15751), so this function used
   to set TSS only once.  Thus, the first thread state created for that
   given OS level thread would "win", which seemed reasonable behaviour.
*/

static void
bind_gilstate_tstate(PyThreadState *tstate)
{
    assert(tstate != NULL);
    assert(tstate_is_alive(tstate));
    assert(tstate_is_bound(tstate));
    // XXX assert(!tstate->_status.active);
    assert(!tstate->_status.bound_gilstate);

    _PyRuntimeState *runtime = tstate->interp->runtime;
    PyThreadState *tcur = gilstate_tss_get(runtime);
    assert(tstate != tcur);

    if (tcur != NULL) {
        tcur->_status.bound_gilstate = 0;
    }
    gilstate_tss_set(runtime, tstate);
    tstate->_status.bound_gilstate = 1;
}

static void
unbind_gilstate_tstate(PyThreadState *tstate)
{
    assert(tstate != NULL);
    // XXX assert(tstate_is_alive(tstate));
    assert(tstate_is_bound(tstate));
    // XXX assert(!tstate->_status.active);
    assert(tstate->_status.bound_gilstate);
    assert(tstate == gilstate_tss_get(tstate->interp->runtime));

    gilstate_tss_clear(tstate->interp->runtime);
    tstate->_status.bound_gilstate = 0;
}


//----------------------------------------------
// the thread state that currently holds the GIL
//----------------------------------------------

/* This is not exported, as it is not reliable!  It can only
   ever be compared to the state for the *current* thread.
   * If not equal, then it doesn't matter that the actual
     value may change immediately after comparison, as it can't
     possibly change to the current thread's state.
   * If equal, then the current thread holds the lock, so the value can't
     change until we yield the lock.
*/
static int
holds_gil(PyThreadState *tstate)
{
    // XXX Fall back to tstate->interp->runtime->ceval.gil.last_holder
    // (and tstate->interp->runtime->ceval.gil.locked).
    assert(tstate != NULL);
    _PyRuntimeState *runtime = tstate->interp->runtime;
    /* Must be the tstate for this thread */
    assert(tstate == gilstate_tss_get(runtime));
    return tstate == current_fast_get(runtime);
}


/****************************/
/* the global runtime state */
/****************************/

//----------
// lifecycle
//----------

/* Suppress deprecation warning for PyBytesObject.ob_shash */
_Py_COMP_DIAG_PUSH
_Py_COMP_DIAG_IGNORE_DEPR_DECLS
/* We use "initial" if the runtime gets re-used
   (e.g. Py_Finalize() followed by Py_Initialize().
   Note that we initialize "initial" relative to _PyRuntime,
   to ensure pre-initialized pointers point to the active
   runtime state (and not "initial"). */
static const _PyRuntimeState initial = _PyRuntimeState_INIT(_PyRuntime);
_Py_COMP_DIAG_POP

#define NUMLOCKS 9
#define LOCKS_INIT(runtime) \
    { \
        &(runtime)->interpreters.mutex, \
        &(runtime)->xidregistry.mutex, \
        &(runtime)->getargs.mutex, \
        &(runtime)->unicode_state.ids.lock, \
        &(runtime)->imports.extensions.mutex, \
        &(runtime)->ceval.pending_mainthread.lock, \
        &(runtime)->atexit.mutex, \
        &(runtime)->audit_hooks.mutex, \
        &(runtime)->allocators.mutex, \
    }

static int
alloc_for_runtime(PyThread_type_lock locks[NUMLOCKS])
{
    /* Force default allocator, since _PyRuntimeState_Fini() must
       use the same allocator than this function. */
    PyMemAllocatorEx old_alloc;
    _PyMem_SetDefaultAllocator(PYMEM_DOMAIN_RAW, &old_alloc);

    for (int i = 0; i < NUMLOCKS; i++) {
        PyThread_type_lock lock = PyThread_allocate_lock();
        if (lock == NULL) {
            for (int j = 0; j < i; j++) {
                PyThread_free_lock(locks[j]);
                locks[j] = NULL;
            }
            break;
        }
        locks[i] = lock;
    }

    PyMem_SetAllocator(PYMEM_DOMAIN_RAW, &old_alloc);
    return 0;
}

static void
init_runtime(_PyRuntimeState *runtime,
             void *open_code_hook, void *open_code_userdata,
             _Py_AuditHookEntry *audit_hook_head,
             Py_ssize_t unicode_next_index,
             PyThread_type_lock locks[NUMLOCKS])
{
    if (runtime->_initialized) {
        Py_FatalError("runtime already initialized");
    }
    assert(!runtime->preinitializing &&
           !runtime->preinitialized &&
           !runtime->core_initialized &&
           !runtime->initialized);

    runtime->open_code_hook = open_code_hook;
    runtime->open_code_userdata = open_code_userdata;
    runtime->audit_hooks.head = audit_hook_head;

    PyPreConfig_InitPythonConfig(&runtime->preconfig);

    PyThread_type_lock *lockptrs[NUMLOCKS] = LOCKS_INIT(runtime);
    for (int i = 0; i < NUMLOCKS; i++) {
        assert(locks[i] != NULL);
        *lockptrs[i] = locks[i];
    }

    // Set it to the ID of the main thread of the main interpreter.
    runtime->main_thread = PyThread_get_thread_ident();

    runtime->unicode_state.ids.next_index = unicode_next_index;

    runtime->_initialized = 1;
}

PyStatus
_PyRuntimeState_Init(_PyRuntimeState *runtime)
{
    /* We preserve the hook across init, because there is
       currently no public API to set it between runtime
       initialization and interpreter initialization. */
    void *open_code_hook = runtime->open_code_hook;
    void *open_code_userdata = runtime->open_code_userdata;
    _Py_AuditHookEntry *audit_hook_head = runtime->audit_hooks.head;
    // bpo-42882: Preserve next_index value if Py_Initialize()/Py_Finalize()
    // is called multiple times.
    Py_ssize_t unicode_next_index = runtime->unicode_state.ids.next_index;

    PyThread_type_lock locks[NUMLOCKS];
    if (alloc_for_runtime(locks) != 0) {
        return _PyStatus_NO_MEMORY();
    }

    if (runtime->_initialized) {
        // Py_Initialize() must be running again.
        // Reset to _PyRuntimeState_INIT.
        memcpy(runtime, &initial, sizeof(*runtime));
    }

    if (gilstate_tss_init(runtime) != 0) {
        _PyRuntimeState_Fini(runtime);
        return _PyStatus_NO_MEMORY();
    }

    if (PyThread_tss_create(&runtime->trashTSSkey) != 0) {
        _PyRuntimeState_Fini(runtime);
        return _PyStatus_NO_MEMORY();
    }

    init_runtime(runtime, open_code_hook, open_code_userdata, audit_hook_head,
                 unicode_next_index, locks);

    return _PyStatus_OK();
}

void
_PyRuntimeState_Fini(_PyRuntimeState *runtime)
{
#ifdef Py_REF_DEBUG
    /* The count is cleared by _Py_FinalizeRefTotal(). */
    assert(runtime->object_state.interpreter_leaks == 0);
#endif

    if (gilstate_tss_initialized(runtime)) {
        gilstate_tss_fini(runtime);
    }

    if (PyThread_tss_is_created(&runtime->trashTSSkey)) {
        PyThread_tss_delete(&runtime->trashTSSkey);
    }

    /* Force the allocator used by _PyRuntimeState_Init(). */
    PyMemAllocatorEx old_alloc;
    _PyMem_SetDefaultAllocator(PYMEM_DOMAIN_RAW, &old_alloc);
#define FREE_LOCK(LOCK) \
    if (LOCK != NULL) { \
        PyThread_free_lock(LOCK); \
        LOCK = NULL; \
    }

    PyThread_type_lock *lockptrs[NUMLOCKS] = LOCKS_INIT(runtime);
    for (int i = 0; i < NUMLOCKS; i++) {
        FREE_LOCK(*lockptrs[i]);
    }

#undef FREE_LOCK
    PyMem_SetAllocator(PYMEM_DOMAIN_RAW, &old_alloc);
}

#ifdef HAVE_FORK
/* This function is called from PyOS_AfterFork_Child to ensure that
   newly created child processes do not share locks with the parent. */
PyStatus
_PyRuntimeState_ReInitThreads(_PyRuntimeState *runtime)
{
    // This was initially set in _PyRuntimeState_Init().
    runtime->main_thread = PyThread_get_thread_ident();

    /* Force default allocator, since _PyRuntimeState_Fini() must
       use the same allocator than this function. */
    PyMemAllocatorEx old_alloc;
    _PyMem_SetDefaultAllocator(PYMEM_DOMAIN_RAW, &old_alloc);

    PyThread_type_lock *lockptrs[NUMLOCKS] = LOCKS_INIT(runtime);
    int reinit_err = 0;
    for (int i = 0; i < NUMLOCKS; i++) {
        reinit_err += _PyThread_at_fork_reinit(lockptrs[i]);
    }

    PyMem_SetAllocator(PYMEM_DOMAIN_RAW, &old_alloc);

    /* bpo-42540: id_mutex is freed by _PyInterpreterState_Delete, which does
     * not force the default allocator. */
    reinit_err += _PyThread_at_fork_reinit(&runtime->interpreters.main->id_mutex);

    if (reinit_err < 0) {
        return _PyStatus_ERR("Failed to reinitialize runtime locks");
    }

    PyStatus status = gilstate_tss_reinit(runtime);
    if (_PyStatus_EXCEPTION(status)) {
        return status;
    }

    if (PyThread_tss_is_created(&runtime->trashTSSkey)) {
        PyThread_tss_delete(&runtime->trashTSSkey);
    }
    if (PyThread_tss_create(&runtime->trashTSSkey) != 0) {
        return _PyStatus_NO_MEMORY();
    }

    return _PyStatus_OK();
}
#endif


/*************************************/
/* the per-interpreter runtime state */
/*************************************/

//----------
// lifecycle
//----------

/* Calling this indicates that the runtime is ready to create interpreters. */

PyStatus
_PyInterpreterState_Enable(_PyRuntimeState *runtime)
{
    struct pyinterpreters *interpreters = &runtime->interpreters;
    interpreters->next_id = 0;

    /* Py_Finalize() calls _PyRuntimeState_Fini() which clears the mutex.
       Create a new mutex if needed. */
    if (interpreters->mutex == NULL) {
        /* Force default allocator, since _PyRuntimeState_Fini() must
           use the same allocator than this function. */
        PyMemAllocatorEx old_alloc;
        _PyMem_SetDefaultAllocator(PYMEM_DOMAIN_RAW, &old_alloc);

        interpreters->mutex = PyThread_allocate_lock();

        PyMem_SetAllocator(PYMEM_DOMAIN_RAW, &old_alloc);

        if (interpreters->mutex == NULL) {
            return _PyStatus_ERR("Can't initialize threads for interpreter");
        }
    }

    return _PyStatus_OK();
}


static PyInterpreterState *
alloc_interpreter(void)
{
    return PyMem_RawCalloc(1, sizeof(PyInterpreterState));
}

static void
free_interpreter(PyInterpreterState *interp)
{
    // The main interpreter is statically allocated so
    // should not be freed.
    if (interp != &_PyRuntime._main_interpreter) {
        PyMem_RawFree(interp);
    }
}

/* Get the interpreter state to a minimal consistent state.
   Further init happens in pylifecycle.c before it can be used.
   All fields not initialized here are expected to be zeroed out,
   e.g. by PyMem_RawCalloc() or memset(), or otherwise pre-initialized.
   The runtime state is not manipulated.  Instead it is assumed that
   the interpreter is getting added to the runtime.

   Note that the main interpreter was statically initialized as part
   of the runtime and most state is already set properly.  That leaves
   a small number of fields to initialize dynamically, as well as some
   that are initialized lazily.

   For subinterpreters we memcpy() the main interpreter in
   PyInterpreterState_New(), leaving it in the same mostly-initialized
   state.  The only difference is that the interpreter has some
   self-referential state that is statically initializexd to the
   main interpreter.  We fix those fields here, in addition
   to the other dynamically initialized fields.
  */
static void
init_interpreter(PyInterpreterState *interp,
                 _PyRuntimeState *runtime, int64_t id,
                 PyInterpreterState *next,
                 PyThread_type_lock pending_lock)
{
    if (interp->_initialized) {
        Py_FatalError("interpreter already initialized");
    }

    assert(runtime != NULL);
    interp->runtime = runtime;

    assert(id > 0 || (id == 0 && interp == runtime->interpreters.main));
    interp->id = id;

    assert(runtime->interpreters.head == interp);
    assert(next != NULL || (interp == runtime->interpreters.main));
    interp->next = next;

    /* Initialize obmalloc, but only for subinterpreters,
       since the main interpreter is initialized statically. */
    if (interp != &runtime->_main_interpreter) {
        poolp temp[OBMALLOC_USED_POOLS_SIZE] = \
                _obmalloc_pools_INIT(interp->obmalloc.pools);
        memcpy(&interp->obmalloc.pools.used, temp, sizeof(temp));
    }

    _PyEval_InitState(interp, pending_lock);
    _PyGC_InitState(&interp->gc);
    PyConfig_InitPythonConfig(&interp->config);
    _PyType_InitCache(interp);
    for (int i = 0; i < _PY_MONITORING_UNGROUPED_EVENTS; i++) {
        interp->monitors.tools[i] = 0;
    }
    for (int t = 0; t < PY_MONITORING_TOOL_IDS; t++) {
        for (int e = 0; e < _PY_MONITORING_EVENTS; e++) {
            interp->monitoring_callables[t][e] = NULL;

        }
    }
    interp->sys_profile_initialized = false;
    interp->sys_trace_initialized = false;
    interp->optimizer = &_PyOptimizer_Default;
    interp->optimizer_backedge_threshold = _PyOptimizer_Default.backedge_threshold;
    interp->optimizer_resume_threshold = _PyOptimizer_Default.backedge_threshold;
    if (interp != &runtime->_main_interpreter) {
        /* Fix the self-referential, statically initialized fields. */
        interp->dtoa = (struct _dtoa_state)_dtoa_state_INIT(interp);
    }
    interp->f_opcode_trace_set = false;
    interp->_initialized = 1;
}

PyInterpreterState *
PyInterpreterState_New(void)
{
    PyInterpreterState *interp;
    _PyRuntimeState *runtime = &_PyRuntime;
    PyThreadState *tstate = current_fast_get(runtime);

    /* tstate is NULL when Py_InitializeFromConfig() calls
       PyInterpreterState_New() to create the main interpreter. */
    if (_PySys_Audit(tstate, "cpython.PyInterpreterState_New", NULL) < 0) {
        return NULL;
    }

    PyThread_type_lock pending_lock = PyThread_allocate_lock();
    if (pending_lock == NULL) {
        if (tstate != NULL) {
            _PyErr_NoMemory(tstate);
        }
        return NULL;
    }

    /* Don't get runtime from tstate since tstate can be NULL. */
    struct pyinterpreters *interpreters = &runtime->interpreters;

    /* We completely serialize creation of multiple interpreters, since
       it simplifies things here and blocking concurrent calls isn't a problem.
       Regardless, we must fully block subinterpreter creation until
       after the main interpreter is created. */
    HEAD_LOCK(runtime);

    int64_t id = interpreters->next_id;
    interpreters->next_id += 1;

    // Allocate the interpreter and add it to the runtime state.
    PyInterpreterState *old_head = interpreters->head;
    if (old_head == NULL) {
        // We are creating the main interpreter.
        assert(interpreters->main == NULL);
        assert(id == 0);

        interp = &runtime->_main_interpreter;
        assert(interp->id == 0);
        assert(interp->next == NULL);

        interpreters->main = interp;
    }
    else {
        assert(interpreters->main != NULL);
        assert(id != 0);

        interp = alloc_interpreter();
        if (interp == NULL) {
            goto error;
        }
        // Set to _PyInterpreterState_INIT.
        memcpy(interp, &initial._main_interpreter,
               sizeof(*interp));

        if (id < 0) {
            /* overflow or Py_Initialize() not called yet! */
            if (tstate != NULL) {
                _PyErr_SetString(tstate, PyExc_RuntimeError,
                                 "failed to get an interpreter ID");
            }
            goto error;
        }
    }
    interpreters->head = interp;

    init_interpreter(interp, runtime, id, old_head, pending_lock);

    HEAD_UNLOCK(runtime);
    return interp;

error:
    HEAD_UNLOCK(runtime);

    PyThread_free_lock(pending_lock);
    if (interp != NULL) {
        free_interpreter(interp);
    }
    return NULL;
}


static void
interpreter_clear(PyInterpreterState *interp, PyThreadState *tstate)
{
    assert(interp != NULL);
    assert(tstate != NULL);
    _PyRuntimeState *runtime = interp->runtime;

    /* XXX Conditions we need to enforce:

       * the GIL must be held by the current thread
       * tstate must be the "current" thread state (current_fast_get())
       * tstate->interp must be interp
       * for the main interpreter, tstate must be the main thread
     */
    // XXX Ideally, we would not rely on any thread state in this function
    // (and we would drop the "tstate" argument).

    if (_PySys_Audit(tstate, "cpython.PyInterpreterState_Clear", NULL) < 0) {
        _PyErr_Clear(tstate);
    }

    // Clear the current/main thread state last.
    HEAD_LOCK(runtime);
    PyThreadState *p = interp->threads.head;
    HEAD_UNLOCK(runtime);
    while (p != NULL) {
        // See https://github.com/python/cpython/issues/102126
        // Must be called without HEAD_LOCK held as it can deadlock
        // if any finalizer tries to acquire that lock.
        PyThreadState_Clear(p);
        HEAD_LOCK(runtime);
        p = p->next;
        HEAD_UNLOCK(runtime);
    }
    if (tstate->interp == interp) {
        /* We fix tstate->_status below when we for sure aren't using it
           (e.g. no longer need the GIL). */
        // XXX Eliminate the need to do this.
        tstate->_status.cleared = 0;
    }

    Py_CLEAR(interp->optimizer);
    interp->optimizer = &_PyOptimizer_Default;
    interp->optimizer_backedge_threshold = _PyOptimizer_Default.backedge_threshold;
    interp->optimizer_resume_threshold = _PyOptimizer_Default.backedge_threshold;

    /* It is possible that any of the objects below have a finalizer
       that runs Python code or otherwise relies on a thread state
       or even the interpreter state.  For now we trust that isn't
       a problem.
     */
    // XXX Make sure we properly deal with problematic finalizers.

    Py_CLEAR(interp->audit_hooks);

    for (int i = 0; i < _PY_MONITORING_UNGROUPED_EVENTS; i++) {
        interp->monitors.tools[i] = 0;
    }
    for (int t = 0; t < PY_MONITORING_TOOL_IDS; t++) {
        for (int e = 0; e < _PY_MONITORING_EVENTS; e++) {
            Py_CLEAR(interp->monitoring_callables[t][e]);
        }
    }
    interp->sys_profile_initialized = false;
    interp->sys_trace_initialized = false;
    for (int t = 0; t < PY_MONITORING_TOOL_IDS; t++) {
        Py_CLEAR(interp->monitoring_tool_names[t]);
    }

    PyConfig_Clear(&interp->config);
    Py_CLEAR(interp->codec_search_path);
    Py_CLEAR(interp->codec_search_cache);
    Py_CLEAR(interp->codec_error_registry);

    assert(interp->imports.modules == NULL);
    assert(interp->imports.modules_by_index == NULL);
    assert(interp->imports.importlib == NULL);
    assert(interp->imports.import_func == NULL);

    Py_CLEAR(interp->sysdict_copy);
    Py_CLEAR(interp->builtins_copy);
    Py_CLEAR(interp->dict);
#ifdef HAVE_FORK
    Py_CLEAR(interp->before_forkers);
    Py_CLEAR(interp->after_forkers_parent);
    Py_CLEAR(interp->after_forkers_child);
#endif

    _PyAST_Fini(interp);
    _PyWarnings_Fini(interp);
    _PyAtExit_Fini(interp);

    // All Python types must be destroyed before the last GC collection. Python
    // types create a reference cycle to themselves in their in their
    // PyTypeObject.tp_mro member (the tuple contains the type).

    /* Last garbage collection on this interpreter */
    _PyGC_CollectNoFail(tstate);
    _PyGC_Fini(interp);

    /* We don't clear sysdict and builtins until the end of this function.
       Because clearing other attributes can execute arbitrary Python code
       which requires sysdict and builtins. */
    PyDict_Clear(interp->sysdict);
    PyDict_Clear(interp->builtins);
    Py_CLEAR(interp->sysdict);
    Py_CLEAR(interp->builtins);

    if (tstate->interp == interp) {
        /* We are now safe to fix tstate->_status.cleared. */
        // XXX Do this (much) earlier?
        tstate->_status.cleared = 1;
    }

    for (int i=0; i < DICT_MAX_WATCHERS; i++) {
        interp->dict_state.watchers[i] = NULL;
    }

    for (int i=0; i < TYPE_MAX_WATCHERS; i++) {
        interp->type_watchers[i] = NULL;
    }

    for (int i=0; i < FUNC_MAX_WATCHERS; i++) {
        interp->func_watchers[i] = NULL;
    }
    interp->active_func_watchers = 0;

    for (int i=0; i < CODE_MAX_WATCHERS; i++) {
        interp->code_watchers[i] = NULL;
    }
    interp->active_code_watchers = 0;
    interp->f_opcode_trace_set = false;
    // XXX Once we have one allocator per interpreter (i.e.
    // per-interpreter GC) we must ensure that all of the interpreter's
    // objects have been cleaned up at the point.
}


void
PyInterpreterState_Clear(PyInterpreterState *interp)
{
    // Use the current Python thread state to call audit hooks and to collect
    // garbage. It can be different than the current Python thread state
    // of 'interp'.
    PyThreadState *current_tstate = current_fast_get(interp->runtime);
    _PyImport_ClearCore(interp);
    interpreter_clear(interp, current_tstate);
}


void
_PyInterpreterState_Clear(PyThreadState *tstate)
{
    _PyImport_ClearCore(tstate->interp);
    interpreter_clear(tstate->interp, tstate);
}


static inline void tstate_deactivate(PyThreadState *tstate);
static void zapthreads(PyInterpreterState *interp);

void
PyInterpreterState_Delete(PyInterpreterState *interp)
{
    _PyRuntimeState *runtime = interp->runtime;
    struct pyinterpreters *interpreters = &runtime->interpreters;

    // XXX Clearing the "current" thread state should happen before
    // we start finalizing the interpreter (or the current thread state).
    PyThreadState *tcur = current_fast_get(runtime);
    if (tcur != NULL && interp == tcur->interp) {
        /* Unset current thread.  After this, many C API calls become crashy. */
        current_fast_clear(runtime);
        tstate_deactivate(tcur);
        _PyEval_ReleaseLock(interp, NULL);
    }

    zapthreads(interp);

    _PyEval_FiniState(&interp->ceval);

    // XXX These two calls should be done at the end of clear_interpreter(),
    // but currently some objects get decref'ed after that.
#ifdef Py_REF_DEBUG
    _PyInterpreterState_FinalizeRefTotal(interp);
#endif
    _PyInterpreterState_FinalizeAllocatedBlocks(interp);

    HEAD_LOCK(runtime);
    PyInterpreterState **p;
    for (p = &interpreters->head; ; p = &(*p)->next) {
        if (*p == NULL) {
            Py_FatalError("NULL interpreter");
        }
        if (*p == interp) {
            break;
        }
    }
    if (interp->threads.head != NULL) {
        Py_FatalError("remaining threads");
    }
    *p = interp->next;

    if (interpreters->main == interp) {
        interpreters->main = NULL;
        if (interpreters->head != NULL) {
            Py_FatalError("remaining subinterpreters");
        }
    }
    HEAD_UNLOCK(runtime);

    if (interp->id_mutex != NULL) {
        PyThread_free_lock(interp->id_mutex);
    }
    free_interpreter(interp);
}


#ifdef HAVE_FORK
/*
 * Delete all interpreter states except the main interpreter.  If there
 * is a current interpreter state, it *must* be the main interpreter.
 */
PyStatus
_PyInterpreterState_DeleteExceptMain(_PyRuntimeState *runtime)
{
    struct pyinterpreters *interpreters = &runtime->interpreters;

    PyThreadState *tstate = _PyThreadState_Swap(runtime, NULL);
    if (tstate != NULL && tstate->interp != interpreters->main) {
        return _PyStatus_ERR("not main interpreter");
    }

    HEAD_LOCK(runtime);
    PyInterpreterState *interp = interpreters->head;
    interpreters->head = NULL;
    while (interp != NULL) {
        if (interp == interpreters->main) {
            interpreters->main->next = NULL;
            interpreters->head = interp;
            interp = interp->next;
            continue;
        }

        // XXX Won't this fail since PyInterpreterState_Clear() requires
        // the "current" tstate to be set?
        PyInterpreterState_Clear(interp);  // XXX must activate?
        zapthreads(interp);
        if (interp->id_mutex != NULL) {
            PyThread_free_lock(interp->id_mutex);
        }
        PyInterpreterState *prev_interp = interp;
        interp = interp->next;
        free_interpreter(prev_interp);
    }
    HEAD_UNLOCK(runtime);

    if (interpreters->head == NULL) {
        return _PyStatus_ERR("missing main interpreter");
    }
    _PyThreadState_Swap(runtime, tstate);
    return _PyStatus_OK();
}
#endif


//----------
// accessors
//----------

int64_t
PyInterpreterState_GetID(PyInterpreterState *interp)
{
    if (interp == NULL) {
        PyErr_SetString(PyExc_RuntimeError, "no interpreter provided");
        return -1;
    }
    return interp->id;
}


int
_PyInterpreterState_IDInitref(PyInterpreterState *interp)
{
    if (interp->id_mutex != NULL) {
        return 0;
    }
    interp->id_mutex = PyThread_allocate_lock();
    if (interp->id_mutex == NULL) {
        PyErr_SetString(PyExc_RuntimeError,
                        "failed to create init interpreter ID mutex");
        return -1;
    }
    interp->id_refcount = 0;
    return 0;
}


int
_PyInterpreterState_IDIncref(PyInterpreterState *interp)
{
    if (_PyInterpreterState_IDInitref(interp) < 0) {
        return -1;
    }

    PyThread_acquire_lock(interp->id_mutex, WAIT_LOCK);
    interp->id_refcount += 1;
    PyThread_release_lock(interp->id_mutex);
    return 0;
}


void
_PyInterpreterState_IDDecref(PyInterpreterState *interp)
{
    assert(interp->id_mutex != NULL);
    _PyRuntimeState *runtime = interp->runtime;

    PyThread_acquire_lock(interp->id_mutex, WAIT_LOCK);
    assert(interp->id_refcount != 0);
    interp->id_refcount -= 1;
    int64_t refcount = interp->id_refcount;
    PyThread_release_lock(interp->id_mutex);

    if (refcount == 0 && interp->requires_idref) {
        // XXX Using the "head" thread isn't strictly correct.
        PyThreadState *tstate = PyInterpreterState_ThreadHead(interp);
        // XXX Possible GILState issues?
        PyThreadState *save_tstate = _PyThreadState_Swap(runtime, tstate);
        Py_EndInterpreter(tstate);
        _PyThreadState_Swap(runtime, save_tstate);
    }
}

int
_PyInterpreterState_RequiresIDRef(PyInterpreterState *interp)
{
    return interp->requires_idref;
}

void
_PyInterpreterState_RequireIDRef(PyInterpreterState *interp, int required)
{
    interp->requires_idref = required ? 1 : 0;
}

PyObject *
_PyInterpreterState_GetMainModule(PyInterpreterState *interp)
{
    PyObject *modules = _PyImport_GetModules(interp);
    if (modules == NULL) {
        PyErr_SetString(PyExc_RuntimeError, "interpreter not initialized");
        return NULL;
    }
    return PyMapping_GetItemString(modules, "__main__");
}

PyObject *
PyInterpreterState_GetDict(PyInterpreterState *interp)
{
    if (interp->dict == NULL) {
        interp->dict = PyDict_New();
        if (interp->dict == NULL) {
            PyErr_Clear();
        }
    }
    /* Returning NULL means no per-interpreter dict is available. */
    return interp->dict;
}


//-----------------------------
// look up an interpreter state
//-----------------------------

/* Return the interpreter associated with the current OS thread.

   The GIL must be held.
  */

PyInterpreterState*
PyInterpreterState_Get(void)
{
    PyThreadState *tstate = current_fast_get(&_PyRuntime);
    _Py_EnsureTstateNotNULL(tstate);
    PyInterpreterState *interp = tstate->interp;
    if (interp == NULL) {
        Py_FatalError("no current interpreter");
    }
    return interp;
}


static PyInterpreterState *
interp_look_up_id(_PyRuntimeState *runtime, int64_t requested_id)
{
    PyInterpreterState *interp = runtime->interpreters.head;
    while (interp != NULL) {
        int64_t id = PyInterpreterState_GetID(interp);
        if (id < 0) {
            return NULL;
        }
        if (requested_id == id) {
            return interp;
        }
        interp = PyInterpreterState_Next(interp);
    }
    return NULL;
}

/* Return the interpreter state with the given ID.

   Fail with RuntimeError if the interpreter is not found. */

PyInterpreterState *
_PyInterpreterState_LookUpID(int64_t requested_id)
{
    PyInterpreterState *interp = NULL;
    if (requested_id >= 0) {
        _PyRuntimeState *runtime = &_PyRuntime;
        HEAD_LOCK(runtime);
        interp = interp_look_up_id(runtime, requested_id);
        HEAD_UNLOCK(runtime);
    }
    if (interp == NULL && !PyErr_Occurred()) {
        PyErr_Format(PyExc_RuntimeError,
                     "unrecognized interpreter ID %lld", requested_id);
    }
    return interp;
}


/********************************/
/* the per-thread runtime state */
/********************************/

#ifndef NDEBUG
static inline int
tstate_is_alive(PyThreadState *tstate)
{
    return (tstate->_status.initialized &&
            !tstate->_status.finalized &&
            !tstate->_status.cleared &&
            !tstate->_status.finalizing);
}
#endif


//----------
// lifecycle
//----------

/* Minimum size of data stack chunk */
#define DATA_STACK_CHUNK_SIZE (16*1024)

static _PyStackChunk*
allocate_chunk(int size_in_bytes, _PyStackChunk* previous)
{
    assert(size_in_bytes % sizeof(PyObject **) == 0);
    _PyStackChunk *res = _PyObject_VirtualAlloc(size_in_bytes);
    if (res == NULL) {
        return NULL;
    }
    res->previous = previous;
    res->size = size_in_bytes;
    res->top = 0;
    return res;
}

static PyThreadState *
alloc_threadstate(void)
{
    return PyMem_RawCalloc(1, sizeof(PyThreadState));
}

static void
free_threadstate(PyThreadState *tstate)
{
    // The initial thread state of the interpreter is allocated
    // as part of the interpreter state so should not be freed.
    if (tstate != &tstate->interp->_initial_thread) {
        PyMem_RawFree(tstate);
    }
}

/* Get the thread state to a minimal consistent state.
   Further init happens in pylifecycle.c before it can be used.
   All fields not initialized here are expected to be zeroed out,
   e.g. by PyMem_RawCalloc() or memset(), or otherwise pre-initialized.
   The interpreter state is not manipulated.  Instead it is assumed that
   the thread is getting added to the interpreter.
  */

static void
init_threadstate(PyThreadState *tstate,
                 PyInterpreterState *interp, uint64_t id)
{
    if (tstate->_status.initialized) {
        Py_FatalError("thread state already initialized");
    }

    assert(interp != NULL);
    tstate->interp = interp;

    // next/prev are set in add_threadstate().
    assert(tstate->next == NULL);
    assert(tstate->prev == NULL);

    assert(id > 0);
    tstate->id = id;

    // thread_id and native_thread_id are set in bind_tstate().

    tstate->py_recursion_limit = interp->ceval.recursion_limit,
    tstate->py_recursion_remaining = interp->ceval.recursion_limit,
    tstate->c_recursion_remaining = C_RECURSION_LIMIT;

    tstate->exc_info = &tstate->exc_state;

    // PyGILState_Release must not try to delete this thread state.
    // This is cleared when PyGILState_Ensure() creates the thread state.
    tstate->gilstate_counter = 1;

    tstate->cframe = &tstate->root_cframe;
    tstate->datastack_chunk = NULL;
    tstate->datastack_top = NULL;
    tstate->datastack_limit = NULL;
    tstate->what_event = -1;

    tstate->_status.initialized = 1;
}

static void
add_threadstate(PyInterpreterState *interp, PyThreadState *tstate,
                PyThreadState *next)
{
    assert(interp->threads.head != tstate);
    assert((next != NULL && tstate->id != 1) ||
           (next == NULL && tstate->id == 1));
    if (next != NULL) {
        assert(next->prev == NULL || next->prev == tstate);
        next->prev = tstate;
    }
    tstate->next = next;
    assert(tstate->prev == NULL);
    interp->threads.head = tstate;
}

static PyThreadState *
new_threadstate(PyInterpreterState *interp)
{
    PyThreadState *tstate;
    _PyRuntimeState *runtime = interp->runtime;
    // We don't need to allocate a thread state for the main interpreter
    // (the common case), but doing it later for the other case revealed a
    // reentrancy problem (deadlock).  So for now we always allocate before
    // taking the interpreters lock.  See GH-96071.
    PyThreadState *new_tstate = alloc_threadstate();
    int used_newtstate;
    if (new_tstate == NULL) {
        return NULL;
    }
    /* We serialize concurrent creation to protect global state. */
    HEAD_LOCK(runtime);

    interp->threads.next_unique_id += 1;
    uint64_t id = interp->threads.next_unique_id;

    // Allocate the thread state and add it to the interpreter.
    PyThreadState *old_head = interp->threads.head;
    if (old_head == NULL) {
        // It's the interpreter's initial thread state.
        assert(id == 1);
        used_newtstate = 0;
        tstate = &interp->_initial_thread;
    }
    else {
        // Every valid interpreter must have at least one thread.
        assert(id > 1);
        assert(old_head->prev == NULL);
        used_newtstate = 1;
        tstate = new_tstate;
        // Set to _PyThreadState_INIT.
        memcpy(tstate,
               &initial._main_interpreter._initial_thread,
               sizeof(*tstate));
    }

    init_threadstate(tstate, interp, id);
    add_threadstate(interp, tstate, old_head);

    HEAD_UNLOCK(runtime);
    if (!used_newtstate) {
        // Must be called with lock unlocked to avoid re-entrancy deadlock.
        PyMem_RawFree(new_tstate);
    }
    return tstate;
}

PyThreadState *
PyThreadState_New(PyInterpreterState *interp)
{
    PyThreadState *tstate = new_threadstate(interp);
    if (tstate) {
        bind_tstate(tstate);
        // This makes sure there's a gilstate tstate bound
        // as soon as possible.
        if (gilstate_tss_get(tstate->interp->runtime) == NULL) {
            bind_gilstate_tstate(tstate);
        }
    }
    return tstate;
}

// This must be followed by a call to _PyThreadState_Bind();
PyThreadState *
_PyThreadState_New(PyInterpreterState *interp)
{
    return new_threadstate(interp);
}

// We keep this for stable ABI compabibility.
PyAPI_FUNC(PyThreadState*)
_PyThreadState_Prealloc(PyInterpreterState *interp)
{
    return _PyThreadState_New(interp);
}

// We keep this around for (accidental) stable ABI compatibility.
// Realistically, no extensions are using it.
PyAPI_FUNC(void)
_PyThreadState_Init(PyThreadState *tstate)
{
    Py_FatalError("_PyThreadState_Init() is for internal use only");
}


static void
clear_datastack(PyThreadState *tstate)
{
    _PyStackChunk *chunk = tstate->datastack_chunk;
    tstate->datastack_chunk = NULL;
    while (chunk != NULL) {
        _PyStackChunk *prev = chunk->previous;
        _PyObject_VirtualFree(chunk, chunk->size);
        chunk = prev;
    }
}

void
PyThreadState_Clear(PyThreadState *tstate)
{
    assert(tstate->_status.initialized && !tstate->_status.cleared);
    // XXX assert(!tstate->_status.bound || tstate->_status.unbound);
    tstate->_status.finalizing = 1;  // just in case

    /* XXX Conditions we need to enforce:

       * the GIL must be held by the current thread
       * current_fast_get()->interp must match tstate->interp
       * for the main interpreter, current_fast_get() must be the main thread
     */

    int verbose = _PyInterpreterState_GetConfig(tstate->interp)->verbose;

    if (verbose && tstate->cframe->current_frame != NULL) {
        /* bpo-20526: After the main thread calls
           _PyInterpreterState_SetFinalizing() in Py_FinalizeEx()
           (or in Py_EndInterpreter() for subinterpreters),
           threads must exit when trying to take the GIL.
           If a thread exit in the middle of _PyEval_EvalFrameDefault(),
           tstate->frame is not reset to its previous value.
           It is more likely with daemon threads, but it can happen
           with regular threads if threading._shutdown() fails
           (ex: interrupted by CTRL+C). */
        fprintf(stderr,
          "PyThreadState_Clear: warning: thread still has a frame\n");
    }

    /* At this point tstate shouldn't be used any more,
       neither to run Python code nor for other uses.

       This is tricky when current_fast_get() == tstate, in the same way
       as noted in interpreter_clear() above.  The below finalizers
       can possibly run Python code or otherwise use the partially
       cleared thread state.  For now we trust that isn't a problem
       in practice.
     */
    // XXX Deal with the possibility of problematic finalizers.

    /* Don't clear tstate->pyframe: it is a borrowed reference */

    Py_CLEAR(tstate->dict);
    Py_CLEAR(tstate->async_exc);

    Py_CLEAR(tstate->current_exception);

    Py_CLEAR(tstate->exc_state.exc_value);

    /* The stack of exception states should contain just this thread. */
    if (verbose && tstate->exc_info != &tstate->exc_state) {
        fprintf(stderr,
          "PyThreadState_Clear: warning: thread still has a generator\n");
    }

    if (tstate->c_profilefunc != NULL) {
        tstate->interp->sys_profiling_threads--;
        tstate->c_profilefunc = NULL;
    }
    if (tstate->c_tracefunc != NULL) {
        tstate->interp->sys_tracing_threads--;
        tstate->c_tracefunc = NULL;
    }
    Py_CLEAR(tstate->c_profileobj);
    Py_CLEAR(tstate->c_traceobj);

    Py_CLEAR(tstate->async_gen_firstiter);
    Py_CLEAR(tstate->async_gen_finalizer);

    Py_CLEAR(tstate->context);

    if (tstate->on_delete != NULL) {
        tstate->on_delete(tstate->on_delete_data);
    }

    tstate->_status.cleared = 1;

    // XXX Call _PyThreadStateSwap(runtime, NULL) here if "current".
    // XXX Do it as early in the function as possible.
}

/* Common code for PyThreadState_Delete() and PyThreadState_DeleteCurrent() */
static void
tstate_delete_common(PyThreadState *tstate)
{
    assert(tstate->_status.cleared && !tstate->_status.finalized);

    PyInterpreterState *interp = tstate->interp;
    if (interp == NULL) {
        Py_FatalError("NULL interpreter");
    }
    _PyRuntimeState *runtime = interp->runtime;

    HEAD_LOCK(runtime);
    if (tstate->prev) {
        tstate->prev->next = tstate->next;
    }
    else {
        interp->threads.head = tstate->next;
    }
    if (tstate->next) {
        tstate->next->prev = tstate->prev;
    }
    HEAD_UNLOCK(runtime);

    // XXX Unbind in PyThreadState_Clear(), or earlier
    // (and assert not-equal here)?
    if (tstate->_status.bound_gilstate) {
        unbind_gilstate_tstate(tstate);
    }
    unbind_tstate(tstate);

    // XXX Move to PyThreadState_Clear()?
    clear_datastack(tstate);

    tstate->_status.finalized = 1;
}

static void
zapthreads(PyInterpreterState *interp)
{
    PyThreadState *tstate;
    /* No need to lock the mutex here because this should only happen
       when the threads are all really dead (XXX famous last words). */
    while ((tstate = interp->threads.head) != NULL) {
        tstate_verify_not_active(tstate);
        tstate_delete_common(tstate);
        free_threadstate(tstate);
    }
}


void
PyThreadState_Delete(PyThreadState *tstate)
{
    _Py_EnsureTstateNotNULL(tstate);
    tstate_verify_not_active(tstate);
    tstate_delete_common(tstate);
    free_threadstate(tstate);
}


void
_PyThreadState_DeleteCurrent(PyThreadState *tstate)
{
    _Py_EnsureTstateNotNULL(tstate);
    tstate_delete_common(tstate);
    current_fast_clear(tstate->interp->runtime);
    _PyEval_ReleaseLock(tstate->interp, NULL);
    free_threadstate(tstate);
}

void
PyThreadState_DeleteCurrent(void)
{
    PyThreadState *tstate = current_fast_get(&_PyRuntime);
    _PyThreadState_DeleteCurrent(tstate);
}


/*
 * Delete all thread states except the one passed as argument.
 * Note that, if there is a current thread state, it *must* be the one
 * passed as argument.  Also, this won't touch any other interpreters
 * than the current one, since we don't know which thread state should
 * be kept in those other interpreters.
 */
void
_PyThreadState_DeleteExcept(PyThreadState *tstate)
{
    assert(tstate != NULL);
    PyInterpreterState *interp = tstate->interp;
    _PyRuntimeState *runtime = interp->runtime;

    HEAD_LOCK(runtime);
    /* Remove all thread states, except tstate, from the linked list of
       thread states.  This will allow calling PyThreadState_Clear()
       without holding the lock. */
    PyThreadState *list = interp->threads.head;
    if (list == tstate) {
        list = tstate->next;
    }
    if (tstate->prev) {
        tstate->prev->next = tstate->next;
    }
    if (tstate->next) {
        tstate->next->prev = tstate->prev;
    }
    tstate->prev = tstate->next = NULL;
    interp->threads.head = tstate;
    HEAD_UNLOCK(runtime);

    /* Clear and deallocate all stale thread states.  Even if this
       executes Python code, we should be safe since it executes
       in the current thread, not one of the stale threads. */
    PyThreadState *p, *next;
    for (p = list; p; p = next) {
        next = p->next;
        PyThreadState_Clear(p);
        free_threadstate(p);
    }
}


//-------------------------
// "detached" thread states
//-------------------------

void
_PyThreadState_InitDetached(PyThreadState *tstate, PyInterpreterState *interp)
{
    _PyRuntimeState *runtime = interp->runtime;

    HEAD_LOCK(runtime);
    interp->threads.next_unique_id += 1;
    uint64_t id = interp->threads.next_unique_id;
    HEAD_UNLOCK(runtime);

    init_threadstate(tstate, interp, id);
    // We do not call add_threadstate().
}

void
_PyThreadState_ClearDetached(PyThreadState *tstate)
{
    assert(!tstate->_status.bound);
    assert(!tstate->_status.bound_gilstate);
    assert(tstate->datastack_chunk == NULL);
    assert(tstate->thread_id == 0);
    assert(tstate->native_thread_id == 0);
    assert(tstate->next == NULL);
    assert(tstate->prev == NULL);

    PyThreadState_Clear(tstate);
    clear_datastack(tstate);
}

void
_PyThreadState_BindDetached(PyThreadState *tstate)
{
    assert(!_Py_IsMainInterpreter(
        current_fast_get(tstate->interp->runtime)->interp));
    assert(_Py_IsMainInterpreter(tstate->interp));
    bind_tstate(tstate);
    /* Unlike _PyThreadState_Bind(), we do not modify gilstate TSS. */
}

void
_PyThreadState_UnbindDetached(PyThreadState *tstate)
{
    assert(!_Py_IsMainInterpreter(
        current_fast_get(tstate->interp->runtime)->interp));
    assert(_Py_IsMainInterpreter(tstate->interp));
    assert(tstate_is_alive(tstate));
    assert(!tstate->_status.active);
    assert(gilstate_tss_get(tstate->interp->runtime) != tstate);

    unbind_tstate(tstate);

    /* This thread state may be bound/unbound repeatedly,
       so we must erase evidence that it was ever bound (or unbound). */
    tstate->_status.bound = 0;
    tstate->_status.unbound = 0;

    /* We must fully unlink the thread state from any OS thread,
       to allow it to be bound more than once. */
    tstate->thread_id = 0;
#ifdef PY_HAVE_THREAD_NATIVE_ID
    tstate->native_thread_id = 0;
#endif
}


//----------
// accessors
//----------

/* An extension mechanism to store arbitrary additional per-thread state.
   PyThreadState_GetDict() returns a dictionary that can be used to hold such
   state; the caller should pick a unique key and store its state there.  If
   PyThreadState_GetDict() returns NULL, an exception has *not* been raised
   and the caller should assume no per-thread state is available. */

PyObject *
_PyThreadState_GetDict(PyThreadState *tstate)
{
    assert(tstate != NULL);
    if (tstate->dict == NULL) {
        tstate->dict = PyDict_New();
        if (tstate->dict == NULL) {
            _PyErr_Clear(tstate);
        }
    }
    return tstate->dict;
}


PyObject *
PyThreadState_GetDict(void)
{
    PyThreadState *tstate = current_fast_get(&_PyRuntime);
    if (tstate == NULL) {
        return NULL;
    }
    return _PyThreadState_GetDict(tstate);
}


PyInterpreterState *
PyThreadState_GetInterpreter(PyThreadState *tstate)
{
    assert(tstate != NULL);
    return tstate->interp;
}


PyFrameObject*
PyThreadState_GetFrame(PyThreadState *tstate)
{
    assert(tstate != NULL);
    _PyInterpreterFrame *f = _PyThreadState_GetFrame(tstate);
    if (f == NULL) {
        return NULL;
    }
    PyFrameObject *frame = _PyFrame_GetFrameObject(f);
    if (frame == NULL) {
        PyErr_Clear();
    }
    return (PyFrameObject*)Py_XNewRef(frame);
}


uint64_t
PyThreadState_GetID(PyThreadState *tstate)
{
    assert(tstate != NULL);
    return tstate->id;
}


static inline void
tstate_activate(PyThreadState *tstate)
{
    assert(tstate != NULL);
    // XXX assert(tstate_is_alive(tstate));
    assert(tstate_is_bound(tstate));
    assert(!tstate->_status.active);

    assert(!tstate->_status.bound_gilstate ||
           tstate == gilstate_tss_get((tstate->interp->runtime)));
    if (!tstate->_status.bound_gilstate) {
        bind_gilstate_tstate(tstate);
    }

    tstate->_status.active = 1;
}

static inline void
tstate_deactivate(PyThreadState *tstate)
{
    assert(tstate != NULL);
    // XXX assert(tstate_is_alive(tstate));
    assert(tstate_is_bound(tstate));
    assert(tstate->_status.active);

    tstate->_status.active = 0;

    // We do not unbind the gilstate tstate here.
    // It will still be used in PyGILState_Ensure().
}


//----------
// other API
//----------

/* Asynchronously raise an exception in a thread.
   Requested by Just van Rossum and Alex Martelli.
   To prevent naive misuse, you must write your own extension
   to call this, or use ctypes.  Must be called with the GIL held.
   Returns the number of tstates modified (normally 1, but 0 if `id` didn't
   match any known thread id).  Can be called with exc=NULL to clear an
   existing async exception.  This raises no exceptions. */

// XXX Move this to Python/ceval_gil.c?
// XXX Deprecate this.
int
PyThreadState_SetAsyncExc(unsigned long id, PyObject *exc)
{
    _PyRuntimeState *runtime = &_PyRuntime;
    PyInterpreterState *interp = _PyInterpreterState_GET();

    /* Although the GIL is held, a few C API functions can be called
     * without the GIL held, and in particular some that create and
     * destroy thread and interpreter states.  Those can mutate the
     * list of thread states we're traversing, so to prevent that we lock
     * head_mutex for the duration.
     */
    HEAD_LOCK(runtime);
    for (PyThreadState *tstate = interp->threads.head; tstate != NULL; tstate = tstate->next) {
        if (tstate->thread_id != id) {
            continue;
        }

        /* Tricky:  we need to decref the current value
         * (if any) in tstate->async_exc, but that can in turn
         * allow arbitrary Python code to run, including
         * perhaps calls to this function.  To prevent
         * deadlock, we need to release head_mutex before
         * the decref.
         */
        PyObject *old_exc = tstate->async_exc;
        tstate->async_exc = Py_XNewRef(exc);
        HEAD_UNLOCK(runtime);

        Py_XDECREF(old_exc);
        _PyEval_SignalAsyncExc(tstate->interp);
        return 1;
    }
    HEAD_UNLOCK(runtime);
    return 0;
}


//---------------------------------
// API for the current thread state
//---------------------------------

PyThreadState *
_PyThreadState_UncheckedGet(void)
{
    return current_fast_get(&_PyRuntime);
}


PyThreadState *
PyThreadState_Get(void)
{
    PyThreadState *tstate = current_fast_get(&_PyRuntime);
    _Py_EnsureTstateNotNULL(tstate);
    return tstate;
}


static void
_swap_thread_states(_PyRuntimeState *runtime,
                    PyThreadState *oldts, PyThreadState *newts)
{
    // XXX Do this only if oldts != NULL?
    current_fast_clear(runtime);

    if (oldts != NULL) {
        // XXX assert(tstate_is_alive(oldts) && tstate_is_bound(oldts));
        tstate_deactivate(oldts);
    }

    if (newts != NULL) {
        // XXX assert(tstate_is_alive(newts));
        assert(tstate_is_bound(newts));
        current_fast_set(runtime, newts);
        tstate_activate(newts);
    }
}

PyThreadState *
_PyThreadState_SwapNoGIL(PyThreadState *newts)
{
#if defined(Py_DEBUG)
    /* This can be called from PyEval_RestoreThread(). Similar
       to it, we need to ensure errno doesn't change.
    */
    int err = errno;
#endif

    PyThreadState *oldts = current_fast_get(&_PyRuntime);
    _swap_thread_states(&_PyRuntime, oldts, newts);

#if defined(Py_DEBUG)
    errno = err;
#endif
    return oldts;
}

PyThreadState *
_PyThreadState_Swap(_PyRuntimeState *runtime, PyThreadState *newts)
{
    PyThreadState *oldts = current_fast_get(runtime);
    if (oldts != NULL) {
        _PyEval_ReleaseLock(oldts->interp, oldts);
    }
    _swap_thread_states(runtime, oldts, newts);
    if (newts != NULL) {
        _PyEval_AcquireLock(newts);
    }
    return oldts;
}

PyThreadState *
PyThreadState_Swap(PyThreadState *newts)
{
    return _PyThreadState_Swap(&_PyRuntime, newts);
}


void
_PyThreadState_Bind(PyThreadState *tstate)
{
    bind_tstate(tstate);
    // This makes sure there's a gilstate tstate bound
    // as soon as possible.
    if (gilstate_tss_get(tstate->interp->runtime) == NULL) {
        bind_gilstate_tstate(tstate);
    }
}


/***********************************/
/* routines for advanced debuggers */
/***********************************/

// (requested by David Beazley)
// Don't use unless you know what you are doing!

PyInterpreterState *
PyInterpreterState_Head(void)
{
    return _PyRuntime.interpreters.head;
}

PyInterpreterState *
PyInterpreterState_Main(void)
{
    return _PyInterpreterState_Main();
}

PyInterpreterState *
PyInterpreterState_Next(PyInterpreterState *interp) {
    return interp->next;
}

PyThreadState *
PyInterpreterState_ThreadHead(PyInterpreterState *interp) {
    return interp->threads.head;
}

PyThreadState *
PyThreadState_Next(PyThreadState *tstate) {
    return tstate->next;
}


/********************************************/
/* reporting execution state of all threads */
/********************************************/

/* The implementation of sys._current_frames().  This is intended to be
   called with the GIL held, as it will be when called via
   sys._current_frames().  It's possible it would work fine even without
   the GIL held, but haven't thought enough about that.
*/
PyObject *
_PyThread_CurrentFrames(void)
{
    _PyRuntimeState *runtime = &_PyRuntime;
    PyThreadState *tstate = current_fast_get(runtime);
    if (_PySys_Audit(tstate, "sys._current_frames", NULL) < 0) {
        return NULL;
    }

    PyObject *result = PyDict_New();
    if (result == NULL) {
        return NULL;
    }

    /* for i in all interpreters:
     *     for t in all of i's thread states:
     *          if t's frame isn't NULL, map t's id to its frame
     * Because these lists can mutate even when the GIL is held, we
     * need to grab head_mutex for the duration.
     */
    HEAD_LOCK(runtime);
    PyInterpreterState *i;
    for (i = runtime->interpreters.head; i != NULL; i = i->next) {
        PyThreadState *t;
        for (t = i->threads.head; t != NULL; t = t->next) {
            _PyInterpreterFrame *frame = t->cframe->current_frame;
            frame = _PyFrame_GetFirstComplete(frame);
            if (frame == NULL) {
                continue;
            }
            PyObject *id = PyLong_FromUnsignedLong(t->thread_id);
            if (id == NULL) {
                goto fail;
            }
            PyObject *frameobj = (PyObject *)_PyFrame_GetFrameObject(frame);
            if (frameobj == NULL) {
                Py_DECREF(id);
                goto fail;
            }
            int stat = PyDict_SetItem(result, id, frameobj);
            Py_DECREF(id);
            if (stat < 0) {
                goto fail;
            }
        }
    }
    goto done;

fail:
    Py_CLEAR(result);

done:
    HEAD_UNLOCK(runtime);
    return result;
}

/* The implementation of sys._current_exceptions().  This is intended to be
   called with the GIL held, as it will be when called via
   sys._current_exceptions().  It's possible it would work fine even without
   the GIL held, but haven't thought enough about that.
*/
PyObject *
_PyThread_CurrentExceptions(void)
{
    _PyRuntimeState *runtime = &_PyRuntime;
    PyThreadState *tstate = current_fast_get(runtime);

    _Py_EnsureTstateNotNULL(tstate);

    if (_PySys_Audit(tstate, "sys._current_exceptions", NULL) < 0) {
        return NULL;
    }

    PyObject *result = PyDict_New();
    if (result == NULL) {
        return NULL;
    }

    /* for i in all interpreters:
     *     for t in all of i's thread states:
     *          if t's frame isn't NULL, map t's id to its frame
     * Because these lists can mutate even when the GIL is held, we
     * need to grab head_mutex for the duration.
     */
    HEAD_LOCK(runtime);
    PyInterpreterState *i;
    for (i = runtime->interpreters.head; i != NULL; i = i->next) {
        PyThreadState *t;
        for (t = i->threads.head; t != NULL; t = t->next) {
            _PyErr_StackItem *err_info = _PyErr_GetTopmostException(t);
            if (err_info == NULL) {
                continue;
            }
            PyObject *id = PyLong_FromUnsignedLong(t->thread_id);
            if (id == NULL) {
                goto fail;
            }
            PyObject *exc = err_info->exc_value;
            assert(exc == NULL ||
                   exc == Py_None ||
                   PyExceptionInstance_Check(exc));

            int stat = PyDict_SetItem(result, id, exc == NULL ? Py_None : exc);
            Py_DECREF(id);
            if (stat < 0) {
                goto fail;
            }
        }
    }
    goto done;

fail:
    Py_CLEAR(result);

done:
    HEAD_UNLOCK(runtime);
    return result;
}


/***********************************/
/* Python "auto thread state" API. */
/***********************************/

/* Internal initialization/finalization functions called by
   Py_Initialize/Py_FinalizeEx
*/
PyStatus
_PyGILState_Init(PyInterpreterState *interp)
{
    if (!_Py_IsMainInterpreter(interp)) {
        /* Currently, PyGILState is shared by all interpreters. The main
         * interpreter is responsible to initialize it. */
        return _PyStatus_OK();
    }
    _PyRuntimeState *runtime = interp->runtime;
    assert(gilstate_tss_get(runtime) == NULL);
    assert(runtime->gilstate.autoInterpreterState == NULL);
    runtime->gilstate.autoInterpreterState = interp;
    return _PyStatus_OK();
}

void
_PyGILState_Fini(PyInterpreterState *interp)
{
    if (!_Py_IsMainInterpreter(interp)) {
        /* Currently, PyGILState is shared by all interpreters. The main
         * interpreter is responsible to initialize it. */
        return;
    }
    interp->runtime->gilstate.autoInterpreterState = NULL;
}


// XXX Drop this.
PyStatus
_PyGILState_SetTstate(PyThreadState *tstate)
{
    /* must init with valid states */
    assert(tstate != NULL);
    assert(tstate->interp != NULL);

    if (!_Py_IsMainInterpreter(tstate->interp)) {
        /* Currently, PyGILState is shared by all interpreters. The main
         * interpreter is responsible to initialize it. */
        return _PyStatus_OK();
    }

#ifndef NDEBUG
    _PyRuntimeState *runtime = tstate->interp->runtime;

    assert(runtime->gilstate.autoInterpreterState == tstate->interp);
    assert(gilstate_tss_get(runtime) == tstate);
    assert(tstate->gilstate_counter == 1);
#endif

    return _PyStatus_OK();
}

PyInterpreterState *
_PyGILState_GetInterpreterStateUnsafe(void)
{
    return _PyRuntime.gilstate.autoInterpreterState;
}

/* The public functions */

PyThreadState *
PyGILState_GetThisThreadState(void)
{
    _PyRuntimeState *runtime = &_PyRuntime;
    if (!gilstate_tss_initialized(runtime)) {
        return NULL;
    }
    return gilstate_tss_get(runtime);
}

int
PyGILState_Check(void)
{
    _PyRuntimeState *runtime = &_PyRuntime;
    if (!runtime->gilstate.check_enabled) {
        return 1;
    }

    if (!gilstate_tss_initialized(runtime)) {
        return 1;
    }

    PyThreadState *tstate = current_fast_get(runtime);
    if (tstate == NULL) {
        return 0;
    }

    return (tstate == gilstate_tss_get(runtime));
}

PyGILState_STATE
PyGILState_Ensure(void)
{
    _PyRuntimeState *runtime = &_PyRuntime;

    /* Note that we do not auto-init Python here - apart from
       potential races with 2 threads auto-initializing, pep-311
       spells out other issues.  Embedders are expected to have
       called Py_Initialize(). */

    /* Ensure that _PyEval_InitThreads() and _PyGILState_Init() have been
       called by Py_Initialize() */
    assert(_PyEval_ThreadsInitialized());
    assert(gilstate_tss_initialized(runtime));
    assert(runtime->gilstate.autoInterpreterState != NULL);

    PyThreadState *tcur = gilstate_tss_get(runtime);
    int has_gil;
    if (tcur == NULL) {
        /* Create a new Python thread state for this thread */
        tcur = new_threadstate(runtime->gilstate.autoInterpreterState);
        if (tcur == NULL) {
            Py_FatalError("Couldn't create thread-state for new thread");
        }
        bind_tstate(tcur);
        bind_gilstate_tstate(tcur);

        /* This is our thread state!  We'll need to delete it in the
           matching call to PyGILState_Release(). */
        assert(tcur->gilstate_counter == 1);
        tcur->gilstate_counter = 0;
        has_gil = 0; /* new thread state is never current */
    }
    else {
        has_gil = holds_gil(tcur);
    }

    if (!has_gil) {
        PyEval_RestoreThread(tcur);
    }

    /* Update our counter in the thread-state - no need for locks:
       - tcur will remain valid as we hold the GIL.
       - the counter is safe as we are the only thread "allowed"
         to modify this value
    */
    ++tcur->gilstate_counter;

    return has_gil ? PyGILState_LOCKED : PyGILState_UNLOCKED;
}

void
PyGILState_Release(PyGILState_STATE oldstate)
{
    _PyRuntimeState *runtime = &_PyRuntime;
    PyThreadState *tstate = gilstate_tss_get(runtime);
    if (tstate == NULL) {
        Py_FatalError("auto-releasing thread-state, "
                      "but no thread-state for this thread");
    }

    /* We must hold the GIL and have our thread state current */
    /* XXX - remove the check - the assert should be fine,
       but while this is very new (April 2003), the extra check
       by release-only users can't hurt.
    */
    if (!holds_gil(tstate)) {
        _Py_FatalErrorFormat(__func__,
                             "thread state %p must be current when releasing",
                             tstate);
    }
    assert(holds_gil(tstate));
    --tstate->gilstate_counter;
    assert(tstate->gilstate_counter >= 0); /* illegal counter value */

    /* If we're going to destroy this thread-state, we must
     * clear it while the GIL is held, as destructors may run.
     */
    if (tstate->gilstate_counter == 0) {
        /* can't have been locked when we created it */
        assert(oldstate == PyGILState_UNLOCKED);
        // XXX Unbind tstate here.
        PyThreadState_Clear(tstate);
        /* Delete the thread-state.  Note this releases the GIL too!
         * It's vital that the GIL be held here, to avoid shutdown
         * races; see bugs 225673 and 1061968 (that nasty bug has a
         * habit of coming back).
         */
        assert(current_fast_get(runtime) == tstate);
        _PyThreadState_DeleteCurrent(tstate);
    }
    /* Release the lock if necessary */
    else if (oldstate == PyGILState_UNLOCKED) {
        PyEval_SaveThread();
    }
}


/**************************/
/* cross-interpreter data */
/**************************/

/* cross-interpreter data */

static inline void
_xidata_init(_PyCrossInterpreterData *data)
{
    // If the value is being reused
    // then _xidata_clear() should have been called already.
    assert(data->data == NULL);
    assert(data->obj == NULL);
    *data = (_PyCrossInterpreterData){0};
    data->interp = -1;
}

static inline void
_xidata_clear(_PyCrossInterpreterData *data)
{
    if (data->free != NULL) {
        data->free(data->data);
    }
    data->data = NULL;
    Py_CLEAR(data->obj);
}

void
_PyCrossInterpreterData_Init(_PyCrossInterpreterData *data,
                             PyInterpreterState *interp,
                             void *shared, PyObject *obj,
                             xid_newobjectfunc new_object)
{
    assert(data != NULL);
    assert(new_object != NULL);
    _xidata_init(data);
    data->data = shared;
    if (obj != NULL) {
        assert(interp != NULL);
        // released in _PyCrossInterpreterData_Clear()
        data->obj = Py_NewRef(obj);
    }
    // Ideally every object would know its owning interpreter.
    // Until then, we have to rely on the caller to identify it
    // (but we don't need it in all cases).
    data->interp = (interp != NULL) ? interp->id : -1;
    data->new_object = new_object;
}

int
_PyCrossInterpreterData_InitWithSize(_PyCrossInterpreterData *data,
                                     PyInterpreterState *interp,
                                     const size_t size, PyObject *obj,
                                     xid_newobjectfunc new_object)
{
    assert(size > 0);
    // For now we always free the shared data in the same interpreter
    // where it was allocated, so the interpreter is required.
    assert(interp != NULL);
    _PyCrossInterpreterData_Init(data, interp, NULL, obj, new_object);
    data->data = PyMem_RawMalloc(size);
    if (data->data == NULL) {
        return -1;
    }
    data->free = PyMem_RawFree;
    return 0;
}

void
_PyCrossInterpreterData_Clear(PyInterpreterState *interp,
                              _PyCrossInterpreterData *data)
{
    assert(data != NULL);
    // This must be called in the owning interpreter.
    assert(interp == NULL || data->interp == interp->id);
    _xidata_clear(data);
}

static int
_check_xidata(PyThreadState *tstate, _PyCrossInterpreterData *data)
{
    // data->data can be anything, including NULL, so we don't check it.

    // data->obj may be NULL, so we don't check it.

    if (data->interp < 0) {
        _PyErr_SetString(tstate, PyExc_SystemError, "missing interp");
        return -1;
    }

    if (data->new_object == NULL) {
        _PyErr_SetString(tstate, PyExc_SystemError, "missing new_object func");
        return -1;
    }

    // data->free may be NULL, so we don't check it.

    return 0;
}

crossinterpdatafunc _PyCrossInterpreterData_Lookup(PyObject *);

/* This is a separate func from _PyCrossInterpreterData_Lookup in order
   to keep the registry code separate. */
static crossinterpdatafunc
_lookup_getdata(PyObject *obj)
{
    crossinterpdatafunc getdata = _PyCrossInterpreterData_Lookup(obj);
    if (getdata == NULL && PyErr_Occurred() == 0)
        PyErr_Format(PyExc_ValueError,
                     "%S does not support cross-interpreter data", obj);
    return getdata;
}

int
_PyObject_CheckCrossInterpreterData(PyObject *obj)
{
    crossinterpdatafunc getdata = _lookup_getdata(obj);
    if (getdata == NULL) {
        return -1;
    }
    return 0;
}

int
_PyObject_GetCrossInterpreterData(PyObject *obj, _PyCrossInterpreterData *data)
{
    _PyRuntimeState *runtime = &_PyRuntime;
    PyThreadState *tstate = current_fast_get(runtime);
#ifdef Py_DEBUG
    // The caller must hold the GIL
    _Py_EnsureTstateNotNULL(tstate);
#endif
    PyInterpreterState *interp = tstate->interp;

    // Reset data before re-populating.
    *data = (_PyCrossInterpreterData){0};
    data->interp = -1;

    // Call the "getdata" func for the object.
    Py_INCREF(obj);
    crossinterpdatafunc getdata = _lookup_getdata(obj);
    if (getdata == NULL) {
        Py_DECREF(obj);
        return -1;
    }
    int res = getdata(tstate, obj, data);
    Py_DECREF(obj);
    if (res != 0) {
        return -1;
    }

    // Fill in the blanks and validate the result.
    data->interp = interp->id;
    if (_check_xidata(tstate, data) != 0) {
        (void)_PyCrossInterpreterData_Release(data);
        return -1;
    }

    return 0;
}

PyObject *
_PyCrossInterpreterData_NewObject(_PyCrossInterpreterData *data)
{
    return data->new_object(data);
}

typedef void (*releasefunc)(PyInterpreterState *, void *);

static void
_call_in_interpreter(PyInterpreterState *interp, releasefunc func, void *arg)
{
    /* We would use Py_AddPendingCall() if it weren't specific to the
     * main interpreter (see bpo-33608).  In the meantime we take a
     * naive approach.
     */
    _PyRuntimeState *runtime = interp->runtime;
    PyThreadState *save_tstate = NULL;
    if (interp != current_fast_get(runtime)->interp) {
        // XXX Using the "head" thread isn't strictly correct.
        PyThreadState *tstate = PyInterpreterState_ThreadHead(interp);
        // XXX Possible GILState issues?
        save_tstate = _PyThreadState_Swap(runtime, tstate);
    }

    // XXX Once the GIL is per-interpreter, this should be called with the
    // calling interpreter's GIL released and the target interpreter's held.
    func(interp, arg);

    // Switch back.
    if (save_tstate != NULL) {
        _PyThreadState_Swap(runtime, save_tstate);
    }
}

int
_PyCrossInterpreterData_Release(_PyCrossInterpreterData *data)
{
    if (data->free == NULL && data->obj == NULL) {
        // Nothing to release!
        data->data = NULL;
        return 0;
    }

    // Switch to the original interpreter.
    PyInterpreterState *interp = _PyInterpreterState_LookUpID(data->interp);
    if (interp == NULL) {
        // The interpreter was already destroyed.
        // This function shouldn't have been called.
        // XXX Someone leaked some memory...
        assert(PyErr_Occurred());
        return -1;
    }

    // "Release" the data and/or the object.
    _call_in_interpreter(interp,
                         (releasefunc)_PyCrossInterpreterData_Clear, data);
    return 0;
}

/* registry of {type -> crossinterpdatafunc} */

/* For now we use a global registry of shareable classes.  An
   alternative would be to add a tp_* slot for a class's
   crossinterpdatafunc. It would be simpler and more efficient. */

static int
_xidregistry_add_type(struct _xidregistry *xidregistry, PyTypeObject *cls,
                 crossinterpdatafunc getdata)
{
    // Note that we effectively replace already registered classes
    // rather than failing.
    struct _xidregitem *newhead = PyMem_RawMalloc(sizeof(struct _xidregitem));
    if (newhead == NULL) {
        return -1;
    }
    // XXX Assign a callback to clear the entry from the registry?
    newhead->cls = PyWeakref_NewRef((PyObject *)cls, NULL);
    if (newhead->cls == NULL) {
        PyMem_RawFree(newhead);
        return -1;
    }
    newhead->getdata = getdata;
    newhead->prev = NULL;
    newhead->next = xidregistry->head;
    if (newhead->next != NULL) {
        newhead->next->prev = newhead;
    }
    xidregistry->head = newhead;
    return 0;
}

static struct _xidregitem *
_xidregistry_remove_entry(struct _xidregistry *xidregistry,
                          struct _xidregitem *entry)
{
    struct _xidregitem *next = entry->next;
    if (entry->prev != NULL) {
        assert(entry->prev->next == entry);
        entry->prev->next = next;
    }
    else {
        assert(xidregistry->head == entry);
        xidregistry->head = next;
    }
    if (next != NULL) {
        next->prev = entry->prev;
    }
    Py_DECREF(entry->cls);
    PyMem_RawFree(entry);
    return next;
}

static struct _xidregitem *
_xidregistry_find_type(struct _xidregistry *xidregistry, PyTypeObject *cls)
{
    struct _xidregitem *cur = xidregistry->head;
    while (cur != NULL) {
        PyObject *registered = _PyWeakref_GET_REF(cur->cls);
        if (registered == NULL) {
            // The weakly ref'ed object was freed.
            cur = _xidregistry_remove_entry(xidregistry, cur);
        }
        else {
            assert(PyType_Check(registered));
            if (registered == (PyObject *)cls) {
                Py_DECREF(registered);
                return cur;
            }
            Py_DECREF(registered);
            cur = cur->next;
        }
    }
    return NULL;
}

static void _register_builtins_for_crossinterpreter_data(struct _xidregistry *xidregistry);

int
_PyCrossInterpreterData_RegisterClass(PyTypeObject *cls,
                                       crossinterpdatafunc getdata)
{
    if (!PyType_Check(cls)) {
        PyErr_Format(PyExc_ValueError, "only classes may be registered");
        return -1;
    }
    if (getdata == NULL) {
        PyErr_Format(PyExc_ValueError, "missing 'getdata' func");
        return -1;
    }

    struct _xidregistry *xidregistry = &_PyRuntime.xidregistry ;
    PyThread_acquire_lock(xidregistry->mutex, WAIT_LOCK);
    if (xidregistry->head == NULL) {
        _register_builtins_for_crossinterpreter_data(xidregistry);
    }
    int res = _xidregistry_add_type(xidregistry, cls, getdata);
    PyThread_release_lock(xidregistry->mutex);
    return res;
}

int
_PyCrossInterpreterData_UnregisterClass(PyTypeObject *cls)
{
    int res = 0;
    struct _xidregistry *xidregistry = &_PyRuntime.xidregistry ;
    PyThread_acquire_lock(xidregistry->mutex, WAIT_LOCK);
    struct _xidregitem *matched = _xidregistry_find_type(xidregistry, cls);
    if (matched != NULL) {
        (void)_xidregistry_remove_entry(xidregistry, matched);
        res = 1;
    }
    PyThread_release_lock(xidregistry->mutex);
    return res;
}


/* Cross-interpreter objects are looked up by exact match on the class.
   We can reassess this policy when we move from a global registry to a
   tp_* slot. */

crossinterpdatafunc
_PyCrossInterpreterData_Lookup(PyObject *obj)
{
    struct _xidregistry *xidregistry = &_PyRuntime.xidregistry ;
    PyObject *cls = PyObject_Type(obj);
    PyThread_acquire_lock(xidregistry->mutex, WAIT_LOCK);
    if (xidregistry->head == NULL) {
        _register_builtins_for_crossinterpreter_data(xidregistry);
    }
    struct _xidregitem *matched = _xidregistry_find_type(xidregistry,
                                                         (PyTypeObject *)cls);
    Py_DECREF(cls);
    PyThread_release_lock(xidregistry->mutex);
    return matched != NULL ? matched->getdata : NULL;
}

/* cross-interpreter data for builtin types */

struct _shared_bytes_data {
    char *bytes;
    Py_ssize_t len;
};

static PyObject *
_new_bytes_object(_PyCrossInterpreterData *data)
{
    struct _shared_bytes_data *shared = (struct _shared_bytes_data *)(data->data);
    return PyBytes_FromStringAndSize(shared->bytes, shared->len);
}

static int
_bytes_shared(PyThreadState *tstate, PyObject *obj,
              _PyCrossInterpreterData *data)
{
    if (_PyCrossInterpreterData_InitWithSize(
            data, tstate->interp, sizeof(struct _shared_bytes_data), obj,
            _new_bytes_object
            ) < 0)
    {
        return -1;
    }
    struct _shared_bytes_data *shared = (struct _shared_bytes_data *)data->data;
    if (PyBytes_AsStringAndSize(obj, &shared->bytes, &shared->len) < 0) {
        _PyCrossInterpreterData_Clear(tstate->interp, data);
        return -1;
    }
    return 0;
}

struct _shared_str_data {
    int kind;
    const void *buffer;
    Py_ssize_t len;
};

static PyObject *
_new_str_object(_PyCrossInterpreterData *data)
{
    struct _shared_str_data *shared = (struct _shared_str_data *)(data->data);
    return PyUnicode_FromKindAndData(shared->kind, shared->buffer, shared->len);
}

static int
_str_shared(PyThreadState *tstate, PyObject *obj,
            _PyCrossInterpreterData *data)
{
    if (_PyCrossInterpreterData_InitWithSize(
            data, tstate->interp, sizeof(struct _shared_str_data), obj,
            _new_str_object
            ) < 0)
    {
        return -1;
    }
    struct _shared_str_data *shared = (struct _shared_str_data *)data->data;
    shared->kind = PyUnicode_KIND(obj);
    shared->buffer = PyUnicode_DATA(obj);
    shared->len = PyUnicode_GET_LENGTH(obj);
    return 0;
}

static PyObject *
_new_long_object(_PyCrossInterpreterData *data)
{
    return PyLong_FromSsize_t((Py_ssize_t)(data->data));
}

static int
_long_shared(PyThreadState *tstate, PyObject *obj,
             _PyCrossInterpreterData *data)
{
    /* Note that this means the size of shareable ints is bounded by
     * sys.maxsize.  Hence on 32-bit architectures that is half the
     * size of maximum shareable ints on 64-bit.
     */
    Py_ssize_t value = PyLong_AsSsize_t(obj);
    if (value == -1 && PyErr_Occurred()) {
        if (PyErr_ExceptionMatches(PyExc_OverflowError)) {
            PyErr_SetString(PyExc_OverflowError, "try sending as bytes");
        }
        return -1;
    }
    _PyCrossInterpreterData_Init(data, tstate->interp, (void *)value, NULL,
            _new_long_object);
    // data->obj and data->free remain NULL
    return 0;
}

static PyObject *
_new_none_object(_PyCrossInterpreterData *data)
{
    // XXX Singleton refcounts are problematic across interpreters...
    return Py_NewRef(Py_None);
}

static int
_none_shared(PyThreadState *tstate, PyObject *obj,
             _PyCrossInterpreterData *data)
{
    _PyCrossInterpreterData_Init(data, tstate->interp, NULL, NULL,
            _new_none_object);
    // data->data, data->obj and data->free remain NULL
    return 0;
}

static void
_register_builtins_for_crossinterpreter_data(struct _xidregistry *xidregistry)
{
    // None
    if (_xidregistry_add_type(xidregistry, (PyTypeObject *)PyObject_Type(Py_None), _none_shared) != 0) {
        Py_FatalError("could not register None for cross-interpreter sharing");
    }

    // int
    if (_xidregistry_add_type(xidregistry, &PyLong_Type, _long_shared) != 0) {
        Py_FatalError("could not register int for cross-interpreter sharing");
    }

    // bytes
    if (_xidregistry_add_type(xidregistry, &PyBytes_Type, _bytes_shared) != 0) {
        Py_FatalError("could not register bytes for cross-interpreter sharing");
    }

    // str
    if (_xidregistry_add_type(xidregistry, &PyUnicode_Type, _str_shared) != 0) {
        Py_FatalError("could not register str for cross-interpreter sharing");
    }
}


_PyFrameEvalFunction
_PyInterpreterState_GetEvalFrameFunc(PyInterpreterState *interp)
{
    if (interp->eval_frame == NULL) {
        return _PyEval_EvalFrameDefault;
    }
    return interp->eval_frame;
}


void
_PyInterpreterState_SetEvalFrameFunc(PyInterpreterState *interp,
                                     _PyFrameEvalFunction eval_frame)
{
    if (eval_frame == _PyEval_EvalFrameDefault) {
        interp->eval_frame = NULL;
    }
    else {
        interp->eval_frame = eval_frame;
    }
}


const PyConfig*
_PyInterpreterState_GetConfig(PyInterpreterState *interp)
{
    return &interp->config;
}


int
_PyInterpreterState_GetConfigCopy(PyConfig *config)
{
    PyInterpreterState *interp = _PyInterpreterState_GET();

    PyStatus status = _PyConfig_Copy(config, &interp->config);
    if (PyStatus_Exception(status)) {
        _PyErr_SetFromPyStatus(status);
        return -1;
    }
    return 0;
}


const PyConfig*
_Py_GetConfig(void)
{
    _PyRuntimeState *runtime = &_PyRuntime;
    assert(PyGILState_Check());
    PyThreadState *tstate = current_fast_get(runtime);
    _Py_EnsureTstateNotNULL(tstate);
    return _PyInterpreterState_GetConfig(tstate->interp);
}


int
_PyInterpreterState_HasFeature(PyInterpreterState *interp, unsigned long feature)
{
    return ((interp->feature_flags & feature) != 0);
}


#define MINIMUM_OVERHEAD 1000

static PyObject **
push_chunk(PyThreadState *tstate, int size)
{
    int allocate_size = DATA_STACK_CHUNK_SIZE;
    while (allocate_size < (int)sizeof(PyObject*)*(size + MINIMUM_OVERHEAD)) {
        allocate_size *= 2;
    }
    _PyStackChunk *new = allocate_chunk(allocate_size, tstate->datastack_chunk);
    if (new == NULL) {
        return NULL;
    }
    if (tstate->datastack_chunk) {
        tstate->datastack_chunk->top = tstate->datastack_top -
                                       &tstate->datastack_chunk->data[0];
    }
    tstate->datastack_chunk = new;
    tstate->datastack_limit = (PyObject **)(((char *)new) + allocate_size);
    // When new is the "root" chunk (i.e. new->previous == NULL), we can keep
    // _PyThreadState_PopFrame from freeing it later by "skipping" over the
    // first element:
    PyObject **res = &new->data[new->previous == NULL];
    tstate->datastack_top = res + size;
    return res;
}

_PyInterpreterFrame *
_PyThreadState_PushFrame(PyThreadState *tstate, size_t size)
{
    assert(size < INT_MAX/sizeof(PyObject *));
    if (_PyThreadState_HasStackSpace(tstate, (int)size)) {
        _PyInterpreterFrame *res = (_PyInterpreterFrame *)tstate->datastack_top;
        tstate->datastack_top += size;
        return res;
    }
    return (_PyInterpreterFrame *)push_chunk(tstate, (int)size);
}

void
_PyThreadState_PopFrame(PyThreadState *tstate, _PyInterpreterFrame * frame)
{
    assert(tstate->datastack_chunk);
    PyObject **base = (PyObject **)frame;
    if (base == &tstate->datastack_chunk->data[0]) {
        _PyStackChunk *chunk = tstate->datastack_chunk;
        _PyStackChunk *previous = chunk->previous;
        // push_chunk ensures that the root chunk is never popped:
        assert(previous);
        tstate->datastack_top = &previous->data[previous->top];
        tstate->datastack_chunk = previous;
        _PyObject_VirtualFree(chunk, chunk->size);
        tstate->datastack_limit = (PyObject **)(((char *)previous) + previous->size);
    }
    else {
        assert(tstate->datastack_top);
        assert(tstate->datastack_top >= base);
        tstate->datastack_top = base;
    }
}


#ifdef __cplusplus
}
#endif