summaryrefslogtreecommitdiffstats
path: root/src/runtime/mheap.go
blob: 1855330da5fa8af1c986fe92947592e87af35cb7 (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
// Copyright 2009 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

// Page heap.
//
// See malloc.go for overview.

package runtime

import (
	"internal/cpu"
	"runtime/internal/atomic"
	"runtime/internal/sys"
	"unsafe"
)

const (
	// minPhysPageSize is a lower-bound on the physical page size. The
	// true physical page size may be larger than this. In contrast,
	// sys.PhysPageSize is an upper-bound on the physical page size.
	minPhysPageSize = 4096

	// maxPhysPageSize is the maximum page size the runtime supports.
	maxPhysPageSize = 512 << 10

	// maxPhysHugePageSize sets an upper-bound on the maximum huge page size
	// that the runtime supports.
	maxPhysHugePageSize = pallocChunkBytes

	// pagesPerReclaimerChunk indicates how many pages to scan from the
	// pageInUse bitmap at a time. Used by the page reclaimer.
	//
	// Higher values reduce contention on scanning indexes (such as
	// h.reclaimIndex), but increase the minimum latency of the
	// operation.
	//
	// The time required to scan this many pages can vary a lot depending
	// on how many spans are actually freed. Experimentally, it can
	// scan for pages at ~300 GB/ms on a 2.6GHz Core i7, but can only
	// free spans at ~32 MB/ms. Using 512 pages bounds this at
	// roughly 100µs.
	//
	// Must be a multiple of the pageInUse bitmap element size and
	// must also evenly divide pagesPerArena.
	pagesPerReclaimerChunk = 512

	// physPageAlignedStacks indicates whether stack allocations must be
	// physical page aligned. This is a requirement for MAP_STACK on
	// OpenBSD.
	physPageAlignedStacks = GOOS == "openbsd"
)

// Main malloc heap.
// The heap itself is the "free" and "scav" treaps,
// but all the other global data is here too.
//
// mheap must not be heap-allocated because it contains mSpanLists,
// which must not be heap-allocated.
//
//go:notinheap
type mheap struct {
	// lock must only be acquired on the system stack, otherwise a g
	// could self-deadlock if its stack grows with the lock held.
	lock      mutex
	pages     pageAlloc // page allocation data structure
	sweepgen  uint32    // sweep generation, see comment in mspan; written during STW
	sweepdone uint32    // all spans are swept
	sweepers  uint32    // number of active sweepone calls

	// allspans is a slice of all mspans ever created. Each mspan
	// appears exactly once.
	//
	// The memory for allspans is manually managed and can be
	// reallocated and move as the heap grows.
	//
	// In general, allspans is protected by mheap_.lock, which
	// prevents concurrent access as well as freeing the backing
	// store. Accesses during STW might not hold the lock, but
	// must ensure that allocation cannot happen around the
	// access (since that may free the backing store).
	allspans []*mspan // all spans out there

	_ uint32 // align uint64 fields on 32-bit for atomics

	// Proportional sweep
	//
	// These parameters represent a linear function from heap_live
	// to page sweep count. The proportional sweep system works to
	// stay in the black by keeping the current page sweep count
	// above this line at the current heap_live.
	//
	// The line has slope sweepPagesPerByte and passes through a
	// basis point at (sweepHeapLiveBasis, pagesSweptBasis). At
	// any given time, the system is at (memstats.heap_live,
	// pagesSwept) in this space.
	//
	// It's important that the line pass through a point we
	// control rather than simply starting at a (0,0) origin
	// because that lets us adjust sweep pacing at any time while
	// accounting for current progress. If we could only adjust
	// the slope, it would create a discontinuity in debt if any
	// progress has already been made.
	pagesInUse         uint64  // pages of spans in stats mSpanInUse; updated atomically
	pagesSwept         uint64  // pages swept this cycle; updated atomically
	pagesSweptBasis    uint64  // pagesSwept to use as the origin of the sweep ratio; updated atomically
	sweepHeapLiveBasis uint64  // value of heap_live to use as the origin of sweep ratio; written with lock, read without
	sweepPagesPerByte  float64 // proportional sweep ratio; written with lock, read without
	// TODO(austin): pagesInUse should be a uintptr, but the 386
	// compiler can't 8-byte align fields.

	// scavengeGoal is the amount of total retained heap memory (measured by
	// heapRetained) that the runtime will try to maintain by returning memory
	// to the OS.
	scavengeGoal uint64

	// Page reclaimer state

	// reclaimIndex is the page index in allArenas of next page to
	// reclaim. Specifically, it refers to page (i %
	// pagesPerArena) of arena allArenas[i / pagesPerArena].
	//
	// If this is >= 1<<63, the page reclaimer is done scanning
	// the page marks.
	//
	// This is accessed atomically.
	reclaimIndex uint64
	// reclaimCredit is spare credit for extra pages swept. Since
	// the page reclaimer works in large chunks, it may reclaim
	// more than requested. Any spare pages released go to this
	// credit pool.
	//
	// This is accessed atomically.
	reclaimCredit uintptr

	// arenas is the heap arena map. It points to the metadata for
	// the heap for every arena frame of the entire usable virtual
	// address space.
	//
	// Use arenaIndex to compute indexes into this array.
	//
	// For regions of the address space that are not backed by the
	// Go heap, the arena map contains nil.
	//
	// Modifications are protected by mheap_.lock. Reads can be
	// performed without locking; however, a given entry can
	// transition from nil to non-nil at any time when the lock
	// isn't held. (Entries never transitions back to nil.)
	//
	// In general, this is a two-level mapping consisting of an L1
	// map and possibly many L2 maps. This saves space when there
	// are a huge number of arena frames. However, on many
	// platforms (even 64-bit), arenaL1Bits is 0, making this
	// effectively a single-level map. In this case, arenas[0]
	// will never be nil.
	arenas [1 << arenaL1Bits]*[1 << arenaL2Bits]*heapArena

	// heapArenaAlloc is pre-reserved space for allocating heapArena
	// objects. This is only used on 32-bit, where we pre-reserve
	// this space to avoid interleaving it with the heap itself.
	heapArenaAlloc linearAlloc

	// arenaHints is a list of addresses at which to attempt to
	// add more heap arenas. This is initially populated with a
	// set of general hint addresses, and grown with the bounds of
	// actual heap arena ranges.
	arenaHints *arenaHint

	// arena is a pre-reserved space for allocating heap arenas
	// (the actual arenas). This is only used on 32-bit.
	arena linearAlloc

	// allArenas is the arenaIndex of every mapped arena. This can
	// be used to iterate through the address space.
	//
	// Access is protected by mheap_.lock. However, since this is
	// append-only and old backing arrays are never freed, it is
	// safe to acquire mheap_.lock, copy the slice header, and
	// then release mheap_.lock.
	allArenas []arenaIdx

	// sweepArenas is a snapshot of allArenas taken at the
	// beginning of the sweep cycle. This can be read safely by
	// simply blocking GC (by disabling preemption).
	sweepArenas []arenaIdx

	// markArenas is a snapshot of allArenas taken at the beginning
	// of the mark cycle. Because allArenas is append-only, neither
	// this slice nor its contents will change during the mark, so
	// it can be read safely.
	markArenas []arenaIdx

	// curArena is the arena that the heap is currently growing
	// into. This should always be physPageSize-aligned.
	curArena struct {
		base, end uintptr
	}

	_ uint32 // ensure 64-bit alignment of central

	// central free lists for small size classes.
	// the padding makes sure that the mcentrals are
	// spaced CacheLinePadSize bytes apart, so that each mcentral.lock
	// gets its own cache line.
	// central is indexed by spanClass.
	central [numSpanClasses]struct {
		mcentral mcentral
		pad      [cpu.CacheLinePadSize - unsafe.Sizeof(mcentral{})%cpu.CacheLinePadSize]byte
	}

	spanalloc             fixalloc // allocator for span*
	cachealloc            fixalloc // allocator for mcache*
	specialfinalizeralloc fixalloc // allocator for specialfinalizer*
	specialprofilealloc   fixalloc // allocator for specialprofile*
	speciallock           mutex    // lock for special record allocators.
	arenaHintAlloc        fixalloc // allocator for arenaHints

	unused *specialfinalizer // never set, just here to force the specialfinalizer type into DWARF
}

var mheap_ mheap

// A heapArena stores metadata for a heap arena. heapArenas are stored
// outside of the Go heap and accessed via the mheap_.arenas index.
//
//go:notinheap
type heapArena struct {
	// bitmap stores the pointer/scalar bitmap for the words in
	// this arena. See mbitmap.go for a description. Use the
	// heapBits type to access this.
	bitmap [heapArenaBitmapBytes]byte

	// spans maps from virtual address page ID within this arena to *mspan.
	// For allocated spans, their pages map to the span itself.
	// For free spans, only the lowest and highest pages map to the span itself.
	// Internal pages map to an arbitrary span.
	// For pages that have never been allocated, spans entries are nil.
	//
	// Modifications are protected by mheap.lock. Reads can be
	// performed without locking, but ONLY from indexes that are
	// known to contain in-use or stack spans. This means there
	// must not be a safe-point between establishing that an
	// address is live and looking it up in the spans array.
	spans [pagesPerArena]*mspan

	// pageInUse is a bitmap that indicates which spans are in
	// state mSpanInUse. This bitmap is indexed by page number,
	// but only the bit corresponding to the first page in each
	// span is used.
	//
	// Reads and writes are atomic.
	pageInUse [pagesPerArena / 8]uint8

	// pageMarks is a bitmap that indicates which spans have any
	// marked objects on them. Like pageInUse, only the bit
	// corresponding to the first page in each span is used.
	//
	// Writes are done atomically during marking. Reads are
	// non-atomic and lock-free since they only occur during
	// sweeping (and hence never race with writes).
	//
	// This is used to quickly find whole spans that can be freed.
	//
	// TODO(austin): It would be nice if this was uint64 for
	// faster scanning, but we don't have 64-bit atomic bit
	// operations.
	pageMarks [pagesPerArena / 8]uint8

	// pageSpecials is a bitmap that indicates which spans have
	// specials (finalizers or other). Like pageInUse, only the bit
	// corresponding to the first page in each span is used.
	//
	// Writes are done atomically whenever a special is added to
	// a span and whenever the last special is removed from a span.
	// Reads are done atomically to find spans containing specials
	// during marking.
	pageSpecials [pagesPerArena / 8]uint8

	// checkmarks stores the debug.gccheckmark state. It is only
	// used if debug.gccheckmark > 0.
	checkmarks *checkmarksMap

	// zeroedBase marks the first byte of the first page in this
	// arena which hasn't been used yet and is therefore already
	// zero. zeroedBase is relative to the arena base.
	// Increases monotonically until it hits heapArenaBytes.
	//
	// This field is sufficient to determine if an allocation
	// needs to be zeroed because the page allocator follows an
	// address-ordered first-fit policy.
	//
	// Read atomically and written with an atomic CAS.
	zeroedBase uintptr
}

// arenaHint is a hint for where to grow the heap arenas. See
// mheap_.arenaHints.
//
//go:notinheap
type arenaHint struct {
	addr uintptr
	down bool
	next *arenaHint
}

// An mspan is a run of pages.
//
// When a mspan is in the heap free treap, state == mSpanFree
// and heapmap(s->start) == span, heapmap(s->start+s->npages-1) == span.
// If the mspan is in the heap scav treap, then in addition to the
// above scavenged == true. scavenged == false in all other cases.
//
// When a mspan is allocated, state == mSpanInUse or mSpanManual
// and heapmap(i) == span for all s->start <= i < s->start+s->npages.

// Every mspan is in one doubly-linked list, either in the mheap's
// busy list or one of the mcentral's span lists.

// An mspan representing actual memory has state mSpanInUse,
// mSpanManual, or mSpanFree. Transitions between these states are
// constrained as follows:
//
// * A span may transition from free to in-use or manual during any GC
//   phase.
//
// * During sweeping (gcphase == _GCoff), a span may transition from
//   in-use to free (as a result of sweeping) or manual to free (as a
//   result of stacks being freed).
//
// * During GC (gcphase != _GCoff), a span *must not* transition from
//   manual or in-use to free. Because concurrent GC may read a pointer
//   and then look up its span, the span state must be monotonic.
//
// Setting mspan.state to mSpanInUse or mSpanManual must be done
// atomically and only after all other span fields are valid.
// Likewise, if inspecting a span is contingent on it being
// mSpanInUse, the state should be loaded atomically and checked
// before depending on other fields. This allows the garbage collector
// to safely deal with potentially invalid pointers, since resolving
// such pointers may race with a span being allocated.
type mSpanState uint8

const (
	mSpanDead   mSpanState = iota
	mSpanInUse             // allocated for garbage collected heap
	mSpanManual            // allocated for manual management (e.g., stack allocator)
)

// mSpanStateNames are the names of the span states, indexed by
// mSpanState.
var mSpanStateNames = []string{
	"mSpanDead",
	"mSpanInUse",
	"mSpanManual",
	"mSpanFree",
}

// mSpanStateBox holds an mSpanState and provides atomic operations on
// it. This is a separate type to disallow accidental comparison or
// assignment with mSpanState.
type mSpanStateBox struct {
	s mSpanState
}

func (b *mSpanStateBox) set(s mSpanState) {
	atomic.Store8((*uint8)(&b.s), uint8(s))
}

func (b *mSpanStateBox) get() mSpanState {
	return mSpanState(atomic.Load8((*uint8)(&b.s)))
}

// mSpanList heads a linked list of spans.
//
//go:notinheap
type mSpanList struct {
	first *mspan // first span in list, or nil if none
	last  *mspan // last span in list, or nil if none
}

//go:notinheap
type mspan struct {
	next *mspan     // next span in list, or nil if none
	prev *mspan     // previous span in list, or nil if none
	list *mSpanList // For debugging. TODO: Remove.

	startAddr uintptr // address of first byte of span aka s.base()
	npages    uintptr // number of pages in span

	manualFreeList gclinkptr // list of free objects in mSpanManual spans

	// freeindex is the slot index between 0 and nelems at which to begin scanning
	// for the next free object in this span.
	// Each allocation scans allocBits starting at freeindex until it encounters a 0
	// indicating a free object. freeindex is then adjusted so that subsequent scans begin
	// just past the newly discovered free object.
	//
	// If freeindex == nelem, this span has no free objects.
	//
	// allocBits is a bitmap of objects in this span.
	// If n >= freeindex and allocBits[n/8] & (1<<(n%8)) is 0
	// then object n is free;
	// otherwise, object n is allocated. Bits starting at nelem are
	// undefined and should never be referenced.
	//
	// Object n starts at address n*elemsize + (start << pageShift).
	freeindex uintptr
	// TODO: Look up nelems from sizeclass and remove this field if it
	// helps performance.
	nelems uintptr // number of object in the span.

	// Cache of the allocBits at freeindex. allocCache is shifted
	// such that the lowest bit corresponds to the bit freeindex.
	// allocCache holds the complement of allocBits, thus allowing
	// ctz (count trailing zero) to use it directly.
	// allocCache may contain bits beyond s.nelems; the caller must ignore
	// these.
	allocCache uint64

	// allocBits and gcmarkBits hold pointers to a span's mark and
	// allocation bits. The pointers are 8 byte aligned.
	// There are three arenas where this data is held.
	// free: Dirty arenas that are no longer accessed
	//       and can be reused.
	// next: Holds information to be used in the next GC cycle.
	// current: Information being used during this GC cycle.
	// previous: Information being used during the last GC cycle.
	// A new GC cycle starts with the call to finishsweep_m.
	// finishsweep_m moves the previous arena to the free arena,
	// the current arena to the previous arena, and
	// the next arena to the current arena.
	// The next arena is populated as the spans request
	// memory to hold gcmarkBits for the next GC cycle as well
	// as allocBits for newly allocated spans.
	//
	// The pointer arithmetic is done "by hand" instead of using
	// arrays to avoid bounds checks along critical performance
	// paths.
	// The sweep will free the old allocBits and set allocBits to the
	// gcmarkBits. The gcmarkBits are replaced with a fresh zeroed
	// out memory.
	allocBits  *gcBits
	gcmarkBits *gcBits

	// sweep generation:
	// if sweepgen == h->sweepgen - 2, the span needs sweeping
	// if sweepgen == h->sweepgen - 1, the span is currently being swept
	// if sweepgen == h->sweepgen, the span is swept and ready to use
	// if sweepgen == h->sweepgen + 1, the span was cached before sweep began and is still cached, and needs sweeping
	// if sweepgen == h->sweepgen + 3, the span was swept and then cached and is still cached
	// h->sweepgen is incremented by 2 after every GC

	sweepgen    uint32
	divMul      uint16        // for divide by elemsize - divMagic.mul
	baseMask    uint16        // if non-0, elemsize is a power of 2, & this will get object allocation base
	allocCount  uint16        // number of allocated objects
	spanclass   spanClass     // size class and noscan (uint8)
	state       mSpanStateBox // mSpanInUse etc; accessed atomically (get/set methods)
	needzero    uint8         // needs to be zeroed before allocation
	divShift    uint8         // for divide by elemsize - divMagic.shift
	divShift2   uint8         // for divide by elemsize - divMagic.shift2
	elemsize    uintptr       // computed from sizeclass or from npages
	limit       uintptr       // end of data in span
	speciallock mutex         // guards specials list
	specials    *special      // linked list of special records sorted by offset.
}

func (s *mspan) base() uintptr {
	return s.startAddr
}

func (s *mspan) layout() (size, n, total uintptr) {
	total = s.npages << _PageShift
	size = s.elemsize
	if size > 0 {
		n = total / size
	}
	return
}

// recordspan adds a newly allocated span to h.allspans.
//
// This only happens the first time a span is allocated from
// mheap.spanalloc (it is not called when a span is reused).
//
// Write barriers are disallowed here because it can be called from
// gcWork when allocating new workbufs. However, because it's an
// indirect call from the fixalloc initializer, the compiler can't see
// this.
//
// The heap lock must be held.
//
//go:nowritebarrierrec
func recordspan(vh unsafe.Pointer, p unsafe.Pointer) {
	h := (*mheap)(vh)
	s := (*mspan)(p)

	assertLockHeld(&h.lock)

	if len(h.allspans) >= cap(h.allspans) {
		n := 64 * 1024 / sys.PtrSize
		if n < cap(h.allspans)*3/2 {
			n = cap(h.allspans) * 3 / 2
		}
		var new []*mspan
		sp := (*slice)(unsafe.Pointer(&new))
		sp.array = sysAlloc(uintptr(n)*sys.PtrSize, &memstats.other_sys)
		if sp.array == nil {
			throw("runtime: cannot allocate memory")
		}
		sp.len = len(h.allspans)
		sp.cap = n
		if len(h.allspans) > 0 {
			copy(new, h.allspans)
		}
		oldAllspans := h.allspans
		*(*notInHeapSlice)(unsafe.Pointer(&h.allspans)) = *(*notInHeapSlice)(unsafe.Pointer(&new))
		if len(oldAllspans) != 0 {
			sysFree(unsafe.Pointer(&oldAllspans[0]), uintptr(cap(oldAllspans))*unsafe.Sizeof(oldAllspans[0]), &memstats.other_sys)
		}
	}
	h.allspans = h.allspans[:len(h.allspans)+1]
	h.allspans[len(h.allspans)-1] = s
}

// A spanClass represents the size class and noscan-ness of a span.
//
// Each size class has a noscan spanClass and a scan spanClass. The
// noscan spanClass contains only noscan objects, which do not contain
// pointers and thus do not need to be scanned by the garbage
// collector.
type spanClass uint8

const (
	numSpanClasses = _NumSizeClasses << 1
	tinySpanClass  = spanClass(tinySizeClass<<1 | 1)
)

func makeSpanClass(sizeclass uint8, noscan bool) spanClass {
	return spanClass(sizeclass<<1) | spanClass(bool2int(noscan))
}

func (sc spanClass) sizeclass() int8 {
	return int8(sc >> 1)
}

func (sc spanClass) noscan() bool {
	return sc&1 != 0
}

// arenaIndex returns the index into mheap_.arenas of the arena
// containing metadata for p. This index combines of an index into the
// L1 map and an index into the L2 map and should be used as
// mheap_.arenas[ai.l1()][ai.l2()].
//
// If p is outside the range of valid heap addresses, either l1() or
// l2() will be out of bounds.
//
// It is nosplit because it's called by spanOf and several other
// nosplit functions.
//
//go:nosplit
func arenaIndex(p uintptr) arenaIdx {
	return arenaIdx((p - arenaBaseOffset) / heapArenaBytes)
}

// arenaBase returns the low address of the region covered by heap
// arena i.
func arenaBase(i arenaIdx) uintptr {
	return uintptr(i)*heapArenaBytes + arenaBaseOffset
}

type arenaIdx uint

func (i arenaIdx) l1() uint {
	if arenaL1Bits == 0 {
		// Let the compiler optimize this away if there's no
		// L1 map.
		return 0
	} else {
		return uint(i) >> arenaL1Shift
	}
}

func (i arenaIdx) l2() uint {
	if arenaL1Bits == 0 {
		return uint(i)
	} else {
		return uint(i) & (1<<arenaL2Bits - 1)
	}
}

// inheap reports whether b is a pointer into a (potentially dead) heap object.
// It returns false for pointers into mSpanManual spans.
// Non-preemptible because it is used by write barriers.
//go:nowritebarrier
//go:nosplit
func inheap(b uintptr) bool {
	return spanOfHeap(b) != nil
}

// inHeapOrStack is a variant of inheap that returns true for pointers
// into any allocated heap span.
//
//go:nowritebarrier
//go:nosplit
func inHeapOrStack(b uintptr) bool {
	s := spanOf(b)
	if s == nil || b < s.base() {
		return false
	}
	switch s.state.get() {
	case mSpanInUse, mSpanManual:
		return b < s.limit
	default:
		return false
	}
}

// spanOf returns the span of p. If p does not point into the heap
// arena or no span has ever contained p, spanOf returns nil.
//
// If p does not point to allocated memory, this may return a non-nil
// span that does *not* contain p. If this is a possibility, the
// caller should either call spanOfHeap or check the span bounds
// explicitly.
//
// Must be nosplit because it has callers that are nosplit.
//
//go:nosplit
func spanOf(p uintptr) *mspan {
	// This function looks big, but we use a lot of constant
	// folding around arenaL1Bits to get it under the inlining
	// budget. Also, many of the checks here are safety checks
	// that Go needs to do anyway, so the generated code is quite
	// short.
	ri := arenaIndex(p)
	if arenaL1Bits == 0 {
		// If there's no L1, then ri.l1() can't be out of bounds but ri.l2() can.
		if ri.l2() >= uint(len(mheap_.arenas[0])) {
			return nil
		}
	} else {
		// If there's an L1, then ri.l1() can be out of bounds but ri.l2() can't.
		if ri.l1() >= uint(len(mheap_.arenas)) {
			return nil
		}
	}
	l2 := mheap_.arenas[ri.l1()]
	if arenaL1Bits != 0 && l2 == nil { // Should never happen if there's no L1.
		return nil
	}
	ha := l2[ri.l2()]
	if ha == nil {
		return nil
	}
	return ha.spans[(p/pageSize)%pagesPerArena]
}

// spanOfUnchecked is equivalent to spanOf, but the caller must ensure
// that p points into an allocated heap arena.
//
// Must be nosplit because it has callers that are nosplit.
//
//go:nosplit
func spanOfUnchecked(p uintptr) *mspan {
	ai := arenaIndex(p)
	return mheap_.arenas[ai.l1()][ai.l2()].spans[(p/pageSize)%pagesPerArena]
}

// spanOfHeap is like spanOf, but returns nil if p does not point to a
// heap object.
//
// Must be nosplit because it has callers that are nosplit.
//
//go:nosplit
func spanOfHeap(p uintptr) *mspan {
	s := spanOf(p)
	// s is nil if it's never been allocated. Otherwise, we check
	// its state first because we don't trust this pointer, so we
	// have to synchronize with span initialization. Then, it's
	// still possible we picked up a stale span pointer, so we
	// have to check the span's bounds.
	if s == nil || s.state.get() != mSpanInUse || p < s.base() || p >= s.limit {
		return nil
	}
	return s
}

// pageIndexOf returns the arena, page index, and page mask for pointer p.
// The caller must ensure p is in the heap.
func pageIndexOf(p uintptr) (arena *heapArena, pageIdx uintptr, pageMask uint8) {
	ai := arenaIndex(p)
	arena = mheap_.arenas[ai.l1()][ai.l2()]
	pageIdx = ((p / pageSize) / 8) % uintptr(len(arena.pageInUse))
	pageMask = byte(1 << ((p / pageSize) % 8))
	return
}

// Initialize the heap.
func (h *mheap) init() {
	lockInit(&h.lock, lockRankMheap)
	lockInit(&h.speciallock, lockRankMheapSpecial)

	h.spanalloc.init(unsafe.Sizeof(mspan{}), recordspan, unsafe.Pointer(h), &memstats.mspan_sys)
	h.cachealloc.init(unsafe.Sizeof(mcache{}), nil, nil, &memstats.mcache_sys)
	h.specialfinalizeralloc.init(unsafe.Sizeof(specialfinalizer{}), nil, nil, &memstats.other_sys)
	h.specialprofilealloc.init(unsafe.Sizeof(specialprofile{}), nil, nil, &memstats.other_sys)
	h.arenaHintAlloc.init(unsafe.Sizeof(arenaHint{}), nil, nil, &memstats.other_sys)

	// Don't zero mspan allocations. Background sweeping can
	// inspect a span concurrently with allocating it, so it's
	// important that the span's sweepgen survive across freeing
	// and re-allocating a span to prevent background sweeping
	// from improperly cas'ing it from 0.
	//
	// This is safe because mspan contains no heap pointers.
	h.spanalloc.zero = false

	// h->mapcache needs no init

	for i := range h.central {
		h.central[i].mcentral.init(spanClass(i))
	}

	h.pages.init(&h.lock, &memstats.gcMiscSys)
}

// reclaim sweeps and reclaims at least npage pages into the heap.
// It is called before allocating npage pages to keep growth in check.
//
// reclaim implements the page-reclaimer half of the sweeper.
//
// h.lock must NOT be held.
func (h *mheap) reclaim(npage uintptr) {
	// TODO(austin): Half of the time spent freeing spans is in
	// locking/unlocking the heap (even with low contention). We
	// could make the slow path here several times faster by
	// batching heap frees.

	// Bail early if there's no more reclaim work.
	if atomic.Load64(&h.reclaimIndex) >= 1<<63 {
		return
	}

	// Disable preemption so the GC can't start while we're
	// sweeping, so we can read h.sweepArenas, and so
	// traceGCSweepStart/Done pair on the P.
	mp := acquirem()

	if trace.enabled {
		traceGCSweepStart()
	}

	arenas := h.sweepArenas
	locked := false
	for npage > 0 {
		// Pull from accumulated credit first.
		if credit := atomic.Loaduintptr(&h.reclaimCredit); credit > 0 {
			take := credit
			if take > npage {
				// Take only what we need.
				take = npage
			}
			if atomic.Casuintptr(&h.reclaimCredit, credit, credit-take) {
				npage -= take
			}
			continue
		}

		// Claim a chunk of work.
		idx := uintptr(atomic.Xadd64(&h.reclaimIndex, pagesPerReclaimerChunk) - pagesPerReclaimerChunk)
		if idx/pagesPerArena >= uintptr(len(arenas)) {
			// Page reclaiming is done.
			atomic.Store64(&h.reclaimIndex, 1<<63)
			break
		}

		if !locked {
			// Lock the heap for reclaimChunk.
			lock(&h.lock)
			locked = true
		}

		// Scan this chunk.
		nfound := h.reclaimChunk(arenas, idx, pagesPerReclaimerChunk)
		if nfound <= npage {
			npage -= nfound
		} else {
			// Put spare pages toward global credit.
			atomic.Xadduintptr(&h.reclaimCredit, nfound-npage)
			npage = 0
		}
	}
	if locked {
		unlock(&h.lock)
	}

	if trace.enabled {
		traceGCSweepDone()
	}
	releasem(mp)
}

// reclaimChunk sweeps unmarked spans that start at page indexes [pageIdx, pageIdx+n).
// It returns the number of pages returned to the heap.
//
// h.lock must be held and the caller must be non-preemptible. Note: h.lock may be
// temporarily unlocked and re-locked in order to do sweeping or if tracing is
// enabled.
func (h *mheap) reclaimChunk(arenas []arenaIdx, pageIdx, n uintptr) uintptr {
	// The heap lock must be held because this accesses the
	// heapArena.spans arrays using potentially non-live pointers.
	// In particular, if a span were freed and merged concurrently
	// with this probing heapArena.spans, it would be possible to
	// observe arbitrary, stale span pointers.
	assertLockHeld(&h.lock)

	n0 := n
	var nFreed uintptr
	sg := h.sweepgen
	for n > 0 {
		ai := arenas[pageIdx/pagesPerArena]
		ha := h.arenas[ai.l1()][ai.l2()]

		// Get a chunk of the bitmap to work on.
		arenaPage := uint(pageIdx % pagesPerArena)
		inUse := ha.pageInUse[arenaPage/8:]
		marked := ha.pageMarks[arenaPage/8:]
		if uintptr(len(inUse)) > n/8 {
			inUse = inUse[:n/8]
			marked = marked[:n/8]
		}

		// Scan this bitmap chunk for spans that are in-use
		// but have no marked objects on them.
		for i := range inUse {
			inUseUnmarked := atomic.Load8(&inUse[i]) &^ marked[i]
			if inUseUnmarked == 0 {
				continue
			}

			for j := uint(0); j < 8; j++ {
				if inUseUnmarked&(1<<j) != 0 {
					s := ha.spans[arenaPage+uint(i)*8+j]
					if atomic.Load(&s.sweepgen) == sg-2 && atomic.Cas(&s.sweepgen, sg-2, sg-1) {
						npages := s.npages
						unlock(&h.lock)
						if s.sweep(false) {
							nFreed += npages
						}
						lock(&h.lock)
						// Reload inUse. It's possible nearby
						// spans were freed when we dropped the
						// lock and we don't want to get stale
						// pointers from the spans array.
						inUseUnmarked = atomic.Load8(&inUse[i]) &^ marked[i]
					}
				}
			}
		}

		// Advance.
		pageIdx += uintptr(len(inUse) * 8)
		n -= uintptr(len(inUse) * 8)
	}
	if trace.enabled {
		unlock(&h.lock)
		// Account for pages scanned but not reclaimed.
		traceGCSweepSpan((n0 - nFreed) * pageSize)
		lock(&h.lock)
	}

	assertLockHeld(&h.lock) // Must be locked on return.
	return nFreed
}

// spanAllocType represents the type of allocation to make, or
// the type of allocation to be freed.
type spanAllocType uint8

const (
	spanAllocHeap          spanAllocType = iota // heap span
	spanAllocStack                              // stack span
	spanAllocPtrScalarBits                      // unrolled GC prog bitmap span
	spanAllocWorkBuf                            // work buf span
)

// manual returns true if the span allocation is manually managed.
func (s spanAllocType) manual() bool {
	return s != spanAllocHeap
}

// alloc allocates a new span of npage pages from the GC'd heap.
//
// spanclass indicates the span's size class and scannability.
//
// If needzero is true, the memory for the returned span will be zeroed.
func (h *mheap) alloc(npages uintptr, spanclass spanClass, needzero bool) *mspan {
	// Don't do any operations that lock the heap on the G stack.
	// It might trigger stack growth, and the stack growth code needs
	// to be able to allocate heap.
	var s *mspan
	systemstack(func() {
		// To prevent excessive heap growth, before allocating n pages
		// we need to sweep and reclaim at least n pages.
		if h.sweepdone == 0 {
			h.reclaim(npages)
		}
		s = h.allocSpan(npages, spanAllocHeap, spanclass)
	})

	if s != nil {
		if needzero && s.needzero != 0 {
			memclrNoHeapPointers(unsafe.Pointer(s.base()), s.npages<<_PageShift)
		}
		s.needzero = 0
	}
	return s
}

// allocManual allocates a manually-managed span of npage pages.
// allocManual returns nil if allocation fails.
//
// allocManual adds the bytes used to *stat, which should be a
// memstats in-use field. Unlike allocations in the GC'd heap, the
// allocation does *not* count toward heap_inuse or heap_sys.
//
// The memory backing the returned span may not be zeroed if
// span.needzero is set.
//
// allocManual must be called on the system stack because it may
// acquire the heap lock via allocSpan. See mheap for details.
//
// If new code is written to call allocManual, do NOT use an
// existing spanAllocType value and instead declare a new one.
//
//go:systemstack
func (h *mheap) allocManual(npages uintptr, typ spanAllocType) *mspan {
	if !typ.manual() {
		throw("manual span allocation called with non-manually-managed type")
	}
	return h.allocSpan(npages, typ, 0)
}

// setSpans modifies the span map so [spanOf(base), spanOf(base+npage*pageSize))
// is s.
func (h *mheap) setSpans(base, npage uintptr, s *mspan) {
	p := base / pageSize
	ai := arenaIndex(base)
	ha := h.arenas[ai.l1()][ai.l2()]
	for n := uintptr(0); n < npage; n++ {
		i := (p + n) % pagesPerArena
		if i == 0 {
			ai = arenaIndex(base + n*pageSize)
			ha = h.arenas[ai.l1()][ai.l2()]
		}
		ha.spans[i] = s
	}
}

// allocNeedsZero checks if the region of address space [base, base+npage*pageSize),
// assumed to be allocated, needs to be zeroed, updating heap arena metadata for
// future allocations.
//
// This must be called each time pages are allocated from the heap, even if the page
// allocator can otherwise prove the memory it's allocating is already zero because
// they're fresh from the operating system. It updates heapArena metadata that is
// critical for future page allocations.
//
// There are no locking constraints on this method.
func (h *mheap) allocNeedsZero(base, npage uintptr) (needZero bool) {
	for npage > 0 {
		ai := arenaIndex(base)
		ha := h.arenas[ai.l1()][ai.l2()]

		zeroedBase := atomic.Loaduintptr(&ha.zeroedBase)
		arenaBase := base % heapArenaBytes
		if arenaBase < zeroedBase {
			// We extended into the non-zeroed part of the
			// arena, so this region needs to be zeroed before use.
			//
			// zeroedBase is monotonically increasing, so if we see this now then
			// we can be sure we need to zero this memory region.
			//
			// We still need to update zeroedBase for this arena, and
			// potentially more arenas.
			needZero = true
		}
		// We may observe arenaBase > zeroedBase if we're racing with one or more
		// allocations which are acquiring memory directly before us in the address
		// space. But, because we know no one else is acquiring *this* memory, it's
		// still safe to not zero.

		// Compute how far into the arena we extend into, capped
		// at heapArenaBytes.
		arenaLimit := arenaBase + npage*pageSize
		if arenaLimit > heapArenaBytes {
			arenaLimit = heapArenaBytes
		}
		// Increase ha.zeroedBase so it's >= arenaLimit.
		// We may be racing with other updates.
		for arenaLimit > zeroedBase {
			if atomic.Casuintptr(&ha.zeroedBase, zeroedBase, arenaLimit) {
				break
			}
			zeroedBase = atomic.Loaduintptr(&ha.zeroedBase)
			// Sanity check zeroedBase.
			if zeroedBase <= arenaLimit && zeroedBase > arenaBase {
				// The zeroedBase moved into the space we were trying to
				// claim. That's very bad, and indicates someone allocated
				// the same region we did.
				throw("potentially overlapping in-use allocations detected")
			}
		}

		// Move base forward and subtract from npage to move into
		// the next arena, or finish.
		base += arenaLimit - arenaBase
		npage -= (arenaLimit - arenaBase) / pageSize
	}
	return
}

// tryAllocMSpan attempts to allocate an mspan object from
// the P-local cache, but may fail.
//
// h.lock need not be held.
//
// This caller must ensure that its P won't change underneath
// it during this function. Currently to ensure that we enforce
// that the function is run on the system stack, because that's
// the only place it is used now. In the future, this requirement
// may be relaxed if its use is necessary elsewhere.
//
//go:systemstack
func (h *mheap) tryAllocMSpan() *mspan {
	pp := getg().m.p.ptr()
	// If we don't have a p or the cache is empty, we can't do
	// anything here.
	if pp == nil || pp.mspancache.len == 0 {
		return nil
	}
	// Pull off the last entry in the cache.
	s := pp.mspancache.buf[pp.mspancache.len-1]
	pp.mspancache.len--
	return s
}

// allocMSpanLocked allocates an mspan object.
//
// h.lock must be held.
//
// allocMSpanLocked must be called on the system stack because
// its caller holds the heap lock. See mheap for details.
// Running on the system stack also ensures that we won't
// switch Ps during this function. See tryAllocMSpan for details.
//
//go:systemstack
func (h *mheap) allocMSpanLocked() *mspan {
	assertLockHeld(&h.lock)

	pp := getg().m.p.ptr()
	if pp == nil {
		// We don't have a p so just do the normal thing.
		return (*mspan)(h.spanalloc.alloc())
	}
	// Refill the cache if necessary.
	if pp.mspancache.len == 0 {
		const refillCount = len(pp.mspancache.buf) / 2
		for i := 0; i < refillCount; i++ {
			pp.mspancache.buf[i] = (*mspan)(h.spanalloc.alloc())
		}
		pp.mspancache.len = refillCount
	}
	// Pull off the last entry in the cache.
	s := pp.mspancache.buf[pp.mspancache.len-1]
	pp.mspancache.len--
	return s
}

// freeMSpanLocked free an mspan object.
//
// h.lock must be held.
//
// freeMSpanLocked must be called on the system stack because
// its caller holds the heap lock. See mheap for details.
// Running on the system stack also ensures that we won't
// switch Ps during this function. See tryAllocMSpan for details.
//
//go:systemstack
func (h *mheap) freeMSpanLocked(s *mspan) {
	assertLockHeld(&h.lock)

	pp := getg().m.p.ptr()
	// First try to free the mspan directly to the cache.
	if pp != nil && pp.mspancache.len < len(pp.mspancache.buf) {
		pp.mspancache.buf[pp.mspancache.len] = s
		pp.mspancache.len++
		return
	}
	// Failing that (or if we don't have a p), just free it to
	// the heap.
	h.spanalloc.free(unsafe.Pointer(s))
}

// allocSpan allocates an mspan which owns npages worth of memory.
//
// If typ.manual() == false, allocSpan allocates a heap span of class spanclass
// and updates heap accounting. If manual == true, allocSpan allocates a
// manually-managed span (spanclass is ignored), and the caller is
// responsible for any accounting related to its use of the span. Either
// way, allocSpan will atomically add the bytes in the newly allocated
// span to *sysStat.
//
// The returned span is fully initialized.
//
// h.lock must not be held.
//
// allocSpan must be called on the system stack both because it acquires
// the heap lock and because it must block GC transitions.
//
//go:systemstack
func (h *mheap) allocSpan(npages uintptr, typ spanAllocType, spanclass spanClass) (s *mspan) {
	// Function-global state.
	gp := getg()
	base, scav := uintptr(0), uintptr(0)

	// On some platforms we need to provide physical page aligned stack
	// allocations. Where the page size is less than the physical page
	// size, we already manage to do this by default.
	needPhysPageAlign := physPageAlignedStacks && typ == spanAllocStack && pageSize < physPageSize

	// If the allocation is small enough, try the page cache!
	// The page cache does not support aligned allocations, so we cannot use
	// it if we need to provide a physical page aligned stack allocation.
	pp := gp.m.p.ptr()
	if !needPhysPageAlign && pp != nil && npages < pageCachePages/4 {
		c := &pp.pcache

		// If the cache is empty, refill it.
		if c.empty() {
			lock(&h.lock)
			*c = h.pages.allocToCache()
			unlock(&h.lock)
		}

		// Try to allocate from the cache.
		base, scav = c.alloc(npages)
		if base != 0 {
			s = h.tryAllocMSpan()
			if s != nil {
				goto HaveSpan
			}
			// We have a base but no mspan, so we need
			// to lock the heap.
		}
	}

	// For one reason or another, we couldn't get the
	// whole job done without the heap lock.
	lock(&h.lock)

	if needPhysPageAlign {
		// Overallocate by a physical page to allow for later alignment.
		npages += physPageSize / pageSize
	}

	if base == 0 {
		// Try to acquire a base address.
		base, scav = h.pages.alloc(npages)
		if base == 0 {
			if !h.grow(npages) {
				unlock(&h.lock)
				return nil
			}
			base, scav = h.pages.alloc(npages)
			if base == 0 {
				throw("grew heap, but no adequate free space found")
			}
		}
	}
	if s == nil {
		// We failed to get an mspan earlier, so grab
		// one now that we have the heap lock.
		s = h.allocMSpanLocked()
	}

	if needPhysPageAlign {
		allocBase, allocPages := base, npages
		base = alignUp(allocBase, physPageSize)
		npages -= physPageSize / pageSize

		// Return memory around the aligned allocation.
		spaceBefore := base - allocBase
		if spaceBefore > 0 {
			h.pages.free(allocBase, spaceBefore/pageSize)
		}
		spaceAfter := (allocPages-npages)*pageSize - spaceBefore
		if spaceAfter > 0 {
			h.pages.free(base+npages*pageSize, spaceAfter/pageSize)
		}
	}

	unlock(&h.lock)

HaveSpan:
	// At this point, both s != nil and base != 0, and the heap
	// lock is no longer held. Initialize the span.
	s.init(base, npages)
	if h.allocNeedsZero(base, npages) {
		s.needzero = 1
	}
	nbytes := npages * pageSize
	if typ.manual() {
		s.manualFreeList = 0
		s.nelems = 0
		s.limit = s.base() + s.npages*pageSize
		s.state.set(mSpanManual)
	} else {
		// We must set span properties before the span is published anywhere
		// since we're not holding the heap lock.
		s.spanclass = spanclass
		if sizeclass := spanclass.sizeclass(); sizeclass == 0 {
			s.elemsize = nbytes
			s.nelems = 1

			s.divShift = 0
			s.divMul = 0
			s.divShift2 = 0
			s.baseMask = 0
		} else {
			s.elemsize = uintptr(class_to_size[sizeclass])
			s.nelems = nbytes / s.elemsize

			m := &class_to_divmagic[sizeclass]
			s.divShift = m.shift
			s.divMul = m.mul
			s.divShift2 = m.shift2
			s.baseMask = m.baseMask
		}

		// Initialize mark and allocation structures.
		s.freeindex = 0
		s.allocCache = ^uint64(0) // all 1s indicating all free.
		s.gcmarkBits = newMarkBits(s.nelems)
		s.allocBits = newAllocBits(s.nelems)

		// It's safe to access h.sweepgen without the heap lock because it's
		// only ever updated with the world stopped and we run on the
		// systemstack which blocks a STW transition.
		atomic.Store(&s.sweepgen, h.sweepgen)

		// Now that the span is filled in, set its state. This
		// is a publication barrier for the other fields in
		// the span. While valid pointers into this span
		// should never be visible until the span is returned,
		// if the garbage collector finds an invalid pointer,
		// access to the span may race with initialization of
		// the span. We resolve this race by atomically
		// setting the state after the span is fully
		// initialized, and atomically checking the state in
		// any situation where a pointer is suspect.
		s.state.set(mSpanInUse)
	}

	// Commit and account for any scavenged memory that the span now owns.
	if scav != 0 {
		// sysUsed all the pages that are actually available
		// in the span since some of them might be scavenged.
		sysUsed(unsafe.Pointer(base), nbytes)
		atomic.Xadd64(&memstats.heap_released, -int64(scav))
	}
	// Update stats.
	if typ == spanAllocHeap {
		atomic.Xadd64(&memstats.heap_inuse, int64(nbytes))
	}
	if typ.manual() {
		// Manually managed memory doesn't count toward heap_sys.
		memstats.heap_sys.add(-int64(nbytes))
	}
	// Update consistent stats.
	stats := memstats.heapStats.acquire()
	atomic.Xaddint64(&stats.committed, int64(scav))
	atomic.Xaddint64(&stats.released, -int64(scav))
	switch typ {
	case spanAllocHeap:
		atomic.Xaddint64(&stats.inHeap, int64(nbytes))
	case spanAllocStack:
		atomic.Xaddint64(&stats.inStacks, int64(nbytes))
	case spanAllocPtrScalarBits:
		atomic.Xaddint64(&stats.inPtrScalarBits, int64(nbytes))
	case spanAllocWorkBuf:
		atomic.Xaddint64(&stats.inWorkBufs, int64(nbytes))
	}
	memstats.heapStats.release()

	// Publish the span in various locations.

	// This is safe to call without the lock held because the slots
	// related to this span will only ever be read or modified by
	// this thread until pointers into the span are published (and
	// we execute a publication barrier at the end of this function
	// before that happens) or pageInUse is updated.
	h.setSpans(s.base(), npages, s)

	if !typ.manual() {
		// Mark in-use span in arena page bitmap.
		//
		// This publishes the span to the page sweeper, so
		// it's imperative that the span be completely initialized
		// prior to this line.
		arena, pageIdx, pageMask := pageIndexOf(s.base())
		atomic.Or8(&arena.pageInUse[pageIdx], pageMask)

		// Update related page sweeper stats.
		atomic.Xadd64(&h.pagesInUse, int64(npages))
	}

	// Make sure the newly allocated span will be observed
	// by the GC before pointers into the span are published.
	publicationBarrier()

	return s
}

// Try to add at least npage pages of memory to the heap,
// returning whether it worked.
//
// h.lock must be held.
func (h *mheap) grow(npage uintptr) bool {
	assertLockHeld(&h.lock)

	// We must grow the heap in whole palloc chunks.
	ask := alignUp(npage, pallocChunkPages) * pageSize

	totalGrowth := uintptr(0)
	// This may overflow because ask could be very large
	// and is otherwise unrelated to h.curArena.base.
	end := h.curArena.base + ask
	nBase := alignUp(end, physPageSize)
	if nBase > h.curArena.end || /* overflow */ end < h.curArena.base {
		// Not enough room in the current arena. Allocate more
		// arena space. This may not be contiguous with the
		// current arena, so we have to request the full ask.
		av, asize := h.sysAlloc(ask)
		if av == nil {
			print("runtime: out of memory: cannot allocate ", ask, "-byte block (", memstats.heap_sys, " in use)\n")
			return false
		}

		if uintptr(av) == h.curArena.end {
			// The new space is contiguous with the old
			// space, so just extend the current space.
			h.curArena.end = uintptr(av) + asize
		} else {
			// The new space is discontiguous. Track what
			// remains of the current space and switch to
			// the new space. This should be rare.
			if size := h.curArena.end - h.curArena.base; size != 0 {
				h.pages.grow(h.curArena.base, size)
				totalGrowth += size
			}
			// Switch to the new space.
			h.curArena.base = uintptr(av)
			h.curArena.end = uintptr(av) + asize
		}

		// The memory just allocated counts as both released
		// and idle, even though it's not yet backed by spans.
		//
		// The allocation is always aligned to the heap arena
		// size which is always > physPageSize, so its safe to
		// just add directly to heap_released.
		atomic.Xadd64(&memstats.heap_released, int64(asize))
		stats := memstats.heapStats.acquire()
		atomic.Xaddint64(&stats.released, int64(asize))
		memstats.heapStats.release()

		// Recalculate nBase.
		// We know this won't overflow, because sysAlloc returned
		// a valid region starting at h.curArena.base which is at
		// least ask bytes in size.
		nBase = alignUp(h.curArena.base+ask, physPageSize)
	}

	// Grow into the current arena.
	v := h.curArena.base
	h.curArena.base = nBase
	h.pages.grow(v, nBase-v)
	totalGrowth += nBase - v

	// We just caused a heap growth, so scavenge down what will soon be used.
	// By scavenging inline we deal with the failure to allocate out of
	// memory fragments by scavenging the memory fragments that are least
	// likely to be re-used.
	if retained := heapRetained(); retained+uint64(totalGrowth) > h.scavengeGoal {
		todo := totalGrowth
		if overage := uintptr(retained + uint64(totalGrowth) - h.scavengeGoal); todo > overage {
			todo = overage
		}
		h.pages.scavenge(todo, false)
	}
	return true
}

// Free the span back into the heap.
func (h *mheap) freeSpan(s *mspan) {
	systemstack(func() {
		lock(&h.lock)
		if msanenabled {
			// Tell msan that this entire span is no longer in use.
			base := unsafe.Pointer(s.base())
			bytes := s.npages << _PageShift
			msanfree(base, bytes)
		}
		h.freeSpanLocked(s, spanAllocHeap)
		unlock(&h.lock)
	})
}

// freeManual frees a manually-managed span returned by allocManual.
// typ must be the same as the spanAllocType passed to the allocManual that
// allocated s.
//
// This must only be called when gcphase == _GCoff. See mSpanState for
// an explanation.
//
// freeManual must be called on the system stack because it acquires
// the heap lock. See mheap for details.
//
//go:systemstack
func (h *mheap) freeManual(s *mspan, typ spanAllocType) {
	s.needzero = 1
	lock(&h.lock)
	h.freeSpanLocked(s, typ)
	unlock(&h.lock)
}

func (h *mheap) freeSpanLocked(s *mspan, typ spanAllocType) {
	assertLockHeld(&h.lock)

	switch s.state.get() {
	case mSpanManual:
		if s.allocCount != 0 {
			throw("mheap.freeSpanLocked - invalid stack free")
		}
	case mSpanInUse:
		if s.allocCount != 0 || s.sweepgen != h.sweepgen {
			print("mheap.freeSpanLocked - span ", s, " ptr ", hex(s.base()), " allocCount ", s.allocCount, " sweepgen ", s.sweepgen, "/", h.sweepgen, "\n")
			throw("mheap.freeSpanLocked - invalid free")
		}
		atomic.Xadd64(&h.pagesInUse, -int64(s.npages))

		// Clear in-use bit in arena page bitmap.
		arena, pageIdx, pageMask := pageIndexOf(s.base())
		atomic.And8(&arena.pageInUse[pageIdx], ^pageMask)
	default:
		throw("mheap.freeSpanLocked - invalid span state")
	}

	// Update stats.
	//
	// Mirrors the code in allocSpan.
	nbytes := s.npages * pageSize
	if typ == spanAllocHeap {
		atomic.Xadd64(&memstats.heap_inuse, -int64(nbytes))
	}
	if typ.manual() {
		// Manually managed memory doesn't count toward heap_sys, so add it back.
		memstats.heap_sys.add(int64(nbytes))
	}
	// Update consistent stats.
	stats := memstats.heapStats.acquire()
	switch typ {
	case spanAllocHeap:
		atomic.Xaddint64(&stats.inHeap, -int64(nbytes))
	case spanAllocStack:
		atomic.Xaddint64(&stats.inStacks, -int64(nbytes))
	case spanAllocPtrScalarBits:
		atomic.Xaddint64(&stats.inPtrScalarBits, -int64(nbytes))
	case spanAllocWorkBuf:
		atomic.Xaddint64(&stats.inWorkBufs, -int64(nbytes))
	}
	memstats.heapStats.release()

	// Mark the space as free.
	h.pages.free(s.base(), s.npages)

	// Free the span structure. We no longer have a use for it.
	s.state.set(mSpanDead)
	h.freeMSpanLocked(s)
}

// scavengeAll acquires the heap lock (blocking any additional
// manipulation of the page allocator) and iterates over the whole
// heap, scavenging every free page available.
func (h *mheap) scavengeAll() {
	// Disallow malloc or panic while holding the heap lock. We do
	// this here because this is a non-mallocgc entry-point to
	// the mheap API.
	gp := getg()
	gp.m.mallocing++
	lock(&h.lock)
	// Start a new scavenge generation so we have a chance to walk
	// over the whole heap.
	h.pages.scavengeStartGen()
	released := h.pages.scavenge(^uintptr(0), false)
	gen := h.pages.scav.gen
	unlock(&h.lock)
	gp.m.mallocing--

	if debug.scavtrace > 0 {
		printScavTrace(gen, released, true)
	}
}

//go:linkname runtime_debug_freeOSMemory runtime/debug.freeOSMemory
func runtime_debug_freeOSMemory() {
	GC()
	systemstack(func() { mheap_.scavengeAll() })
}

// Initialize a new span with the given start and npages.
func (span *mspan) init(base uintptr, npages uintptr) {
	// span is *not* zeroed.
	span.next = nil
	span.prev = nil
	span.list = nil
	span.startAddr = base
	span.npages = npages
	span.allocCount = 0
	span.spanclass = 0
	span.elemsize = 0
	span.speciallock.key = 0
	span.specials = nil
	span.needzero = 0
	span.freeindex = 0
	span.allocBits = nil
	span.gcmarkBits = nil
	span.state.set(mSpanDead)
	lockInit(&span.speciallock, lockRankMspanSpecial)
}

func (span *mspan) inList() bool {
	return span.list != nil
}

// Initialize an empty doubly-linked list.
func (list *mSpanList) init() {
	list.first = nil
	list.last = nil
}

func (list *mSpanList) remove(span *mspan) {
	if span.list != list {
		print("runtime: failed mSpanList.remove span.npages=", span.npages,
			" span=", span, " prev=", span.prev, " span.list=", span.list, " list=", list, "\n")
		throw("mSpanList.remove")
	}
	if list.first == span {
		list.first = span.next
	} else {
		span.prev.next = span.next
	}
	if list.last == span {
		list.last = span.prev
	} else {
		span.next.prev = span.prev
	}
	span.next = nil
	span.prev = nil
	span.list = nil
}

func (list *mSpanList) isEmpty() bool {
	return list.first == nil
}

func (list *mSpanList) insert(span *mspan) {
	if span.next != nil || span.prev != nil || span.list != nil {
		println("runtime: failed mSpanList.insert", span, span.next, span.prev, span.list)
		throw("mSpanList.insert")
	}
	span.next = list.first
	if list.first != nil {
		// The list contains at least one span; link it in.
		// The last span in the list doesn't change.
		list.first.prev = span
	} else {
		// The list contains no spans, so this is also the last span.
		list.last = span
	}
	list.first = span
	span.list = list
}

func (list *mSpanList) insertBack(span *mspan) {
	if span.next != nil || span.prev != nil || span.list != nil {
		println("runtime: failed mSpanList.insertBack", span, span.next, span.prev, span.list)
		throw("mSpanList.insertBack")
	}
	span.prev = list.last
	if list.last != nil {
		// The list contains at least one span.
		list.last.next = span
	} else {
		// The list contains no spans, so this is also the first span.
		list.first = span
	}
	list.last = span
	span.list = list
}

// takeAll removes all spans from other and inserts them at the front
// of list.
func (list *mSpanList) takeAll(other *mSpanList) {
	if other.isEmpty() {
		return
	}

	// Reparent everything in other to list.
	for s := other.first; s != nil; s = s.next {
		s.list = list
	}

	// Concatenate the lists.
	if list.isEmpty() {
		*list = *other
	} else {
		// Neither list is empty. Put other before list.
		other.last.next = list.first
		list.first.prev = other.last
		list.first = other.first
	}

	other.first, other.last = nil, nil
}

const (
	_KindSpecialFinalizer = 1
	_KindSpecialProfile   = 2
	// Note: The finalizer special must be first because if we're freeing
	// an object, a finalizer special will cause the freeing operation
	// to abort, and we want to keep the other special records around
	// if that happens.
)

//go:notinheap
type special struct {
	next   *special // linked list in span
	offset uint16   // span offset of object
	kind   byte     // kind of special
}

// spanHasSpecials marks a span as having specials in the arena bitmap.
func spanHasSpecials(s *mspan) {
	arenaPage := (s.base() / pageSize) % pagesPerArena
	ai := arenaIndex(s.base())
	ha := mheap_.arenas[ai.l1()][ai.l2()]
	atomic.Or8(&ha.pageSpecials[arenaPage/8], uint8(1)<<(arenaPage%8))
}

// spanHasNoSpecials marks a span as having no specials in the arena bitmap.
func spanHasNoSpecials(s *mspan) {
	arenaPage := (s.base() / pageSize) % pagesPerArena
	ai := arenaIndex(s.base())
	ha := mheap_.arenas[ai.l1()][ai.l2()]
	atomic.And8(&ha.pageSpecials[arenaPage/8], ^(uint8(1) << (arenaPage % 8)))
}

// Adds the special record s to the list of special records for
// the object p. All fields of s should be filled in except for
// offset & next, which this routine will fill in.
// Returns true if the special was successfully added, false otherwise.
// (The add will fail only if a record with the same p and s->kind
//  already exists.)
func addspecial(p unsafe.Pointer, s *special) bool {
	span := spanOfHeap(uintptr(p))
	if span == nil {
		throw("addspecial on invalid pointer")
	}

	// Ensure that the span is swept.
	// Sweeping accesses the specials list w/o locks, so we have
	// to synchronize with it. And it's just much safer.
	mp := acquirem()
	span.ensureSwept()

	offset := uintptr(p) - span.base()
	kind := s.kind

	lock(&span.speciallock)

	// Find splice point, check for existing record.
	t := &span.specials
	for {
		x := *t
		if x == nil {
			break
		}
		if offset == uintptr(x.offset) && kind == x.kind {
			unlock(&span.speciallock)
			releasem(mp)
			return false // already exists
		}
		if offset < uintptr(x.offset) || (offset == uintptr(x.offset) && kind < x.kind) {
			break
		}
		t = &x.next
	}

	// Splice in record, fill in offset.
	s.offset = uint16(offset)
	s.next = *t
	*t = s
	spanHasSpecials(span)
	unlock(&span.speciallock)
	releasem(mp)

	return true
}

// Removes the Special record of the given kind for the object p.
// Returns the record if the record existed, nil otherwise.
// The caller must FixAlloc_Free the result.
func removespecial(p unsafe.Pointer, kind uint8) *special {
	span := spanOfHeap(uintptr(p))
	if span == nil {
		throw("removespecial on invalid pointer")
	}

	// Ensure that the span is swept.
	// Sweeping accesses the specials list w/o locks, so we have
	// to synchronize with it. And it's just much safer.
	mp := acquirem()
	span.ensureSwept()

	offset := uintptr(p) - span.base()

	var result *special
	lock(&span.speciallock)
	t := &span.specials
	for {
		s := *t
		if s == nil {
			break
		}
		// This function is used for finalizers only, so we don't check for
		// "interior" specials (p must be exactly equal to s->offset).
		if offset == uintptr(s.offset) && kind == s.kind {
			*t = s.next
			result = s
			break
		}
		t = &s.next
	}
	if span.specials == nil {
		spanHasNoSpecials(span)
	}
	unlock(&span.speciallock)
	releasem(mp)
	return result
}

// The described object has a finalizer set for it.
//
// specialfinalizer is allocated from non-GC'd memory, so any heap
// pointers must be specially handled.
//
//go:notinheap
type specialfinalizer struct {
	special special
	fn      *funcval // May be a heap pointer.
	nret    uintptr
	fint    *_type   // May be a heap pointer, but always live.
	ot      *ptrtype // May be a heap pointer, but always live.
}

// Adds a finalizer to the object p. Returns true if it succeeded.
func addfinalizer(p unsafe.Pointer, f *funcval, nret uintptr, fint *_type, ot *ptrtype) bool {
	lock(&mheap_.speciallock)
	s := (*specialfinalizer)(mheap_.specialfinalizeralloc.alloc())
	unlock(&mheap_.speciallock)
	s.special.kind = _KindSpecialFinalizer
	s.fn = f
	s.nret = nret
	s.fint = fint
	s.ot = ot
	if addspecial(p, &s.special) {
		// This is responsible for maintaining the same
		// GC-related invariants as markrootSpans in any
		// situation where it's possible that markrootSpans
		// has already run but mark termination hasn't yet.
		if gcphase != _GCoff {
			base, _, _ := findObject(uintptr(p), 0, 0)
			mp := acquirem()
			gcw := &mp.p.ptr().gcw
			// Mark everything reachable from the object
			// so it's retained for the finalizer.
			scanobject(base, gcw)
			// Mark the finalizer itself, since the
			// special isn't part of the GC'd heap.
			scanblock(uintptr(unsafe.Pointer(&s.fn)), sys.PtrSize, &oneptrmask[0], gcw, nil)
			releasem(mp)
		}
		return true
	}

	// There was an old finalizer
	lock(&mheap_.speciallock)
	mheap_.specialfinalizeralloc.free(unsafe.Pointer(s))
	unlock(&mheap_.speciallock)
	return false
}

// Removes the finalizer (if any) from the object p.
func removefinalizer(p unsafe.Pointer) {
	s := (*specialfinalizer)(unsafe.Pointer(removespecial(p, _KindSpecialFinalizer)))
	if s == nil {
		return // there wasn't a finalizer to remove
	}
	lock(&mheap_.speciallock)
	mheap_.specialfinalizeralloc.free(unsafe.Pointer(s))
	unlock(&mheap_.speciallock)
}

// The described object is being heap profiled.
//
//go:notinheap
type specialprofile struct {
	special special
	b       *bucket
}

// Set the heap profile bucket associated with addr to b.
func setprofilebucket(p unsafe.Pointer, b *bucket) {
	lock(&mheap_.speciallock)
	s := (*specialprofile)(mheap_.specialprofilealloc.alloc())
	unlock(&mheap_.speciallock)
	s.special.kind = _KindSpecialProfile
	s.b = b
	if !addspecial(p, &s.special) {
		throw("setprofilebucket: profile already set")
	}
}

// Do whatever cleanup needs to be done to deallocate s. It has
// already been unlinked from the mspan specials list.
func freespecial(s *special, p unsafe.Pointer, size uintptr) {
	switch s.kind {
	case _KindSpecialFinalizer:
		sf := (*specialfinalizer)(unsafe.Pointer(s))
		queuefinalizer(p, sf.fn, sf.nret, sf.fint, sf.ot)
		lock(&mheap_.speciallock)
		mheap_.specialfinalizeralloc.free(unsafe.Pointer(sf))
		unlock(&mheap_.speciallock)
	case _KindSpecialProfile:
		sp := (*specialprofile)(unsafe.Pointer(s))
		mProf_Free(sp.b, size)
		lock(&mheap_.speciallock)
		mheap_.specialprofilealloc.free(unsafe.Pointer(sp))
		unlock(&mheap_.speciallock)
	default:
		throw("bad special kind")
		panic("not reached")
	}
}

// gcBits is an alloc/mark bitmap. This is always used as *gcBits.
//
//go:notinheap
type gcBits uint8

// bytep returns a pointer to the n'th byte of b.
func (b *gcBits) bytep(n uintptr) *uint8 {
	return addb((*uint8)(b), n)
}

// bitp returns a pointer to the byte containing bit n and a mask for
// selecting that bit from *bytep.
func (b *gcBits) bitp(n uintptr) (bytep *uint8, mask uint8) {
	return b.bytep(n / 8), 1 << (n % 8)
}

const gcBitsChunkBytes = uintptr(64 << 10)
const gcBitsHeaderBytes = unsafe.Sizeof(gcBitsHeader{})

type gcBitsHeader struct {
	free uintptr // free is the index into bits of the next free byte.
	next uintptr // *gcBits triggers recursive type bug. (issue 14620)
}

//go:notinheap
type gcBitsArena struct {
	// gcBitsHeader // side step recursive type bug (issue 14620) by including fields by hand.
	free uintptr // free is the index into bits of the next free byte; read/write atomically
	next *gcBitsArena
	bits [gcBitsChunkBytes - gcBitsHeaderBytes]gcBits
}

var gcBitsArenas struct {
	lock     mutex
	free     *gcBitsArena
	next     *gcBitsArena // Read atomically. Write atomically under lock.
	current  *gcBitsArena
	previous *gcBitsArena
}

// tryAlloc allocates from b or returns nil if b does not have enough room.
// This is safe to call concurrently.
func (b *gcBitsArena) tryAlloc(bytes uintptr) *gcBits {
	if b == nil || atomic.Loaduintptr(&b.free)+bytes > uintptr(len(b.bits)) {
		return nil
	}
	// Try to allocate from this block.
	end := atomic.Xadduintptr(&b.free, bytes)
	if end > uintptr(len(b.bits)) {
		return nil
	}
	// There was enough room.
	start := end - bytes
	return &b.bits[start]
}

// newMarkBits returns a pointer to 8 byte aligned bytes
// to be used for a span's mark bits.
func newMarkBits(nelems uintptr) *gcBits {
	blocksNeeded := uintptr((nelems + 63) / 64)
	bytesNeeded := blocksNeeded * 8

	// Try directly allocating from the current head arena.
	head := (*gcBitsArena)(atomic.Loadp(unsafe.Pointer(&gcBitsArenas.next)))
	if p := head.tryAlloc(bytesNeeded); p != nil {
		return p
	}

	// There's not enough room in the head arena. We may need to
	// allocate a new arena.
	lock(&gcBitsArenas.lock)
	// Try the head arena again, since it may have changed. Now
	// that we hold the lock, the list head can't change, but its
	// free position still can.
	if p := gcBitsArenas.next.tryAlloc(bytesNeeded); p != nil {
		unlock(&gcBitsArenas.lock)
		return p
	}

	// Allocate a new arena. This may temporarily drop the lock.
	fresh := newArenaMayUnlock()
	// If newArenaMayUnlock dropped the lock, another thread may
	// have put a fresh arena on the "next" list. Try allocating
	// from next again.
	if p := gcBitsArenas.next.tryAlloc(bytesNeeded); p != nil {
		// Put fresh back on the free list.
		// TODO: Mark it "already zeroed"
		fresh.next = gcBitsArenas.free
		gcBitsArenas.free = fresh
		unlock(&gcBitsArenas.lock)
		return p
	}

	// Allocate from the fresh arena. We haven't linked it in yet, so
	// this cannot race and is guaranteed to succeed.
	p := fresh.tryAlloc(bytesNeeded)
	if p == nil {
		throw("markBits overflow")
	}

	// Add the fresh arena to the "next" list.
	fresh.next = gcBitsArenas.next
	atomic.StorepNoWB(unsafe.Pointer(&gcBitsArenas.next), unsafe.Pointer(fresh))

	unlock(&gcBitsArenas.lock)
	return p
}

// newAllocBits returns a pointer to 8 byte aligned bytes
// to be used for this span's alloc bits.
// newAllocBits is used to provide newly initialized spans
// allocation bits. For spans not being initialized the
// mark bits are repurposed as allocation bits when
// the span is swept.
func newAllocBits(nelems uintptr) *gcBits {
	return newMarkBits(nelems)
}

// nextMarkBitArenaEpoch establishes a new epoch for the arenas
// holding the mark bits. The arenas are named relative to the
// current GC cycle which is demarcated by the call to finishweep_m.
//
// All current spans have been swept.
// During that sweep each span allocated room for its gcmarkBits in
// gcBitsArenas.next block. gcBitsArenas.next becomes the gcBitsArenas.current
// where the GC will mark objects and after each span is swept these bits
// will be used to allocate objects.
// gcBitsArenas.current becomes gcBitsArenas.previous where the span's
// gcAllocBits live until all the spans have been swept during this GC cycle.
// The span's sweep extinguishes all the references to gcBitsArenas.previous
// by pointing gcAllocBits into the gcBitsArenas.current.
// The gcBitsArenas.previous is released to the gcBitsArenas.free list.
func nextMarkBitArenaEpoch() {
	lock(&gcBitsArenas.lock)
	if gcBitsArenas.previous != nil {
		if gcBitsArenas.free == nil {
			gcBitsArenas.free = gcBitsArenas.previous
		} else {
			// Find end of previous arenas.
			last := gcBitsArenas.previous
			for last = gcBitsArenas.previous; last.next != nil; last = last.next {
			}
			last.next = gcBitsArenas.free
			gcBitsArenas.free = gcBitsArenas.previous
		}
	}
	gcBitsArenas.previous = gcBitsArenas.current
	gcBitsArenas.current = gcBitsArenas.next
	atomic.StorepNoWB(unsafe.Pointer(&gcBitsArenas.next), nil) // newMarkBits calls newArena when needed
	unlock(&gcBitsArenas.lock)
}

// newArenaMayUnlock allocates and zeroes a gcBits arena.
// The caller must hold gcBitsArena.lock. This may temporarily release it.
func newArenaMayUnlock() *gcBitsArena {
	var result *gcBitsArena
	if gcBitsArenas.free == nil {
		unlock(&gcBitsArenas.lock)
		result = (*gcBitsArena)(sysAlloc(gcBitsChunkBytes, &memstats.gcMiscSys))
		if result == nil {
			throw("runtime: cannot allocate memory")
		}
		lock(&gcBitsArenas.lock)
	} else {
		result = gcBitsArenas.free
		gcBitsArenas.free = gcBitsArenas.free.next
		memclrNoHeapPointers(unsafe.Pointer(result), gcBitsChunkBytes)
	}
	result.next = nil
	// If result.bits is not 8 byte aligned adjust index so
	// that &result.bits[result.free] is 8 byte aligned.
	if uintptr(unsafe.Offsetof(gcBitsArena{}.bits))&7 == 0 {
		result.free = 0
	} else {
		result.free = 8 - (uintptr(unsafe.Pointer(&result.bits[0])) & 7)
	}
	return result
}