summaryrefslogtreecommitdiffstats
path: root/src/runtime/malloc.go
blob: b2026ad0dc72d7e9b5ed9f83a78cdb9f6cf39b04 (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
// Copyright 2014 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.

// Memory allocator.
//
// This was originally based on tcmalloc, but has diverged quite a bit.
// http://goog-perftools.sourceforge.net/doc/tcmalloc.html

// The main allocator works in runs of pages.
// Small allocation sizes (up to and including 32 kB) are
// rounded to one of about 70 size classes, each of which
// has its own free set of objects of exactly that size.
// Any free page of memory can be split into a set of objects
// of one size class, which are then managed using a free bitmap.
//
// The allocator's data structures are:
//
//	fixalloc: a free-list allocator for fixed-size off-heap objects,
//		used to manage storage used by the allocator.
//	mheap: the malloc heap, managed at page (8192-byte) granularity.
//	mspan: a run of in-use pages managed by the mheap.
//	mcentral: collects all spans of a given size class.
//	mcache: a per-P cache of mspans with free space.
//	mstats: allocation statistics.
//
// Allocating a small object proceeds up a hierarchy of caches:
//
//	1. Round the size up to one of the small size classes
//	   and look in the corresponding mspan in this P's mcache.
//	   Scan the mspan's free bitmap to find a free slot.
//	   If there is a free slot, allocate it.
//	   This can all be done without acquiring a lock.
//
//	2. If the mspan has no free slots, obtain a new mspan
//	   from the mcentral's list of mspans of the required size
//	   class that have free space.
//	   Obtaining a whole span amortizes the cost of locking
//	   the mcentral.
//
//	3. If the mcentral's mspan list is empty, obtain a run
//	   of pages from the mheap to use for the mspan.
//
//	4. If the mheap is empty or has no page runs large enough,
//	   allocate a new group of pages (at least 1MB) from the
//	   operating system. Allocating a large run of pages
//	   amortizes the cost of talking to the operating system.
//
// Sweeping an mspan and freeing objects on it proceeds up a similar
// hierarchy:
//
//	1. If the mspan is being swept in response to allocation, it
//	   is returned to the mcache to satisfy the allocation.
//
//	2. Otherwise, if the mspan still has allocated objects in it,
//	   it is placed on the mcentral free list for the mspan's size
//	   class.
//
//	3. Otherwise, if all objects in the mspan are free, the mspan's
//	   pages are returned to the mheap and the mspan is now dead.
//
// Allocating and freeing a large object uses the mheap
// directly, bypassing the mcache and mcentral.
//
// If mspan.needzero is false, then free object slots in the mspan are
// already zeroed. Otherwise if needzero is true, objects are zeroed as
// they are allocated. There are various benefits to delaying zeroing
// this way:
//
//	1. Stack frame allocation can avoid zeroing altogether.
//
//	2. It exhibits better temporal locality, since the program is
//	   probably about to write to the memory.
//
//	3. We don't zero pages that never get reused.

// Virtual memory layout
//
// The heap consists of a set of arenas, which are 64MB on 64-bit and
// 4MB on 32-bit (heapArenaBytes). Each arena's start address is also
// aligned to the arena size.
//
// Each arena has an associated heapArena object that stores the
// metadata for that arena: the heap bitmap for all words in the arena
// and the span map for all pages in the arena. heapArena objects are
// themselves allocated off-heap.
//
// Since arenas are aligned, the address space can be viewed as a
// series of arena frames. The arena map (mheap_.arenas) maps from
// arena frame number to *heapArena, or nil for parts of the address
// space not backed by the Go heap. The arena map is structured as a
// two-level array consisting of a "L1" arena map and many "L2" arena
// maps; however, since arenas are large, on many architectures, the
// arena map consists of a single, large L2 map.
//
// The arena map covers the entire possible address space, allowing
// the Go heap to use any part of the address space. The allocator
// attempts to keep arenas contiguous so that large spans (and hence
// large objects) can cross arenas.

package runtime

import (
	"internal/goarch"
	"internal/goos"
	"runtime/internal/atomic"
	"runtime/internal/math"
	"runtime/internal/sys"
	"unsafe"
)

const (
	maxTinySize   = _TinySize
	tinySizeClass = _TinySizeClass
	maxSmallSize  = _MaxSmallSize

	pageShift = _PageShift
	pageSize  = _PageSize

	concurrentSweep = _ConcurrentSweep

	_PageSize = 1 << _PageShift
	_PageMask = _PageSize - 1

	// _64bit = 1 on 64-bit systems, 0 on 32-bit systems
	_64bit = 1 << (^uintptr(0) >> 63) / 2

	// Tiny allocator parameters, see "Tiny allocator" comment in malloc.go.
	_TinySize      = 16
	_TinySizeClass = int8(2)

	_FixAllocChunk = 16 << 10 // Chunk size for FixAlloc

	// Per-P, per order stack segment cache size.
	_StackCacheSize = 32 * 1024

	// Number of orders that get caching. Order 0 is FixedStack
	// and each successive order is twice as large.
	// We want to cache 2KB, 4KB, 8KB, and 16KB stacks. Larger stacks
	// will be allocated directly.
	// Since FixedStack is different on different systems, we
	// must vary NumStackOrders to keep the same maximum cached size.
	//   OS               | FixedStack | NumStackOrders
	//   -----------------+------------+---------------
	//   linux/darwin/bsd | 2KB        | 4
	//   windows/32       | 4KB        | 3
	//   windows/64       | 8KB        | 2
	//   plan9            | 4KB        | 3
	_NumStackOrders = 4 - goarch.PtrSize/4*goos.IsWindows - 1*goos.IsPlan9

	// heapAddrBits is the number of bits in a heap address. On
	// amd64, addresses are sign-extended beyond heapAddrBits. On
	// other arches, they are zero-extended.
	//
	// On most 64-bit platforms, we limit this to 48 bits based on a
	// combination of hardware and OS limitations.
	//
	// amd64 hardware limits addresses to 48 bits, sign-extended
	// to 64 bits. Addresses where the top 16 bits are not either
	// all 0 or all 1 are "non-canonical" and invalid. Because of
	// these "negative" addresses, we offset addresses by 1<<47
	// (arenaBaseOffset) on amd64 before computing indexes into
	// the heap arenas index. In 2017, amd64 hardware added
	// support for 57 bit addresses; however, currently only Linux
	// supports this extension and the kernel will never choose an
	// address above 1<<47 unless mmap is called with a hint
	// address above 1<<47 (which we never do).
	//
	// arm64 hardware (as of ARMv8) limits user addresses to 48
	// bits, in the range [0, 1<<48).
	//
	// ppc64, mips64, and s390x support arbitrary 64 bit addresses
	// in hardware. On Linux, Go leans on stricter OS limits. Based
	// on Linux's processor.h, the user address space is limited as
	// follows on 64-bit architectures:
	//
	// Architecture  Name              Maximum Value (exclusive)
	// ---------------------------------------------------------------------
	// amd64         TASK_SIZE_MAX     0x007ffffffff000 (47 bit addresses)
	// arm64         TASK_SIZE_64      0x01000000000000 (48 bit addresses)
	// ppc64{,le}    TASK_SIZE_USER64  0x00400000000000 (46 bit addresses)
	// mips64{,le}   TASK_SIZE64       0x00010000000000 (40 bit addresses)
	// s390x         TASK_SIZE         1<<64 (64 bit addresses)
	//
	// These limits may increase over time, but are currently at
	// most 48 bits except on s390x. On all architectures, Linux
	// starts placing mmap'd regions at addresses that are
	// significantly below 48 bits, so even if it's possible to
	// exceed Go's 48 bit limit, it's extremely unlikely in
	// practice.
	//
	// On 32-bit platforms, we accept the full 32-bit address
	// space because doing so is cheap.
	// mips32 only has access to the low 2GB of virtual memory, so
	// we further limit it to 31 bits.
	//
	// On ios/arm64, although 64-bit pointers are presumably
	// available, pointers are truncated to 33 bits in iOS <14.
	// Furthermore, only the top 4 GiB of the address space are
	// actually available to the application. In iOS >=14, more
	// of the address space is available, and the OS can now
	// provide addresses outside of those 33 bits. Pick 40 bits
	// as a reasonable balance between address space usage by the
	// page allocator, and flexibility for what mmap'd regions
	// we'll accept for the heap. We can't just move to the full
	// 48 bits because this uses too much address space for older
	// iOS versions.
	// TODO(mknyszek): Once iOS <14 is deprecated, promote ios/arm64
	// to a 48-bit address space like every other arm64 platform.
	//
	// WebAssembly currently has a limit of 4GB linear memory.
	heapAddrBits = (_64bit*(1-goarch.IsWasm)*(1-goos.IsIos*goarch.IsArm64))*48 + (1-_64bit+goarch.IsWasm)*(32-(goarch.IsMips+goarch.IsMipsle)) + 40*goos.IsIos*goarch.IsArm64

	// maxAlloc is the maximum size of an allocation. On 64-bit,
	// it's theoretically possible to allocate 1<<heapAddrBits bytes. On
	// 32-bit, however, this is one less than 1<<32 because the
	// number of bytes in the address space doesn't actually fit
	// in a uintptr.
	maxAlloc = (1 << heapAddrBits) - (1-_64bit)*1

	// The number of bits in a heap address, the size of heap
	// arenas, and the L1 and L2 arena map sizes are related by
	//
	//   (1 << addr bits) = arena size * L1 entries * L2 entries
	//
	// Currently, we balance these as follows:
	//
	//       Platform  Addr bits  Arena size  L1 entries   L2 entries
	// --------------  ---------  ----------  ----------  -----------
	//       */64-bit         48        64MB           1    4M (32MB)
	// windows/64-bit         48         4MB          64    1M  (8MB)
	//      ios/arm64         33         4MB           1  2048  (8KB)
	//       */32-bit         32         4MB           1  1024  (4KB)
	//     */mips(le)         31         4MB           1   512  (2KB)

	// heapArenaBytes is the size of a heap arena. The heap
	// consists of mappings of size heapArenaBytes, aligned to
	// heapArenaBytes. The initial heap mapping is one arena.
	//
	// This is currently 64MB on 64-bit non-Windows and 4MB on
	// 32-bit and on Windows. We use smaller arenas on Windows
	// because all committed memory is charged to the process,
	// even if it's not touched. Hence, for processes with small
	// heaps, the mapped arena space needs to be commensurate.
	// This is particularly important with the race detector,
	// since it significantly amplifies the cost of committed
	// memory.
	heapArenaBytes = 1 << logHeapArenaBytes

	heapArenaWords = heapArenaBytes / goarch.PtrSize

	// logHeapArenaBytes is log_2 of heapArenaBytes. For clarity,
	// prefer using heapArenaBytes where possible (we need the
	// constant to compute some other constants).
	logHeapArenaBytes = (6+20)*(_64bit*(1-goos.IsWindows)*(1-goarch.IsWasm)*(1-goos.IsIos*goarch.IsArm64)) + (2+20)*(_64bit*goos.IsWindows) + (2+20)*(1-_64bit) + (2+20)*goarch.IsWasm + (2+20)*goos.IsIos*goarch.IsArm64

	// heapArenaBitmapWords is the size of each heap arena's bitmap in uintptrs.
	heapArenaBitmapWords = heapArenaWords / (8 * goarch.PtrSize)

	pagesPerArena = heapArenaBytes / pageSize

	// arenaL1Bits is the number of bits of the arena number
	// covered by the first level arena map.
	//
	// This number should be small, since the first level arena
	// map requires PtrSize*(1<<arenaL1Bits) of space in the
	// binary's BSS. It can be zero, in which case the first level
	// index is effectively unused. There is a performance benefit
	// to this, since the generated code can be more efficient,
	// but comes at the cost of having a large L2 mapping.
	//
	// We use the L1 map on 64-bit Windows because the arena size
	// is small, but the address space is still 48 bits, and
	// there's a high cost to having a large L2.
	arenaL1Bits = 6 * (_64bit * goos.IsWindows)

	// arenaL2Bits is the number of bits of the arena number
	// covered by the second level arena index.
	//
	// The size of each arena map allocation is proportional to
	// 1<<arenaL2Bits, so it's important that this not be too
	// large. 48 bits leads to 32MB arena index allocations, which
	// is about the practical threshold.
	arenaL2Bits = heapAddrBits - logHeapArenaBytes - arenaL1Bits

	// arenaL1Shift is the number of bits to shift an arena frame
	// number by to compute an index into the first level arena map.
	arenaL1Shift = arenaL2Bits

	// arenaBits is the total bits in a combined arena map index.
	// This is split between the index into the L1 arena map and
	// the L2 arena map.
	arenaBits = arenaL1Bits + arenaL2Bits

	// arenaBaseOffset is the pointer value that corresponds to
	// index 0 in the heap arena map.
	//
	// On amd64, the address space is 48 bits, sign extended to 64
	// bits. This offset lets us handle "negative" addresses (or
	// high addresses if viewed as unsigned).
	//
	// On aix/ppc64, this offset allows to keep the heapAddrBits to
	// 48. Otherwise, it would be 60 in order to handle mmap addresses
	// (in range 0x0a00000000000000 - 0x0afffffffffffff). But in this
	// case, the memory reserved in (s *pageAlloc).init for chunks
	// is causing important slowdowns.
	//
	// On other platforms, the user address space is contiguous
	// and starts at 0, so no offset is necessary.
	arenaBaseOffset = 0xffff800000000000*goarch.IsAmd64 + 0x0a00000000000000*goos.IsAix
	// A typed version of this constant that will make it into DWARF (for viewcore).
	arenaBaseOffsetUintptr = uintptr(arenaBaseOffset)

	// Max number of threads to run garbage collection.
	// 2, 3, and 4 are all plausible maximums depending
	// on the hardware details of the machine. The garbage
	// collector scales well to 32 cpus.
	_MaxGcproc = 32

	// minLegalPointer is the smallest possible legal pointer.
	// This is the smallest possible architectural page size,
	// since we assume that the first page is never mapped.
	//
	// This should agree with minZeroPage in the compiler.
	minLegalPointer uintptr = 4096

	// minHeapForMetadataHugePages sets a threshold on when certain kinds of
	// heap metadata, currently the arenas map L2 entries and page alloc bitmap
	// mappings, are allowed to be backed by huge pages. If the heap goal ever
	// exceeds this threshold, then huge pages are enabled.
	//
	// These numbers are chosen with the assumption that huge pages are on the
	// order of a few MiB in size.
	//
	// The kind of metadata this applies to has a very low overhead when compared
	// to address space used, but their constant overheads for small heaps would
	// be very high if they were to be backed by huge pages (e.g. a few MiB makes
	// a huge difference for an 8 MiB heap, but barely any difference for a 1 GiB
	// heap). The benefit of huge pages is also not worth it for small heaps,
	// because only a very, very small part of the metadata is used for small heaps.
	//
	// N.B. If the heap goal exceeds the threshold then shrinks to a very small size
	// again, then huge pages will still be enabled for this mapping. The reason is that
	// there's no point unless we're also returning the physical memory for these
	// metadata mappings back to the OS. That would be quite complex to do in general
	// as the heap is likely fragmented after a reduction in heap size.
	minHeapForMetadataHugePages = 1 << 30
)

// physPageSize is the size in bytes of the OS's physical pages.
// Mapping and unmapping operations must be done at multiples of
// physPageSize.
//
// This must be set by the OS init code (typically in osinit) before
// mallocinit.
var physPageSize uintptr

// physHugePageSize is the size in bytes of the OS's default physical huge
// page size whose allocation is opaque to the application. It is assumed
// and verified to be a power of two.
//
// If set, this must be set by the OS init code (typically in osinit) before
// mallocinit. However, setting it at all is optional, and leaving the default
// value is always safe (though potentially less efficient).
//
// Since physHugePageSize is always assumed to be a power of two,
// physHugePageShift is defined as physHugePageSize == 1 << physHugePageShift.
// The purpose of physHugePageShift is to avoid doing divisions in
// performance critical functions.
var (
	physHugePageSize  uintptr
	physHugePageShift uint
)

func mallocinit() {
	if class_to_size[_TinySizeClass] != _TinySize {
		throw("bad TinySizeClass")
	}

	if heapArenaBitmapWords&(heapArenaBitmapWords-1) != 0 {
		// heapBits expects modular arithmetic on bitmap
		// addresses to work.
		throw("heapArenaBitmapWords not a power of 2")
	}

	// Check physPageSize.
	if physPageSize == 0 {
		// The OS init code failed to fetch the physical page size.
		throw("failed to get system page size")
	}
	if physPageSize > maxPhysPageSize {
		print("system page size (", physPageSize, ") is larger than maximum page size (", maxPhysPageSize, ")\n")
		throw("bad system page size")
	}
	if physPageSize < minPhysPageSize {
		print("system page size (", physPageSize, ") is smaller than minimum page size (", minPhysPageSize, ")\n")
		throw("bad system page size")
	}
	if physPageSize&(physPageSize-1) != 0 {
		print("system page size (", physPageSize, ") must be a power of 2\n")
		throw("bad system page size")
	}
	if physHugePageSize&(physHugePageSize-1) != 0 {
		print("system huge page size (", physHugePageSize, ") must be a power of 2\n")
		throw("bad system huge page size")
	}
	if physHugePageSize > maxPhysHugePageSize {
		// physHugePageSize is greater than the maximum supported huge page size.
		// Don't throw here, like in the other cases, since a system configured
		// in this way isn't wrong, we just don't have the code to support them.
		// Instead, silently set the huge page size to zero.
		physHugePageSize = 0
	}
	if physHugePageSize != 0 {
		// Since physHugePageSize is a power of 2, it suffices to increase
		// physHugePageShift until 1<<physHugePageShift == physHugePageSize.
		for 1<<physHugePageShift != physHugePageSize {
			physHugePageShift++
		}
	}
	if pagesPerArena%pagesPerSpanRoot != 0 {
		print("pagesPerArena (", pagesPerArena, ") is not divisible by pagesPerSpanRoot (", pagesPerSpanRoot, ")\n")
		throw("bad pagesPerSpanRoot")
	}
	if pagesPerArena%pagesPerReclaimerChunk != 0 {
		print("pagesPerArena (", pagesPerArena, ") is not divisible by pagesPerReclaimerChunk (", pagesPerReclaimerChunk, ")\n")
		throw("bad pagesPerReclaimerChunk")
	}

	if minTagBits > taggedPointerBits {
		throw("taggedPointerbits too small")
	}

	// Initialize the heap.
	mheap_.init()
	mcache0 = allocmcache()
	lockInit(&gcBitsArenas.lock, lockRankGcBitsArenas)
	lockInit(&profInsertLock, lockRankProfInsert)
	lockInit(&profBlockLock, lockRankProfBlock)
	lockInit(&profMemActiveLock, lockRankProfMemActive)
	for i := range profMemFutureLock {
		lockInit(&profMemFutureLock[i], lockRankProfMemFuture)
	}
	lockInit(&globalAlloc.mutex, lockRankGlobalAlloc)

	// Create initial arena growth hints.
	if goarch.PtrSize == 8 {
		// On a 64-bit machine, we pick the following hints
		// because:
		//
		// 1. Starting from the middle of the address space
		// makes it easier to grow out a contiguous range
		// without running in to some other mapping.
		//
		// 2. This makes Go heap addresses more easily
		// recognizable when debugging.
		//
		// 3. Stack scanning in gccgo is still conservative,
		// so it's important that addresses be distinguishable
		// from other data.
		//
		// Starting at 0x00c0 means that the valid memory addresses
		// will begin 0x00c0, 0x00c1, ...
		// In little-endian, that's c0 00, c1 00, ... None of those are valid
		// UTF-8 sequences, and they are otherwise as far away from
		// ff (likely a common byte) as possible. If that fails, we try other 0xXXc0
		// addresses. An earlier attempt to use 0x11f8 caused out of memory errors
		// on OS X during thread allocations.  0x00c0 causes conflicts with
		// AddressSanitizer which reserves all memory up to 0x0100.
		// These choices reduce the odds of a conservative garbage collector
		// not collecting memory because some non-pointer block of memory
		// had a bit pattern that matched a memory address.
		//
		// However, on arm64, we ignore all this advice above and slam the
		// allocation at 0x40 << 32 because when using 4k pages with 3-level
		// translation buffers, the user address space is limited to 39 bits
		// On ios/arm64, the address space is even smaller.
		//
		// On AIX, mmaps starts at 0x0A00000000000000 for 64-bit.
		// processes.
		//
		// Space mapped for user arenas comes immediately after the range
		// originally reserved for the regular heap when race mode is not
		// enabled because user arena chunks can never be used for regular heap
		// allocations and we want to avoid fragmenting the address space.
		//
		// In race mode we have no choice but to just use the same hints because
		// the race detector requires that the heap be mapped contiguously.
		for i := 0x7f; i >= 0; i-- {
			var p uintptr
			switch {
			case raceenabled:
				// The TSAN runtime requires the heap
				// to be in the range [0x00c000000000,
				// 0x00e000000000).
				p = uintptr(i)<<32 | uintptrMask&(0x00c0<<32)
				if p >= uintptrMask&0x00e000000000 {
					continue
				}
			case GOARCH == "arm64" && GOOS == "ios":
				p = uintptr(i)<<40 | uintptrMask&(0x0013<<28)
			case GOARCH == "arm64":
				p = uintptr(i)<<40 | uintptrMask&(0x0040<<32)
			case GOOS == "aix":
				if i == 0 {
					// We don't use addresses directly after 0x0A00000000000000
					// to avoid collisions with others mmaps done by non-go programs.
					continue
				}
				p = uintptr(i)<<40 | uintptrMask&(0xa0<<52)
			default:
				p = uintptr(i)<<40 | uintptrMask&(0x00c0<<32)
			}
			// Switch to generating hints for user arenas if we've gone
			// through about half the hints. In race mode, take only about
			// a quarter; we don't have very much space to work with.
			hintList := &mheap_.arenaHints
			if (!raceenabled && i > 0x3f) || (raceenabled && i > 0x5f) {
				hintList = &mheap_.userArena.arenaHints
			}
			hint := (*arenaHint)(mheap_.arenaHintAlloc.alloc())
			hint.addr = p
			hint.next, *hintList = *hintList, hint
		}
	} else {
		// On a 32-bit machine, we're much more concerned
		// about keeping the usable heap contiguous.
		// Hence:
		//
		// 1. We reserve space for all heapArenas up front so
		// they don't get interleaved with the heap. They're
		// ~258MB, so this isn't too bad. (We could reserve a
		// smaller amount of space up front if this is a
		// problem.)
		//
		// 2. We hint the heap to start right above the end of
		// the binary so we have the best chance of keeping it
		// contiguous.
		//
		// 3. We try to stake out a reasonably large initial
		// heap reservation.

		const arenaMetaSize = (1 << arenaBits) * unsafe.Sizeof(heapArena{})
		meta := uintptr(sysReserve(nil, arenaMetaSize))
		if meta != 0 {
			mheap_.heapArenaAlloc.init(meta, arenaMetaSize, true)
		}

		// We want to start the arena low, but if we're linked
		// against C code, it's possible global constructors
		// have called malloc and adjusted the process' brk.
		// Query the brk so we can avoid trying to map the
		// region over it (which will cause the kernel to put
		// the region somewhere else, likely at a high
		// address).
		procBrk := sbrk0()

		// If we ask for the end of the data segment but the
		// operating system requires a little more space
		// before we can start allocating, it will give out a
		// slightly higher pointer. Except QEMU, which is
		// buggy, as usual: it won't adjust the pointer
		// upward. So adjust it upward a little bit ourselves:
		// 1/4 MB to get away from the running binary image.
		p := firstmoduledata.end
		if p < procBrk {
			p = procBrk
		}
		if mheap_.heapArenaAlloc.next <= p && p < mheap_.heapArenaAlloc.end {
			p = mheap_.heapArenaAlloc.end
		}
		p = alignUp(p+(256<<10), heapArenaBytes)
		// Because we're worried about fragmentation on
		// 32-bit, we try to make a large initial reservation.
		arenaSizes := []uintptr{
			512 << 20,
			256 << 20,
			128 << 20,
		}
		for _, arenaSize := range arenaSizes {
			a, size := sysReserveAligned(unsafe.Pointer(p), arenaSize, heapArenaBytes)
			if a != nil {
				mheap_.arena.init(uintptr(a), size, false)
				p = mheap_.arena.end // For hint below
				break
			}
		}
		hint := (*arenaHint)(mheap_.arenaHintAlloc.alloc())
		hint.addr = p
		hint.next, mheap_.arenaHints = mheap_.arenaHints, hint

		// Place the hint for user arenas just after the large reservation.
		//
		// While this potentially competes with the hint above, in practice we probably
		// aren't going to be getting this far anyway on 32-bit platforms.
		userArenaHint := (*arenaHint)(mheap_.arenaHintAlloc.alloc())
		userArenaHint.addr = p
		userArenaHint.next, mheap_.userArena.arenaHints = mheap_.userArena.arenaHints, userArenaHint
	}
	// Initialize the memory limit here because the allocator is going to look at it
	// but we haven't called gcinit yet and we're definitely going to allocate memory before then.
	gcController.memoryLimit.Store(maxInt64)
}

// sysAlloc allocates heap arena space for at least n bytes. The
// returned pointer is always heapArenaBytes-aligned and backed by
// h.arenas metadata. The returned size is always a multiple of
// heapArenaBytes. sysAlloc returns nil on failure.
// There is no corresponding free function.
//
// hintList is a list of hint addresses for where to allocate new
// heap arenas. It must be non-nil.
//
// register indicates whether the heap arena should be registered
// in allArenas.
//
// sysAlloc returns a memory region in the Reserved state. This region must
// be transitioned to Prepared and then Ready before use.
//
// h must be locked.
func (h *mheap) sysAlloc(n uintptr, hintList **arenaHint, register bool) (v unsafe.Pointer, size uintptr) {
	assertLockHeld(&h.lock)

	n = alignUp(n, heapArenaBytes)

	if hintList == &h.arenaHints {
		// First, try the arena pre-reservation.
		// Newly-used mappings are considered released.
		//
		// Only do this if we're using the regular heap arena hints.
		// This behavior is only for the heap.
		v = h.arena.alloc(n, heapArenaBytes, &gcController.heapReleased)
		if v != nil {
			size = n
			goto mapped
		}
	}

	// Try to grow the heap at a hint address.
	for *hintList != nil {
		hint := *hintList
		p := hint.addr
		if hint.down {
			p -= n
		}
		if p+n < p {
			// We can't use this, so don't ask.
			v = nil
		} else if arenaIndex(p+n-1) >= 1<<arenaBits {
			// Outside addressable heap. Can't use.
			v = nil
		} else {
			v = sysReserve(unsafe.Pointer(p), n)
		}
		if p == uintptr(v) {
			// Success. Update the hint.
			if !hint.down {
				p += n
			}
			hint.addr = p
			size = n
			break
		}
		// Failed. Discard this hint and try the next.
		//
		// TODO: This would be cleaner if sysReserve could be
		// told to only return the requested address. In
		// particular, this is already how Windows behaves, so
		// it would simplify things there.
		if v != nil {
			sysFreeOS(v, n)
		}
		*hintList = hint.next
		h.arenaHintAlloc.free(unsafe.Pointer(hint))
	}

	if size == 0 {
		if raceenabled {
			// The race detector assumes the heap lives in
			// [0x00c000000000, 0x00e000000000), but we
			// just ran out of hints in this region. Give
			// a nice failure.
			throw("too many address space collisions for -race mode")
		}

		// All of the hints failed, so we'll take any
		// (sufficiently aligned) address the kernel will give
		// us.
		v, size = sysReserveAligned(nil, n, heapArenaBytes)
		if v == nil {
			return nil, 0
		}

		// Create new hints for extending this region.
		hint := (*arenaHint)(h.arenaHintAlloc.alloc())
		hint.addr, hint.down = uintptr(v), true
		hint.next, mheap_.arenaHints = mheap_.arenaHints, hint
		hint = (*arenaHint)(h.arenaHintAlloc.alloc())
		hint.addr = uintptr(v) + size
		hint.next, mheap_.arenaHints = mheap_.arenaHints, hint
	}

	// Check for bad pointers or pointers we can't use.
	{
		var bad string
		p := uintptr(v)
		if p+size < p {
			bad = "region exceeds uintptr range"
		} else if arenaIndex(p) >= 1<<arenaBits {
			bad = "base outside usable address space"
		} else if arenaIndex(p+size-1) >= 1<<arenaBits {
			bad = "end outside usable address space"
		}
		if bad != "" {
			// This should be impossible on most architectures,
			// but it would be really confusing to debug.
			print("runtime: memory allocated by OS [", hex(p), ", ", hex(p+size), ") not in usable address space: ", bad, "\n")
			throw("memory reservation exceeds address space limit")
		}
	}

	if uintptr(v)&(heapArenaBytes-1) != 0 {
		throw("misrounded allocation in sysAlloc")
	}

mapped:
	// Create arena metadata.
	for ri := arenaIndex(uintptr(v)); ri <= arenaIndex(uintptr(v)+size-1); ri++ {
		l2 := h.arenas[ri.l1()]
		if l2 == nil {
			// Allocate an L2 arena map.
			//
			// Use sysAllocOS instead of sysAlloc or persistentalloc because there's no
			// statistic we can comfortably account for this space in. With this structure,
			// we rely on demand paging to avoid large overheads, but tracking which memory
			// is paged in is too expensive. Trying to account for the whole region means
			// that it will appear like an enormous memory overhead in statistics, even though
			// it is not.
			l2 = (*[1 << arenaL2Bits]*heapArena)(sysAllocOS(unsafe.Sizeof(*l2)))
			if l2 == nil {
				throw("out of memory allocating heap arena map")
			}
			if h.arenasHugePages {
				sysHugePage(unsafe.Pointer(l2), unsafe.Sizeof(*l2))
			} else {
				sysNoHugePage(unsafe.Pointer(l2), unsafe.Sizeof(*l2))
			}
			atomic.StorepNoWB(unsafe.Pointer(&h.arenas[ri.l1()]), unsafe.Pointer(l2))
		}

		if l2[ri.l2()] != nil {
			throw("arena already initialized")
		}
		var r *heapArena
		r = (*heapArena)(h.heapArenaAlloc.alloc(unsafe.Sizeof(*r), goarch.PtrSize, &memstats.gcMiscSys))
		if r == nil {
			r = (*heapArena)(persistentalloc(unsafe.Sizeof(*r), goarch.PtrSize, &memstats.gcMiscSys))
			if r == nil {
				throw("out of memory allocating heap arena metadata")
			}
		}

		// Register the arena in allArenas if requested.
		if register {
			if len(h.allArenas) == cap(h.allArenas) {
				size := 2 * uintptr(cap(h.allArenas)) * goarch.PtrSize
				if size == 0 {
					size = physPageSize
				}
				newArray := (*notInHeap)(persistentalloc(size, goarch.PtrSize, &memstats.gcMiscSys))
				if newArray == nil {
					throw("out of memory allocating allArenas")
				}
				oldSlice := h.allArenas
				*(*notInHeapSlice)(unsafe.Pointer(&h.allArenas)) = notInHeapSlice{newArray, len(h.allArenas), int(size / goarch.PtrSize)}
				copy(h.allArenas, oldSlice)
				// Do not free the old backing array because
				// there may be concurrent readers. Since we
				// double the array each time, this can lead
				// to at most 2x waste.
			}
			h.allArenas = h.allArenas[:len(h.allArenas)+1]
			h.allArenas[len(h.allArenas)-1] = ri
		}

		// Store atomically just in case an object from the
		// new heap arena becomes visible before the heap lock
		// is released (which shouldn't happen, but there's
		// little downside to this).
		atomic.StorepNoWB(unsafe.Pointer(&l2[ri.l2()]), unsafe.Pointer(r))
	}

	// Tell the race detector about the new heap memory.
	if raceenabled {
		racemapshadow(v, size)
	}

	return
}

// sysReserveAligned is like sysReserve, but the returned pointer is
// aligned to align bytes. It may reserve either n or n+align bytes,
// so it returns the size that was reserved.
func sysReserveAligned(v unsafe.Pointer, size, align uintptr) (unsafe.Pointer, uintptr) {
	// Since the alignment is rather large in uses of this
	// function, we're not likely to get it by chance, so we ask
	// for a larger region and remove the parts we don't need.
	retries := 0
retry:
	p := uintptr(sysReserve(v, size+align))
	switch {
	case p == 0:
		return nil, 0
	case p&(align-1) == 0:
		return unsafe.Pointer(p), size + align
	case GOOS == "windows":
		// On Windows we can't release pieces of a
		// reservation, so we release the whole thing and
		// re-reserve the aligned sub-region. This may race,
		// so we may have to try again.
		sysFreeOS(unsafe.Pointer(p), size+align)
		p = alignUp(p, align)
		p2 := sysReserve(unsafe.Pointer(p), size)
		if p != uintptr(p2) {
			// Must have raced. Try again.
			sysFreeOS(p2, size)
			if retries++; retries == 100 {
				throw("failed to allocate aligned heap memory; too many retries")
			}
			goto retry
		}
		// Success.
		return p2, size
	default:
		// Trim off the unaligned parts.
		pAligned := alignUp(p, align)
		sysFreeOS(unsafe.Pointer(p), pAligned-p)
		end := pAligned + size
		endLen := (p + size + align) - end
		if endLen > 0 {
			sysFreeOS(unsafe.Pointer(end), endLen)
		}
		return unsafe.Pointer(pAligned), size
	}
}

// enableMetadataHugePages enables huge pages for various sources of heap metadata.
//
// A note on latency: for sufficiently small heaps (<10s of GiB) this function will take constant
// time, but may take time proportional to the size of the mapped heap beyond that.
//
// This function is idempotent.
//
// The heap lock must not be held over this operation, since it will briefly acquire
// the heap lock.
//
// Must be called on the system stack because it acquires the heap lock.
//
//go:systemstack
func (h *mheap) enableMetadataHugePages() {
	// Enable huge pages for page structure.
	h.pages.enableChunkHugePages()

	// Grab the lock and set arenasHugePages if it's not.
	//
	// Once arenasHugePages is set, all new L2 entries will be eligible for
	// huge pages. We'll set all the old entries after we release the lock.
	lock(&h.lock)
	if h.arenasHugePages {
		unlock(&h.lock)
		return
	}
	h.arenasHugePages = true
	unlock(&h.lock)

	// N.B. The arenas L1 map is quite small on all platforms, so it's fine to
	// just iterate over the whole thing.
	for i := range h.arenas {
		l2 := (*[1 << arenaL2Bits]*heapArena)(atomic.Loadp(unsafe.Pointer(&h.arenas[i])))
		if l2 == nil {
			continue
		}
		sysHugePage(unsafe.Pointer(l2), unsafe.Sizeof(*l2))
	}
}

// base address for all 0-byte allocations
var zerobase uintptr

// nextFreeFast returns the next free object if one is quickly available.
// Otherwise it returns 0.
func nextFreeFast(s *mspan) gclinkptr {
	theBit := sys.TrailingZeros64(s.allocCache) // Is there a free object in the allocCache?
	if theBit < 64 {
		result := s.freeindex + uintptr(theBit)
		if result < s.nelems {
			freeidx := result + 1
			if freeidx%64 == 0 && freeidx != s.nelems {
				return 0
			}
			s.allocCache >>= uint(theBit + 1)
			s.freeindex = freeidx
			s.allocCount++
			return gclinkptr(result*s.elemsize + s.base())
		}
	}
	return 0
}

// nextFree returns the next free object from the cached span if one is available.
// Otherwise it refills the cache with a span with an available object and
// returns that object along with a flag indicating that this was a heavy
// weight allocation. If it is a heavy weight allocation the caller must
// determine whether a new GC cycle needs to be started or if the GC is active
// whether this goroutine needs to assist the GC.
//
// Must run in a non-preemptible context since otherwise the owner of
// c could change.
func (c *mcache) nextFree(spc spanClass) (v gclinkptr, s *mspan, shouldhelpgc bool) {
	s = c.alloc[spc]
	shouldhelpgc = false
	freeIndex := s.nextFreeIndex()
	if freeIndex == s.nelems {
		// The span is full.
		if uintptr(s.allocCount) != s.nelems {
			println("runtime: s.allocCount=", s.allocCount, "s.nelems=", s.nelems)
			throw("s.allocCount != s.nelems && freeIndex == s.nelems")
		}
		c.refill(spc)
		shouldhelpgc = true
		s = c.alloc[spc]

		freeIndex = s.nextFreeIndex()
	}

	if freeIndex >= s.nelems {
		throw("freeIndex is not valid")
	}

	v = gclinkptr(freeIndex*s.elemsize + s.base())
	s.allocCount++
	if uintptr(s.allocCount) > s.nelems {
		println("s.allocCount=", s.allocCount, "s.nelems=", s.nelems)
		throw("s.allocCount > s.nelems")
	}
	return
}

// Allocate an object of size bytes.
// Small objects are allocated from the per-P cache's free lists.
// Large objects (> 32 kB) are allocated straight from the heap.
func mallocgc(size uintptr, typ *_type, needzero bool) unsafe.Pointer {
	if gcphase == _GCmarktermination {
		throw("mallocgc called with gcphase == _GCmarktermination")
	}

	if size == 0 {
		return unsafe.Pointer(&zerobase)
	}

	// It's possible for any malloc to trigger sweeping, which may in
	// turn queue finalizers. Record this dynamic lock edge.
	lockRankMayQueueFinalizer()

	userSize := size
	if asanenabled {
		// Refer to ASAN runtime library, the malloc() function allocates extra memory,
		// the redzone, around the user requested memory region. And the redzones are marked
		// as unaddressable. We perform the same operations in Go to detect the overflows or
		// underflows.
		size += computeRZlog(size)
	}

	if debug.malloc {
		if debug.sbrk != 0 {
			align := uintptr(16)
			if typ != nil {
				// TODO(austin): This should be just
				//   align = uintptr(typ.align)
				// but that's only 4 on 32-bit platforms,
				// even if there's a uint64 field in typ (see #599).
				// This causes 64-bit atomic accesses to panic.
				// Hence, we use stricter alignment that matches
				// the normal allocator better.
				if size&7 == 0 {
					align = 8
				} else if size&3 == 0 {
					align = 4
				} else if size&1 == 0 {
					align = 2
				} else {
					align = 1
				}
			}
			return persistentalloc(size, align, &memstats.other_sys)
		}

		if inittrace.active && inittrace.id == getg().goid {
			// Init functions are executed sequentially in a single goroutine.
			inittrace.allocs += 1
		}
	}

	// assistG is the G to charge for this allocation, or nil if
	// GC is not currently active.
	assistG := deductAssistCredit(size)

	// Set mp.mallocing to keep from being preempted by GC.
	mp := acquirem()
	if mp.mallocing != 0 {
		throw("malloc deadlock")
	}
	if mp.gsignal == getg() {
		throw("malloc during signal")
	}
	mp.mallocing = 1

	shouldhelpgc := false
	dataSize := userSize
	c := getMCache(mp)
	if c == nil {
		throw("mallocgc called without a P or outside bootstrapping")
	}
	var span *mspan
	var x unsafe.Pointer
	noscan := typ == nil || typ.PtrBytes == 0
	// In some cases block zeroing can profitably (for latency reduction purposes)
	// be delayed till preemption is possible; delayedZeroing tracks that state.
	delayedZeroing := false
	if size <= maxSmallSize {
		if noscan && size < maxTinySize {
			// Tiny allocator.
			//
			// Tiny allocator combines several tiny allocation requests
			// into a single memory block. The resulting memory block
			// is freed when all subobjects are unreachable. The subobjects
			// must be noscan (don't have pointers), this ensures that
			// the amount of potentially wasted memory is bounded.
			//
			// Size of the memory block used for combining (maxTinySize) is tunable.
			// Current setting is 16 bytes, which relates to 2x worst case memory
			// wastage (when all but one subobjects are unreachable).
			// 8 bytes would result in no wastage at all, but provides less
			// opportunities for combining.
			// 32 bytes provides more opportunities for combining,
			// but can lead to 4x worst case wastage.
			// The best case winning is 8x regardless of block size.
			//
			// Objects obtained from tiny allocator must not be freed explicitly.
			// So when an object will be freed explicitly, we ensure that
			// its size >= maxTinySize.
			//
			// SetFinalizer has a special case for objects potentially coming
			// from tiny allocator, it such case it allows to set finalizers
			// for an inner byte of a memory block.
			//
			// The main targets of tiny allocator are small strings and
			// standalone escaping variables. On a json benchmark
			// the allocator reduces number of allocations by ~12% and
			// reduces heap size by ~20%.
			off := c.tinyoffset
			// Align tiny pointer for required (conservative) alignment.
			if size&7 == 0 {
				off = alignUp(off, 8)
			} else if goarch.PtrSize == 4 && size == 12 {
				// Conservatively align 12-byte objects to 8 bytes on 32-bit
				// systems so that objects whose first field is a 64-bit
				// value is aligned to 8 bytes and does not cause a fault on
				// atomic access. See issue 37262.
				// TODO(mknyszek): Remove this workaround if/when issue 36606
				// is resolved.
				off = alignUp(off, 8)
			} else if size&3 == 0 {
				off = alignUp(off, 4)
			} else if size&1 == 0 {
				off = alignUp(off, 2)
			}
			if off+size <= maxTinySize && c.tiny != 0 {
				// The object fits into existing tiny block.
				x = unsafe.Pointer(c.tiny + off)
				c.tinyoffset = off + size
				c.tinyAllocs++
				mp.mallocing = 0
				releasem(mp)
				return x
			}
			// Allocate a new maxTinySize block.
			span = c.alloc[tinySpanClass]
			v := nextFreeFast(span)
			if v == 0 {
				v, span, shouldhelpgc = c.nextFree(tinySpanClass)
			}
			x = unsafe.Pointer(v)
			(*[2]uint64)(x)[0] = 0
			(*[2]uint64)(x)[1] = 0
			// See if we need to replace the existing tiny block with the new one
			// based on amount of remaining free space.
			if !raceenabled && (size < c.tinyoffset || c.tiny == 0) {
				// Note: disabled when race detector is on, see comment near end of this function.
				c.tiny = uintptr(x)
				c.tinyoffset = size
			}
			size = maxTinySize
		} else {
			var sizeclass uint8
			if size <= smallSizeMax-8 {
				sizeclass = size_to_class8[divRoundUp(size, smallSizeDiv)]
			} else {
				sizeclass = size_to_class128[divRoundUp(size-smallSizeMax, largeSizeDiv)]
			}
			size = uintptr(class_to_size[sizeclass])
			spc := makeSpanClass(sizeclass, noscan)
			span = c.alloc[spc]
			v := nextFreeFast(span)
			if v == 0 {
				v, span, shouldhelpgc = c.nextFree(spc)
			}
			x = unsafe.Pointer(v)
			if needzero && span.needzero != 0 {
				memclrNoHeapPointers(x, size)
			}
		}
	} else {
		shouldhelpgc = true
		// For large allocations, keep track of zeroed state so that
		// bulk zeroing can be happen later in a preemptible context.
		span = c.allocLarge(size, noscan)
		span.freeindex = 1
		span.allocCount = 1
		size = span.elemsize
		x = unsafe.Pointer(span.base())
		if needzero && span.needzero != 0 {
			if noscan {
				delayedZeroing = true
			} else {
				memclrNoHeapPointers(x, size)
				// We've in theory cleared almost the whole span here,
				// and could take the extra step of actually clearing
				// the whole thing. However, don't. Any GC bits for the
				// uncleared parts will be zero, and it's just going to
				// be needzero = 1 once freed anyway.
			}
		}
	}

	if !noscan {
		var scanSize uintptr
		heapBitsSetType(uintptr(x), size, dataSize, typ)
		if dataSize > typ.Size_ {
			// Array allocation. If there are any
			// pointers, GC has to scan to the last
			// element.
			if typ.PtrBytes != 0 {
				scanSize = dataSize - typ.Size_ + typ.PtrBytes
			}
		} else {
			scanSize = typ.PtrBytes
		}
		c.scanAlloc += scanSize
	}

	// Ensure that the stores above that initialize x to
	// type-safe memory and set the heap bits occur before
	// the caller can make x observable to the garbage
	// collector. Otherwise, on weakly ordered machines,
	// the garbage collector could follow a pointer to x,
	// but see uninitialized memory or stale heap bits.
	publicationBarrier()
	// As x and the heap bits are initialized, update
	// freeIndexForScan now so x is seen by the GC
	// (including conservative scan) as an allocated object.
	// While this pointer can't escape into user code as a
	// _live_ pointer until we return, conservative scanning
	// may find a dead pointer that happens to point into this
	// object. Delaying this update until now ensures that
	// conservative scanning considers this pointer dead until
	// this point.
	span.freeIndexForScan = span.freeindex

	// Allocate black during GC.
	// All slots hold nil so no scanning is needed.
	// This may be racing with GC so do it atomically if there can be
	// a race marking the bit.
	if gcphase != _GCoff {
		gcmarknewobject(span, uintptr(x), size)
	}

	if raceenabled {
		racemalloc(x, size)
	}

	if msanenabled {
		msanmalloc(x, size)
	}

	if asanenabled {
		// We should only read/write the memory with the size asked by the user.
		// The rest of the allocated memory should be poisoned, so that we can report
		// errors when accessing poisoned memory.
		// The allocated memory is larger than required userSize, it will also include
		// redzone and some other padding bytes.
		rzBeg := unsafe.Add(x, userSize)
		asanpoison(rzBeg, size-userSize)
		asanunpoison(x, userSize)
	}

	if rate := MemProfileRate; rate > 0 {
		// Note cache c only valid while m acquired; see #47302
		if rate != 1 && size < c.nextSample {
			c.nextSample -= size
		} else {
			profilealloc(mp, x, size)
		}
	}
	mp.mallocing = 0
	releasem(mp)

	// Pointerfree data can be zeroed late in a context where preemption can occur.
	// x will keep the memory alive.
	if delayedZeroing {
		if !noscan {
			throw("delayed zeroing on data that may contain pointers")
		}
		memclrNoHeapPointersChunked(size, x) // This is a possible preemption point: see #47302
	}

	if debug.malloc {
		if debug.allocfreetrace != 0 {
			tracealloc(x, size, typ)
		}

		if inittrace.active && inittrace.id == getg().goid {
			// Init functions are executed sequentially in a single goroutine.
			inittrace.bytes += uint64(size)
		}
	}

	if assistG != nil {
		// Account for internal fragmentation in the assist
		// debt now that we know it.
		assistG.gcAssistBytes -= int64(size - dataSize)
	}

	if shouldhelpgc {
		if t := (gcTrigger{kind: gcTriggerHeap}); t.test() {
			gcStart(t)
		}
	}

	if raceenabled && noscan && dataSize < maxTinySize {
		// Pad tinysize allocations so they are aligned with the end
		// of the tinyalloc region. This ensures that any arithmetic
		// that goes off the top end of the object will be detectable
		// by checkptr (issue 38872).
		// Note that we disable tinyalloc when raceenabled for this to work.
		// TODO: This padding is only performed when the race detector
		// is enabled. It would be nice to enable it if any package
		// was compiled with checkptr, but there's no easy way to
		// detect that (especially at compile time).
		// TODO: enable this padding for all allocations, not just
		// tinyalloc ones. It's tricky because of pointer maps.
		// Maybe just all noscan objects?
		x = add(x, size-dataSize)
	}

	return x
}

// deductAssistCredit reduces the current G's assist credit
// by size bytes, and assists the GC if necessary.
//
// Caller must be preemptible.
//
// Returns the G for which the assist credit was accounted.
func deductAssistCredit(size uintptr) *g {
	var assistG *g
	if gcBlackenEnabled != 0 {
		// Charge the current user G for this allocation.
		assistG = getg()
		if assistG.m.curg != nil {
			assistG = assistG.m.curg
		}
		// Charge the allocation against the G. We'll account
		// for internal fragmentation at the end of mallocgc.
		assistG.gcAssistBytes -= int64(size)

		if assistG.gcAssistBytes < 0 {
			// This G is in debt. Assist the GC to correct
			// this before allocating. This must happen
			// before disabling preemption.
			gcAssistAlloc(assistG)
		}
	}
	return assistG
}

// memclrNoHeapPointersChunked repeatedly calls memclrNoHeapPointers
// on chunks of the buffer to be zeroed, with opportunities for preemption
// along the way.  memclrNoHeapPointers contains no safepoints and also
// cannot be preemptively scheduled, so this provides a still-efficient
// block copy that can also be preempted on a reasonable granularity.
//
// Use this with care; if the data being cleared is tagged to contain
// pointers, this allows the GC to run before it is all cleared.
func memclrNoHeapPointersChunked(size uintptr, x unsafe.Pointer) {
	v := uintptr(x)
	// got this from benchmarking. 128k is too small, 512k is too large.
	const chunkBytes = 256 * 1024
	vsize := v + size
	for voff := v; voff < vsize; voff = voff + chunkBytes {
		if getg().preempt {
			// may hold locks, e.g., profiling
			goschedguarded()
		}
		// clear min(avail, lump) bytes
		n := vsize - voff
		if n > chunkBytes {
			n = chunkBytes
		}
		memclrNoHeapPointers(unsafe.Pointer(voff), n)
	}
}

// implementation of new builtin
// compiler (both frontend and SSA backend) knows the signature
// of this function.
func newobject(typ *_type) unsafe.Pointer {
	return mallocgc(typ.Size_, typ, true)
}

//go:linkname reflect_unsafe_New reflect.unsafe_New
func reflect_unsafe_New(typ *_type) unsafe.Pointer {
	return mallocgc(typ.Size_, typ, true)
}

//go:linkname reflectlite_unsafe_New internal/reflectlite.unsafe_New
func reflectlite_unsafe_New(typ *_type) unsafe.Pointer {
	return mallocgc(typ.Size_, typ, true)
}

// newarray allocates an array of n elements of type typ.
func newarray(typ *_type, n int) unsafe.Pointer {
	if n == 1 {
		return mallocgc(typ.Size_, typ, true)
	}
	mem, overflow := math.MulUintptr(typ.Size_, uintptr(n))
	if overflow || mem > maxAlloc || n < 0 {
		panic(plainError("runtime: allocation size out of range"))
	}
	return mallocgc(mem, typ, true)
}

//go:linkname reflect_unsafe_NewArray reflect.unsafe_NewArray
func reflect_unsafe_NewArray(typ *_type, n int) unsafe.Pointer {
	return newarray(typ, n)
}

func profilealloc(mp *m, x unsafe.Pointer, size uintptr) {
	c := getMCache(mp)
	if c == nil {
		throw("profilealloc called without a P or outside bootstrapping")
	}
	c.nextSample = nextSample()
	mProf_Malloc(x, size)
}

// nextSample returns the next sampling point for heap profiling. The goal is
// to sample allocations on average every MemProfileRate bytes, but with a
// completely random distribution over the allocation timeline; this
// corresponds to a Poisson process with parameter MemProfileRate. In Poisson
// processes, the distance between two samples follows the exponential
// distribution (exp(MemProfileRate)), so the best return value is a random
// number taken from an exponential distribution whose mean is MemProfileRate.
func nextSample() uintptr {
	if MemProfileRate == 1 {
		// Callers assign our return value to
		// mcache.next_sample, but next_sample is not used
		// when the rate is 1. So avoid the math below and
		// just return something.
		return 0
	}
	if GOOS == "plan9" {
		// Plan 9 doesn't support floating point in note handler.
		if gp := getg(); gp == gp.m.gsignal {
			return nextSampleNoFP()
		}
	}

	return uintptr(fastexprand(MemProfileRate))
}

// fastexprand returns a random number from an exponential distribution with
// the specified mean.
func fastexprand(mean int) int32 {
	// Avoid overflow. Maximum possible step is
	// -ln(1/(1<<randomBitCount)) * mean, approximately 20 * mean.
	switch {
	case mean > 0x7000000:
		mean = 0x7000000
	case mean == 0:
		return 0
	}

	// Take a random sample of the exponential distribution exp(-mean*x).
	// The probability distribution function is mean*exp(-mean*x), so the CDF is
	// p = 1 - exp(-mean*x), so
	// q = 1 - p == exp(-mean*x)
	// log_e(q) = -mean*x
	// -log_e(q)/mean = x
	// x = -log_e(q) * mean
	// x = log_2(q) * (-log_e(2)) * mean    ; Using log_2 for efficiency
	const randomBitCount = 26
	q := fastrandn(1<<randomBitCount) + 1
	qlog := fastlog2(float64(q)) - randomBitCount
	if qlog > 0 {
		qlog = 0
	}
	const minusLog2 = -0.6931471805599453 // -ln(2)
	return int32(qlog*(minusLog2*float64(mean))) + 1
}

// nextSampleNoFP is similar to nextSample, but uses older,
// simpler code to avoid floating point.
func nextSampleNoFP() uintptr {
	// Set first allocation sample size.
	rate := MemProfileRate
	if rate > 0x3fffffff { // make 2*rate not overflow
		rate = 0x3fffffff
	}
	if rate != 0 {
		return uintptr(fastrandn(uint32(2 * rate)))
	}
	return 0
}

type persistentAlloc struct {
	base *notInHeap
	off  uintptr
}

var globalAlloc struct {
	mutex
	persistentAlloc
}

// persistentChunkSize is the number of bytes we allocate when we grow
// a persistentAlloc.
const persistentChunkSize = 256 << 10

// persistentChunks is a list of all the persistent chunks we have
// allocated. The list is maintained through the first word in the
// persistent chunk. This is updated atomically.
var persistentChunks *notInHeap

// Wrapper around sysAlloc that can allocate small chunks.
// There is no associated free operation.
// Intended for things like function/type/debug-related persistent data.
// If align is 0, uses default align (currently 8).
// The returned memory will be zeroed.
// sysStat must be non-nil.
//
// Consider marking persistentalloc'd types not in heap by embedding
// runtime/internal/sys.NotInHeap.
func persistentalloc(size, align uintptr, sysStat *sysMemStat) unsafe.Pointer {
	var p *notInHeap
	systemstack(func() {
		p = persistentalloc1(size, align, sysStat)
	})
	return unsafe.Pointer(p)
}

// Must run on system stack because stack growth can (re)invoke it.
// See issue 9174.
//
//go:systemstack
func persistentalloc1(size, align uintptr, sysStat *sysMemStat) *notInHeap {
	const (
		maxBlock = 64 << 10 // VM reservation granularity is 64K on windows
	)

	if size == 0 {
		throw("persistentalloc: size == 0")
	}
	if align != 0 {
		if align&(align-1) != 0 {
			throw("persistentalloc: align is not a power of 2")
		}
		if align > _PageSize {
			throw("persistentalloc: align is too large")
		}
	} else {
		align = 8
	}

	if size >= maxBlock {
		return (*notInHeap)(sysAlloc(size, sysStat))
	}

	mp := acquirem()
	var persistent *persistentAlloc
	if mp != nil && mp.p != 0 {
		persistent = &mp.p.ptr().palloc
	} else {
		lock(&globalAlloc.mutex)
		persistent = &globalAlloc.persistentAlloc
	}
	persistent.off = alignUp(persistent.off, align)
	if persistent.off+size > persistentChunkSize || persistent.base == nil {
		persistent.base = (*notInHeap)(sysAlloc(persistentChunkSize, &memstats.other_sys))
		if persistent.base == nil {
			if persistent == &globalAlloc.persistentAlloc {
				unlock(&globalAlloc.mutex)
			}
			throw("runtime: cannot allocate memory")
		}

		// Add the new chunk to the persistentChunks list.
		for {
			chunks := uintptr(unsafe.Pointer(persistentChunks))
			*(*uintptr)(unsafe.Pointer(persistent.base)) = chunks
			if atomic.Casuintptr((*uintptr)(unsafe.Pointer(&persistentChunks)), chunks, uintptr(unsafe.Pointer(persistent.base))) {
				break
			}
		}
		persistent.off = alignUp(goarch.PtrSize, align)
	}
	p := persistent.base.add(persistent.off)
	persistent.off += size
	releasem(mp)
	if persistent == &globalAlloc.persistentAlloc {
		unlock(&globalAlloc.mutex)
	}

	if sysStat != &memstats.other_sys {
		sysStat.add(int64(size))
		memstats.other_sys.add(-int64(size))
	}
	return p
}

// inPersistentAlloc reports whether p points to memory allocated by
// persistentalloc. This must be nosplit because it is called by the
// cgo checker code, which is called by the write barrier code.
//
//go:nosplit
func inPersistentAlloc(p uintptr) bool {
	chunk := atomic.Loaduintptr((*uintptr)(unsafe.Pointer(&persistentChunks)))
	for chunk != 0 {
		if p >= chunk && p < chunk+persistentChunkSize {
			return true
		}
		chunk = *(*uintptr)(unsafe.Pointer(chunk))
	}
	return false
}

// linearAlloc is a simple linear allocator that pre-reserves a region
// of memory and then optionally maps that region into the Ready state
// as needed.
//
// The caller is responsible for locking.
type linearAlloc struct {
	next   uintptr // next free byte
	mapped uintptr // one byte past end of mapped space
	end    uintptr // end of reserved space

	mapMemory bool // transition memory from Reserved to Ready if true
}

func (l *linearAlloc) init(base, size uintptr, mapMemory bool) {
	if base+size < base {
		// Chop off the last byte. The runtime isn't prepared
		// to deal with situations where the bounds could overflow.
		// Leave that memory reserved, though, so we don't map it
		// later.
		size -= 1
	}
	l.next, l.mapped = base, base
	l.end = base + size
	l.mapMemory = mapMemory
}

func (l *linearAlloc) alloc(size, align uintptr, sysStat *sysMemStat) unsafe.Pointer {
	p := alignUp(l.next, align)
	if p+size > l.end {
		return nil
	}
	l.next = p + size
	if pEnd := alignUp(l.next-1, physPageSize); pEnd > l.mapped {
		if l.mapMemory {
			// Transition from Reserved to Prepared to Ready.
			n := pEnd - l.mapped
			sysMap(unsafe.Pointer(l.mapped), n, sysStat)
			sysUsed(unsafe.Pointer(l.mapped), n, n)
		}
		l.mapped = pEnd
	}
	return unsafe.Pointer(p)
}

// notInHeap is off-heap memory allocated by a lower-level allocator
// like sysAlloc or persistentAlloc.
//
// In general, it's better to use real types which embed
// runtime/internal/sys.NotInHeap, but this serves as a generic type
// for situations where that isn't possible (like in the allocators).
//
// TODO: Use this as the return type of sysAlloc, persistentAlloc, etc?
type notInHeap struct{ _ sys.NotInHeap }

func (p *notInHeap) add(bytes uintptr) *notInHeap {
	return (*notInHeap)(unsafe.Pointer(uintptr(unsafe.Pointer(p)) + bytes))
}

// computeRZlog computes the size of the redzone.
// Refer to the implementation of the compiler-rt.
func computeRZlog(userSize uintptr) uintptr {
	switch {
	case userSize <= (64 - 16):
		return 16 << 0
	case userSize <= (128 - 32):
		return 16 << 1
	case userSize <= (512 - 64):
		return 16 << 2
	case userSize <= (4096 - 128):
		return 16 << 3
	case userSize <= (1<<14)-256:
		return 16 << 4
	case userSize <= (1<<15)-512:
		return 16 << 5
	case userSize <= (1<<16)-1024:
		return 16 << 6
	default:
		return 16 << 7
	}
}