summaryrefslogtreecommitdiffhomepage
path: root/pkg/tcpip/network/ipv6/ipv6.go
blob: 94043ed4e8fadd87ddd562e836c97747507e063a (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
// Copyright 2020 The gVisor Authors.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

// Package ipv6 contains the implementation of the ipv6 network protocol.
package ipv6

import (
	"encoding/binary"
	"fmt"
	"hash/fnv"
	"math"
	"reflect"
	"sort"
	"sync/atomic"
	"time"

	"gvisor.dev/gvisor/pkg/sync"
	"gvisor.dev/gvisor/pkg/tcpip"
	"gvisor.dev/gvisor/pkg/tcpip/buffer"
	"gvisor.dev/gvisor/pkg/tcpip/header"
	"gvisor.dev/gvisor/pkg/tcpip/header/parse"
	"gvisor.dev/gvisor/pkg/tcpip/network/fragmentation"
	"gvisor.dev/gvisor/pkg/tcpip/network/hash"
	"gvisor.dev/gvisor/pkg/tcpip/stack"
)

const (
	// ReassembleTimeout controls how long a fragment will be held.
	// As per RFC 8200 section 4.5:
	//
	//   If insufficient fragments are received to complete reassembly of a packet
	//   within 60 seconds of the reception of the first-arriving fragment of that
	//   packet, reassembly of that packet must be abandoned.
	//
	// Linux also uses 60 seconds for reassembly timeout:
	// https://github.com/torvalds/linux/blob/47ec5303d73ea344e84f46660fff693c57641386/include/net/ipv6.h#L456
	ReassembleTimeout = 60 * time.Second

	// ProtocolNumber is the ipv6 protocol number.
	ProtocolNumber = header.IPv6ProtocolNumber

	// maxPayloadSize is the maximum size that can be encoded in the 16-bit
	// PayloadLength field of the ipv6 header.
	maxPayloadSize = 0xffff

	// DefaultTTL is the default hop limit for IPv6 Packets egressed by
	// Netstack.
	DefaultTTL = 64

	// buckets for fragment identifiers
	buckets = 2048
)

// policyTable is the default policy table defined in RFC 6724 section 2.1.
//
// A more human-readable version:
//
//  Prefix        Precedence Label
//  ::1/128               50     0
//  ::/0                  40     1
//  ::ffff:0:0/96         35     4
//  2002::/16             30     2
//  2001::/32              5     5
//  fc00::/7               3    13
//  ::/96                  1     3
//  fec0::/10              1    11
//  3ffe::/16              1    12
//
// The table is sorted by prefix length so longest-prefix match can be easily
// achieved.
//
// We willingly left out ::/96, fec0::/10 and 3ffe::/16 since those prefix
// assignments are deprecated.
//
// As per RFC 4291 section 2.5.5.1 (for ::/96),
//
//   The "IPv4-Compatible IPv6 address" is now deprecated because the
//   current IPv6 transition mechanisms no longer use these addresses.
//   New or updated implementations are not required to support this
//   address type.
//
// As per RFC 3879 section 4 (for fec0::/10),
//
//    This document formally deprecates the IPv6 site-local unicast prefix
//    defined in [RFC3513], i.e., 1111111011 binary or FEC0::/10.
//
// As per RFC 3701 section 1 (for 3ffe::/16),
//
//   As clearly stated in [TEST-NEW], the addresses for the 6bone are
//   temporary and will be reclaimed in the future. It further states
//   that all users of these addresses (within the 3FFE::/16 prefix) will
//   be required to renumber at some time in the future.
//
// and section 2,
//
//   Thus after the pTLA allocation cutoff date January 1, 2004, it is
//   REQUIRED that no new 6bone 3FFE pTLAs be allocated.
//
// MUST NOT BE MODIFIED.
var policyTable = [...]struct {
	subnet tcpip.Subnet

	label uint8
}{
	// ::1/128
	{
		subnet: header.IPv6Loopback.WithPrefix().Subnet(),
		label:  0,
	},
	// ::ffff:0:0/96
	{
		subnet: header.IPv4MappedIPv6Subnet,
		label:  4,
	},
	// 2001::/32 (Teredo prefix as per RFC 4380 section 2.6).
	{
		subnet: tcpip.AddressWithPrefix{
			Address:   "\x20\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00",
			PrefixLen: 32,
		}.Subnet(),
		label: 5,
	},
	// 2002::/16 (6to4 prefix as per RFC 3056 section 2).
	{
		subnet: tcpip.AddressWithPrefix{
			Address:   "\x20\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00",
			PrefixLen: 16,
		}.Subnet(),
		label: 2,
	},
	// fc00::/7 (Unique local addresses as per RFC 4193 section 3.1).
	{
		subnet: tcpip.AddressWithPrefix{
			Address:   "\xfc\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00",
			PrefixLen: 7,
		}.Subnet(),
		label: 13,
	},
	// ::/0
	{
		subnet: header.IPv6EmptySubnet,
		label:  1,
	},
}

func getLabel(addr tcpip.Address) uint8 {
	for _, p := range policyTable {
		if p.subnet.Contains(addr) {
			return p.label
		}
	}

	panic(fmt.Sprintf("should have a label for address = %s", addr))
}

var _ stack.LinkResolvableNetworkEndpoint = (*endpoint)(nil)
var _ stack.GroupAddressableEndpoint = (*endpoint)(nil)
var _ stack.AddressableEndpoint = (*endpoint)(nil)
var _ stack.NetworkEndpoint = (*endpoint)(nil)
var _ stack.NDPEndpoint = (*endpoint)(nil)
var _ NDPEndpoint = (*endpoint)(nil)

type endpoint struct {
	nic           stack.NetworkInterface
	linkAddrCache stack.LinkAddressCache
	nud           stack.NUDHandler
	dispatcher    stack.TransportDispatcher
	protocol      *protocol
	stack         *stack.Stack
	stats         sharedStats

	// enabled is set to 1 when the endpoint is enabled and 0 when it is
	// disabled.
	//
	// Must be accessed using atomic operations.
	enabled uint32

	mu struct {
		sync.RWMutex

		addressableEndpointState stack.AddressableEndpointState
		ndp                      ndpState
		mld                      mldState
	}
}

// NICNameFromID is a function that returns a stable name for the specified NIC,
// even if different NIC IDs are used to refer to the same NIC in different
// program runs. It is used when generating opaque interface identifiers (IIDs).
// If the NIC was created with a name, it is passed to NICNameFromID.
//
// NICNameFromID SHOULD return unique NIC names so unique opaque IIDs are
// generated for the same prefix on differnt NICs.
type NICNameFromID func(tcpip.NICID, string) string

// OpaqueInterfaceIdentifierOptions holds the options related to the generation
// of opaque interface indentifiers (IIDs) as defined by RFC 7217.
type OpaqueInterfaceIdentifierOptions struct {
	// NICNameFromID is a function that returns a stable name for a specified NIC,
	// even if the NIC ID changes over time.
	//
	// Must be specified to generate the opaque IID.
	NICNameFromID NICNameFromID

	// SecretKey is a pseudo-random number used as the secret key when generating
	// opaque IIDs as defined by RFC 7217. The key SHOULD be at least
	// header.OpaqueIIDSecretKeyMinBytes bytes and MUST follow minimum randomness
	// requirements for security as outlined by RFC 4086. SecretKey MUST NOT
	// change between program runs, unless explicitly changed.
	//
	// OpaqueInterfaceIdentifierOptions takes ownership of SecretKey. SecretKey
	// MUST NOT be modified after Stack is created.
	//
	// May be nil, but a nil value is highly discouraged to maintain
	// some level of randomness between nodes.
	SecretKey []byte
}

// HandleLinkResolutionFailure implements stack.LinkResolvableNetworkEndpoint.
func (e *endpoint) HandleLinkResolutionFailure(pkt *stack.PacketBuffer) {
	// handleControl expects the entire offending packet to be in the packet
	// buffer's data field.
	pkt = stack.NewPacketBuffer(stack.PacketBufferOptions{
		Data: buffer.NewVectorisedView(pkt.Size(), pkt.Views()),
	})
	pkt.NICID = e.nic.ID()
	pkt.NetworkProtocolNumber = ProtocolNumber
	e.handleControl(stack.ControlAddressUnreachable, 0, pkt)
}

// onAddressAssignedLocked handles an address being assigned.
//
// Precondition: e.mu must be exclusively locked.
func (e *endpoint) onAddressAssignedLocked(addr tcpip.Address) {
	// As per RFC 2710 section 3,
	//
	//   All MLD  messages described in this document are sent with a link-local
	//   IPv6 Source Address, ...
	//
	// If we just completed DAD for a link-local address, then attempt to send any
	// queued MLD reports. Note, we may have sent reports already for some of the
	// groups before we had a valid link-local address to use as the source for
	// the MLD messages, but that was only so that MLD snooping switches are aware
	// of our membership to groups - routers would not have handled those reports.
	//
	// As per RFC 3590 section 4,
	//
	//   MLD Report and Done messages are sent with a link-local address as
	//   the IPv6 source address, if a valid address is available on the
	//   interface. If a valid link-local address is not available (e.g., one
	//   has not been configured), the message is sent with the unspecified
	//   address (::) as the IPv6 source address.
	//
	//   Once a valid link-local address is available, a node SHOULD generate
	//   new MLD Report messages for all multicast addresses joined on the
	//   interface.
	//
	//   Routers receiving an MLD Report or Done message with the unspecified
	//   address as the IPv6 source address MUST silently discard the packet
	//   without taking any action on the packets contents.
	//
	//   Snooping switches MUST manage multicast forwarding state based on MLD
	//   Report and Done messages sent with the unspecified address as the
	//   IPv6 source address.
	if header.IsV6LinkLocalAddress(addr) {
		e.mu.mld.sendQueuedReports()
	}
}

// InvalidateDefaultRouter implements stack.NDPEndpoint.
func (e *endpoint) InvalidateDefaultRouter(rtr tcpip.Address) {
	e.mu.Lock()
	defer e.mu.Unlock()
	e.mu.ndp.invalidateDefaultRouter(rtr)
}

// SetNDPConfigurations implements NDPEndpoint.
func (e *endpoint) SetNDPConfigurations(c NDPConfigurations) {
	c.validate()
	e.mu.Lock()
	defer e.mu.Unlock()
	e.mu.ndp.configs = c
}

// hasTentativeAddr returns true if addr is tentative on e.
func (e *endpoint) hasTentativeAddr(addr tcpip.Address) bool {
	e.mu.RLock()
	addressEndpoint := e.getAddressRLocked(addr)
	e.mu.RUnlock()
	return addressEndpoint != nil && addressEndpoint.GetKind() == stack.PermanentTentative
}

// dupTentativeAddrDetected attempts to inform e that a tentative addr is a
// duplicate on a link.
//
// dupTentativeAddrDetected removes the tentative address if it exists. If the
// address was generated via SLAAC, an attempt is made to generate a new
// address.
func (e *endpoint) dupTentativeAddrDetected(addr tcpip.Address) tcpip.Error {
	e.mu.Lock()
	defer e.mu.Unlock()

	addressEndpoint := e.getAddressRLocked(addr)
	if addressEndpoint == nil {
		return &tcpip.ErrBadAddress{}
	}

	if addressEndpoint.GetKind() != stack.PermanentTentative {
		return &tcpip.ErrInvalidEndpointState{}
	}

	// If the address is a SLAAC address, do not invalidate its SLAAC prefix as an
	// attempt will be made to generate a new address for it.
	if err := e.removePermanentEndpointLocked(addressEndpoint, false /* allowSLAACInvalidation */); err != nil {
		return err
	}

	prefix := addressEndpoint.Subnet()

	switch t := addressEndpoint.ConfigType(); t {
	case stack.AddressConfigStatic:
	case stack.AddressConfigSlaac:
		e.mu.ndp.regenerateSLAACAddr(prefix)
	case stack.AddressConfigSlaacTemp:
		// Do not reset the generation attempts counter for the prefix as the
		// temporary address is being regenerated in response to a DAD conflict.
		e.mu.ndp.regenerateTempSLAACAddr(prefix, false /* resetGenAttempts */)
	default:
		panic(fmt.Sprintf("unrecognized address config type = %d", t))
	}

	return nil
}

// transitionForwarding transitions the endpoint's forwarding status to
// forwarding.
//
// Must only be called when the forwarding status changes.
func (e *endpoint) transitionForwarding(forwarding bool) {
	e.mu.Lock()
	defer e.mu.Unlock()

	if !e.Enabled() {
		return
	}

	if forwarding {
		// When transitioning into an IPv6 router, host-only state (NDP discovered
		// routers, discovered on-link prefixes, and auto-generated addresses) is
		// cleaned up/invalidated and NDP router solicitations are stopped.
		e.mu.ndp.stopSolicitingRouters()
		e.mu.ndp.cleanupState(true /* hostOnly */)
	} else {
		// When transitioning into an IPv6 host, NDP router solicitations are
		// started.
		e.mu.ndp.startSolicitingRouters()
	}
}

// Enable implements stack.NetworkEndpoint.
func (e *endpoint) Enable() tcpip.Error {
	e.mu.Lock()
	defer e.mu.Unlock()

	// If the NIC is not enabled, the endpoint can't do anything meaningful so
	// don't enable the endpoint.
	if !e.nic.Enabled() {
		return &tcpip.ErrNotPermitted{}
	}

	// If the endpoint is already enabled, there is nothing for it to do.
	if !e.setEnabled(true) {
		return nil
	}

	// Groups may have been joined when the endpoint was disabled, or the
	// endpoint may have left groups from the perspective of MLD when the
	// endpoint was disabled. Either way, we need to let routers know to
	// send us multicast traffic.
	e.mu.mld.initializeAll()

	// Join the IPv6 All-Nodes Multicast group if the stack is configured to
	// use IPv6. This is required to ensure that this node properly receives
	// and responds to the various NDP messages that are destined to the
	// all-nodes multicast address. An example is the Neighbor Advertisement
	// when we perform Duplicate Address Detection, or Router Advertisement
	// when we do Router Discovery. See RFC 4862, section 5.4.2 and RFC 4861
	// section 4.2 for more information.
	//
	// Also auto-generate an IPv6 link-local address based on the endpoint's
	// link address if it is configured to do so. Note, each interface is
	// required to have IPv6 link-local unicast address, as per RFC 4291
	// section 2.1.

	// Join the All-Nodes multicast group before starting DAD as responses to DAD
	// (NDP NS) messages may be sent to the All-Nodes multicast group if the
	// source address of the NDP NS is the unspecified address, as per RFC 4861
	// section 7.2.4.
	if err := e.joinGroupLocked(header.IPv6AllNodesMulticastAddress); err != nil {
		// joinGroupLocked only returns an error if the group address is not a valid
		// IPv6 multicast address.
		panic(fmt.Sprintf("e.joinGroupLocked(%s): %s", header.IPv6AllNodesMulticastAddress, err))
	}

	// Perform DAD on the all the unicast IPv6 endpoints that are in the permanent
	// state.
	//
	// Addresses may have aleady completed DAD but in the time since the endpoint
	// was last enabled, other devices may have acquired the same addresses.
	var err tcpip.Error
	e.mu.addressableEndpointState.ForEachEndpoint(func(addressEndpoint stack.AddressEndpoint) bool {
		addr := addressEndpoint.AddressWithPrefix().Address
		if !header.IsV6UnicastAddress(addr) {
			return true
		}

		switch addressEndpoint.GetKind() {
		case stack.Permanent:
			addressEndpoint.SetKind(stack.PermanentTentative)
			fallthrough
		case stack.PermanentTentative:
			err = e.mu.ndp.startDuplicateAddressDetection(addr, addressEndpoint)
			return err == nil
		default:
			return true
		}
	})
	if err != nil {
		return err
	}

	// Do not auto-generate an IPv6 link-local address for loopback devices.
	if e.protocol.options.AutoGenLinkLocal && !e.nic.IsLoopback() {
		// The valid and preferred lifetime is infinite for the auto-generated
		// link-local address.
		e.mu.ndp.doSLAAC(header.IPv6LinkLocalPrefix.Subnet(), header.NDPInfiniteLifetime, header.NDPInfiniteLifetime)
	}

	// If we are operating as a router, then do not solicit routers since we
	// won't process the RAs anyway.
	//
	// Routers do not process Router Advertisements (RA) the same way a host
	// does. That is, routers do not learn from RAs (e.g. on-link prefixes
	// and default routers). Therefore, soliciting RAs from other routers on
	// a link is unnecessary for routers.
	if !e.protocol.Forwarding() {
		e.mu.ndp.startSolicitingRouters()
	}

	return nil
}

// Enabled implements stack.NetworkEndpoint.
func (e *endpoint) Enabled() bool {
	return e.nic.Enabled() && e.isEnabled()
}

// isEnabled returns true if the endpoint is enabled, regardless of the
// enabled status of the NIC.
func (e *endpoint) isEnabled() bool {
	return atomic.LoadUint32(&e.enabled) == 1
}

// setEnabled sets the enabled status for the endpoint.
//
// Returns true if the enabled status was updated.
func (e *endpoint) setEnabled(v bool) bool {
	if v {
		return atomic.SwapUint32(&e.enabled, 1) == 0
	}
	return atomic.SwapUint32(&e.enabled, 0) == 1
}

// Disable implements stack.NetworkEndpoint.
func (e *endpoint) Disable() {
	e.mu.Lock()
	defer e.mu.Unlock()
	e.disableLocked()
}

func (e *endpoint) disableLocked() {
	if !e.Enabled() {
		return
	}

	e.mu.ndp.stopSolicitingRouters()
	e.mu.ndp.cleanupState(false /* hostOnly */)
	e.stopDADForPermanentAddressesLocked()

	// The endpoint may have already left the multicast group.
	switch err := e.leaveGroupLocked(header.IPv6AllNodesMulticastAddress); err.(type) {
	case nil, *tcpip.ErrBadLocalAddress:
	default:
		panic(fmt.Sprintf("unexpected error when leaving group = %s: %s", header.IPv6AllNodesMulticastAddress, err))
	}

	// Leave groups from the perspective of MLD so that routers know that
	// we are no longer interested in the group.
	e.mu.mld.softLeaveAll()

	if !e.setEnabled(false) {
		panic("should have only done work to disable the endpoint if it was enabled")
	}
}

// stopDADForPermanentAddressesLocked stops DAD for all permaneent addresses.
//
// Precondition: e.mu must be write locked.
func (e *endpoint) stopDADForPermanentAddressesLocked() {
	// Stop DAD for all the tentative unicast addresses.
	e.mu.addressableEndpointState.ForEachEndpoint(func(addressEndpoint stack.AddressEndpoint) bool {
		if addressEndpoint.GetKind() != stack.PermanentTentative {
			return true
		}

		addr := addressEndpoint.AddressWithPrefix().Address
		if header.IsV6UnicastAddress(addr) {
			e.mu.ndp.stopDuplicateAddressDetection(addr)
		}

		return true
	})
}

// DefaultTTL is the default hop limit for this endpoint.
func (e *endpoint) DefaultTTL() uint8 {
	return e.protocol.DefaultTTL()
}

// MTU implements stack.NetworkEndpoint.MTU. It returns the link-layer MTU minus
// the network layer max header length.
func (e *endpoint) MTU() uint32 {
	networkMTU, err := calculateNetworkMTU(e.nic.MTU(), header.IPv6MinimumSize)
	if err != nil {
		return 0
	}
	return networkMTU
}

// MaxHeaderLength returns the maximum length needed by ipv6 headers (and
// underlying protocols).
func (e *endpoint) MaxHeaderLength() uint16 {
	// TODO(gvisor.dev/issues/5035): The maximum header length returned here does
	// not open the possibility for the caller to know about size required for
	// extension headers.
	return e.nic.MaxHeaderLength() + header.IPv6MinimumSize
}

func addIPHeader(srcAddr, dstAddr tcpip.Address, pkt *stack.PacketBuffer, params stack.NetworkHeaderParams, extensionHeaders header.IPv6ExtHdrSerializer) tcpip.Error {
	extHdrsLen := extensionHeaders.Length()
	length := pkt.Size() + extensionHeaders.Length()
	if length > math.MaxUint16 {
		return &tcpip.ErrMessageTooLong{}
	}
	ip := header.IPv6(pkt.NetworkHeader().Push(header.IPv6MinimumSize + extHdrsLen))
	ip.Encode(&header.IPv6Fields{
		PayloadLength:     uint16(length),
		TransportProtocol: params.Protocol,
		HopLimit:          params.TTL,
		TrafficClass:      params.TOS,
		SrcAddr:           srcAddr,
		DstAddr:           dstAddr,
		ExtensionHeaders:  extensionHeaders,
	})
	pkt.NetworkProtocolNumber = ProtocolNumber
	return nil
}

func packetMustBeFragmented(pkt *stack.PacketBuffer, networkMTU uint32, gso *stack.GSO) bool {
	payload := pkt.TransportHeader().View().Size() + pkt.Data.Size()
	return (gso == nil || gso.Type == stack.GSONone) && uint32(payload) > networkMTU
}

// handleFragments fragments pkt and calls the handler function on each
// fragment. It returns the number of fragments handled and the number of
// fragments left to be processed. The IP header must already be present in the
// original packet. The transport header protocol number is required to avoid
// parsing the IPv6 extension headers.
func (e *endpoint) handleFragments(r *stack.Route, gso *stack.GSO, networkMTU uint32, pkt *stack.PacketBuffer, transProto tcpip.TransportProtocolNumber, handler func(*stack.PacketBuffer) tcpip.Error) (int, int, tcpip.Error) {
	networkHeader := header.IPv6(pkt.NetworkHeader().View())

	// TODO(gvisor.dev/issue/3912): Once the Authentication or ESP Headers are
	// supported for outbound packets, their length should not affect the fragment
	// maximum payload length because they should only be transmitted once.
	fragmentPayloadLen := (networkMTU - header.IPv6FragmentHeaderSize) &^ 7
	if fragmentPayloadLen < header.IPv6FragmentExtHdrFragmentOffsetBytesPerUnit {
		// We need at least 8 bytes of space left for the fragmentable part because
		// the fragment payload must obviously be non-zero and must be a multiple
		// of 8 as per RFC 8200 section 4.5:
		//   Each complete fragment, except possibly the last ("rightmost") one, is
		//   an integer multiple of 8 octets long.
		return 0, 1, &tcpip.ErrMessageTooLong{}
	}

	if fragmentPayloadLen < uint32(pkt.TransportHeader().View().Size()) {
		// As per RFC 8200 Section 4.5, the Transport Header is expected to be small
		// enough to fit in the first fragment.
		return 0, 1, &tcpip.ErrMessageTooLong{}
	}

	pf := fragmentation.MakePacketFragmenter(pkt, fragmentPayloadLen, calculateFragmentReserve(pkt))
	id := atomic.AddUint32(&e.protocol.ids[hashRoute(r, e.protocol.hashIV)%buckets], 1)

	var n int
	for {
		fragPkt, more := buildNextFragment(&pf, networkHeader, transProto, id)
		if err := handler(fragPkt); err != nil {
			return n, pf.RemainingFragmentCount() + 1, err
		}
		n++
		if !more {
			return n, pf.RemainingFragmentCount(), nil
		}
	}
}

// WritePacket writes a packet to the given destination address and protocol.
func (e *endpoint) WritePacket(r *stack.Route, gso *stack.GSO, params stack.NetworkHeaderParams, pkt *stack.PacketBuffer) tcpip.Error {
	if err := addIPHeader(r.LocalAddress, r.RemoteAddress, pkt, params, nil /* extensionHeaders */); err != nil {
		return err
	}

	// iptables filtering. All packets that reach here are locally
	// generated.
	outNicName := e.protocol.stack.FindNICNameFromID(e.nic.ID())
	if ok := e.protocol.stack.IPTables().Check(stack.Output, pkt, gso, r, "" /* preroutingAddr */, "" /* inNicName */, outNicName); !ok {
		// iptables is telling us to drop the packet.
		e.stats.ip.IPTablesOutputDropped.Increment()
		return nil
	}

	// If the packet is manipulated as per NAT Output rules, handle packet
	// based on destination address and do not send the packet to link
	// layer.
	//
	// TODO(gvisor.dev/issue/170): We should do this for every
	// packet, rather than only NATted packets, but removing this check
	// short circuits broadcasts before they are sent out to other hosts.
	if pkt.NatDone {
		netHeader := header.IPv6(pkt.NetworkHeader().View())
		if ep, err := e.protocol.stack.FindNetworkEndpoint(ProtocolNumber, netHeader.DestinationAddress()); err == nil {
			pkt := pkt.CloneToInbound()
			if e.protocol.stack.ParsePacketBuffer(ProtocolNumber, pkt) == stack.ParsedOK {
				// Since we rewrote the packet but it is being routed back to us, we can
				// safely assume the checksum is valid.
				pkt.RXTransportChecksumValidated = true
				ep.(*endpoint).handlePacket(pkt)
			}
			return nil
		}
	}

	return e.writePacket(r, gso, pkt, params.Protocol, false /* headerIncluded */)
}

func (e *endpoint) writePacket(r *stack.Route, gso *stack.GSO, pkt *stack.PacketBuffer, protocol tcpip.TransportProtocolNumber, headerIncluded bool) tcpip.Error {
	if r.Loop&stack.PacketLoop != 0 {
		pkt := pkt.CloneToInbound()
		if e.protocol.stack.ParsePacketBuffer(ProtocolNumber, pkt) == stack.ParsedOK {
			// If the packet was generated by the stack (not a raw/packet endpoint
			// where a packet may be written with the header included), then we can
			// safely assume the checksum is valid.
			pkt.RXTransportChecksumValidated = !headerIncluded
			e.handlePacket(pkt)
		}
	}
	if r.Loop&stack.PacketOut == 0 {
		return nil
	}

	stats := e.stats.ip
	networkMTU, err := calculateNetworkMTU(e.nic.MTU(), uint32(pkt.NetworkHeader().View().Size()))
	if err != nil {
		stats.OutgoingPacketErrors.Increment()
		return err
	}

	if packetMustBeFragmented(pkt, networkMTU, gso) {
		sent, remain, err := e.handleFragments(r, gso, networkMTU, pkt, protocol, func(fragPkt *stack.PacketBuffer) tcpip.Error {
			// TODO(gvisor.dev/issue/3884): Evaluate whether we want to send each
			// fragment one by one using WritePacket() (current strategy) or if we
			// want to create a PacketBufferList from the fragments and feed it to
			// WritePackets(). It'll be faster but cost more memory.
			return e.nic.WritePacket(r, gso, ProtocolNumber, fragPkt)
		})
		stats.PacketsSent.IncrementBy(uint64(sent))
		stats.OutgoingPacketErrors.IncrementBy(uint64(remain))
		return err
	}

	if err := e.nic.WritePacket(r, gso, ProtocolNumber, pkt); err != nil {
		stats.OutgoingPacketErrors.Increment()
		return err
	}

	stats.PacketsSent.Increment()
	return nil
}

// WritePackets implements stack.NetworkEndpoint.WritePackets.
func (e *endpoint) WritePackets(r *stack.Route, gso *stack.GSO, pkts stack.PacketBufferList, params stack.NetworkHeaderParams) (int, tcpip.Error) {
	if r.Loop&stack.PacketLoop != 0 {
		panic("not implemented")
	}
	if r.Loop&stack.PacketOut == 0 {
		return pkts.Len(), nil
	}

	stats := e.stats.ip
	linkMTU := e.nic.MTU()
	for pb := pkts.Front(); pb != nil; pb = pb.Next() {
		if err := addIPHeader(r.LocalAddress, r.RemoteAddress, pb, params, nil /* extensionHeaders */); err != nil {
			return 0, err
		}

		networkMTU, err := calculateNetworkMTU(linkMTU, uint32(pb.NetworkHeader().View().Size()))
		if err != nil {
			stats.OutgoingPacketErrors.IncrementBy(uint64(pkts.Len()))
			return 0, err
		}
		if packetMustBeFragmented(pb, networkMTU, gso) {
			// Keep track of the packet that is about to be fragmented so it can be
			// removed once the fragmentation is done.
			originalPkt := pb
			if _, _, err := e.handleFragments(r, gso, networkMTU, pb, params.Protocol, func(fragPkt *stack.PacketBuffer) tcpip.Error {
				// Modify the packet list in place with the new fragments.
				pkts.InsertAfter(pb, fragPkt)
				pb = fragPkt
				return nil
			}); err != nil {
				stats.OutgoingPacketErrors.IncrementBy(uint64(pkts.Len()))
				return 0, err
			}
			// Remove the packet that was just fragmented and process the rest.
			pkts.Remove(originalPkt)
		}
	}

	// iptables filtering. All packets that reach here are locally
	// generated.
	outNicName := e.protocol.stack.FindNICNameFromID(e.nic.ID())
	dropped, natPkts := e.protocol.stack.IPTables().CheckPackets(stack.Output, pkts, gso, r, "" /* inNicName */, outNicName)
	if len(dropped) == 0 && len(natPkts) == 0 {
		// Fast path: If no packets are to be dropped then we can just invoke the
		// faster WritePackets API directly.
		n, err := e.nic.WritePackets(r, gso, pkts, ProtocolNumber)
		stats.PacketsSent.IncrementBy(uint64(n))
		if err != nil {
			stats.OutgoingPacketErrors.IncrementBy(uint64(pkts.Len() - n))
		}
		return n, err
	}
	stats.IPTablesOutputDropped.IncrementBy(uint64(len(dropped)))

	// Slow path as we are dropping some packets in the batch degrade to
	// emitting one packet at a time.
	n := 0
	for pkt := pkts.Front(); pkt != nil; pkt = pkt.Next() {
		if _, ok := dropped[pkt]; ok {
			continue
		}
		if _, ok := natPkts[pkt]; ok {
			netHeader := header.IPv6(pkt.NetworkHeader().View())
			if ep, err := e.protocol.stack.FindNetworkEndpoint(ProtocolNumber, netHeader.DestinationAddress()); err == nil {
				pkt := pkt.CloneToInbound()
				if e.protocol.stack.ParsePacketBuffer(ProtocolNumber, pkt) == stack.ParsedOK {
					// Since we rewrote the packet but it is being routed back to us, we
					// can safely assume the checksum is valid.
					pkt.RXTransportChecksumValidated = true
					ep.(*endpoint).handlePacket(pkt)
				}
				n++
				continue
			}
		}
		if err := e.nic.WritePacket(r, gso, ProtocolNumber, pkt); err != nil {
			stats.PacketsSent.IncrementBy(uint64(n))
			stats.OutgoingPacketErrors.IncrementBy(uint64(pkts.Len() - n + len(dropped)))
			// Dropped packets aren't errors, so include them in
			// the return value.
			return n + len(dropped), err
		}
		n++
	}

	stats.PacketsSent.IncrementBy(uint64(n))
	// Dropped packets aren't errors, so include them in the return value.
	return n + len(dropped), nil
}

// WriteHeaderIncludedPacket implements stack.NetworkEndpoint.
func (e *endpoint) WriteHeaderIncludedPacket(r *stack.Route, pkt *stack.PacketBuffer) tcpip.Error {
	// The packet already has an IP header, but there are a few required checks.
	h, ok := pkt.Data.PullUp(header.IPv6MinimumSize)
	if !ok {
		return &tcpip.ErrMalformedHeader{}
	}
	ip := header.IPv6(h)

	// Always set the payload length.
	pktSize := pkt.Data.Size()
	ip.SetPayloadLength(uint16(pktSize - header.IPv6MinimumSize))

	// Set the source address when zero.
	if ip.SourceAddress() == header.IPv6Any {
		ip.SetSourceAddress(r.LocalAddress)
	}

	// Set the destination. If the packet already included a destination, it will
	// be part of the route anyways.
	ip.SetDestinationAddress(r.RemoteAddress)

	// Populate the packet buffer's network header and don't allow an invalid
	// packet to be sent.
	//
	// Note that parsing only makes sure that the packet is well formed as per the
	// wire format. We also want to check if the header's fields are valid before
	// sending the packet.
	proto, _, _, _, ok := parse.IPv6(pkt)
	if !ok || !header.IPv6(pkt.NetworkHeader().View()).IsValid(pktSize) {
		return &tcpip.ErrMalformedHeader{}
	}

	return e.writePacket(r, nil /* gso */, pkt, proto, true /* headerIncluded */)
}

// forwardPacket attempts to forward a packet to its final destination.
func (e *endpoint) forwardPacket(pkt *stack.PacketBuffer) tcpip.Error {
	h := header.IPv6(pkt.NetworkHeader().View())
	hopLimit := h.HopLimit()
	if hopLimit <= 1 {
		// As per RFC 4443 section 3.3,
		//
		//   If a router receives a packet with a Hop Limit of zero, or if a
		//   router decrements a packet's Hop Limit to zero, it MUST discard the
		//   packet and originate an ICMPv6 Time Exceeded message with Code 0 to
		//   the source of the packet.  This indicates either a routing loop or
		//   too small an initial Hop Limit value.
		return e.protocol.returnError(&icmpReasonHopLimitExceeded{}, pkt)
	}

	dstAddr := h.DestinationAddress()

	// Check if the destination is owned by the stack.
	networkEndpoint, err := e.protocol.stack.FindNetworkEndpoint(ProtocolNumber, dstAddr)
	if err == nil {
		networkEndpoint.(*endpoint).handlePacket(pkt)
		return nil
	}
	if _, ok := err.(*tcpip.ErrBadAddress); !ok {
		return err
	}

	r, err := e.protocol.stack.FindRoute(0, "", dstAddr, ProtocolNumber, false /* multicastLoop */)
	if err != nil {
		return err
	}
	defer r.Release()

	// We need to do a deep copy of the IP packet because
	// WriteHeaderIncludedPacket takes ownership of the packet buffer, but we do
	// not own it.
	newHdr := header.IPv6(stack.PayloadSince(pkt.NetworkHeader()))

	// As per RFC 8200 section 3,
	//
	//   Hop Limit           8-bit unsigned integer. Decremented by 1 by
	//                       each node that forwards the packet.
	newHdr.SetHopLimit(hopLimit - 1)

	return r.WriteHeaderIncludedPacket(stack.NewPacketBuffer(stack.PacketBufferOptions{
		ReserveHeaderBytes: int(r.MaxHeaderLength()),
		Data:               buffer.View(newHdr).ToVectorisedView(),
	}))
}

// HandlePacket is called by the link layer when new ipv6 packets arrive for
// this endpoint.
func (e *endpoint) HandlePacket(pkt *stack.PacketBuffer) {
	stats := e.stats.ip

	stats.PacketsReceived.Increment()

	if !e.isEnabled() {
		stats.DisabledPacketsReceived.Increment()
		return
	}

	// Loopback traffic skips the prerouting chain.
	if !e.nic.IsLoopback() {
		inNicName := e.protocol.stack.FindNICNameFromID(e.nic.ID())
		if ok := e.protocol.stack.IPTables().Check(stack.Prerouting, pkt, nil, nil, e.MainAddress().Address, inNicName, "" /* outNicName */); !ok {
			// iptables is telling us to drop the packet.
			stats.IPTablesPreroutingDropped.Increment()
			return
		}
	}

	e.handlePacket(pkt)
}

// handlePacket is like HandlePacket except it does not perform the prerouting
// iptables hook.
func (e *endpoint) handlePacket(pkt *stack.PacketBuffer) {
	pkt.NICID = e.nic.ID()
	stats := e.stats.ip

	h := header.IPv6(pkt.NetworkHeader().View())
	if !h.IsValid(pkt.Data.Size() + pkt.NetworkHeader().View().Size() + pkt.TransportHeader().View().Size()) {
		stats.MalformedPacketsReceived.Increment()
		return
	}
	srcAddr := h.SourceAddress()
	dstAddr := h.DestinationAddress()

	// As per RFC 4291 section 2.7:
	//   Multicast addresses must not be used as source addresses in IPv6
	//   packets or appear in any Routing header.
	if header.IsV6MulticastAddress(srcAddr) {
		stats.InvalidSourceAddressesReceived.Increment()
		return
	}

	// The destination address should be an address we own or a group we joined
	// for us to receive the packet. Otherwise, attempt to forward the packet.
	if addressEndpoint := e.AcquireAssignedAddress(dstAddr, e.nic.Promiscuous(), stack.CanBePrimaryEndpoint); addressEndpoint != nil {
		addressEndpoint.DecRef()
	} else if !e.IsInGroup(dstAddr) {
		if !e.protocol.Forwarding() {
			stats.InvalidDestinationAddressesReceived.Increment()
			return
		}

		_ = e.forwardPacket(pkt)
		return
	}

	// vv consists of:
	// - Any IPv6 header bytes after the first 40 (i.e. extensions).
	// - The transport header, if present.
	// - Any other payload data.
	vv := pkt.NetworkHeader().View()[header.IPv6MinimumSize:].ToVectorisedView()
	vv.AppendView(pkt.TransportHeader().View())
	vv.Append(pkt.Data)
	it := header.MakeIPv6PayloadIterator(header.IPv6ExtensionHeaderIdentifier(h.NextHeader()), vv)
	hasFragmentHeader := false

	// iptables filtering. All packets that reach here are intended for
	// this machine and need not be forwarded.
	inNicName := e.protocol.stack.FindNICNameFromID(e.nic.ID())
	if ok := e.protocol.stack.IPTables().Check(stack.Input, pkt, nil, nil, "" /* preroutingAddr */, inNicName, "" /* outNicName */); !ok {
		// iptables is telling us to drop the packet.
		stats.IPTablesInputDropped.Increment()
		return
	}

	for {
		// Keep track of the start of the previous header so we can report the
		// special case of a Hop by Hop at a location other than at the start.
		previousHeaderStart := it.HeaderOffset()
		extHdr, done, err := it.Next()
		if err != nil {
			stats.MalformedPacketsReceived.Increment()
			return
		}
		if done {
			break
		}

		switch extHdr := extHdr.(type) {
		case header.IPv6HopByHopOptionsExtHdr:
			// As per RFC 8200 section 4.1, the Hop By Hop extension header is
			// restricted to appear immediately after an IPv6 fixed header.
			if previousHeaderStart != 0 {
				_ = e.protocol.returnError(&icmpReasonParameterProblem{
					code:    header.ICMPv6UnknownHeader,
					pointer: previousHeaderStart,
				}, pkt)
				return
			}

			optsIt := extHdr.Iter()

			for {
				opt, done, err := optsIt.Next()
				if err != nil {
					stats.MalformedPacketsReceived.Increment()
					return
				}
				if done {
					break
				}

				// We currently do not support any IPv6 Hop By Hop extension header
				// options.
				switch opt.UnknownAction() {
				case header.IPv6OptionUnknownActionSkip:
				case header.IPv6OptionUnknownActionDiscard:
					return
				case header.IPv6OptionUnknownActionDiscardSendICMPNoMulticastDest:
					if header.IsV6MulticastAddress(dstAddr) {
						return
					}
					fallthrough
				case header.IPv6OptionUnknownActionDiscardSendICMP:
					// This case satisfies a requirement of RFC 8200 section 4.2
					// which states that an unknown option starting with bits [10] should:
					//
					//    discard the packet and, regardless of whether or not the
					//    packet's Destination Address was a multicast address, send an
					//    ICMP Parameter Problem, Code 2, message to the packet's
					//    Source Address, pointing to the unrecognized Option Type.
					//
					_ = e.protocol.returnError(&icmpReasonParameterProblem{
						code:               header.ICMPv6UnknownOption,
						pointer:            it.ParseOffset() + optsIt.OptionOffset(),
						respondToMulticast: true,
					}, pkt)
					return
				default:
					panic(fmt.Sprintf("unrecognized action for an unrecognized Hop By Hop extension header option = %d", opt))
				}
			}

		case header.IPv6RoutingExtHdr:
			// As per RFC 8200 section 4.4, if a node encounters a routing header with
			// an unrecognized routing type value, with a non-zero Segments Left
			// value, the node must discard the packet and send an ICMP Parameter
			// Problem, Code 0 to the packet's Source Address, pointing to the
			// unrecognized Routing Type.
			//
			// If the Segments Left is 0, the node must ignore the Routing extension
			// header and process the next header in the packet.
			//
			// Note, the stack does not yet handle any type of routing extension
			// header, so we just make sure Segments Left is zero before processing
			// the next extension header.
			if extHdr.SegmentsLeft() != 0 {
				_ = e.protocol.returnError(&icmpReasonParameterProblem{
					code:    header.ICMPv6ErroneousHeader,
					pointer: it.ParseOffset(),
				}, pkt)
				return
			}

		case header.IPv6FragmentExtHdr:
			hasFragmentHeader = true

			if extHdr.IsAtomic() {
				// This fragment extension header indicates that this packet is an
				// atomic fragment. An atomic fragment is a fragment that contains
				// all the data required to reassemble a full packet. As per RFC 6946,
				// atomic fragments must not interfere with "normal" fragmented traffic
				// so we skip processing the fragment instead of feeding it through the
				// reassembly process below.
				continue
			}

			fragmentFieldOffset := it.ParseOffset()

			// Don't consume the iterator if we have the first fragment because we
			// will use it to validate that the first fragment holds the upper layer
			// header.
			rawPayload := it.AsRawHeader(extHdr.FragmentOffset() != 0 /* consume */)

			if extHdr.FragmentOffset() == 0 {
				// Check that the iterator ends with a raw payload as the first fragment
				// should include all headers up to and including any upper layer
				// headers, as per RFC 8200 section 4.5; only upper layer data
				// (non-headers) should follow the fragment extension header.
				var lastHdr header.IPv6PayloadHeader

				for {
					it, done, err := it.Next()
					if err != nil {
						stats.MalformedPacketsReceived.Increment()
						stats.MalformedFragmentsReceived.Increment()
						return
					}
					if done {
						break
					}

					lastHdr = it
				}

				// If the last header is a raw header, then the last portion of the IPv6
				// payload is not a known IPv6 extension header. Note, this does not
				// mean that the last portion is an upper layer header or not an
				// extension header because:
				//  1) we do not yet support all extension headers
				//  2) we do not validate the upper layer header before reassembling.
				//
				// This check makes sure that a known IPv6 extension header is not
				// present after the Fragment extension header in a non-initial
				// fragment.
				//
				// TODO(#2196): Support IPv6 Authentication and Encapsulated
				// Security Payload extension headers.
				// TODO(#2333): Validate that the upper layer header is valid.
				switch lastHdr.(type) {
				case header.IPv6RawPayloadHeader:
				default:
					stats.MalformedPacketsReceived.Increment()
					stats.MalformedFragmentsReceived.Increment()
					return
				}
			}

			fragmentPayloadLen := rawPayload.Buf.Size()
			if fragmentPayloadLen == 0 {
				// Drop the packet as it's marked as a fragment but has no payload.
				stats.MalformedPacketsReceived.Increment()
				stats.MalformedFragmentsReceived.Increment()
				return
			}

			// As per RFC 2460 Section 4.5:
			//
			//    If the length of a fragment, as derived from the fragment packet's
			//    Payload Length field, is not a multiple of 8 octets and the M flag
			//    of that fragment is 1, then that fragment must be discarded and an
			//    ICMP Parameter Problem, Code 0, message should be sent to the source
			//    of the fragment, pointing to the Payload Length field of the
			//    fragment packet.
			if extHdr.More() && fragmentPayloadLen%header.IPv6FragmentExtHdrFragmentOffsetBytesPerUnit != 0 {
				stats.MalformedPacketsReceived.Increment()
				stats.MalformedFragmentsReceived.Increment()
				_ = e.protocol.returnError(&icmpReasonParameterProblem{
					code:    header.ICMPv6ErroneousHeader,
					pointer: header.IPv6PayloadLenOffset,
				}, pkt)
				return
			}

			// The packet is a fragment, let's try to reassemble it.
			start := extHdr.FragmentOffset() * header.IPv6FragmentExtHdrFragmentOffsetBytesPerUnit

			// As per RFC 2460 Section 4.5:
			//
			//    If the length and offset of a fragment are such that the Payload
			//    Length of the packet reassembled from that fragment would exceed
			//    65,535 octets, then that fragment must be discarded and an ICMP
			//    Parameter Problem, Code 0, message should be sent to the source of
			//    the fragment, pointing to the Fragment Offset field of the fragment
			//    packet.
			if int(start)+fragmentPayloadLen > header.IPv6MaximumPayloadSize {
				stats.MalformedPacketsReceived.Increment()
				stats.MalformedFragmentsReceived.Increment()
				_ = e.protocol.returnError(&icmpReasonParameterProblem{
					code:    header.ICMPv6ErroneousHeader,
					pointer: fragmentFieldOffset,
				}, pkt)
				return
			}

			// Note that pkt doesn't have its transport header set after reassembly,
			// and won't until DeliverNetworkPacket sets it.
			data, proto, ready, err := e.protocol.fragmentation.Process(
				// IPv6 ignores the Protocol field since the ID only needs to be unique
				// across source-destination pairs, as per RFC 8200 section 4.5.
				fragmentation.FragmentID{
					Source:      srcAddr,
					Destination: dstAddr,
					ID:          extHdr.ID(),
				},
				start,
				start+uint16(fragmentPayloadLen)-1,
				extHdr.More(),
				uint8(rawPayload.Identifier),
				pkt,
			)
			if err != nil {
				stats.MalformedPacketsReceived.Increment()
				stats.MalformedFragmentsReceived.Increment()
				return
			}

			if ready {
				pkt.Data = data

				// We create a new iterator with the reassembled packet because we could
				// have more extension headers in the reassembled payload, as per RFC
				// 8200 section 4.5. We also use the NextHeader value from the first
				// fragment.
				it = header.MakeIPv6PayloadIterator(header.IPv6ExtensionHeaderIdentifier(proto), pkt.Data)
			}

		case header.IPv6DestinationOptionsExtHdr:
			optsIt := extHdr.Iter()

			for {
				opt, done, err := optsIt.Next()
				if err != nil {
					stats.MalformedPacketsReceived.Increment()
					return
				}
				if done {
					break
				}

				// We currently do not support any IPv6 Destination extension header
				// options.
				switch opt.UnknownAction() {
				case header.IPv6OptionUnknownActionSkip:
				case header.IPv6OptionUnknownActionDiscard:
					return
				case header.IPv6OptionUnknownActionDiscardSendICMPNoMulticastDest:
					if header.IsV6MulticastAddress(dstAddr) {
						return
					}
					fallthrough
				case header.IPv6OptionUnknownActionDiscardSendICMP:
					// This case satisfies a requirement of RFC 8200 section 4.2
					// which states that an unknown option starting with bits [10] should:
					//
					//    discard the packet and, regardless of whether or not the
					//    packet's Destination Address was a multicast address, send an
					//    ICMP Parameter Problem, Code 2, message to the packet's
					//    Source Address, pointing to the unrecognized Option Type.
					//
					_ = e.protocol.returnError(&icmpReasonParameterProblem{
						code:               header.ICMPv6UnknownOption,
						pointer:            it.ParseOffset() + optsIt.OptionOffset(),
						respondToMulticast: true,
					}, pkt)
					return
				default:
					panic(fmt.Sprintf("unrecognized action for an unrecognized Destination extension header option = %d", opt))
				}
			}

		case header.IPv6RawPayloadHeader:
			// If the last header in the payload isn't a known IPv6 extension header,
			// handle it as if it is transport layer data.

			// For unfragmented packets, extHdr still contains the transport header.
			// Get rid of it.
			//
			// For reassembled fragments, pkt.TransportHeader is unset, so this is a
			// no-op and pkt.Data begins with the transport header.
			extHdr.Buf.TrimFront(pkt.TransportHeader().View().Size())
			pkt.Data = extHdr.Buf

			stats.PacketsDelivered.Increment()
			if p := tcpip.TransportProtocolNumber(extHdr.Identifier); p == header.ICMPv6ProtocolNumber {
				pkt.TransportProtocolNumber = p
				e.handleICMP(pkt, hasFragmentHeader)
			} else {
				stats.PacketsDelivered.Increment()
				switch res := e.dispatcher.DeliverTransportPacket(p, pkt); res {
				case stack.TransportPacketHandled:
				case stack.TransportPacketDestinationPortUnreachable:
					// As per RFC 4443 section 3.1:
					//   A destination node SHOULD originate a Destination Unreachable
					//   message with Code 4 in response to a packet for which the
					//   transport protocol (e.g., UDP) has no listener, if that transport
					//   protocol has no alternative means to inform the sender.
					_ = e.protocol.returnError(&icmpReasonPortUnreachable{}, pkt)
				case stack.TransportPacketProtocolUnreachable:
					// As per RFC 8200 section 4. (page 7):
					//   Extension headers are numbered from IANA IP Protocol Numbers
					//   [IANA-PN], the same values used for IPv4 and IPv6.  When
					//   processing a sequence of Next Header values in a packet, the
					//   first one that is not an extension header [IANA-EH] indicates
					//   that the next item in the packet is the corresponding upper-layer
					//   header.
					// With more related information on page 8:
					//   If, as a result of processing a header, the destination node is
					//   required to proceed to the next header but the Next Header value
					//   in the current header is unrecognized by the node, it should
					//   discard the packet and send an ICMP Parameter Problem message to
					//   the source of the packet, with an ICMP Code value of 1
					//   ("unrecognized Next Header type encountered") and the ICMP
					//   Pointer field containing the offset of the unrecognized value
					//   within the original packet.
					//
					// Which when taken together indicate that an unknown protocol should
					// be treated as an unrecognized next header value.
					// The location of the Next Header field is in a different place in
					// the initial IPv6 header than it is in the extension headers so
					// treat it specially.
					prevHdrIDOffset := uint32(header.IPv6NextHeaderOffset)
					if previousHeaderStart != 0 {
						prevHdrIDOffset = previousHeaderStart
					}
					_ = e.protocol.returnError(&icmpReasonParameterProblem{
						code:    header.ICMPv6UnknownHeader,
						pointer: prevHdrIDOffset,
					}, pkt)
				default:
					panic(fmt.Sprintf("unrecognized result from DeliverTransportPacket = %d", res))
				}
			}

		default:
			// Since the iterator returns IPv6RawPayloadHeader for unknown Extension
			// Header IDs this should never happen unless we missed a supported type
			// here.
			panic(fmt.Sprintf("unrecognized type from it.Next() = %T", extHdr))

		}
	}
}

// Close cleans up resources associated with the endpoint.
func (e *endpoint) Close() {
	e.mu.Lock()
	e.disableLocked()
	e.mu.ndp.removeSLAACAddresses(false /* keepLinkLocal */)
	e.stopDADForPermanentAddressesLocked()
	e.mu.addressableEndpointState.Cleanup()
	e.mu.Unlock()

	e.protocol.forgetEndpoint(e.nic.ID())
}

// NetworkProtocolNumber implements stack.NetworkEndpoint.NetworkProtocolNumber.
func (e *endpoint) NetworkProtocolNumber() tcpip.NetworkProtocolNumber {
	return e.protocol.Number()
}

// AddAndAcquirePermanentAddress implements stack.AddressableEndpoint.
func (e *endpoint) AddAndAcquirePermanentAddress(addr tcpip.AddressWithPrefix, peb stack.PrimaryEndpointBehavior, configType stack.AddressConfigType, deprecated bool) (stack.AddressEndpoint, tcpip.Error) {
	// TODO(b/169350103): add checks here after making sure we no longer receive
	// an empty address.
	e.mu.Lock()
	defer e.mu.Unlock()
	return e.addAndAcquirePermanentAddressLocked(addr, peb, configType, deprecated)
}

// addAndAcquirePermanentAddressLocked is like AddAndAcquirePermanentAddress but
// with locking requirements.
//
// addAndAcquirePermanentAddressLocked also joins the passed address's
// solicited-node multicast group and start duplicate address detection.
//
// Precondition: e.mu must be write locked.
func (e *endpoint) addAndAcquirePermanentAddressLocked(addr tcpip.AddressWithPrefix, peb stack.PrimaryEndpointBehavior, configType stack.AddressConfigType, deprecated bool) (stack.AddressEndpoint, tcpip.Error) {
	addressEndpoint, err := e.mu.addressableEndpointState.AddAndAcquirePermanentAddress(addr, peb, configType, deprecated)
	if err != nil {
		return nil, err
	}

	if !header.IsV6UnicastAddress(addr.Address) {
		return addressEndpoint, nil
	}

	addressEndpoint.SetKind(stack.PermanentTentative)

	if e.Enabled() {
		if err := e.mu.ndp.startDuplicateAddressDetection(addr.Address, addressEndpoint); err != nil {
			return nil, err
		}
	}

	snmc := header.SolicitedNodeAddr(addr.Address)
	if err := e.joinGroupLocked(snmc); err != nil {
		// joinGroupLocked only returns an error if the group address is not a valid
		// IPv6 multicast address.
		panic(fmt.Sprintf("e.joinGroupLocked(%s): %s", snmc, err))
	}

	return addressEndpoint, nil
}

// RemovePermanentAddress implements stack.AddressableEndpoint.
func (e *endpoint) RemovePermanentAddress(addr tcpip.Address) tcpip.Error {
	e.mu.Lock()
	defer e.mu.Unlock()

	addressEndpoint := e.getAddressRLocked(addr)
	if addressEndpoint == nil || !addressEndpoint.GetKind().IsPermanent() {
		return &tcpip.ErrBadLocalAddress{}
	}

	return e.removePermanentEndpointLocked(addressEndpoint, true)
}

// removePermanentEndpointLocked is like removePermanentAddressLocked except
// it works with a stack.AddressEndpoint.
//
// Precondition: e.mu must be write locked.
func (e *endpoint) removePermanentEndpointLocked(addressEndpoint stack.AddressEndpoint, allowSLAACInvalidation bool) tcpip.Error {
	addr := addressEndpoint.AddressWithPrefix()
	unicast := header.IsV6UnicastAddress(addr.Address)
	if unicast {
		e.mu.ndp.stopDuplicateAddressDetection(addr.Address)

		// If we are removing an address generated via SLAAC, cleanup
		// its SLAAC resources and notify the integrator.
		switch addressEndpoint.ConfigType() {
		case stack.AddressConfigSlaac:
			e.mu.ndp.cleanupSLAACAddrResourcesAndNotify(addr, allowSLAACInvalidation)
		case stack.AddressConfigSlaacTemp:
			e.mu.ndp.cleanupTempSLAACAddrResourcesAndNotify(addr, allowSLAACInvalidation)
		}
	}

	if err := e.mu.addressableEndpointState.RemovePermanentEndpoint(addressEndpoint); err != nil {
		return err
	}

	if !unicast {
		return nil
	}

	snmc := header.SolicitedNodeAddr(addr.Address)
	err := e.leaveGroupLocked(snmc)
	// The endpoint may have already left the multicast group.
	if _, ok := err.(*tcpip.ErrBadLocalAddress); ok {
		err = nil
	}
	return err
}

// hasPermanentAddressLocked returns true if the endpoint has a permanent
// address equal to the passed address.
//
// Precondition: e.mu must be read or write locked.
func (e *endpoint) hasPermanentAddressRLocked(addr tcpip.Address) bool {
	addressEndpoint := e.getAddressRLocked(addr)
	if addressEndpoint == nil {
		return false
	}
	return addressEndpoint.GetKind().IsPermanent()
}

// getAddressRLocked returns the endpoint for the passed address.
//
// Precondition: e.mu must be read or write locked.
func (e *endpoint) getAddressRLocked(localAddr tcpip.Address) stack.AddressEndpoint {
	return e.mu.addressableEndpointState.GetAddress(localAddr)
}

// MainAddress implements stack.AddressableEndpoint.
func (e *endpoint) MainAddress() tcpip.AddressWithPrefix {
	e.mu.RLock()
	defer e.mu.RUnlock()
	return e.mu.addressableEndpointState.MainAddress()
}

// AcquireAssignedAddress implements stack.AddressableEndpoint.
func (e *endpoint) AcquireAssignedAddress(localAddr tcpip.Address, allowTemp bool, tempPEB stack.PrimaryEndpointBehavior) stack.AddressEndpoint {
	e.mu.Lock()
	defer e.mu.Unlock()
	return e.acquireAddressOrCreateTempLocked(localAddr, allowTemp, tempPEB)
}

// acquireAddressOrCreateTempLocked is like AcquireAssignedAddress but with
// locking requirements.
//
// Precondition: e.mu must be write locked.
func (e *endpoint) acquireAddressOrCreateTempLocked(localAddr tcpip.Address, allowTemp bool, tempPEB stack.PrimaryEndpointBehavior) stack.AddressEndpoint {
	return e.mu.addressableEndpointState.AcquireAssignedAddress(localAddr, allowTemp, tempPEB)
}

// AcquireOutgoingPrimaryAddress implements stack.AddressableEndpoint.
func (e *endpoint) AcquireOutgoingPrimaryAddress(remoteAddr tcpip.Address, allowExpired bool) stack.AddressEndpoint {
	e.mu.RLock()
	defer e.mu.RUnlock()
	return e.acquireOutgoingPrimaryAddressRLocked(remoteAddr, allowExpired)
}

// getLinkLocalAddressRLocked returns a link-local address from the primary list
// of addresses, if one is available.
//
// See stack.PrimaryEndpointBehavior for more details about the primary list.
//
// Precondition: e.mu must be read locked.
func (e *endpoint) getLinkLocalAddressRLocked() tcpip.Address {
	var linkLocalAddr tcpip.Address
	e.mu.addressableEndpointState.ForEachPrimaryEndpoint(func(addressEndpoint stack.AddressEndpoint) bool {
		if addressEndpoint.IsAssigned(false /* allowExpired */) {
			if addr := addressEndpoint.AddressWithPrefix().Address; header.IsV6LinkLocalAddress(addr) {
				linkLocalAddr = addr
				return false
			}
		}
		return true
	})
	return linkLocalAddr
}

// acquireOutgoingPrimaryAddressRLocked is like AcquireOutgoingPrimaryAddress
// but with locking requirements.
//
// Precondition: e.mu must be read locked.
func (e *endpoint) acquireOutgoingPrimaryAddressRLocked(remoteAddr tcpip.Address, allowExpired bool) stack.AddressEndpoint {
	// addrCandidate is a candidate for Source Address Selection, as per
	// RFC 6724 section 5.
	type addrCandidate struct {
		addressEndpoint stack.AddressEndpoint
		addr            tcpip.Address
		scope           header.IPv6AddressScope

		label          uint8
		matchingPrefix uint8
	}

	if len(remoteAddr) == 0 {
		return e.mu.addressableEndpointState.AcquireOutgoingPrimaryAddress(remoteAddr, allowExpired)
	}

	// Create a candidate set of available addresses we can potentially use as a
	// source address.
	var cs []addrCandidate
	e.mu.addressableEndpointState.ForEachPrimaryEndpoint(func(addressEndpoint stack.AddressEndpoint) bool {
		// If r is not valid for outgoing connections, it is not a valid endpoint.
		if !addressEndpoint.IsAssigned(allowExpired) {
			return true
		}

		addr := addressEndpoint.AddressWithPrefix().Address
		scope, err := header.ScopeForIPv6Address(addr)
		if err != nil {
			// Should never happen as we got r from the primary IPv6 endpoint list and
			// ScopeForIPv6Address only returns an error if addr is not an IPv6
			// address.
			panic(fmt.Sprintf("header.ScopeForIPv6Address(%s): %s", addr, err))
		}

		cs = append(cs, addrCandidate{
			addressEndpoint: addressEndpoint,
			addr:            addr,
			scope:           scope,
			label:           getLabel(addr),
			matchingPrefix:  remoteAddr.MatchingPrefix(addr),
		})

		return true
	})

	remoteScope, err := header.ScopeForIPv6Address(remoteAddr)
	if err != nil {
		// primaryIPv6Endpoint should never be called with an invalid IPv6 address.
		panic(fmt.Sprintf("header.ScopeForIPv6Address(%s): %s", remoteAddr, err))
	}

	remoteLabel := getLabel(remoteAddr)

	// Sort the addresses as per RFC 6724 section 5 rules 1-3.
	//
	// TODO(b/146021396): Implement rules 4, 5 of RFC 6724 section 5.
	sort.Slice(cs, func(i, j int) bool {
		sa := cs[i]
		sb := cs[j]

		// Prefer same address as per RFC 6724 section 5 rule 1.
		if sa.addr == remoteAddr {
			return true
		}
		if sb.addr == remoteAddr {
			return false
		}

		// Prefer appropriate scope as per RFC 6724 section 5 rule 2.
		if sa.scope < sb.scope {
			return sa.scope >= remoteScope
		} else if sb.scope < sa.scope {
			return sb.scope < remoteScope
		}

		// Avoid deprecated addresses as per RFC 6724 section 5 rule 3.
		if saDep, sbDep := sa.addressEndpoint.Deprecated(), sb.addressEndpoint.Deprecated(); saDep != sbDep {
			// If sa is not deprecated, it is preferred over sb.
			return sbDep
		}

		// Prefer matching label as per RFC 6724 section 5 rule 6.
		if sa, sb := sa.label == remoteLabel, sb.label == remoteLabel; sa != sb {
			if sa {
				return true
			}
			if sb {
				return false
			}
		}

		// Prefer temporary addresses as per RFC 6724 section 5 rule 7.
		if saTemp, sbTemp := sa.addressEndpoint.ConfigType() == stack.AddressConfigSlaacTemp, sb.addressEndpoint.ConfigType() == stack.AddressConfigSlaacTemp; saTemp != sbTemp {
			return saTemp
		}

		// Use longest matching prefix as per RFC 6724 section 5 rule 8.
		if sa.matchingPrefix > sb.matchingPrefix {
			return true
		}
		if sb.matchingPrefix > sa.matchingPrefix {
			return false
		}

		// sa and sb are equal, return the endpoint that is closest to the front of
		// the primary endpoint list.
		return i < j
	})

	// Return the most preferred address that can have its reference count
	// incremented.
	for _, c := range cs {
		if c.addressEndpoint.IncRef() {
			return c.addressEndpoint
		}
	}

	return nil
}

// PrimaryAddresses implements stack.AddressableEndpoint.
func (e *endpoint) PrimaryAddresses() []tcpip.AddressWithPrefix {
	e.mu.RLock()
	defer e.mu.RUnlock()
	return e.mu.addressableEndpointState.PrimaryAddresses()
}

// PermanentAddresses implements stack.AddressableEndpoint.
func (e *endpoint) PermanentAddresses() []tcpip.AddressWithPrefix {
	e.mu.RLock()
	defer e.mu.RUnlock()
	return e.mu.addressableEndpointState.PermanentAddresses()
}

// JoinGroup implements stack.GroupAddressableEndpoint.
func (e *endpoint) JoinGroup(addr tcpip.Address) tcpip.Error {
	e.mu.Lock()
	defer e.mu.Unlock()
	return e.joinGroupLocked(addr)
}

// joinGroupLocked is like JoinGroup but with locking requirements.
//
// Precondition: e.mu must be locked.
func (e *endpoint) joinGroupLocked(addr tcpip.Address) tcpip.Error {
	if !header.IsV6MulticastAddress(addr) {
		return &tcpip.ErrBadAddress{}
	}

	e.mu.mld.joinGroup(addr)
	return nil
}

// LeaveGroup implements stack.GroupAddressableEndpoint.
func (e *endpoint) LeaveGroup(addr tcpip.Address) tcpip.Error {
	e.mu.Lock()
	defer e.mu.Unlock()
	return e.leaveGroupLocked(addr)
}

// leaveGroupLocked is like LeaveGroup but with locking requirements.
//
// Precondition: e.mu must be locked.
func (e *endpoint) leaveGroupLocked(addr tcpip.Address) tcpip.Error {
	return e.mu.mld.leaveGroup(addr)
}

// IsInGroup implements stack.GroupAddressableEndpoint.
func (e *endpoint) IsInGroup(addr tcpip.Address) bool {
	e.mu.RLock()
	defer e.mu.RUnlock()
	return e.mu.mld.isInGroup(addr)
}

// Stats implements stack.NetworkEndpoint.
func (e *endpoint) Stats() stack.NetworkEndpointStats {
	return &e.stats.localStats
}

var _ stack.ForwardingNetworkProtocol = (*protocol)(nil)
var _ stack.NetworkProtocol = (*protocol)(nil)
var _ fragmentation.TimeoutHandler = (*protocol)(nil)

type protocol struct {
	stack   *stack.Stack
	options Options

	mu struct {
		sync.RWMutex

		// eps is keyed by NICID to allow protocol methods to retrieve an endpoint
		// when handling a packet, by looking at which NIC handled the packet.
		eps map[tcpip.NICID]*endpoint
	}

	ids    []uint32
	hashIV uint32

	// defaultTTL is the current default TTL for the protocol. Only the
	// uint8 portion of it is meaningful.
	//
	// Must be accessed using atomic operations.
	defaultTTL uint32

	// forwarding is set to 1 when the protocol has forwarding enabled and 0
	// when it is disabled.
	//
	// Must be accessed using atomic operations.
	forwarding uint32

	fragmentation *fragmentation.Fragmentation
}

// Number returns the ipv6 protocol number.
func (p *protocol) Number() tcpip.NetworkProtocolNumber {
	return ProtocolNumber
}

// MinimumPacketSize returns the minimum valid ipv6 packet size.
func (p *protocol) MinimumPacketSize() int {
	return header.IPv6MinimumSize
}

// DefaultPrefixLen returns the IPv6 default prefix length.
func (p *protocol) DefaultPrefixLen() int {
	return header.IPv6AddressSize * 8
}

// ParseAddresses implements NetworkProtocol.ParseAddresses.
func (*protocol) ParseAddresses(v buffer.View) (src, dst tcpip.Address) {
	h := header.IPv6(v)
	return h.SourceAddress(), h.DestinationAddress()
}

// NewEndpoint creates a new ipv6 endpoint.
func (p *protocol) NewEndpoint(nic stack.NetworkInterface, linkAddrCache stack.LinkAddressCache, nud stack.NUDHandler, dispatcher stack.TransportDispatcher) stack.NetworkEndpoint {
	e := &endpoint{
		nic:           nic,
		linkAddrCache: linkAddrCache,
		nud:           nud,
		dispatcher:    dispatcher,
		protocol:      p,
	}
	e.mu.Lock()
	e.mu.addressableEndpointState.Init(e)
	e.mu.ndp.init(e)
	e.mu.mld.init(e)
	e.mu.Unlock()

	stackStats := p.stack.Stats()
	tcpip.InitStatCounters(reflect.ValueOf(&e.stats.localStats).Elem())
	e.stats.ip.Init(&e.stats.localStats.IP, &stackStats.IP)
	e.stats.icmp.init(&e.stats.localStats.ICMP, &stackStats.ICMP.V6)

	p.mu.Lock()
	defer p.mu.Unlock()
	p.mu.eps[nic.ID()] = e
	return e
}

func (p *protocol) forgetEndpoint(nicID tcpip.NICID) {
	p.mu.Lock()
	defer p.mu.Unlock()
	delete(p.mu.eps, nicID)
}

// SetOption implements NetworkProtocol.SetOption.
func (p *protocol) SetOption(option tcpip.SettableNetworkProtocolOption) tcpip.Error {
	switch v := option.(type) {
	case *tcpip.DefaultTTLOption:
		p.SetDefaultTTL(uint8(*v))
		return nil
	default:
		return &tcpip.ErrUnknownProtocolOption{}
	}
}

// Option implements NetworkProtocol.Option.
func (p *protocol) Option(option tcpip.GettableNetworkProtocolOption) tcpip.Error {
	switch v := option.(type) {
	case *tcpip.DefaultTTLOption:
		*v = tcpip.DefaultTTLOption(p.DefaultTTL())
		return nil
	default:
		return &tcpip.ErrUnknownProtocolOption{}
	}
}

// SetDefaultTTL sets the default TTL for endpoints created with this protocol.
func (p *protocol) SetDefaultTTL(ttl uint8) {
	atomic.StoreUint32(&p.defaultTTL, uint32(ttl))
}

// DefaultTTL returns the default TTL for endpoints created with this protocol.
func (p *protocol) DefaultTTL() uint8 {
	return uint8(atomic.LoadUint32(&p.defaultTTL))
}

// Close implements stack.TransportProtocol.Close.
func (*protocol) Close() {}

// Wait implements stack.TransportProtocol.Wait.
func (*protocol) Wait() {}

// Parse implements stack.NetworkProtocol.Parse.
func (*protocol) Parse(pkt *stack.PacketBuffer) (proto tcpip.TransportProtocolNumber, hasTransportHdr bool, ok bool) {
	proto, _, fragOffset, fragMore, ok := parse.IPv6(pkt)
	if !ok {
		return 0, false, false
	}

	return proto, !fragMore && fragOffset == 0, true
}

// Forwarding implements stack.ForwardingNetworkProtocol.
func (p *protocol) Forwarding() bool {
	return uint8(atomic.LoadUint32(&p.forwarding)) == 1
}

// setForwarding sets the forwarding status for the protocol.
//
// Returns true if the forwarding status was updated.
func (p *protocol) setForwarding(v bool) bool {
	if v {
		return atomic.SwapUint32(&p.forwarding, 1) == 0
	}
	return atomic.SwapUint32(&p.forwarding, 0) == 1
}

// SetForwarding implements stack.ForwardingNetworkProtocol.
func (p *protocol) SetForwarding(v bool) {
	p.mu.Lock()
	defer p.mu.Unlock()

	if !p.setForwarding(v) {
		return
	}

	for _, ep := range p.mu.eps {
		ep.transitionForwarding(v)
	}
}

// calculateNetworkMTU calculates the network-layer payload MTU based on the
// link-layer payload MTU and the length of every IPv6 header.
// Note that this is different than the Payload Length field of the IPv6 header,
// which includes the length of the extension headers.
func calculateNetworkMTU(linkMTU, networkHeadersLen uint32) (uint32, tcpip.Error) {
	if linkMTU < header.IPv6MinimumMTU {
		return 0, &tcpip.ErrInvalidEndpointState{}
	}

	// As per RFC 7112 section 5, we should discard packets if their IPv6 header
	// is bigger than 1280 bytes (ie, the minimum link MTU) since we do not
	// support PMTU discovery:
	//   Hosts that do not discover the Path MTU MUST limit the IPv6 Header Chain
	//   length to 1280 bytes.  Limiting the IPv6 Header Chain length to 1280
	//   bytes ensures that the header chain length does not exceed the IPv6
	//   minimum MTU.
	if networkHeadersLen > header.IPv6MinimumMTU {
		return 0, &tcpip.ErrMalformedHeader{}
	}

	networkMTU := linkMTU - uint32(networkHeadersLen)
	if networkMTU > maxPayloadSize {
		networkMTU = maxPayloadSize
	}
	return networkMTU, nil
}

// Options holds options to configure a new protocol.
type Options struct {
	// NDPConfigs is the default NDP configurations used by interfaces.
	NDPConfigs NDPConfigurations

	// AutoGenLinkLocal determines whether or not the stack attempts to
	// auto-generate a link-local address for newly enabled non-loopback
	// NICs.
	//
	// Note, setting this to true does not mean that a link-local address is
	// assigned right away, or at all. If Duplicate Address Detection is enabled,
	// an address is only assigned if it successfully resolves. If it fails, no
	// further attempts are made to auto-generate a link-local adddress.
	//
	// The generated link-local address follows RFC 4291 Appendix A guidelines.
	AutoGenLinkLocal bool

	// NDPDisp is the NDP event dispatcher that an integrator can provide to
	// receive NDP related events.
	NDPDisp NDPDispatcher

	// OpaqueIIDOpts hold the options for generating opaque interface
	// identifiers (IIDs) as outlined by RFC 7217.
	OpaqueIIDOpts OpaqueInterfaceIdentifierOptions

	// TempIIDSeed is used to seed the initial temporary interface identifier
	// history value used to generate IIDs for temporary SLAAC addresses.
	//
	// Temporary SLAAC adresses are short-lived addresses which are unpredictable
	// and random from the perspective of other nodes on the network. It is
	// recommended that the seed be a random byte buffer of at least
	// header.IIDSize bytes to make sure that temporary SLAAC addresses are
	// sufficiently random. It should follow minimum randomness requirements for
	// security as outlined by RFC 4086.
	//
	// Note: using a nil value, the same seed across netstack program runs, or a
	// seed that is too small would reduce randomness and increase predictability,
	// defeating the purpose of temporary SLAAC addresses.
	TempIIDSeed []byte

	// MLD holds options for MLD.
	MLD MLDOptions
}

// NewProtocolWithOptions returns an IPv6 network protocol.
func NewProtocolWithOptions(opts Options) stack.NetworkProtocolFactory {
	opts.NDPConfigs.validate()

	ids := hash.RandN32(buckets)
	hashIV := hash.RandN32(1)[0]

	return func(s *stack.Stack) stack.NetworkProtocol {
		p := &protocol{
			stack:   s,
			options: opts,

			ids:    ids,
			hashIV: hashIV,
		}
		p.fragmentation = fragmentation.NewFragmentation(header.IPv6FragmentExtHdrFragmentOffsetBytesPerUnit, fragmentation.HighFragThreshold, fragmentation.LowFragThreshold, ReassembleTimeout, s.Clock(), p)
		p.mu.eps = make(map[tcpip.NICID]*endpoint)
		p.SetDefaultTTL(DefaultTTL)
		return p
	}
}

// NewProtocol is equivalent to NewProtocolWithOptions with an empty Options.
func NewProtocol(s *stack.Stack) stack.NetworkProtocol {
	return NewProtocolWithOptions(Options{})(s)
}

func calculateFragmentReserve(pkt *stack.PacketBuffer) int {
	return pkt.AvailableHeaderBytes() + pkt.NetworkHeader().View().Size() + header.IPv6FragmentHeaderSize
}

// hashRoute calculates a hash value for the given route. It uses the source &
// destination address and 32-bit number to generate the hash.
func hashRoute(r *stack.Route, hashIV uint32) uint32 {
	// The FNV-1a was chosen because it is a fast hashing algorithm, and
	// cryptographic properties are not needed here.
	h := fnv.New32a()
	if _, err := h.Write([]byte(r.LocalAddress)); err != nil {
		panic(fmt.Sprintf("Hash.Write: %s, but Hash' implementation of Write is not expected to ever return an error", err))
	}

	if _, err := h.Write([]byte(r.RemoteAddress)); err != nil {
		panic(fmt.Sprintf("Hash.Write: %s, but Hash' implementation of Write is not expected to ever return an error", err))
	}

	s := make([]byte, 4)
	binary.LittleEndian.PutUint32(s, hashIV)
	if _, err := h.Write(s); err != nil {
		panic(fmt.Sprintf("Hash.Write: %s, but Hash' implementation of Write is not expected ever to return an error", err))
	}

	return h.Sum32()
}

func buildNextFragment(pf *fragmentation.PacketFragmenter, originalIPHeaders header.IPv6, transportProto tcpip.TransportProtocolNumber, id uint32) (*stack.PacketBuffer, bool) {
	fragPkt, offset, copied, more := pf.BuildNextFragment()
	fragPkt.NetworkProtocolNumber = ProtocolNumber

	originalIPHeadersLength := len(originalIPHeaders)

	s := header.IPv6ExtHdrSerializer{&header.IPv6SerializableFragmentExtHdr{
		FragmentOffset: uint16(offset / header.IPv6FragmentExtHdrFragmentOffsetBytesPerUnit),
		M:              more,
		Identification: id,
	}}

	fragmentIPHeadersLength := originalIPHeadersLength + s.Length()
	fragmentIPHeaders := header.IPv6(fragPkt.NetworkHeader().Push(fragmentIPHeadersLength))

	// Copy the IPv6 header and any extension headers already populated.
	if copied := copy(fragmentIPHeaders, originalIPHeaders); copied != originalIPHeadersLength {
		panic(fmt.Sprintf("wrong number of bytes copied into fragmentIPHeaders: got %d, want %d", copied, originalIPHeadersLength))
	}

	nextHeader, _ := s.Serialize(transportProto, fragmentIPHeaders[originalIPHeadersLength:])

	fragmentIPHeaders.SetNextHeader(nextHeader)
	fragmentIPHeaders.SetPayloadLength(uint16(copied + fragmentIPHeadersLength - header.IPv6MinimumSize))

	return fragPkt, more
}