summaryrefslogtreecommitdiffstats
path: root/sys/netinet/tcp_syncache.c
blob: 781e01b978978bcd9705a32928cdbce7b2393fde (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
/*-
 * Copyright (c) 2001 McAfee, Inc.
 * Copyright (c) 2006,2013 Andre Oppermann, Internet Business Solutions AG
 * All rights reserved.
 *
 * This software was developed for the FreeBSD Project by Jonathan Lemon
 * and McAfee Research, the Security Research Division of McAfee, Inc. under
 * DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as part of the
 * DARPA CHATS research program. [2001 McAfee, Inc.]
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 */

#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");

#include "opt_inet.h"
#include "opt_inet6.h"
#include "opt_ipsec.h"
#include "opt_pcbgroup.h"

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/limits.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/proc.h>		/* for proc0 declaration */
#include <sys/random.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/syslog.h>
#include <sys/ucred.h>

#include <sys/md5.h>
#include <crypto/siphash/siphash.h>

#include <vm/uma.h>

#include <net/if.h>
#include <net/if_var.h>
#include <net/route.h>
#include <net/vnet.h>

#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet/ip.h>
#include <netinet/in_var.h>
#include <netinet/in_pcb.h>
#include <netinet/ip_var.h>
#include <netinet/ip_options.h>
#ifdef INET6
#include <netinet/ip6.h>
#include <netinet/icmp6.h>
#include <netinet6/nd6.h>
#include <netinet6/ip6_var.h>
#include <netinet6/in6_pcb.h>
#endif
#include <netinet/tcp.h>
#include <netinet/tcp_fsm.h>
#include <netinet/tcp_seq.h>
#include <netinet/tcp_timer.h>
#include <netinet/tcp_var.h>
#include <netinet/tcp_syncache.h>
#ifdef INET6
#include <netinet6/tcp6_var.h>
#endif
#ifdef TCP_OFFLOAD
#include <netinet/toecore.h>
#endif

#ifdef IPSEC
#include <netipsec/ipsec.h>
#ifdef INET6
#include <netipsec/ipsec6.h>
#endif
#include <netipsec/key.h>
#endif /*IPSEC*/

#include <machine/in_cksum.h>

#include <security/mac/mac_framework.h>

static VNET_DEFINE(int, tcp_syncookies) = 1;
#define	V_tcp_syncookies		VNET(tcp_syncookies)
SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, syncookies, CTLFLAG_RW,
    &VNET_NAME(tcp_syncookies), 0,
    "Use TCP SYN cookies if the syncache overflows");

static VNET_DEFINE(int, tcp_syncookiesonly) = 0;
#define	V_tcp_syncookiesonly		VNET(tcp_syncookiesonly)
SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, syncookies_only, CTLFLAG_RW,
    &VNET_NAME(tcp_syncookiesonly), 0,
    "Use only TCP SYN cookies");

#ifdef TCP_OFFLOAD
#define ADDED_BY_TOE(sc) ((sc)->sc_tod != NULL)
#endif

static void	 syncache_drop(struct syncache *, struct syncache_head *);
static void	 syncache_free(struct syncache *);
static void	 syncache_insert(struct syncache *, struct syncache_head *);
struct syncache *syncache_lookup(struct in_conninfo *, struct syncache_head **);
static int	 syncache_respond(struct syncache *);
static struct	 socket *syncache_socket(struct syncache *, struct socket *,
		    struct mbuf *m);
static void	 syncache_timeout(struct syncache *sc, struct syncache_head *sch,
		    int docallout);
static void	 syncache_timer(void *);

static uint32_t	 syncookie_mac(struct in_conninfo *, tcp_seq, uint8_t,
		    uint8_t *, uintptr_t);
static tcp_seq	 syncookie_generate(struct syncache_head *, struct syncache *);
static struct syncache
		*syncookie_lookup(struct in_conninfo *, struct syncache_head *,
		    struct syncache *, struct tcphdr *, struct tcpopt *,
		    struct socket *);
static void	 syncookie_reseed(void *);
#ifdef INVARIANTS
static int	 syncookie_cmp(struct in_conninfo *inc, struct syncache_head *sch,
		    struct syncache *sc, struct tcphdr *th, struct tcpopt *to,
		    struct socket *lso);
#endif

/*
 * Transmit the SYN,ACK fewer times than TCP_MAXRXTSHIFT specifies.
 * 3 retransmits corresponds to a timeout of 3 * (1 + 2 + 4 + 8) == 45 seconds,
 * the odds are that the user has given up attempting to connect by then.
 */
#define SYNCACHE_MAXREXMTS		3

/* Arbitrary values */
#define TCP_SYNCACHE_HASHSIZE		512
#define TCP_SYNCACHE_BUCKETLIMIT	30

static VNET_DEFINE(struct tcp_syncache, tcp_syncache);
#define	V_tcp_syncache			VNET(tcp_syncache)

static SYSCTL_NODE(_net_inet_tcp, OID_AUTO, syncache, CTLFLAG_RW, 0,
    "TCP SYN cache");

SYSCTL_VNET_UINT(_net_inet_tcp_syncache, OID_AUTO, bucketlimit, CTLFLAG_RDTUN,
    &VNET_NAME(tcp_syncache.bucket_limit), 0,
    "Per-bucket hash limit for syncache");

SYSCTL_VNET_UINT(_net_inet_tcp_syncache, OID_AUTO, cachelimit, CTLFLAG_RDTUN,
    &VNET_NAME(tcp_syncache.cache_limit), 0,
    "Overall entry limit for syncache");

SYSCTL_UMA_CUR(_net_inet_tcp_syncache, OID_AUTO, count, CTLFLAG_VNET,
    &VNET_NAME(tcp_syncache.zone), "Current number of entries in syncache");

SYSCTL_VNET_UINT(_net_inet_tcp_syncache, OID_AUTO, hashsize, CTLFLAG_RDTUN,
    &VNET_NAME(tcp_syncache.hashsize), 0,
    "Size of TCP syncache hashtable");

SYSCTL_VNET_UINT(_net_inet_tcp_syncache, OID_AUTO, rexmtlimit, CTLFLAG_RW,
    &VNET_NAME(tcp_syncache.rexmt_limit), 0,
    "Limit on SYN/ACK retransmissions");

VNET_DEFINE(int, tcp_sc_rst_sock_fail) = 1;
SYSCTL_VNET_INT(_net_inet_tcp_syncache, OID_AUTO, rst_on_sock_fail,
    CTLFLAG_RW, &VNET_NAME(tcp_sc_rst_sock_fail), 0,
    "Send reset on socket allocation failure");

static MALLOC_DEFINE(M_SYNCACHE, "syncache", "TCP syncache");

#define SYNCACHE_HASH(inc, mask)					\
	((V_tcp_syncache.hash_secret ^					\
	  (inc)->inc_faddr.s_addr ^					\
	  ((inc)->inc_faddr.s_addr >> 16) ^				\
	  (inc)->inc_fport ^ (inc)->inc_lport) & mask)

#define SYNCACHE_HASH6(inc, mask)					\
	((V_tcp_syncache.hash_secret ^					\
	  (inc)->inc6_faddr.s6_addr32[0] ^				\
	  (inc)->inc6_faddr.s6_addr32[3] ^				\
	  (inc)->inc_fport ^ (inc)->inc_lport) & mask)

#define ENDPTS_EQ(a, b) (						\
	(a)->ie_fport == (b)->ie_fport &&				\
	(a)->ie_lport == (b)->ie_lport &&				\
	(a)->ie_faddr.s_addr == (b)->ie_faddr.s_addr &&			\
	(a)->ie_laddr.s_addr == (b)->ie_laddr.s_addr			\
)

#define ENDPTS6_EQ(a, b) (memcmp(a, b, sizeof(*a)) == 0)

#define	SCH_LOCK(sch)		mtx_lock(&(sch)->sch_mtx)
#define	SCH_UNLOCK(sch)		mtx_unlock(&(sch)->sch_mtx)
#define	SCH_LOCK_ASSERT(sch)	mtx_assert(&(sch)->sch_mtx, MA_OWNED)

/*
 * Requires the syncache entry to be already removed from the bucket list.
 */
static void
syncache_free(struct syncache *sc)
{

	if (sc->sc_ipopts)
		(void) m_free(sc->sc_ipopts);
	if (sc->sc_cred)
		crfree(sc->sc_cred);
#ifdef MAC
	mac_syncache_destroy(&sc->sc_label);
#endif

	uma_zfree(V_tcp_syncache.zone, sc);
}

void
syncache_init(void)
{
	int i;

	V_tcp_syncache.hashsize = TCP_SYNCACHE_HASHSIZE;
	V_tcp_syncache.bucket_limit = TCP_SYNCACHE_BUCKETLIMIT;
	V_tcp_syncache.rexmt_limit = SYNCACHE_MAXREXMTS;
	V_tcp_syncache.hash_secret = arc4random();

	TUNABLE_INT_FETCH("net.inet.tcp.syncache.hashsize",
	    &V_tcp_syncache.hashsize);
	TUNABLE_INT_FETCH("net.inet.tcp.syncache.bucketlimit",
	    &V_tcp_syncache.bucket_limit);
	if (!powerof2(V_tcp_syncache.hashsize) ||
	    V_tcp_syncache.hashsize == 0) {
		printf("WARNING: syncache hash size is not a power of 2.\n");
		V_tcp_syncache.hashsize = TCP_SYNCACHE_HASHSIZE;
	}
	V_tcp_syncache.hashmask = V_tcp_syncache.hashsize - 1;

	/* Set limits. */
	V_tcp_syncache.cache_limit =
	    V_tcp_syncache.hashsize * V_tcp_syncache.bucket_limit;
	TUNABLE_INT_FETCH("net.inet.tcp.syncache.cachelimit",
	    &V_tcp_syncache.cache_limit);

	/* Allocate the hash table. */
	V_tcp_syncache.hashbase = malloc(V_tcp_syncache.hashsize *
	    sizeof(struct syncache_head), M_SYNCACHE, M_WAITOK | M_ZERO);

#ifdef VIMAGE
	V_tcp_syncache.vnet = curvnet;
#endif

	/* Initialize the hash buckets. */
	for (i = 0; i < V_tcp_syncache.hashsize; i++) {
		TAILQ_INIT(&V_tcp_syncache.hashbase[i].sch_bucket);
		mtx_init(&V_tcp_syncache.hashbase[i].sch_mtx, "tcp_sc_head",
			 NULL, MTX_DEF);
		callout_init_mtx(&V_tcp_syncache.hashbase[i].sch_timer,
			 &V_tcp_syncache.hashbase[i].sch_mtx, 0);
		V_tcp_syncache.hashbase[i].sch_length = 0;
		V_tcp_syncache.hashbase[i].sch_sc = &V_tcp_syncache;
	}

	/* Create the syncache entry zone. */
	V_tcp_syncache.zone = uma_zcreate("syncache", sizeof(struct syncache),
	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
	V_tcp_syncache.cache_limit = uma_zone_set_max(V_tcp_syncache.zone,
	    V_tcp_syncache.cache_limit);

	/* Start the SYN cookie reseeder callout. */
	callout_init(&V_tcp_syncache.secret.reseed, 1);
	arc4rand(V_tcp_syncache.secret.key[0], SYNCOOKIE_SECRET_SIZE, 0);
	arc4rand(V_tcp_syncache.secret.key[1], SYNCOOKIE_SECRET_SIZE, 0);
	callout_reset(&V_tcp_syncache.secret.reseed, SYNCOOKIE_LIFETIME * hz,
	    syncookie_reseed, &V_tcp_syncache);
}

#ifdef VIMAGE
void
syncache_destroy(void)
{
	struct syncache_head *sch;
	struct syncache *sc, *nsc;
	int i;

	/* Cleanup hash buckets: stop timers, free entries, destroy locks. */
	for (i = 0; i < V_tcp_syncache.hashsize; i++) {

		sch = &V_tcp_syncache.hashbase[i];
		callout_drain(&sch->sch_timer);

		SCH_LOCK(sch);
		TAILQ_FOREACH_SAFE(sc, &sch->sch_bucket, sc_hash, nsc)
			syncache_drop(sc, sch);
		SCH_UNLOCK(sch);
		KASSERT(TAILQ_EMPTY(&sch->sch_bucket),
		    ("%s: sch->sch_bucket not empty", __func__));
		KASSERT(sch->sch_length == 0, ("%s: sch->sch_length %d not 0",
		    __func__, sch->sch_length));
		mtx_destroy(&sch->sch_mtx);
	}

	KASSERT(uma_zone_get_cur(V_tcp_syncache.zone) == 0,
	    ("%s: cache_count not 0", __func__));

	/* Free the allocated global resources. */
	uma_zdestroy(V_tcp_syncache.zone);
	free(V_tcp_syncache.hashbase, M_SYNCACHE);

	callout_drain(&V_tcp_syncache.secret.reseed);
}
#endif

/*
 * Inserts a syncache entry into the specified bucket row.
 * Locks and unlocks the syncache_head autonomously.
 */
static void
syncache_insert(struct syncache *sc, struct syncache_head *sch)
{
	struct syncache *sc2;

	SCH_LOCK(sch);

	/*
	 * Make sure that we don't overflow the per-bucket limit.
	 * If the bucket is full, toss the oldest element.
	 */
	if (sch->sch_length >= V_tcp_syncache.bucket_limit) {
		KASSERT(!TAILQ_EMPTY(&sch->sch_bucket),
			("sch->sch_length incorrect"));
		sc2 = TAILQ_LAST(&sch->sch_bucket, sch_head);
		syncache_drop(sc2, sch);
		TCPSTAT_INC(tcps_sc_bucketoverflow);
	}

	/* Put it into the bucket. */
	TAILQ_INSERT_HEAD(&sch->sch_bucket, sc, sc_hash);
	sch->sch_length++;

#ifdef TCP_OFFLOAD
	if (ADDED_BY_TOE(sc)) {
		struct toedev *tod = sc->sc_tod;

		tod->tod_syncache_added(tod, sc->sc_todctx);
	}
#endif

	/* Reinitialize the bucket row's timer. */
	if (sch->sch_length == 1)
		sch->sch_nextc = ticks + INT_MAX;
	syncache_timeout(sc, sch, 1);

	SCH_UNLOCK(sch);

	TCPSTAT_INC(tcps_sc_added);
}

/*
 * Remove and free entry from syncache bucket row.
 * Expects locked syncache head.
 */
static void
syncache_drop(struct syncache *sc, struct syncache_head *sch)
{

	SCH_LOCK_ASSERT(sch);

	TAILQ_REMOVE(&sch->sch_bucket, sc, sc_hash);
	sch->sch_length--;

#ifdef TCP_OFFLOAD
	if (ADDED_BY_TOE(sc)) {
		struct toedev *tod = sc->sc_tod;

		tod->tod_syncache_removed(tod, sc->sc_todctx);
	}
#endif

	syncache_free(sc);
}

/*
 * Engage/reengage time on bucket row.
 */
static void
syncache_timeout(struct syncache *sc, struct syncache_head *sch, int docallout)
{
	sc->sc_rxttime = ticks +
		TCPTV_RTOBASE * (tcp_syn_backoff[sc->sc_rxmits]);
	sc->sc_rxmits++;
	if (TSTMP_LT(sc->sc_rxttime, sch->sch_nextc)) {
		sch->sch_nextc = sc->sc_rxttime;
		if (docallout)
			callout_reset(&sch->sch_timer, sch->sch_nextc - ticks,
			    syncache_timer, (void *)sch);
	}
}

/*
 * Walk the timer queues, looking for SYN,ACKs that need to be retransmitted.
 * If we have retransmitted an entry the maximum number of times, expire it.
 * One separate timer for each bucket row.
 */
static void
syncache_timer(void *xsch)
{
	struct syncache_head *sch = (struct syncache_head *)xsch;
	struct syncache *sc, *nsc;
	int tick = ticks;
	char *s;

	CURVNET_SET(sch->sch_sc->vnet);

	/* NB: syncache_head has already been locked by the callout. */
	SCH_LOCK_ASSERT(sch);

	/*
	 * In the following cycle we may remove some entries and/or
	 * advance some timeouts, so re-initialize the bucket timer.
	 */
	sch->sch_nextc = tick + INT_MAX;

	TAILQ_FOREACH_SAFE(sc, &sch->sch_bucket, sc_hash, nsc) {
		/*
		 * We do not check if the listen socket still exists
		 * and accept the case where the listen socket may be
		 * gone by the time we resend the SYN/ACK.  We do
		 * not expect this to happens often. If it does,
		 * then the RST will be sent by the time the remote
		 * host does the SYN/ACK->ACK.
		 */
		if (TSTMP_GT(sc->sc_rxttime, tick)) {
			if (TSTMP_LT(sc->sc_rxttime, sch->sch_nextc))
				sch->sch_nextc = sc->sc_rxttime;
			continue;
		}
		if (sc->sc_rxmits > V_tcp_syncache.rexmt_limit) {
			if ((s = tcp_log_addrs(&sc->sc_inc, NULL, NULL, NULL))) {
				log(LOG_DEBUG, "%s; %s: Retransmits exhausted, "
				    "giving up and removing syncache entry\n",
				    s, __func__);
				free(s, M_TCPLOG);
			}
			syncache_drop(sc, sch);
			TCPSTAT_INC(tcps_sc_stale);
			continue;
		}
		if ((s = tcp_log_addrs(&sc->sc_inc, NULL, NULL, NULL))) {
			log(LOG_DEBUG, "%s; %s: Response timeout, "
			    "retransmitting (%u) SYN|ACK\n",
			    s, __func__, sc->sc_rxmits);
			free(s, M_TCPLOG);
		}

		(void) syncache_respond(sc);
		TCPSTAT_INC(tcps_sc_retransmitted);
		syncache_timeout(sc, sch, 0);
	}
	if (!TAILQ_EMPTY(&(sch)->sch_bucket))
		callout_reset(&(sch)->sch_timer, (sch)->sch_nextc - tick,
			syncache_timer, (void *)(sch));
	CURVNET_RESTORE();
}

/*
 * Find an entry in the syncache.
 * Returns always with locked syncache_head plus a matching entry or NULL.
 */
struct syncache *
syncache_lookup(struct in_conninfo *inc, struct syncache_head **schp)
{
	struct syncache *sc;
	struct syncache_head *sch;

#ifdef INET6
	if (inc->inc_flags & INC_ISIPV6) {
		sch = &V_tcp_syncache.hashbase[
		    SYNCACHE_HASH6(inc, V_tcp_syncache.hashmask)];
		*schp = sch;

		SCH_LOCK(sch);

		/* Circle through bucket row to find matching entry. */
		TAILQ_FOREACH(sc, &sch->sch_bucket, sc_hash) {
			if (ENDPTS6_EQ(&inc->inc_ie, &sc->sc_inc.inc_ie))
				return (sc);
		}
	} else
#endif
	{
		sch = &V_tcp_syncache.hashbase[
		    SYNCACHE_HASH(inc, V_tcp_syncache.hashmask)];
		*schp = sch;

		SCH_LOCK(sch);

		/* Circle through bucket row to find matching entry. */
		TAILQ_FOREACH(sc, &sch->sch_bucket, sc_hash) {
#ifdef INET6
			if (sc->sc_inc.inc_flags & INC_ISIPV6)
				continue;
#endif
			if (ENDPTS_EQ(&inc->inc_ie, &sc->sc_inc.inc_ie))
				return (sc);
		}
	}
	SCH_LOCK_ASSERT(*schp);
	return (NULL);			/* always returns with locked sch */
}

/*
 * This function is called when we get a RST for a
 * non-existent connection, so that we can see if the
 * connection is in the syn cache.  If it is, zap it.
 */
void
syncache_chkrst(struct in_conninfo *inc, struct tcphdr *th)
{
	struct syncache *sc;
	struct syncache_head *sch;
	char *s = NULL;

	sc = syncache_lookup(inc, &sch);	/* returns locked sch */
	SCH_LOCK_ASSERT(sch);

	/*
	 * Any RST to our SYN|ACK must not carry ACK, SYN or FIN flags.
	 * See RFC 793 page 65, section SEGMENT ARRIVES.
	 */
	if (th->th_flags & (TH_ACK|TH_SYN|TH_FIN)) {
		if ((s = tcp_log_addrs(inc, th, NULL, NULL)))
			log(LOG_DEBUG, "%s; %s: Spurious RST with ACK, SYN or "
			    "FIN flag set, segment ignored\n", s, __func__);
		TCPSTAT_INC(tcps_badrst);
		goto done;
	}

	/*
	 * No corresponding connection was found in syncache.
	 * If syncookies are enabled and possibly exclusively
	 * used, or we are under memory pressure, a valid RST
	 * may not find a syncache entry.  In that case we're
	 * done and no SYN|ACK retransmissions will happen.
	 * Otherwise the RST was misdirected or spoofed.
	 */
	if (sc == NULL) {
		if ((s = tcp_log_addrs(inc, th, NULL, NULL)))
			log(LOG_DEBUG, "%s; %s: Spurious RST without matching "
			    "syncache entry (possibly syncookie only), "
			    "segment ignored\n", s, __func__);
		TCPSTAT_INC(tcps_badrst);
		goto done;
	}

	/*
	 * If the RST bit is set, check the sequence number to see
	 * if this is a valid reset segment.
	 * RFC 793 page 37:
	 *   In all states except SYN-SENT, all reset (RST) segments
	 *   are validated by checking their SEQ-fields.  A reset is
	 *   valid if its sequence number is in the window.
	 *
	 *   The sequence number in the reset segment is normally an
	 *   echo of our outgoing acknowlegement numbers, but some hosts
	 *   send a reset with the sequence number at the rightmost edge
	 *   of our receive window, and we have to handle this case.
	 */
	if (SEQ_GEQ(th->th_seq, sc->sc_irs) &&
	    SEQ_LEQ(th->th_seq, sc->sc_irs + sc->sc_wnd)) {
		syncache_drop(sc, sch);
		if ((s = tcp_log_addrs(inc, th, NULL, NULL)))
			log(LOG_DEBUG, "%s; %s: Our SYN|ACK was rejected, "
			    "connection attempt aborted by remote endpoint\n",
			    s, __func__);
		TCPSTAT_INC(tcps_sc_reset);
	} else {
		if ((s = tcp_log_addrs(inc, th, NULL, NULL)))
			log(LOG_DEBUG, "%s; %s: RST with invalid SEQ %u != "
			    "IRS %u (+WND %u), segment ignored\n",
			    s, __func__, th->th_seq, sc->sc_irs, sc->sc_wnd);
		TCPSTAT_INC(tcps_badrst);
	}

done:
	if (s != NULL)
		free(s, M_TCPLOG);
	SCH_UNLOCK(sch);
}

void
syncache_badack(struct in_conninfo *inc)
{
	struct syncache *sc;
	struct syncache_head *sch;

	sc = syncache_lookup(inc, &sch);	/* returns locked sch */
	SCH_LOCK_ASSERT(sch);
	if (sc != NULL) {
		syncache_drop(sc, sch);
		TCPSTAT_INC(tcps_sc_badack);
	}
	SCH_UNLOCK(sch);
}

void
syncache_unreach(struct in_conninfo *inc, struct tcphdr *th)
{
	struct syncache *sc;
	struct syncache_head *sch;

	sc = syncache_lookup(inc, &sch);	/* returns locked sch */
	SCH_LOCK_ASSERT(sch);
	if (sc == NULL)
		goto done;

	/* If the sequence number != sc_iss, then it's a bogus ICMP msg */
	if (ntohl(th->th_seq) != sc->sc_iss)
		goto done;

	/*
	 * If we've rertransmitted 3 times and this is our second error,
	 * we remove the entry.  Otherwise, we allow it to continue on.
	 * This prevents us from incorrectly nuking an entry during a
	 * spurious network outage.
	 *
	 * See tcp_notify().
	 */
	if ((sc->sc_flags & SCF_UNREACH) == 0 || sc->sc_rxmits < 3 + 1) {
		sc->sc_flags |= SCF_UNREACH;
		goto done;
	}
	syncache_drop(sc, sch);
	TCPSTAT_INC(tcps_sc_unreach);
done:
	SCH_UNLOCK(sch);
}

/*
 * Build a new TCP socket structure from a syncache entry.
 */
static struct socket *
syncache_socket(struct syncache *sc, struct socket *lso, struct mbuf *m)
{
	struct inpcb *inp = NULL;
	struct socket *so;
	struct tcpcb *tp;
	int error;
	char *s;

	INP_INFO_WLOCK_ASSERT(&V_tcbinfo);

	/*
	 * Ok, create the full blown connection, and set things up
	 * as they would have been set up if we had created the
	 * connection when the SYN arrived.  If we can't create
	 * the connection, abort it.
	 */
	so = sonewconn(lso, 0);
	if (so == NULL) {
		/*
		 * Drop the connection; we will either send a RST or
		 * have the peer retransmit its SYN again after its
		 * RTO and try again.
		 */
		TCPSTAT_INC(tcps_listendrop);
		if ((s = tcp_log_addrs(&sc->sc_inc, NULL, NULL, NULL))) {
			log(LOG_DEBUG, "%s; %s: Socket create failed "
			    "due to limits or memory shortage\n",
			    s, __func__);
			free(s, M_TCPLOG);
		}
		goto abort2;
	}
#ifdef MAC
	mac_socketpeer_set_from_mbuf(m, so);
#endif

	inp = sotoinpcb(so);
	inp->inp_inc.inc_fibnum = so->so_fibnum;
	INP_WLOCK(inp);
	INP_HASH_WLOCK(&V_tcbinfo);

	/* Insert new socket into PCB hash list. */
	inp->inp_inc.inc_flags = sc->sc_inc.inc_flags;
#ifdef INET6
	if (sc->sc_inc.inc_flags & INC_ISIPV6) {
		inp->in6p_laddr = sc->sc_inc.inc6_laddr;
	} else {
		inp->inp_vflag &= ~INP_IPV6;
		inp->inp_vflag |= INP_IPV4;
#endif
		inp->inp_laddr = sc->sc_inc.inc_laddr;
#ifdef INET6
	}
#endif

	/*
	 * If there's an mbuf and it has a flowid, then let's initialise the
	 * inp with that particular flowid.
	 */
	if (m != NULL && m->m_flags & M_FLOWID) {
		inp->inp_flags |= INP_HW_FLOWID;
		inp->inp_flags &= ~INP_SW_FLOWID;
		inp->inp_flowid = m->m_pkthdr.flowid;
	}

	/*
	 * Install in the reservation hash table for now, but don't yet
	 * install a connection group since the full 4-tuple isn't yet
	 * configured.
	 */
	inp->inp_lport = sc->sc_inc.inc_lport;
	if ((error = in_pcbinshash_nopcbgroup(inp)) != 0) {
		/*
		 * Undo the assignments above if we failed to
		 * put the PCB on the hash lists.
		 */
#ifdef INET6
		if (sc->sc_inc.inc_flags & INC_ISIPV6)
			inp->in6p_laddr = in6addr_any;
		else
#endif
			inp->inp_laddr.s_addr = INADDR_ANY;
		inp->inp_lport = 0;
		if ((s = tcp_log_addrs(&sc->sc_inc, NULL, NULL, NULL))) {
			log(LOG_DEBUG, "%s; %s: in_pcbinshash failed "
			    "with error %i\n",
			    s, __func__, error);
			free(s, M_TCPLOG);
		}
		INP_HASH_WUNLOCK(&V_tcbinfo);
		goto abort;
	}
#ifdef IPSEC
	/* Copy old policy into new socket's. */
	if (ipsec_copy_policy(sotoinpcb(lso)->inp_sp, inp->inp_sp))
		printf("syncache_socket: could not copy policy\n");
#endif
#ifdef INET6
	if (sc->sc_inc.inc_flags & INC_ISIPV6) {
		struct inpcb *oinp = sotoinpcb(lso);
		struct in6_addr laddr6;
		struct sockaddr_in6 sin6;
		/*
		 * Inherit socket options from the listening socket.
		 * Note that in6p_inputopts are not (and should not be)
		 * copied, since it stores previously received options and is
		 * used to detect if each new option is different than the
		 * previous one and hence should be passed to a user.
		 * If we copied in6p_inputopts, a user would not be able to
		 * receive options just after calling the accept system call.
		 */
		inp->inp_flags |= oinp->inp_flags & INP_CONTROLOPTS;
		if (oinp->in6p_outputopts)
			inp->in6p_outputopts =
			    ip6_copypktopts(oinp->in6p_outputopts, M_NOWAIT);

		sin6.sin6_family = AF_INET6;
		sin6.sin6_len = sizeof(sin6);
		sin6.sin6_addr = sc->sc_inc.inc6_faddr;
		sin6.sin6_port = sc->sc_inc.inc_fport;
		sin6.sin6_flowinfo = sin6.sin6_scope_id = 0;
		laddr6 = inp->in6p_laddr;
		if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
			inp->in6p_laddr = sc->sc_inc.inc6_laddr;
		if ((error = in6_pcbconnect_mbuf(inp, (struct sockaddr *)&sin6,
		    thread0.td_ucred, m)) != 0) {
			inp->in6p_laddr = laddr6;
			if ((s = tcp_log_addrs(&sc->sc_inc, NULL, NULL, NULL))) {
				log(LOG_DEBUG, "%s; %s: in6_pcbconnect failed "
				    "with error %i\n",
				    s, __func__, error);
				free(s, M_TCPLOG);
			}
			INP_HASH_WUNLOCK(&V_tcbinfo);
			goto abort;
		}
		/* Override flowlabel from in6_pcbconnect. */
		inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
		inp->inp_flow |= sc->sc_flowlabel;
	}
#endif /* INET6 */
#if defined(INET) && defined(INET6)
	else
#endif
#ifdef INET
	{
		struct in_addr laddr;
		struct sockaddr_in sin;

		inp->inp_options = (m) ? ip_srcroute(m) : NULL;
		
		if (inp->inp_options == NULL) {
			inp->inp_options = sc->sc_ipopts;
			sc->sc_ipopts = NULL;
		}

		sin.sin_family = AF_INET;
		sin.sin_len = sizeof(sin);
		sin.sin_addr = sc->sc_inc.inc_faddr;
		sin.sin_port = sc->sc_inc.inc_fport;
		bzero((caddr_t)sin.sin_zero, sizeof(sin.sin_zero));
		laddr = inp->inp_laddr;
		if (inp->inp_laddr.s_addr == INADDR_ANY)
			inp->inp_laddr = sc->sc_inc.inc_laddr;
		if ((error = in_pcbconnect_mbuf(inp, (struct sockaddr *)&sin,
		    thread0.td_ucred, m)) != 0) {
			inp->inp_laddr = laddr;
			if ((s = tcp_log_addrs(&sc->sc_inc, NULL, NULL, NULL))) {
				log(LOG_DEBUG, "%s; %s: in_pcbconnect failed "
				    "with error %i\n",
				    s, __func__, error);
				free(s, M_TCPLOG);
			}
			INP_HASH_WUNLOCK(&V_tcbinfo);
			goto abort;
		}
	}
#endif /* INET */
	INP_HASH_WUNLOCK(&V_tcbinfo);
	tp = intotcpcb(inp);
	tcp_state_change(tp, TCPS_SYN_RECEIVED);
	tp->iss = sc->sc_iss;
	tp->irs = sc->sc_irs;
	tcp_rcvseqinit(tp);
	tcp_sendseqinit(tp);
	tp->snd_wl1 = sc->sc_irs;
	tp->snd_max = tp->iss + 1;
	tp->snd_nxt = tp->iss + 1;
	tp->rcv_up = sc->sc_irs + 1;
	tp->rcv_wnd = sc->sc_wnd;
	tp->rcv_adv += tp->rcv_wnd;
	tp->last_ack_sent = tp->rcv_nxt;

	tp->t_flags = sototcpcb(lso)->t_flags & (TF_NOPUSH|TF_NODELAY);
	if (sc->sc_flags & SCF_NOOPT)
		tp->t_flags |= TF_NOOPT;
	else {
		if (sc->sc_flags & SCF_WINSCALE) {
			tp->t_flags |= TF_REQ_SCALE|TF_RCVD_SCALE;
			tp->snd_scale = sc->sc_requested_s_scale;
			tp->request_r_scale = sc->sc_requested_r_scale;
		}
		if (sc->sc_flags & SCF_TIMESTAMP) {
			tp->t_flags |= TF_REQ_TSTMP|TF_RCVD_TSTMP;
			tp->ts_recent = sc->sc_tsreflect;
			tp->ts_recent_age = tcp_ts_getticks();
			tp->ts_offset = sc->sc_tsoff;
		}
#ifdef TCP_SIGNATURE
		if (sc->sc_flags & SCF_SIGNATURE)
			tp->t_flags |= TF_SIGNATURE;
#endif
		if (sc->sc_flags & SCF_SACK)
			tp->t_flags |= TF_SACK_PERMIT;
	}

	if (sc->sc_flags & SCF_ECN)
		tp->t_flags |= TF_ECN_PERMIT;

	/*
	 * Set up MSS and get cached values from tcp_hostcache.
	 * This might overwrite some of the defaults we just set.
	 */
	tcp_mss(tp, sc->sc_peer_mss);

	/*
	 * If the SYN,ACK was retransmitted, indicate that CWND to be
	 * limited to one segment in cc_conn_init().
	 * NB: sc_rxmits counts all SYN,ACK transmits, not just retransmits.
	 */
	if (sc->sc_rxmits > 1)
		tp->snd_cwnd = 1;

#ifdef TCP_OFFLOAD
	/*
	 * Allow a TOE driver to install its hooks.  Note that we hold the
	 * pcbinfo lock too and that prevents tcp_usr_accept from accepting a
	 * new connection before the TOE driver has done its thing.
	 */
	if (ADDED_BY_TOE(sc)) {
		struct toedev *tod = sc->sc_tod;

		tod->tod_offload_socket(tod, sc->sc_todctx, so);
	}
#endif
	/*
	 * Copy and activate timers.
	 */
	tp->t_keepinit = sototcpcb(lso)->t_keepinit;
	tp->t_keepidle = sototcpcb(lso)->t_keepidle;
	tp->t_keepintvl = sototcpcb(lso)->t_keepintvl;
	tp->t_keepcnt = sototcpcb(lso)->t_keepcnt;
	tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));

	INP_WUNLOCK(inp);

	soisconnected(so);

	TCPSTAT_INC(tcps_accepts);
	return (so);

abort:
	INP_WUNLOCK(inp);
abort2:
	if (so != NULL)
		soabort(so);
	return (NULL);
}

/*
 * This function gets called when we receive an ACK for a
 * socket in the LISTEN state.  We look up the connection
 * in the syncache, and if its there, we pull it out of
 * the cache and turn it into a full-blown connection in
 * the SYN-RECEIVED state.
 */
int
syncache_expand(struct in_conninfo *inc, struct tcpopt *to, struct tcphdr *th,
    struct socket **lsop, struct mbuf *m)
{
	struct syncache *sc;
	struct syncache_head *sch;
	struct syncache scs;
	char *s;

	/*
	 * Global TCP locks are held because we manipulate the PCB lists
	 * and create a new socket.
	 */
	INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
	KASSERT((th->th_flags & (TH_RST|TH_ACK|TH_SYN)) == TH_ACK,
	    ("%s: can handle only ACK", __func__));

	sc = syncache_lookup(inc, &sch);	/* returns locked sch */
	SCH_LOCK_ASSERT(sch);

#ifdef INVARIANTS
	/*
	 * Test code for syncookies comparing the syncache stored
	 * values with the reconstructed values from the cookie.
	 */
	if (sc != NULL)
		syncookie_cmp(inc, sch, sc, th, to, *lsop);
#endif

	if (sc == NULL) {
		/*
		 * There is no syncache entry, so see if this ACK is
		 * a returning syncookie.  To do this, first:
		 *  A. See if this socket has had a syncache entry dropped in
		 *     the past.  We don't want to accept a bogus syncookie
		 *     if we've never received a SYN.
		 *  B. check that the syncookie is valid.  If it is, then
		 *     cobble up a fake syncache entry, and return.
		 */
		if (!V_tcp_syncookies) {
			SCH_UNLOCK(sch);
			if ((s = tcp_log_addrs(inc, th, NULL, NULL)))
				log(LOG_DEBUG, "%s; %s: Spurious ACK, "
				    "segment rejected (syncookies disabled)\n",
				    s, __func__);
			goto failed;
		}
		bzero(&scs, sizeof(scs));
		sc = syncookie_lookup(inc, sch, &scs, th, to, *lsop);
		SCH_UNLOCK(sch);
		if (sc == NULL) {
			if ((s = tcp_log_addrs(inc, th, NULL, NULL)))
				log(LOG_DEBUG, "%s; %s: Segment failed "
				    "SYNCOOKIE authentication, segment rejected "
				    "(probably spoofed)\n", s, __func__);
			goto failed;
		}
	} else {
		/* Pull out the entry to unlock the bucket row. */
		TAILQ_REMOVE(&sch->sch_bucket, sc, sc_hash);
		sch->sch_length--;
#ifdef TCP_OFFLOAD
		if (ADDED_BY_TOE(sc)) {
			struct toedev *tod = sc->sc_tod;

			tod->tod_syncache_removed(tod, sc->sc_todctx);
		}
#endif
		SCH_UNLOCK(sch);
	}

	/*
	 * Segment validation:
	 * ACK must match our initial sequence number + 1 (the SYN|ACK).
	 */
	if (th->th_ack != sc->sc_iss + 1) {
		if ((s = tcp_log_addrs(inc, th, NULL, NULL)))
			log(LOG_DEBUG, "%s; %s: ACK %u != ISS+1 %u, segment "
			    "rejected\n", s, __func__, th->th_ack, sc->sc_iss);
		goto failed;
	}

	/*
	 * The SEQ must fall in the window starting at the received
	 * initial receive sequence number + 1 (the SYN).
	 */
	if (SEQ_LEQ(th->th_seq, sc->sc_irs) ||
	    SEQ_GT(th->th_seq, sc->sc_irs + sc->sc_wnd)) {
		if ((s = tcp_log_addrs(inc, th, NULL, NULL)))
			log(LOG_DEBUG, "%s; %s: SEQ %u != IRS+1 %u, segment "
			    "rejected\n", s, __func__, th->th_seq, sc->sc_irs);
		goto failed;
	}

	/*
	 * If timestamps were not negotiated during SYN/ACK they
	 * must not appear on any segment during this session.
	 */
	if (!(sc->sc_flags & SCF_TIMESTAMP) && (to->to_flags & TOF_TS)) {
		if ((s = tcp_log_addrs(inc, th, NULL, NULL)))
			log(LOG_DEBUG, "%s; %s: Timestamp not expected, "
			    "segment rejected\n", s, __func__);
		goto failed;
	}

	/*
	 * If timestamps were negotiated during SYN/ACK they should
	 * appear on every segment during this session.
	 * XXXAO: This is only informal as there have been unverified
	 * reports of non-compliants stacks.
	 */
	if ((sc->sc_flags & SCF_TIMESTAMP) && !(to->to_flags & TOF_TS)) {
		if ((s = tcp_log_addrs(inc, th, NULL, NULL))) {
			log(LOG_DEBUG, "%s; %s: Timestamp missing, "
			    "no action\n", s, __func__);
			free(s, M_TCPLOG);
			s = NULL;
		}
	}

	/*
	 * If timestamps were negotiated the reflected timestamp
	 * must be equal to what we actually sent in the SYN|ACK.
	 */
	if ((to->to_flags & TOF_TS) && to->to_tsecr != sc->sc_ts) {
		if ((s = tcp_log_addrs(inc, th, NULL, NULL)))
			log(LOG_DEBUG, "%s; %s: TSECR %u != TS %u, "
			    "segment rejected\n",
			    s, __func__, to->to_tsecr, sc->sc_ts);
		goto failed;
	}

	*lsop = syncache_socket(sc, *lsop, m);

	if (*lsop == NULL)
		TCPSTAT_INC(tcps_sc_aborted);
	else
		TCPSTAT_INC(tcps_sc_completed);

/* how do we find the inp for the new socket? */
	if (sc != &scs)
		syncache_free(sc);
	return (1);
failed:
	if (sc != NULL && sc != &scs)
		syncache_free(sc);
	if (s != NULL)
		free(s, M_TCPLOG);
	*lsop = NULL;
	return (0);
}

/*
 * Given a LISTEN socket and an inbound SYN request, add
 * this to the syn cache, and send back a segment:
 *	<SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
 * to the source.
 *
 * IMPORTANT NOTE: We do _NOT_ ACK data that might accompany the SYN.
 * Doing so would require that we hold onto the data and deliver it
 * to the application.  However, if we are the target of a SYN-flood
 * DoS attack, an attacker could send data which would eventually
 * consume all available buffer space if it were ACKed.  By not ACKing
 * the data, we avoid this DoS scenario.
 */
void
syncache_add(struct in_conninfo *inc, struct tcpopt *to, struct tcphdr *th,
    struct inpcb *inp, struct socket **lsop, struct mbuf *m, void *tod,
    void *todctx)
{
	struct tcpcb *tp;
	struct socket *so;
	struct syncache *sc = NULL;
	struct syncache_head *sch;
	struct mbuf *ipopts = NULL;
	u_int ltflags;
	int win, sb_hiwat, ip_ttl, ip_tos;
	char *s;
#ifdef INET6
	int autoflowlabel = 0;
#endif
#ifdef MAC
	struct label *maclabel;
#endif
	struct syncache scs;
	struct ucred *cred;

	INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
	INP_WLOCK_ASSERT(inp);			/* listen socket */
	KASSERT((th->th_flags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN,
	    ("%s: unexpected tcp flags", __func__));

	/*
	 * Combine all so/tp operations very early to drop the INP lock as
	 * soon as possible.
	 */
	so = *lsop;
	tp = sototcpcb(so);
	cred = crhold(so->so_cred);

#ifdef INET6
	if ((inc->inc_flags & INC_ISIPV6) &&
	    (inp->inp_flags & IN6P_AUTOFLOWLABEL))
		autoflowlabel = 1;
#endif
	ip_ttl = inp->inp_ip_ttl;
	ip_tos = inp->inp_ip_tos;
	win = sbspace(&so->so_rcv);
	sb_hiwat = so->so_rcv.sb_hiwat;
	ltflags = (tp->t_flags & (TF_NOOPT | TF_SIGNATURE));

	/* By the time we drop the lock these should no longer be used. */
	so = NULL;
	tp = NULL;

#ifdef MAC
	if (mac_syncache_init(&maclabel) != 0) {
		INP_WUNLOCK(inp);
		INP_INFO_WUNLOCK(&V_tcbinfo);
		goto done;
	} else
		mac_syncache_create(maclabel, inp);
#endif
	INP_WUNLOCK(inp);
	INP_INFO_WUNLOCK(&V_tcbinfo);

	/*
	 * Remember the IP options, if any.
	 */
#ifdef INET6
	if (!(inc->inc_flags & INC_ISIPV6))
#endif
#ifdef INET
		ipopts = (m) ? ip_srcroute(m) : NULL;
#else
		ipopts = NULL;
#endif

	/*
	 * See if we already have an entry for this connection.
	 * If we do, resend the SYN,ACK, and reset the retransmit timer.
	 *
	 * XXX: should the syncache be re-initialized with the contents
	 * of the new SYN here (which may have different options?)
	 *
	 * XXX: We do not check the sequence number to see if this is a
	 * real retransmit or a new connection attempt.  The question is
	 * how to handle such a case; either ignore it as spoofed, or
	 * drop the current entry and create a new one?
	 */
	sc = syncache_lookup(inc, &sch);	/* returns locked entry */
	SCH_LOCK_ASSERT(sch);
	if (sc != NULL) {
		TCPSTAT_INC(tcps_sc_dupsyn);
		if (ipopts) {
			/*
			 * If we were remembering a previous source route,
			 * forget it and use the new one we've been given.
			 */
			if (sc->sc_ipopts)
				(void) m_free(sc->sc_ipopts);
			sc->sc_ipopts = ipopts;
		}
		/*
		 * Update timestamp if present.
		 */
		if ((sc->sc_flags & SCF_TIMESTAMP) && (to->to_flags & TOF_TS))
			sc->sc_tsreflect = to->to_tsval;
		else
			sc->sc_flags &= ~SCF_TIMESTAMP;
#ifdef MAC
		/*
		 * Since we have already unconditionally allocated label
		 * storage, free it up.  The syncache entry will already
		 * have an initialized label we can use.
		 */
		mac_syncache_destroy(&maclabel);
#endif
		/* Retransmit SYN|ACK and reset retransmit count. */
		if ((s = tcp_log_addrs(&sc->sc_inc, th, NULL, NULL))) {
			log(LOG_DEBUG, "%s; %s: Received duplicate SYN, "
			    "resetting timer and retransmitting SYN|ACK\n",
			    s, __func__);
			free(s, M_TCPLOG);
		}
		if (syncache_respond(sc) == 0) {
			sc->sc_rxmits = 0;
			syncache_timeout(sc, sch, 1);
			TCPSTAT_INC(tcps_sndacks);
			TCPSTAT_INC(tcps_sndtotal);
		}
		SCH_UNLOCK(sch);
		goto done;
	}

	sc = uma_zalloc(V_tcp_syncache.zone, M_NOWAIT | M_ZERO);
	if (sc == NULL) {
		/*
		 * The zone allocator couldn't provide more entries.
		 * Treat this as if the cache was full; drop the oldest
		 * entry and insert the new one.
		 */
		TCPSTAT_INC(tcps_sc_zonefail);
		if ((sc = TAILQ_LAST(&sch->sch_bucket, sch_head)) != NULL)
			syncache_drop(sc, sch);
		sc = uma_zalloc(V_tcp_syncache.zone, M_NOWAIT | M_ZERO);
		if (sc == NULL) {
			if (V_tcp_syncookies) {
				bzero(&scs, sizeof(scs));
				sc = &scs;
			} else {
				SCH_UNLOCK(sch);
				if (ipopts)
					(void) m_free(ipopts);
				goto done;
			}
		}
	}
	
	/*
	 * Fill in the syncache values.
	 */
#ifdef MAC
	sc->sc_label = maclabel;
#endif
	sc->sc_cred = cred;
	cred = NULL;
	sc->sc_ipopts = ipopts;
	bcopy(inc, &sc->sc_inc, sizeof(struct in_conninfo));
#ifdef INET6
	if (!(inc->inc_flags & INC_ISIPV6))
#endif
	{
		sc->sc_ip_tos = ip_tos;
		sc->sc_ip_ttl = ip_ttl;
	}
#ifdef TCP_OFFLOAD
	sc->sc_tod = tod;
	sc->sc_todctx = todctx;
#endif
	sc->sc_irs = th->th_seq;
	sc->sc_iss = arc4random();
	sc->sc_flags = 0;
	sc->sc_flowlabel = 0;

	/*
	 * Initial receive window: clip sbspace to [0 .. TCP_MAXWIN].
	 * win was derived from socket earlier in the function.
	 */
	win = imax(win, 0);
	win = imin(win, TCP_MAXWIN);
	sc->sc_wnd = win;

	if (V_tcp_do_rfc1323) {
		/*
		 * A timestamp received in a SYN makes
		 * it ok to send timestamp requests and replies.
		 */
		if (to->to_flags & TOF_TS) {
			sc->sc_tsreflect = to->to_tsval;
			sc->sc_ts = tcp_ts_getticks();
			sc->sc_flags |= SCF_TIMESTAMP;
		}
		if (to->to_flags & TOF_SCALE) {
			int wscale = 0;

			/*
			 * Pick the smallest possible scaling factor that
			 * will still allow us to scale up to sb_max, aka
			 * kern.ipc.maxsockbuf.
			 *
			 * We do this because there are broken firewalls that
			 * will corrupt the window scale option, leading to
			 * the other endpoint believing that our advertised
			 * window is unscaled.  At scale factors larger than
			 * 5 the unscaled window will drop below 1500 bytes,
			 * leading to serious problems when traversing these
			 * broken firewalls.
			 *
			 * With the default maxsockbuf of 256K, a scale factor
			 * of 3 will be chosen by this algorithm.  Those who
			 * choose a larger maxsockbuf should watch out
			 * for the compatiblity problems mentioned above.
			 *
			 * RFC1323: The Window field in a SYN (i.e., a <SYN>
			 * or <SYN,ACK>) segment itself is never scaled.
			 */
			while (wscale < TCP_MAX_WINSHIFT &&
			    (TCP_MAXWIN << wscale) < sb_max)
				wscale++;
			sc->sc_requested_r_scale = wscale;
			sc->sc_requested_s_scale = to->to_wscale;
			sc->sc_flags |= SCF_WINSCALE;
		}
	}
#ifdef TCP_SIGNATURE
	/*
	 * If listening socket requested TCP digests, and received SYN
	 * contains the option, flag this in the syncache so that
	 * syncache_respond() will do the right thing with the SYN+ACK.
	 * XXX: Currently we always record the option by default and will
	 * attempt to use it in syncache_respond().
	 */
	if (to->to_flags & TOF_SIGNATURE || ltflags & TF_SIGNATURE)
		sc->sc_flags |= SCF_SIGNATURE;
#endif
	if (to->to_flags & TOF_SACKPERM)
		sc->sc_flags |= SCF_SACK;
	if (to->to_flags & TOF_MSS)
		sc->sc_peer_mss = to->to_mss;	/* peer mss may be zero */
	if (ltflags & TF_NOOPT)
		sc->sc_flags |= SCF_NOOPT;
	if ((th->th_flags & (TH_ECE|TH_CWR)) && V_tcp_do_ecn)
		sc->sc_flags |= SCF_ECN;

	if (V_tcp_syncookies)
		sc->sc_iss = syncookie_generate(sch, sc);
#ifdef INET6
	if (autoflowlabel) {
		if (V_tcp_syncookies)
			sc->sc_flowlabel = sc->sc_iss;
		else
			sc->sc_flowlabel = ip6_randomflowlabel();
		sc->sc_flowlabel = htonl(sc->sc_flowlabel) & IPV6_FLOWLABEL_MASK;
	}
#endif
	SCH_UNLOCK(sch);

	/*
	 * Do a standard 3-way handshake.
	 */
	if (syncache_respond(sc) == 0) {
		if (V_tcp_syncookies && V_tcp_syncookiesonly && sc != &scs)
			syncache_free(sc);
		else if (sc != &scs)
			syncache_insert(sc, sch);   /* locks and unlocks sch */
		TCPSTAT_INC(tcps_sndacks);
		TCPSTAT_INC(tcps_sndtotal);
	} else {
		if (sc != &scs)
			syncache_free(sc);
		TCPSTAT_INC(tcps_sc_dropped);
	}

done:
	if (cred != NULL)
		crfree(cred);
#ifdef MAC
	if (sc == &scs)
		mac_syncache_destroy(&maclabel);
#endif
	if (m) {
		
		*lsop = NULL;
		m_freem(m);
	}
}

static int
syncache_respond(struct syncache *sc)
{
	struct ip *ip = NULL;
	struct mbuf *m;
	struct tcphdr *th = NULL;
	int optlen, error = 0;	/* Make compiler happy */
	u_int16_t hlen, tlen, mssopt;
	struct tcpopt to;
#ifdef INET6
	struct ip6_hdr *ip6 = NULL;
#endif

	hlen =
#ifdef INET6
	       (sc->sc_inc.inc_flags & INC_ISIPV6) ? sizeof(struct ip6_hdr) :
#endif
		sizeof(struct ip);
	tlen = hlen + sizeof(struct tcphdr);

	/* Determine MSS we advertize to other end of connection. */
	mssopt = tcp_mssopt(&sc->sc_inc);
	if (sc->sc_peer_mss)
		mssopt = max( min(sc->sc_peer_mss, mssopt), V_tcp_minmss);

	/* XXX: Assume that the entire packet will fit in a header mbuf. */
	KASSERT(max_linkhdr + tlen + TCP_MAXOLEN <= MHLEN,
	    ("syncache: mbuf too small"));

	/* Create the IP+TCP header from scratch. */
	m = m_gethdr(M_NOWAIT, MT_DATA);
	if (m == NULL)
		return (ENOBUFS);
#ifdef MAC
	mac_syncache_create_mbuf(sc->sc_label, m);
#endif
	m->m_data += max_linkhdr;
	m->m_len = tlen;
	m->m_pkthdr.len = tlen;
	m->m_pkthdr.rcvif = NULL;

#ifdef INET6
	if (sc->sc_inc.inc_flags & INC_ISIPV6) {
		ip6 = mtod(m, struct ip6_hdr *);
		ip6->ip6_vfc = IPV6_VERSION;
		ip6->ip6_nxt = IPPROTO_TCP;
		ip6->ip6_src = sc->sc_inc.inc6_laddr;
		ip6->ip6_dst = sc->sc_inc.inc6_faddr;
		ip6->ip6_plen = htons(tlen - hlen);
		/* ip6_hlim is set after checksum */
		ip6->ip6_flow &= ~IPV6_FLOWLABEL_MASK;
		ip6->ip6_flow |= sc->sc_flowlabel;

		th = (struct tcphdr *)(ip6 + 1);
	}
#endif
#if defined(INET6) && defined(INET)
	else
#endif
#ifdef INET
	{
		ip = mtod(m, struct ip *);
		ip->ip_v = IPVERSION;
		ip->ip_hl = sizeof(struct ip) >> 2;
		ip->ip_len = htons(tlen);
		ip->ip_id = 0;
		ip->ip_off = 0;
		ip->ip_sum = 0;
		ip->ip_p = IPPROTO_TCP;
		ip->ip_src = sc->sc_inc.inc_laddr;
		ip->ip_dst = sc->sc_inc.inc_faddr;
		ip->ip_ttl = sc->sc_ip_ttl;
		ip->ip_tos = sc->sc_ip_tos;

		/*
		 * See if we should do MTU discovery.  Route lookups are
		 * expensive, so we will only unset the DF bit if:
		 *
		 *	1) path_mtu_discovery is disabled
		 *	2) the SCF_UNREACH flag has been set
		 */
		if (V_path_mtu_discovery && ((sc->sc_flags & SCF_UNREACH) == 0))
		       ip->ip_off |= htons(IP_DF);

		th = (struct tcphdr *)(ip + 1);
	}
#endif /* INET */
	th->th_sport = sc->sc_inc.inc_lport;
	th->th_dport = sc->sc_inc.inc_fport;

	th->th_seq = htonl(sc->sc_iss);
	th->th_ack = htonl(sc->sc_irs + 1);
	th->th_off = sizeof(struct tcphdr) >> 2;
	th->th_x2 = 0;
	th->th_flags = TH_SYN|TH_ACK;
	th->th_win = htons(sc->sc_wnd);
	th->th_urp = 0;

	if (sc->sc_flags & SCF_ECN) {
		th->th_flags |= TH_ECE;
		TCPSTAT_INC(tcps_ecn_shs);
	}

	/* Tack on the TCP options. */
	if ((sc->sc_flags & SCF_NOOPT) == 0) {
		to.to_flags = 0;

		to.to_mss = mssopt;
		to.to_flags = TOF_MSS;
		if (sc->sc_flags & SCF_WINSCALE) {
			to.to_wscale = sc->sc_requested_r_scale;
			to.to_flags |= TOF_SCALE;
		}
		if (sc->sc_flags & SCF_TIMESTAMP) {
			/* Virgin timestamp or TCP cookie enhanced one. */
			to.to_tsval = sc->sc_ts;
			to.to_tsecr = sc->sc_tsreflect;
			to.to_flags |= TOF_TS;
		}
		if (sc->sc_flags & SCF_SACK)
			to.to_flags |= TOF_SACKPERM;
#ifdef TCP_SIGNATURE
		if (sc->sc_flags & SCF_SIGNATURE)
			to.to_flags |= TOF_SIGNATURE;
#endif
		optlen = tcp_addoptions(&to, (u_char *)(th + 1));

		/* Adjust headers by option size. */
		th->th_off = (sizeof(struct tcphdr) + optlen) >> 2;
		m->m_len += optlen;
		m->m_pkthdr.len += optlen;

#ifdef TCP_SIGNATURE
		if (sc->sc_flags & SCF_SIGNATURE)
			tcp_signature_compute(m, 0, 0, optlen,
			    to.to_signature, IPSEC_DIR_OUTBOUND);
#endif
#ifdef INET6
		if (sc->sc_inc.inc_flags & INC_ISIPV6)
			ip6->ip6_plen = htons(ntohs(ip6->ip6_plen) + optlen);
		else
#endif
			ip->ip_len = htons(ntohs(ip->ip_len) + optlen);
	} else
		optlen = 0;

	M_SETFIB(m, sc->sc_inc.inc_fibnum);
	m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
#ifdef INET6
	if (sc->sc_inc.inc_flags & INC_ISIPV6) {
		m->m_pkthdr.csum_flags = CSUM_TCP_IPV6;
		th->th_sum = in6_cksum_pseudo(ip6, tlen + optlen - hlen,
		    IPPROTO_TCP, 0);
		ip6->ip6_hlim = in6_selecthlim(NULL, NULL);
#ifdef TCP_OFFLOAD
		if (ADDED_BY_TOE(sc)) {
			struct toedev *tod = sc->sc_tod;

			error = tod->tod_syncache_respond(tod, sc->sc_todctx, m);

			return (error);
		}
#endif
		error = ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
	}
#endif
#if defined(INET6) && defined(INET)
	else
#endif
#ifdef INET
	{
		m->m_pkthdr.csum_flags = CSUM_TCP;
		th->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr,
		    htons(tlen + optlen - hlen + IPPROTO_TCP));
#ifdef TCP_OFFLOAD
		if (ADDED_BY_TOE(sc)) {
			struct toedev *tod = sc->sc_tod;

			error = tod->tod_syncache_respond(tod, sc->sc_todctx, m);

			return (error);
		}
#endif
		error = ip_output(m, sc->sc_ipopts, NULL, 0, NULL, NULL);
	}
#endif
	return (error);
}

/*
 * The purpose of syncookies is to handle spoofed SYN flooding DoS attacks
 * that exceed the capacity of the syncache by avoiding the storage of any
 * of the SYNs we receive.  Syncookies defend against blind SYN flooding
 * attacks where the attacker does not have access to our responses.
 *
 * Syncookies encode and include all necessary information about the
 * connection setup within the SYN|ACK that we send back.  That way we
 * can avoid keeping any local state until the ACK to our SYN|ACK returns
 * (if ever).  Normally the syncache and syncookies are running in parallel
 * with the latter taking over when the former is exhausted.  When matching
 * syncache entry is found the syncookie is ignored.
 *
 * The only reliable information persisting the 3WHS is our inital sequence
 * number ISS of 32 bits.  Syncookies embed a cryptographically sufficient
 * strong hash (MAC) value and a few bits of TCP SYN options in the ISS
 * of our SYN|ACK.  The MAC can be recomputed when the ACK to our SYN|ACK
 * returns and signifies a legitimate connection if it matches the ACK.
 *
 * The available space of 32 bits to store the hash and to encode the SYN
 * option information is very tight and we should have at least 24 bits for
 * the MAC to keep the number of guesses by blind spoofing reasonably high.
 *
 * SYN option information we have to encode to fully restore a connection:
 * MSS: is imporant to chose an optimal segment size to avoid IP level
 *   fragmentation along the path.  The common MSS values can be encoded
 *   in a 3-bit table.  Uncommon values are captured by the next lower value
 *   in the table leading to a slight increase in packetization overhead.
 * WSCALE: is necessary to allow large windows to be used for high delay-
 *   bandwidth product links.  Not scaling the window when it was initially
 *   negotiated is bad for performance as lack of scaling further decreases
 *   the apparent available send window.  We only need to encode the WSCALE
 *   we received from the remote end.  Our end can be recalculated at any
 *   time.  The common WSCALE values can be encoded in a 3-bit table.
 *   Uncommon values are captured by the next lower value in the table
 *   making us under-estimate the available window size halving our
 *   theoretically possible maximum throughput for that connection.
 * SACK: Greatly assists in packet loss recovery and requires 1 bit.
 * TIMESTAMP and SIGNATURE is not encoded because they are permanent options
 *   that are included in all segments on a connection.  We enable them when
 *   the ACK has them.
 *
 * Security of syncookies and attack vectors:
 *
 * The MAC is computed over (faddr||laddr||fport||lport||irs||flags||secmod)
 * together with the gloabl secret to make it unique per connection attempt.
 * Thus any change of any of those parameters results in a different MAC output
 * in an unpredictable way unless a collision is encountered.  24 bits of the
 * MAC are embedded into the ISS.
 *
 * To prevent replay attacks two rotating global secrets are updated with a
 * new random value every 15 seconds.  The life-time of a syncookie is thus
 * 15-30 seconds.
 *
 * Vector 1: Attacking the secret.  This requires finding a weakness in the
 * MAC itself or the way it is used here.  The attacker can do a chosen plain
 * text attack by varying and testing the all parameters under his control.
 * The strength depends on the size and randomness of the secret, and the
 * cryptographic security of the MAC function.  Due to the constant updating
 * of the secret the attacker has at most 29.999 seconds to find the secret
 * and launch spoofed connections.  After that he has to start all over again.
 *
 * Vector 2: Collision attack on the MAC of a single ACK.  With a 24 bit MAC
 * size an average of 4,823 attempts are required for a 50% chance of success
 * to spoof a single syncookie (birthday collision paradox).  However the
 * attacker is blind and doesn't know if one of his attempts succeeded unless
 * he has a side channel to interfere success from.  A single connection setup
 * success average of 90% requires 8,790 packets, 99.99% requires 17,578 packets.
 * This many attempts are required for each one blind spoofed connection.  For
 * every additional spoofed connection he has to launch another N attempts.
 * Thus for a sustained rate 100 spoofed connections per second approximately
 * 1,800,000 packets per second would have to be sent.
 *
 * NB: The MAC function should be fast so that it doesn't become a CPU
 * exhaustion attack vector itself.
 *
 * References:
 *  RFC4987 TCP SYN Flooding Attacks and Common Mitigations
 *  SYN cookies were first proposed by cryptographer Dan J. Bernstein in 1996
 *   http://cr.yp.to/syncookies.html    (overview)
 *   http://cr.yp.to/syncookies/archive (details)
 *
 *
 * Schematic construction of a syncookie enabled Initial Sequence Number:
 *  0        1         2         3
 *  12345678901234567890123456789012
 * |xxxxxxxxxxxxxxxxxxxxxxxxWWWMMMSP|
 *
 *  x 24 MAC (truncated)
 *  W  3 Send Window Scale index
 *  M  3 MSS index
 *  S  1 SACK permitted
 *  P  1 Odd/even secret
 */

/*
 * Distribution and probability of certain MSS values.  Those in between are
 * rounded down to the next lower one.
 * [An Analysis of TCP Maximum Segment Sizes, S. Alcock and R. Nelson, 2011]
 *                            .2%  .3%   5%    7%    7%    20%   15%   45%
 */
static int tcp_sc_msstab[] = { 216, 536, 1200, 1360, 1400, 1440, 1452, 1460 };

/*
 * Distribution and probability of certain WSCALE values.  We have to map the
 * (send) window scale (shift) option with a range of 0-14 from 4 bits into 3
 * bits based on prevalence of certain values.  Where we don't have an exact
 * match for are rounded down to the next lower one letting us under-estimate
 * the true available window.  At the moment this would happen only for the
 * very uncommon values 3, 5 and those above 8 (more than 16MB socket buffer
 * and window size).  The absence of the WSCALE option (no scaling in either
 * direction) is encoded with index zero.
 * [WSCALE values histograms, Allman, 2012]
 *                            X 10 10 35  5  6 14 10%   by host
 *                            X 11  4  5  5 18 49  3%   by connections
 */
static int tcp_sc_wstab[] = { 0, 0, 1, 2, 4, 6, 7, 8 };

/*
 * Compute the MAC for the SYN cookie.  SIPHASH-2-4 is chosen for its speed
 * and good cryptographic properties.
 */
static uint32_t
syncookie_mac(struct in_conninfo *inc, tcp_seq irs, uint8_t flags,
    uint8_t *secbits, uintptr_t secmod)
{
	SIPHASH_CTX ctx;
	uint32_t siphash[2];

	SipHash24_Init(&ctx);
	SipHash_SetKey(&ctx, secbits);
	switch (inc->inc_flags & INC_ISIPV6) {
#ifdef INET
	case 0:
		SipHash_Update(&ctx, &inc->inc_faddr, sizeof(inc->inc_faddr));
		SipHash_Update(&ctx, &inc->inc_laddr, sizeof(inc->inc_laddr));
		break;
#endif
#ifdef INET6
	case INC_ISIPV6:
		SipHash_Update(&ctx, &inc->inc6_faddr, sizeof(inc->inc6_faddr));
		SipHash_Update(&ctx, &inc->inc6_laddr, sizeof(inc->inc6_laddr));
		break;
#endif
	}
	SipHash_Update(&ctx, &inc->inc_fport, sizeof(inc->inc_fport));
	SipHash_Update(&ctx, &inc->inc_lport, sizeof(inc->inc_lport));
	SipHash_Update(&ctx, &flags, sizeof(flags));
	SipHash_Update(&ctx, &secmod, sizeof(secmod));
	SipHash_Final((u_int8_t *)&siphash, &ctx);

	return (siphash[0] ^ siphash[1]);
}

static tcp_seq
syncookie_generate(struct syncache_head *sch, struct syncache *sc)
{
	u_int i, mss, secbit, wscale;
	uint32_t iss, hash;
	uint8_t *secbits;
	union syncookie cookie;

	SCH_LOCK_ASSERT(sch);

	cookie.cookie = 0;

	/* Map our computed MSS into the 3-bit index. */
	mss = min(tcp_mssopt(&sc->sc_inc), max(sc->sc_peer_mss, V_tcp_minmss));
	for (i = sizeof(tcp_sc_msstab) / sizeof(*tcp_sc_msstab) - 1;
	     tcp_sc_msstab[i] > mss && i > 0;
	     i--)
		;
	cookie.flags.mss_idx = i;

	/*
	 * Map the send window scale into the 3-bit index but only if
	 * the wscale option was received.
	 */
	if (sc->sc_flags & SCF_WINSCALE) {
		wscale = sc->sc_requested_s_scale;
		for (i = sizeof(tcp_sc_wstab) / sizeof(*tcp_sc_wstab) - 1;
		     tcp_sc_wstab[i] > wscale && i > 0;
		     i--)
			;
		cookie.flags.wscale_idx = i;
	}

	/* Can we do SACK? */
	if (sc->sc_flags & SCF_SACK)
		cookie.flags.sack_ok = 1;

	/* Which of the two secrets to use. */
	secbit = sch->sch_sc->secret.oddeven & 0x1;
	cookie.flags.odd_even = secbit;

	secbits = sch->sch_sc->secret.key[secbit];
	hash = syncookie_mac(&sc->sc_inc, sc->sc_irs, cookie.cookie, secbits,
	    (uintptr_t)sch);

	/*
	 * Put the flags into the hash and XOR them to get better ISS number
	 * variance.  This doesn't enhance the cryptographic strength and is
	 * done to prevent the 8 cookie bits from showing up directly on the
	 * wire.
	 */
	iss = hash & ~0xff;
	iss |= cookie.cookie ^ (hash >> 24);

	/* Randomize the timestamp. */
	if (sc->sc_flags & SCF_TIMESTAMP) {
		sc->sc_ts = arc4random();
		sc->sc_tsoff = sc->sc_ts - tcp_ts_getticks();
	}

	TCPSTAT_INC(tcps_sc_sendcookie);
	return (iss);
}

static struct syncache *
syncookie_lookup(struct in_conninfo *inc, struct syncache_head *sch, 
    struct syncache *sc, struct tcphdr *th, struct tcpopt *to,
    struct socket *lso)
{
	uint32_t hash;
	uint8_t *secbits;
	tcp_seq ack, seq;
	int wnd, wscale = 0;
	union syncookie cookie;

	SCH_LOCK_ASSERT(sch);

	/*
	 * Pull information out of SYN-ACK/ACK and revert sequence number
	 * advances.
	 */
	ack = th->th_ack - 1;
	seq = th->th_seq - 1;

	/*
	 * Unpack the flags containing enough information to restore the
	 * connection.
	 */
	cookie.cookie = (ack & 0xff) ^ (ack >> 24);

	/* Which of the two secrets to use. */
	secbits = sch->sch_sc->secret.key[cookie.flags.odd_even];

	hash = syncookie_mac(inc, seq, cookie.cookie, secbits, (uintptr_t)sch);

	/* The recomputed hash matches the ACK if this was a genuine cookie. */
	if ((ack & ~0xff) != (hash & ~0xff))
		return (NULL);

	/* Fill in the syncache values. */
	sc->sc_flags = 0;
	bcopy(inc, &sc->sc_inc, sizeof(struct in_conninfo));
	sc->sc_ipopts = NULL;
	
	sc->sc_irs = seq;
	sc->sc_iss = ack;

	switch (inc->inc_flags & INC_ISIPV6) {
#ifdef INET
	case 0:
		sc->sc_ip_ttl = sotoinpcb(lso)->inp_ip_ttl;
		sc->sc_ip_tos = sotoinpcb(lso)->inp_ip_tos;
		break;
#endif
#ifdef INET6
	case INC_ISIPV6:
		if (sotoinpcb(lso)->inp_flags & IN6P_AUTOFLOWLABEL)
			sc->sc_flowlabel = sc->sc_iss & IPV6_FLOWLABEL_MASK;
		break;
#endif
	}

	sc->sc_peer_mss = tcp_sc_msstab[cookie.flags.mss_idx];

	/* We can simply recompute receive window scale we sent earlier. */
	while (wscale < TCP_MAX_WINSHIFT && (TCP_MAXWIN << wscale) < sb_max)
		wscale++;

	/* Only use wscale if it was enabled in the orignal SYN. */
	if (cookie.flags.wscale_idx > 0) {
		sc->sc_requested_r_scale = wscale;
		sc->sc_requested_s_scale = tcp_sc_wstab[cookie.flags.wscale_idx];
		sc->sc_flags |= SCF_WINSCALE;
	}

	wnd = sbspace(&lso->so_rcv);
	wnd = imax(wnd, 0);
	wnd = imin(wnd, TCP_MAXWIN);
	sc->sc_wnd = wnd;

	if (cookie.flags.sack_ok)
		sc->sc_flags |= SCF_SACK;

	if (to->to_flags & TOF_TS) {
		sc->sc_flags |= SCF_TIMESTAMP;
		sc->sc_tsreflect = to->to_tsval;
		sc->sc_ts = to->to_tsecr;
		sc->sc_tsoff = to->to_tsecr - tcp_ts_getticks();
	}

	if (to->to_flags & TOF_SIGNATURE)
		sc->sc_flags |= SCF_SIGNATURE;

	sc->sc_rxmits = 0;

	TCPSTAT_INC(tcps_sc_recvcookie);
	return (sc);
}

#ifdef INVARIANTS
static int
syncookie_cmp(struct in_conninfo *inc, struct syncache_head *sch,
    struct syncache *sc, struct tcphdr *th, struct tcpopt *to,
    struct socket *lso)
{
	struct syncache scs, *scx;
	char *s;

	bzero(&scs, sizeof(scs));
	scx = syncookie_lookup(inc, sch, &scs, th, to, lso);

	if ((s = tcp_log_addrs(inc, th, NULL, NULL)) == NULL)
		return (0);

	if (scx != NULL) {
		if (sc->sc_peer_mss != scx->sc_peer_mss)
			log(LOG_DEBUG, "%s; %s: mss different %i vs %i\n",
			    s, __func__, sc->sc_peer_mss, scx->sc_peer_mss);

		if (sc->sc_requested_r_scale != scx->sc_requested_r_scale)
			log(LOG_DEBUG, "%s; %s: rwscale different %i vs %i\n",
			    s, __func__, sc->sc_requested_r_scale,
			    scx->sc_requested_r_scale);

		if (sc->sc_requested_s_scale != scx->sc_requested_s_scale)
			log(LOG_DEBUG, "%s; %s: swscale different %i vs %i\n",
			    s, __func__, sc->sc_requested_s_scale,
			    scx->sc_requested_s_scale);

		if ((sc->sc_flags & SCF_SACK) != (scx->sc_flags & SCF_SACK))
			log(LOG_DEBUG, "%s; %s: SACK different\n", s, __func__);
	}

	if (s != NULL)
		free(s, M_TCPLOG);
	return (0);
}
#endif /* INVARIANTS */

static void
syncookie_reseed(void *arg)
{
	struct tcp_syncache *sc = arg;
	uint8_t *secbits;
	int secbit;

	/*
	 * Reseeding the secret doesn't have to be protected by a lock.
	 * It only must be ensured that the new random values are visible
	 * to all CPUs in a SMP environment.  The atomic with release
	 * semantics ensures that.
	 */
	secbit = (sc->secret.oddeven & 0x1) ? 0 : 1;
	secbits = sc->secret.key[secbit];
	arc4rand(secbits, SYNCOOKIE_SECRET_SIZE, 0);
	atomic_add_rel_int(&sc->secret.oddeven, 1);

	/* Reschedule ourself. */
	callout_schedule(&sc->secret.reseed, SYNCOOKIE_LIFETIME * hz);
}

/*
 * Returns the current number of syncache entries.  This number
 * will probably change before you get around to calling 
 * syncache_pcblist.
 */
int
syncache_pcbcount(void)
{
	struct syncache_head *sch;
	int count, i;

	for (count = 0, i = 0; i < V_tcp_syncache.hashsize; i++) {
		/* No need to lock for a read. */
		sch = &V_tcp_syncache.hashbase[i];
		count += sch->sch_length;
	}
	return count;
}

/*
 * Exports the syncache entries to userland so that netstat can display
 * them alongside the other sockets.  This function is intended to be
 * called only from tcp_pcblist.
 *
 * Due to concurrency on an active system, the number of pcbs exported
 * may have no relation to max_pcbs.  max_pcbs merely indicates the
 * amount of space the caller allocated for this function to use.
 */
int
syncache_pcblist(struct sysctl_req *req, int max_pcbs, int *pcbs_exported)
{
	struct xtcpcb xt;
	struct syncache *sc;
	struct syncache_head *sch;
	int count, error, i;

	for (count = 0, error = 0, i = 0; i < V_tcp_syncache.hashsize; i++) {
		sch = &V_tcp_syncache.hashbase[i];
		SCH_LOCK(sch);
		TAILQ_FOREACH(sc, &sch->sch_bucket, sc_hash) {
			if (count >= max_pcbs) {
				SCH_UNLOCK(sch);
				goto exit;
			}
			if (cr_cansee(req->td->td_ucred, sc->sc_cred) != 0)
				continue;
			bzero(&xt, sizeof(xt));
			xt.xt_len = sizeof(xt);
			if (sc->sc_inc.inc_flags & INC_ISIPV6)
				xt.xt_inp.inp_vflag = INP_IPV6;
			else
				xt.xt_inp.inp_vflag = INP_IPV4;
			bcopy(&sc->sc_inc, &xt.xt_inp.inp_inc, sizeof (struct in_conninfo));
			xt.xt_tp.t_inpcb = &xt.xt_inp;
			xt.xt_tp.t_state = TCPS_SYN_RECEIVED;
			xt.xt_socket.xso_protocol = IPPROTO_TCP;
			xt.xt_socket.xso_len = sizeof (struct xsocket);
			xt.xt_socket.so_type = SOCK_STREAM;
			xt.xt_socket.so_state = SS_ISCONNECTING;
			error = SYSCTL_OUT(req, &xt, sizeof xt);
			if (error) {
				SCH_UNLOCK(sch);
				goto exit;
			}
			count++;
		}
		SCH_UNLOCK(sch);
	}
exit:
	*pcbs_exported = count;
	return error;
}
OpenPOWER on IntegriCloud