summaryrefslogtreecommitdiffstats
path: root/src/plugins/mdata/mdata_doc.rst
blob: 95746bd3d0ea4298f06a2d74fe7a7474f4324469 (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
Buffer metadata change tracker
==============================

Introduction
------------

The mdata plugin uses the vlib main loop “before” performance counter
hook to snapshoot buffer metadata before calling the node dispatch
function. Similarly, the plugin uses the main loop “after” hook to
compare a vectors’ worth of buffer metadata after the fact.

The comparison function is a simple octet-by-octet A != B check. We
accumulate changed octets per-node across the entire run, using a single
spinlock-protected accumulator.

The “show buffer metadata” command produces a report of all fields whose
values are changed by nodes visited during a given run.

Since many fields in the vnet_buffer_opaque_t are union members, it may
appear that a certain node changes numerous fields. The entire point of
the exercise is to warn developers that if a packet visits node N, data
placed into opaque union field F *will* be affected.

One should never assume much about buffer metadata immutability across
arbitrary subgraphs. This tool generates accurate reports, to the extent
that one exercises the required subgraph trajectories.
340'>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
/*
 * Copyright (c) 2016 Cisco and/or its affiliates.
 * 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.
 */

#include <vnet/tcp/tcp.h>
#include <vnet/lisp-cp/packets.h>

vlib_node_registration_t tcp4_output_node;
vlib_node_registration_t tcp6_output_node;

typedef enum _tcp_output_nect
{
  TCP_OUTPUT_NEXT_DROP,
  TCP_OUTPUT_NEXT_IP_LOOKUP,
  TCP_OUTPUT_N_NEXT
} tcp_output_next_t;

#define foreach_tcp4_output_next              	\
  _ (DROP, "error-drop")                        \
  _ (IP_LOOKUP, "ip4-lookup")

#define foreach_tcp6_output_next              	\
  _ (DROP, "error-drop")                        \
  _ (IP_LOOKUP, "ip6-lookup")

static char *tcp_error_strings[] = {
#define tcp_error(n,s) s,
#include <vnet/tcp/tcp_error.def>
#undef tcp_error
};

typedef struct
{
  tcp_header_t tcp_header;
  tcp_connection_t tcp_connection;
} tcp_tx_trace_t;

u16 dummy_mtu = 1460;

u8 *
format_tcp_tx_trace (u8 * s, va_list * args)
{
  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
  tcp_tx_trace_t *t = va_arg (*args, tcp_tx_trace_t *);
  uword indent = format_get_indent (s);

  s = format (s, "%U\n%U%U",
	      format_tcp_header, &t->tcp_header, 128,
	      format_white_space, indent,
	      format_tcp_connection, &t->tcp_connection, 1);

  return s;
}

static u8
tcp_window_compute_scale (u32 available_space)
{
  u8 wnd_scale = 0;
  while (wnd_scale < TCP_MAX_WND_SCALE
	 && (available_space >> wnd_scale) > TCP_WND_MAX)
    wnd_scale++;
  return wnd_scale;
}

/**
 * TCP's IW as recommended by RFC6928
 */
always_inline u32
tcp_initial_wnd_unscaled (tcp_connection_t * tc)
{
  return TCP_IW_N_SEGMENTS * tc->mss;
}

/**
 * Compute initial window and scale factor. As per RFC1323, window field in
 * SYN and SYN-ACK segments is never scaled.
 */
u32
tcp_initial_window_to_advertise (tcp_connection_t * tc)
{
  u32 max_fifo;

  /* Initial wnd for SYN. Fifos are not allocated yet.
   * Use some predefined value. For SYN-ACK we still want the
   * scale to be computed in the same way */
  max_fifo = TCP_MAX_RX_FIFO_SIZE;

  tc->rcv_wscale = tcp_window_compute_scale (max_fifo);
  tc->rcv_wnd = tcp_initial_wnd_unscaled (tc);

  return clib_min (tc->rcv_wnd, TCP_WND_MAX);
}

/**
 * Compute and return window to advertise, scaled as per RFC1323
 */
u32
tcp_window_to_advertise (tcp_connection_t * tc, tcp_state_t state)
{
  if (state < TCP_STATE_ESTABLISHED)
    return tcp_initial_window_to_advertise (tc);

  tcp_update_rcv_wnd (tc);

  if (tc->rcv_wnd == 0)
    {
      tc->flags |= TCP_CONN_SENT_RCV_WND0;
    }
  else
    {
      tc->flags &= ~TCP_CONN_SENT_RCV_WND0;
    }

  return tc->rcv_wnd >> tc->rcv_wscale;
}

void
tcp_update_rcv_wnd (tcp_connection_t * tc)
{
  i32 observed_wnd;
  u32 available_space, max_fifo, wnd;

  /*
   * Figure out how much space we have available
   */
  available_space = stream_session_max_rx_enqueue (&tc->connection);
  max_fifo = stream_session_rx_fifo_size (&tc->connection);

  ASSERT (tc->rcv_opts.mss < max_fifo);
  if (available_space < tc->rcv_opts.mss && available_space < max_fifo >> 3)
    available_space = 0;

  /*
   * Use the above and what we know about what we've previously advertised
   * to compute the new window
   */
  observed_wnd = (i32) tc->rcv_wnd - (tc->rcv_nxt - tc->rcv_las);
  if (observed_wnd < 0)
    observed_wnd = 0;

  /* Bad. Thou shalt not shrink */
  if (available_space < observed_wnd)
    {
      wnd = observed_wnd;
      TCP_EVT_DBG (TCP_EVT_RCV_WND_SHRUNK, tc, observed_wnd, available_space);
    }
  else
    {
      wnd = available_space;
    }

  /* Make sure we have a multiple of rcv_wscale */
  if (wnd && tc->rcv_wscale)
    {
      wnd &= ~(1 << tc->rcv_wscale);
      if (wnd == 0)
	wnd = 1 << tc->rcv_wscale;
    }

  tc->rcv_wnd = clib_min (wnd, TCP_WND_MAX << tc->rcv_wscale);
}

/**
 * Write TCP options to segment.
 */
u32
tcp_options_write (u8 * data, tcp_options_t * opts)
{
  u32 opts_len = 0;
  u32 buf, seq_len = 4;

  if (tcp_opts_mss (opts))
    {
      *data++ = TCP_OPTION_MSS;
      *data++ = TCP_OPTION_LEN_MSS;
      buf = clib_host_to_net_u16 (opts->mss);
      clib_memcpy (data, &buf, sizeof (opts->mss));
      data += sizeof (opts->mss);
      opts_len += TCP_OPTION_LEN_MSS;
    }

  if (tcp_opts_wscale (opts))
    {
      *data++ = TCP_OPTION_WINDOW_SCALE;
      *data++ = TCP_OPTION_LEN_WINDOW_SCALE;
      *data++ = opts->wscale;
      opts_len += TCP_OPTION_LEN_WINDOW_SCALE;
    }

  if (tcp_opts_sack_permitted (opts))
    {
      *data++ = TCP_OPTION_SACK_PERMITTED;
      *data++ = TCP_OPTION_LEN_SACK_PERMITTED;
      opts_len += TCP_OPTION_LEN_SACK_PERMITTED;
    }

  if (tcp_opts_tstamp (opts))
    {
      *data++ = TCP_OPTION_TIMESTAMP;
      *data++ = TCP_OPTION_LEN_TIMESTAMP;
      buf = clib_host_to_net_u32 (opts->tsval);
      clib_memcpy (data, &buf, sizeof (opts->tsval));
      data += sizeof (opts->tsval);
      buf = clib_host_to_net_u32 (opts->tsecr);
      clib_memcpy (data, &buf, sizeof (opts->tsecr));
      data += sizeof (opts->tsecr);
      opts_len += TCP_OPTION_LEN_TIMESTAMP;
    }

  if (tcp_opts_sack (opts))
    {
      int i;
      u32 n_sack_blocks = clib_min (vec_len (opts->sacks),
				    TCP_OPTS_MAX_SACK_BLOCKS);

      if (n_sack_blocks != 0)
	{
	  *data++ = TCP_OPTION_SACK_BLOCK;
	  *data++ = 2 + n_sack_blocks * TCP_OPTION_LEN_SACK_BLOCK;
	  for (i = 0; i < n_sack_blocks; i++)
	    {
	      buf = clib_host_to_net_u32 (opts->sacks[i].start);
	      clib_memcpy (data, &buf, seq_len);
	      data += seq_len;
	      buf = clib_host_to_net_u32 (opts->sacks[i].end);
	      clib_memcpy (data, &buf, seq_len);
	      data += seq_len;
	    }
	  opts_len += 2 + n_sack_blocks * TCP_OPTION_LEN_SACK_BLOCK;
	}
    }

  /* Terminate TCP options */
  if (opts_len % 4)
    {
      *data++ = TCP_OPTION_EOL;
      opts_len += TCP_OPTION_LEN_EOL;
    }

  /* Pad with zeroes to a u32 boundary */
  while (opts_len % 4)
    {
      *data++ = TCP_OPTION_NOOP;
      opts_len += TCP_OPTION_LEN_NOOP;
    }
  return opts_len;
}

always_inline int
tcp_make_syn_options (tcp_options_t * opts, u8 wnd_scale)
{
  u8 len = 0;

  opts->flags |= TCP_OPTS_FLAG_MSS;
  opts->mss = dummy_mtu;	/*XXX discover that */
  len += TCP_OPTION_LEN_MSS;

  opts->flags |= TCP_OPTS_FLAG_WSCALE;
  opts->wscale = wnd_scale;
  len += TCP_OPTION_LEN_WINDOW_SCALE;

  opts->flags |= TCP_OPTS_FLAG_TSTAMP;
  opts->tsval = tcp_time_now ();
  opts->tsecr = 0;
  len += TCP_OPTION_LEN_TIMESTAMP;

  if (TCP_USE_SACKS)
    {
      opts->flags |= TCP_OPTS_FLAG_SACK_PERMITTED;
      len += TCP_OPTION_LEN_SACK_PERMITTED;
    }

  /* Align to needed boundary */
  len += (TCP_OPTS_ALIGN - len % TCP_OPTS_ALIGN) % TCP_OPTS_ALIGN;
  return len;
}

always_inline int
tcp_make_synack_options (tcp_connection_t * tc, tcp_options_t * opts)
{
  u8 len = 0;

  opts->flags |= TCP_OPTS_FLAG_MSS;
  opts->mss = tc->mss;
  len += TCP_OPTION_LEN_MSS;

  if (tcp_opts_wscale (&tc->rcv_opts))
    {
      opts->flags |= TCP_OPTS_FLAG_WSCALE;
      opts->wscale = tc->rcv_wscale;
      len += TCP_OPTION_LEN_WINDOW_SCALE;
    }

  if (tcp_opts_tstamp (&tc->rcv_opts))
    {
      opts->flags |= TCP_OPTS_FLAG_TSTAMP;
      opts->tsval = tcp_time_now ();
      opts->tsecr = tc->tsval_recent;
      len += TCP_OPTION_LEN_TIMESTAMP;
    }

  if (tcp_opts_sack_permitted (&tc->rcv_opts))
    {
      opts->flags |= TCP_OPTS_FLAG_SACK_PERMITTED;
      len += TCP_OPTION_LEN_SACK_PERMITTED;
    }

  /* Align to needed boundary */
  len += (TCP_OPTS_ALIGN - len % TCP_OPTS_ALIGN) % TCP_OPTS_ALIGN;
  return len;
}

always_inline int
tcp_make_established_options (tcp_connection_t * tc, tcp_options_t * opts)
{
  u8 len = 0;

  opts->flags = 0;

  if (tcp_opts_tstamp (&tc->rcv_opts))
    {
      opts->flags |= TCP_OPTS_FLAG_TSTAMP;
      opts->tsval = tcp_time_now ();
      opts->tsecr = tc->tsval_recent;
      len += TCP_OPTION_LEN_TIMESTAMP;
    }
  if (tcp_opts_sack_permitted (&tc->rcv_opts))
    {
      if (vec_len (tc->snd_sacks))
	{
	  opts->flags |= TCP_OPTS_FLAG_SACK;
	  opts->sacks = tc->snd_sacks;
	  opts->n_sack_blocks = clib_min (vec_len (tc->snd_sacks),
					  TCP_OPTS_MAX_SACK_BLOCKS);
	  len += 2 + TCP_OPTION_LEN_SACK_BLOCK * opts->n_sack_blocks;
	}
    }

  /* Align to needed boundary */
  len += (TCP_OPTS_ALIGN - len % TCP_OPTS_ALIGN) % TCP_OPTS_ALIGN;
  return len;
}

always_inline int
tcp_make_options (tcp_connection_t * tc, tcp_options_t * opts,
		  tcp_state_t state)
{
  switch (state)
    {
    case TCP_STATE_ESTABLISHED:
    case TCP_STATE_FIN_WAIT_1:
      return tcp_make_established_options (tc, opts);
    case TCP_STATE_SYN_RCVD:
      return tcp_make_synack_options (tc, opts);
    case TCP_STATE_SYN_SENT:
      return tcp_make_syn_options (opts, tc->rcv_wscale);
    default:
      clib_warning ("Not handled!");
      return 0;
    }
}

/**
 * Update max segment size we're able to process.
 *
 * The value is constrained by our interface's MTU and IP options. It is
 * also what we advertise to our peer.
 */
void
tcp_update_rcv_mss (tcp_connection_t * tc)
{
  /* TODO find our iface MTU */
  tc->mss = dummy_mtu;
}

/**
 * Update snd_mss to reflect the effective segment size that we can send
 * by taking into account all TCP options, including SACKs
 */
void
tcp_update_snd_mss (tcp_connection_t * tc)
{
  /* Compute options to be used for connection. These may be reused when
   * sending data or to compute the effective mss (snd_mss) */
  tc->snd_opts_len =
    tcp_make_options (tc, &tc->snd_opts, TCP_STATE_ESTABLISHED);

  /* XXX check if MTU has been updated */
  tc->snd_mss = clib_min (tc->mss, tc->rcv_opts.mss) - tc->snd_opts_len;
  ASSERT (tc->snd_mss > 0);
}

void
tcp_init_mss (tcp_connection_t * tc)
{
  u16 default_min_mss = 536;
  tcp_update_rcv_mss (tc);

  /* TODO cache mss and consider PMTU discovery */
  tc->snd_mss = clib_min (tc->rcv_opts.mss, tc->mss);

  if (tc->snd_mss < 45)
    {
      clib_warning ("snd mss is 0");
      /* Assume that at least the min default mss works */
      tc->snd_mss = default_min_mss;
      tc->rcv_opts.mss = default_min_mss;
    }

  /* We should have enough space for 40 bytes of options */
  ASSERT (tc->snd_mss > 45);

  /* If we use timestamp option, account for it */
  if (tcp_opts_tstamp (&tc->rcv_opts))
    tc->snd_mss -= TCP_OPTION_LEN_TIMESTAMP;
}

#define tcp_get_free_buffer_index(tm, bidx)                             \
do {                                                                    \
  u32 *my_tx_buffers, n_free_buffers;                                   \
  u32 thread_index = vlib_get_thread_index();                             	\
  my_tx_buffers = tm->tx_buffers[thread_index];                            \
  if (PREDICT_FALSE(vec_len (my_tx_buffers) == 0))                      \
    {                                                                   \
      n_free_buffers = 32;      /* TODO config or macro */              \
      vec_validate (my_tx_buffers, n_free_buffers - 1);                 \
      _vec_len(my_tx_buffers) = vlib_buffer_alloc_from_free_list (      \
          tm->vlib_main, my_tx_buffers, n_free_buffers,                 \
          VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX);                         \
      tm->tx_buffers[thread_index] = my_tx_buffers;                        \
    }                                                                   \
  /* buffer shortage */                                                 \
  if (PREDICT_FALSE (vec_len (my_tx_buffers) == 0))                     \
    return;                                                             \
  *bidx = my_tx_buffers[_vec_len (my_tx_buffers)-1];                    \
  _vec_len (my_tx_buffers) -= 1;                                        \
} while (0)

#define tcp_return_buffer(tm)						\
do {									\
  u32 *my_tx_buffers;							\
  u32 thread_index = vlib_get_thread_index();                             	\
  my_tx_buffers = tm->tx_buffers[thread_index];                          	\
  _vec_len (my_tx_buffers) +=1;						\
} while (0)

always_inline void
tcp_reuse_buffer (vlib_main_t * vm, vlib_buffer_t * b)
{
  vlib_buffer_t *it = b;
  do
    {
      it->current_data = 0;
      it->current_length = 0;
      it->total_length_not_including_first_buffer = 0;
    }
  while ((it->flags & VLIB_BUFFER_NEXT_PRESENT)
	 && (it = vlib_get_buffer (vm, it->next_buffer)));

  /* Leave enough space for headers */
  vlib_buffer_make_headroom (b, MAX_HDRS_LEN);
  vnet_buffer (b)->tcp.flags = 0;
}

/**
 * Prepare ACK
 */
void
tcp_make_ack_i (tcp_connection_t * tc, vlib_buffer_t * b, tcp_state_t state,
		u8 flags)
{
  tcp_options_t _snd_opts, *snd_opts = &_snd_opts;
  u8 tcp_opts_len, tcp_hdr_opts_len;
  tcp_header_t *th;
  u16 wnd;

  wnd = tcp_window_to_advertise (tc, state);

  /* Make and write options */
  tcp_opts_len = tcp_make_established_options (tc, snd_opts);
  tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t);

  th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->snd_nxt,
			     tc->rcv_nxt, tcp_hdr_opts_len, flags, wnd);

  tcp_options_write ((u8 *) (th + 1), snd_opts);
  vnet_buffer (b)->tcp.connection_index = tc->c_c_index;
}

/**
 * Convert buffer to ACK
 */
void
tcp_make_ack (tcp_connection_t * tc, vlib_buffer_t * b)
{
  vlib_main_t *vm = vlib_get_main ();

  tcp_reuse_buffer (vm, b);
  tcp_make_ack_i (tc, b, TCP_STATE_ESTABLISHED, TCP_FLAG_ACK);
  TCP_EVT_DBG (TCP_EVT_ACK_SENT, tc);
  vnet_buffer (b)->tcp.flags = TCP_BUF_FLAG_ACK;
  tc->rcv_las = tc->rcv_nxt;
}

/**
 * Convert buffer to FIN-ACK
 */
void
tcp_make_fin (tcp_connection_t * tc, vlib_buffer_t * b)
{
  vlib_main_t *vm = vlib_get_main ();
  u8 flags = 0;

  tcp_reuse_buffer (vm, b);

  flags = TCP_FLAG_FIN | TCP_FLAG_ACK;
  tcp_make_ack_i (tc, b, TCP_STATE_ESTABLISHED, flags);

  /* Reset flags, make sure ack is sent */
  vnet_buffer (b)->tcp.flags &= ~TCP_BUF_FLAG_DUPACK;

  tc->snd_nxt += 1;
}

/**
 * Convert buffer to SYN-ACK
 */
void
tcp_make_synack (tcp_connection_t * tc, vlib_buffer_t * b)
{
  vlib_main_t *vm = vlib_get_main ();
  tcp_options_t _snd_opts, *snd_opts = &_snd_opts;
  u8 tcp_opts_len, tcp_hdr_opts_len;
  tcp_header_t *th;
  u16 initial_wnd;
  u32 time_now;

  memset (snd_opts, 0, sizeof (*snd_opts));

  tcp_reuse_buffer (vm, b);

  /* Set random initial sequence */
  time_now = tcp_time_now ();

  tc->iss = random_u32 (&time_now);
  tc->snd_una = tc->iss;
  tc->snd_nxt = tc->iss + 1;
  tc->snd_una_max = tc->snd_nxt;

  initial_wnd = tcp_initial_window_to_advertise (tc);

  /* Make and write options */
  tcp_opts_len = tcp_make_synack_options (tc, snd_opts);
  tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t);

  th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->iss,
			     tc->rcv_nxt, tcp_hdr_opts_len,
			     TCP_FLAG_SYN | TCP_FLAG_ACK, initial_wnd);

  tcp_options_write ((u8 *) (th + 1), snd_opts);

  vnet_buffer (b)->tcp.connection_index = tc->c_c_index;
  vnet_buffer (b)->tcp.flags = TCP_BUF_FLAG_ACK;

  /* Init retransmit timer */
  tcp_retransmit_timer_set (tc);
}

always_inline void
tcp_enqueue_to_ip_lookup (vlib_main_t * vm, vlib_buffer_t * b, u32 bi,
			  u8 is_ip4)
{
  u32 *to_next, next_index;
  vlib_frame_t *f;

  b->flags |= VNET_BUFFER_LOCALLY_ORIGINATED;
  b->error = 0;

  /* Default FIB for now */
  vnet_buffer (b)->sw_if_index[VLIB_TX] = 0;

  /* Send to IP lookup */
  next_index = is_ip4 ? ip4_lookup_node.index : ip6_lookup_node.index;
  f = vlib_get_frame_to_node (vm, next_index);

  /* Enqueue the packet */
  to_next = vlib_frame_vector_args (f);
  to_next[0] = bi;
  f->n_vectors = 1;
  vlib_put_frame_to_node (vm, next_index, f);
}

int
tcp_make_reset_in_place (vlib_main_t * vm, vlib_buffer_t * b0,
			 tcp_state_t state, u8 thread_index, u8 is_ip4)
{
  ip4_header_t *ih4;
  ip6_header_t *ih6;
  tcp_header_t *th0;
  ip4_address_t src_ip40, dst_ip40;
  ip6_address_t src_ip60, dst_ip60;
  u16 src_port, dst_port;
  u32 tmp;
  u32 seq, ack;
  u8 flags;

  /* Find IP and TCP headers */
  th0 = tcp_buffer_hdr (b0);

  /* Save src and dst ip */
  if (is_ip4)
    {
      ih4 = vlib_buffer_get_current (b0);
      ASSERT ((ih4->ip_version_and_header_length & 0xF0) == 0x40);
      src_ip40.as_u32 = ih4->src_address.as_u32;
      dst_ip40.as_u32 = ih4->dst_address.as_u32;
    }
  else
    {
      ih6 = vlib_buffer_get_current (b0);
      ASSERT ((ih6->ip_version_traffic_class_and_flow_label & 0xF0) == 0x60);
      clib_memcpy (&src_ip60, &ih6->src_address, sizeof (ip6_address_t));
      clib_memcpy (&dst_ip60, &ih6->dst_address, sizeof (ip6_address_t));
    }

  src_port = th0->src_port;
  dst_port = th0->dst_port;

  /* Try to determine what/why we're actually resetting */
  if (state == TCP_STATE_CLOSED)
    {
      if (!tcp_syn (th0))
	return -1;

      tmp = clib_net_to_host_u32 (th0->seq_number);

      /* Got a SYN for no listener. */
      flags = TCP_FLAG_RST | TCP_FLAG_ACK;
      ack = clib_host_to_net_u32 (tmp + 1);
      seq = 0;
    }
  else
    {
      flags = TCP_FLAG_RST;
      seq = th0->ack_number;
      ack = 0;
    }

  tcp_reuse_buffer (vm, b0);
  th0 = vlib_buffer_push_tcp_net_order (b0, dst_port, src_port, seq, ack,
					sizeof (tcp_header_t), flags, 0);

  if (is_ip4)
    {
      ih4 = vlib_buffer_push_ip4 (vm, b0, &dst_ip40, &src_ip40,
				  IP_PROTOCOL_TCP);
      th0->checksum = ip4_tcp_udp_compute_checksum (vm, b0, ih4);
    }
  else
    {
      int bogus = ~0;
      ih6 = vlib_buffer_push_ip6 (vm, b0, &dst_ip60, &src_ip60,
				  IP_PROTOCOL_TCP);
      th0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b0, ih6, &bogus);
      ASSERT (!bogus);
    }

  return 0;
}

/**
 *  Send reset without reusing existing buffer
 */
void
tcp_send_reset (vlib_buffer_t * pkt, u8 is_ip4)
{
  vlib_buffer_t *b;
  u32 bi;
  tcp_main_t *tm = vnet_get_tcp_main ();
  vlib_main_t *vm = vlib_get_main ();
  u8 tcp_hdr_len, flags = 0;
  tcp_header_t *th, *pkt_th;
  u32 seq, ack;
  ip4_header_t *ih4, *pkt_ih4;
  ip6_header_t *ih6, *pkt_ih6;

  tcp_get_free_buffer_index (tm, &bi);
  b = vlib_get_buffer (vm, bi);

  /* Leave enough space for headers */
  vlib_buffer_make_headroom (b, MAX_HDRS_LEN);

  /* Make and write options */
  tcp_hdr_len = sizeof (tcp_header_t);

  if (is_ip4)
    {
      pkt_ih4 = vlib_buffer_get_current (pkt);
      pkt_th = ip4_next_header (pkt_ih4);
    }
  else
    {
      pkt_ih6 = vlib_buffer_get_current (pkt);
      pkt_th = ip6_next_header (pkt_ih6);
    }

  if (tcp_ack (pkt_th))
    {
      flags = TCP_FLAG_RST;
      seq = pkt_th->ack_number;
      ack = 0;
    }
  else
    {
      flags = TCP_FLAG_RST | TCP_FLAG_ACK;
      seq = 0;
      ack = clib_host_to_net_u32 (vnet_buffer (pkt)->tcp.seq_end);
    }

  th = vlib_buffer_push_tcp_net_order (b, pkt_th->dst_port, pkt_th->src_port,
				       seq, ack, tcp_hdr_len, flags, 0);

  /* Swap src and dst ip */
  if (is_ip4)
    {
      ASSERT ((pkt_ih4->ip_version_and_header_length & 0xF0) == 0x40);
      ih4 = vlib_buffer_push_ip4 (vm, b, &pkt_ih4->dst_address,
				  &pkt_ih4->src_address, IP_PROTOCOL_TCP);
      th->checksum = ip4_tcp_udp_compute_checksum (vm, b, ih4);
    }
  else
    {
      int bogus = ~0;
      pkt_ih6 = (ip6_header_t *) (pkt_th - 1);
      ASSERT ((pkt_ih6->ip_version_traffic_class_and_flow_label & 0xF0) ==
	      0x60);
      ih6 =
	vlib_buffer_push_ip6 (vm, b, &pkt_ih6->dst_address,
			      &pkt_ih6->src_address, IP_PROTOCOL_TCP);
      th->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b, ih6, &bogus);
      ASSERT (!bogus);
    }

  tcp_enqueue_to_ip_lookup (vm, b, bi, is_ip4);
}

void
tcp_push_ip_hdr (tcp_main_t * tm, tcp_connection_t * tc, vlib_buffer_t * b)
{
  tcp_header_t *th = vlib_buffer_get_current (b);

  if (tc->c_is_ip4)
    {
      ip4_header_t *ih;
      ih = vlib_buffer_push_ip4 (tm->vlib_main, b, &tc->c_lcl_ip4,
				 &tc->c_rmt_ip4, IP_PROTOCOL_TCP);
      th->checksum = ip4_tcp_udp_compute_checksum (tm->vlib_main, b, ih);
    }
  else
    {
      ip6_header_t *ih;
      int bogus = ~0;

      ih = vlib_buffer_push_ip6 (tm->vlib_main, b, &tc->c_lcl_ip6,
				 &tc->c_rmt_ip6, IP_PROTOCOL_TCP);
      th->checksum = ip6_tcp_udp_icmp_compute_checksum (tm->vlib_main, b, ih,
							&bogus);
      ASSERT (!bogus);
    }
}

/**
 *  Send SYN
 *
 *  Builds a SYN packet for a half-open connection and sends it to ipx_lookup.
 *  The packet is not forwarded through tcpx_output to avoid doing lookups
 *  in the half_open pool.
 */
void
tcp_send_syn (tcp_connection_t * tc)
{
  vlib_buffer_t *b;
  u32 bi;
  tcp_main_t *tm = vnet_get_tcp_main ();
  vlib_main_t *vm = vlib_get_main ();
  u8 tcp_hdr_opts_len, tcp_opts_len;
  tcp_header_t *th;
  u32 time_now;
  u16 initial_wnd;
  tcp_options_t snd_opts;

  tcp_get_free_buffer_index (tm, &bi);
  b = vlib_get_buffer (vm, bi);

  /* Leave enough space for headers */
  vlib_buffer_make_headroom (b, MAX_HDRS_LEN);

  /* Set random initial sequence */
  time_now = tcp_time_now ();

  tc->iss = random_u32 (&time_now);
  tc->snd_una = tc->iss;
  tc->snd_una_max = tc->snd_nxt = tc->iss + 1;

  initial_wnd = tcp_initial_window_to_advertise (tc);

  /* Make and write options */
  memset (&snd_opts, 0, sizeof (snd_opts));
  tcp_opts_len = tcp_make_syn_options (&snd_opts, tc->rcv_wscale);
  tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t);

  th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->iss,
			     tc->rcv_nxt, tcp_hdr_opts_len, TCP_FLAG_SYN,
			     initial_wnd);

  tcp_options_write ((u8 *) (th + 1), &snd_opts);

  /* Measure RTT with this */
  tc->rtt_ts = tcp_time_now ();
  tc->rtt_seq = tc->snd_nxt;

  /* Start retransmit trimer  */
  tcp_timer_set (tc, TCP_TIMER_RETRANSMIT_SYN, tc->rto * TCP_TO_TIMER_TICK);
  tc->rto_boff = 0;

  /* Set the connection establishment timer */
  tcp_timer_set (tc, TCP_TIMER_ESTABLISH, TCP_ESTABLISH_TIME);

  tcp_push_ip_hdr (tm, tc, b);
  tcp_enqueue_to_ip_lookup (vm, b, bi, tc->c_is_ip4);
}

always_inline void
tcp_enqueue_to_output (vlib_main_t * vm, vlib_buffer_t * b, u32 bi, u8 is_ip4)
{
  u32 *to_next, next_index;
  vlib_frame_t *f;

  b->flags |= VNET_BUFFER_LOCALLY_ORIGINATED;
  b->error = 0;

  /* Decide where to send the packet */
  next_index = is_ip4 ? tcp4_output_node.index : tcp6_output_node.index;

  /* Enqueue the packet */
  f = vlib_get_frame_to_node (vm, next_index);
  to_next = vlib_frame_vector_args (f);
  to_next[0] = bi;
  f->n_vectors = 1;
  vlib_put_frame_to_node (vm, next_index, f);
}

/**
 *  Send FIN
 */
void
tcp_send_fin (tcp_connection_t * tc)
{
  vlib_buffer_t *b;
  u32 bi;
  tcp_main_t *tm = vnet_get_tcp_main ();
  vlib_main_t *vm = vlib_get_main ();

  tcp_get_free_buffer_index (tm, &bi);
  b = vlib_get_buffer (vm, bi);

  /* Leave enough space for headers */
  vlib_buffer_make_headroom (b, MAX_HDRS_LEN);

  tcp_make_fin (tc, b);
  tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4);
  tc->flags |= TCP_CONN_FINSNT;
  tcp_retransmit_timer_force_update (tc);
  TCP_EVT_DBG (TCP_EVT_FIN_SENT, tc);
}

always_inline u8
tcp_make_state_flags (tcp_state_t next_state)
{
  switch (next_state)
    {
    case TCP_STATE_ESTABLISHED:
      return TCP_FLAG_ACK;
    case TCP_STATE_SYN_RCVD:
      return TCP_FLAG_SYN | TCP_FLAG_ACK;
    case TCP_STATE_SYN_SENT:
      return TCP_FLAG_SYN;
    case TCP_STATE_LAST_ACK:
    case TCP_STATE_FIN_WAIT_1:
      return TCP_FLAG_FIN;
    default:
      clib_warning ("Shouldn't be here!");
    }
  return 0;
}

/**
 * Push TCP header and update connection variables
 */
static void
tcp_push_hdr_i (tcp_connection_t * tc, vlib_buffer_t * b,
		tcp_state_t next_state, u8 compute_opts)
{
  u32 advertise_wnd, data_len;
  u8 tcp_hdr_opts_len, opts_write_len, flags;
  tcp_header_t *th;

  data_len = b->current_length + b->total_length_not_including_first_buffer;
  vnet_buffer (b)->tcp.flags = 0;

  if (compute_opts)
    tc->snd_opts_len = tcp_make_options (tc, &tc->snd_opts, tc->state);

  tcp_hdr_opts_len = tc->snd_opts_len + sizeof (tcp_header_t);
  advertise_wnd = tcp_window_to_advertise (tc, next_state);
  flags = tcp_make_state_flags (next_state);

  /* Push header and options */
  th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->snd_nxt,
			     tc->rcv_nxt, tcp_hdr_opts_len, flags,
			     advertise_wnd);
  opts_write_len = tcp_options_write ((u8 *) (th + 1), &tc->snd_opts);

  ASSERT (opts_write_len == tc->snd_opts_len);
  vnet_buffer (b)->tcp.connection_index = tc->c_c_index;

  /*
   * Update connection variables
   */

  tc->snd_nxt += data_len;
  tc->rcv_las = tc->rcv_nxt;

  /* TODO this is updated in output as well ... */
  if (seq_gt (tc->snd_nxt, tc->snd_una_max))
    {
      tc->snd_una_max = tc->snd_nxt;
      tcp_validate_txf_size (tc, tc->snd_una_max - tc->snd_una);
    }

  TCP_EVT_DBG (TCP_EVT_PKTIZE, tc);
}

void
tcp_send_ack (tcp_connection_t * tc)
{
  tcp_main_t *tm = vnet_get_tcp_main ();
  vlib_main_t *vm = vlib_get_main ();

  vlib_buffer_t *b;
  u32 bi;

  /* Get buffer */
  tcp_get_free_buffer_index (tm, &bi);
  b = vlib_get_buffer (vm, bi);

  /* Fill in the ACK */
  tcp_make_ack (tc, b);
  tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4);
}

/* Send delayed ACK when timer expires */
void
tcp_timer_delack_handler (u32 index)
{
  u32 thread_index = vlib_get_thread_index ();
  tcp_connection_t *tc;

  tc = tcp_connection_get (index, thread_index);
  tc->timers[TCP_TIMER_DELACK] = TCP_TIMER_HANDLE_INVALID;
  tcp_send_ack (tc);
}

/** Build a retransmit segment
 *
 * @return the number of bytes in the segment or 0 if there's nothing to
 *         retransmit
 */
u32
tcp_prepare_retransmit_segment (tcp_connection_t * tc, vlib_buffer_t * b,
				u32 offset, u32 max_bytes)
{
  vlib_main_t *vm = vlib_get_main ();
  int n_bytes = 0;
  u32 start;

  tcp_reuse_buffer (vm, b);

  ASSERT (tc->state >= TCP_STATE_ESTABLISHED);
  ASSERT (max_bytes != 0);

  max_bytes = clib_min (tc->snd_mss, max_bytes);
  start = tc->snd_una + offset;

  /* Start is beyond snd_congestion */
  if (seq_geq (start, tc->snd_congestion))
    goto done;

  /* Don't overshoot snd_congestion */
  if (seq_gt (start + max_bytes, tc->snd_congestion))
    {
      max_bytes = tc->snd_congestion - start;
      if (max_bytes == 0)
	goto done;
    }

  tc->snd_opts_len = tcp_make_options (tc, &tc->snd_opts, tc->state);

  ASSERT (max_bytes <= tc->snd_mss);

  n_bytes = stream_session_peek_bytes (&tc->connection,
				       vlib_buffer_get_current (b), offset,
				       max_bytes);
  ASSERT (n_bytes > 0);
  b->current_length = n_bytes;
  tcp_push_hdr_i (tc, b, tc->state, 0);

  if (tcp_in_fastrecovery (tc))
    tc->snd_rxt_bytes += n_bytes;

done:
  TCP_EVT_DBG (TCP_EVT_CC_RTX, tc, offset, n_bytes);
  return n_bytes;
}

/**
 * Reset congestion control, switch cwnd to loss window and try again.
 */
static void
tcp_rtx_timeout_cc (tcp_connection_t * tc)
{
  tc->prev_ssthresh = tc->ssthresh;
  tc->prev_cwnd = tc->cwnd;

  /* Cleanly recover cc (also clears up fast retransmit) */
  if (tcp_in_fastrecovery (tc))
    tcp_cc_fastrecovery_exit (tc);

  /* Start again from the beginning */
  tc->ssthresh = clib_max (tcp_flight_size (tc) / 2, 2 * tc->snd_mss);
  tc->cwnd = tcp_loss_wnd (tc);
  tc->snd_congestion = tc->snd_una_max;

  tcp_recovery_on (tc);
}

static void
tcp_timer_retransmit_handler_i (u32 index, u8 is_syn)
{
  tcp_main_t *tm = vnet_get_tcp_main ();
  vlib_main_t *vm = vlib_get_main ();
  u32 thread_index = vlib_get_thread_index ();
  tcp_connection_t *tc;
  vlib_buffer_t *b;
  u32 bi, n_bytes;

  if (is_syn)
    {
      tc = tcp_half_open_connection_get (index);
    }
  else
    {
      tc = tcp_connection_get (index, thread_index);
    }

  /* Make sure timer handle is set to invalid */
  tc->timers[TCP_TIMER_RETRANSMIT] = TCP_TIMER_HANDLE_INVALID;

  if (!tcp_in_recovery (tc) && tc->rto_boff > 0
      && tc->state >= TCP_STATE_ESTABLISHED)
    {
      tc->rto_boff = 0;
      tcp_update_rto (tc);
    }

  /* Increment RTO backoff (also equal to number of retries) */
  tc->rto_boff += 1;

  /* Go back to first un-acked byte */
  tc->snd_nxt = tc->snd_una;

  tcp_get_free_buffer_index (tm, &bi);
  b = vlib_get_buffer (vm, bi);

  if (tc->state >= TCP_STATE_ESTABLISHED)
    {
      /* Lost FIN, retransmit and return */
      if (tc->flags & TCP_CONN_FINSNT)
	{
	  tcp_send_fin (tc);
	  return;
	}

      /* First retransmit timeout */
      if (tc->rto_boff == 1)
	tcp_rtx_timeout_cc (tc);

      /* Exponential backoff */
      tc->rto = clib_min (tc->rto << 1, TCP_RTO_MAX);

      TCP_EVT_DBG (TCP_EVT_CC_EVT, tc, 1);

      /* Send one segment */
      n_bytes = tcp_prepare_retransmit_segment (tc, b, 0, tc->snd_mss);
      /* TODO be less aggressive about this */
      scoreboard_clear (&tc->sack_sb);

      if (n_bytes == 0)
	{
	  clib_warning ("could not retransmit anything");
	  clib_warning ("%U", format_tcp_connection, tc, 2);

	  /* Try again eventually */
	  tcp_retransmit_timer_set (tc);
	  ASSERT (0 || (tc->rto_boff > 1
			&& tc->snd_una == tc->snd_congestion));
	  return;
	}

      /* For first retransmit, record timestamp (Eifel detection RFC3522) */
      if (tc->rto_boff == 1)
	tc->snd_rxt_ts = tcp_time_now ();
    }
  /* Retransmit for SYN/SYNACK */
  else if (tc->state == TCP_STATE_SYN_RCVD || tc->state == TCP_STATE_SYN_SENT)
    {
      /* Try without increasing RTO a number of times. If this fails,
       * start growing RTO exponentially */
      if (tc->rto_boff > TCP_RTO_SYN_RETRIES)
	tc->rto = clib_min (tc->rto << 1, TCP_RTO_MAX);

      vlib_buffer_make_headroom (b, MAX_HDRS_LEN);

      tcp_push_hdr_i (tc, b, tc->state, 1);

      /* Account for the SYN */
      tc->snd_nxt += 1;
    }
  else
    {
      ASSERT (tc->state == TCP_STATE_CLOSED);
      clib_warning ("connection closed ...");
      return;
    }

  if (!is_syn)
    {
      tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4);

      /* Re-enable retransmit timer */
      tcp_retransmit_timer_set (tc);
    }
  else
    {
      ASSERT (tc->state == TCP_STATE_SYN_SENT);

      TCP_EVT_DBG (TCP_EVT_SYN_RTX, tc);

      /* This goes straight to ipx_lookup */
      tcp_push_ip_hdr (tm, tc, b);
      tcp_enqueue_to_ip_lookup (vm, b, bi, tc->c_is_ip4);

      /* Re-enable retransmit timer */
      tcp_timer_set (tc, TCP_TIMER_RETRANSMIT_SYN,
		     tc->rto * TCP_TO_TIMER_TICK);
    }
}

void
tcp_timer_retransmit_handler (u32 index)
{
  tcp_timer_retransmit_handler_i (index, 0);
}

void
tcp_timer_retransmit_syn_handler (u32 index)
{
  tcp_timer_retransmit_handler_i (index, 1);
}

/**
 * Got 0 snd_wnd from peer, try to do something about it.
 *
 */
void
tcp_timer_persist_handler (u32 index)
{
  tcp_main_t *tm = vnet_get_tcp_main ();
  vlib_main_t *vm = vlib_get_main ();
  u32 thread_index = vlib_get_thread_index ();
  tcp_connection_t *tc;
  vlib_buffer_t *b;
  u32 bi, old_snd_nxt;
  int n_bytes = 0;

  tc = tcp_connection_get_if_valid (index, thread_index);

  if (!tc)
    return;

  /* Make sure timer handle is set to invalid */
  tc->timers[TCP_TIMER_PERSIST] = TCP_TIMER_HANDLE_INVALID;

  /* Problem already solved or worse */
  if (tc->state == TCP_STATE_CLOSED || tc->state > TCP_STATE_ESTABLISHED
      || tc->snd_wnd > tc->snd_mss || tcp_in_recovery (tc))
    return;

  /* Increment RTO backoff */
  tc->rto_boff += 1;
  tc->rto = clib_min (tc->rto << 1, TCP_RTO_MAX);

  /* Try to force the first unsent segment  */
  tcp_get_free_buffer_index (tm, &bi);
  b = vlib_get_buffer (vm, bi);

  tcp_validate_txf_size (tc, tc->snd_una_max - tc->snd_una);
  tc->snd_opts_len = tcp_make_options (tc, &tc->snd_opts, tc->state);
  n_bytes = stream_session_peek_bytes (&tc->connection,
				       vlib_buffer_get_current (b),
				       tc->snd_una_max - tc->snd_una,
				       tc->snd_mss);
  /* Nothing to send */
  if (n_bytes <= 0)
    {
      clib_warning ("persist found nothing to send");
      tcp_return_buffer (tm);
      return;
    }

  b->current_length = n_bytes;
  ASSERT (tc->snd_nxt == tc->snd_una_max || tc->rto_boff > 1
	  || tcp_timer_is_active (tc, TCP_TIMER_RETRANSMIT));

  /* Allow updating of snd_una_max but don't update snd_nxt */
  old_snd_nxt = tc->snd_nxt;
  tcp_push_hdr_i (tc, b, tc->state, 0);
  tc->snd_nxt = old_snd_nxt;
  tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4);

  /* Re-enable persist timer */
  tcp_persist_timer_set (tc);
}

/**
 * Retransmit first unacked segment
 */
void
tcp_retransmit_first_unacked (tcp_connection_t * tc)
{
  tcp_main_t *tm = vnet_get_tcp_main ();
  vlib_main_t *vm = vlib_get_main ();
  vlib_buffer_t *b;
  u32 bi, n_bytes, old_snd_nxt;

  old_snd_nxt = tc->snd_nxt;
  tc->snd_nxt = tc->snd_una;

  /* Get buffer */
  tcp_get_free_buffer_index (tm, &bi);
  b = vlib_get_buffer (vm, bi);

  TCP_EVT_DBG (TCP_EVT_CC_EVT, tc, 2);

  n_bytes = tcp_prepare_retransmit_segment (tc, b, 0, tc->snd_mss);
  if (n_bytes == 0)
    {
      tcp_return_buffer (tm);
      goto done;
    }

  tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4);

done:
  tc->snd_nxt = old_snd_nxt;
}

/**
 * Do fast retransmit with SACKs
 */
void
tcp_fast_retransmit_sack (tcp_connection_t * tc)
{
  tcp_main_t *tm = vnet_get_tcp_main ();
  vlib_main_t *vm = vlib_get_main ();
  u32 n_written = 0, offset = 0, max_bytes;
  vlib_buffer_t *b;
  sack_scoreboard_hole_t *hole;
  sack_scoreboard_t *sb;
  u32 bi, old_snd_nxt;
  int snd_space;
  u8 snd_limited = 0, can_rescue = 0;

  ASSERT (tcp_in_fastrecovery (tc));
  TCP_EVT_DBG (TCP_EVT_CC_EVT, tc, 0);

  old_snd_nxt = tc->snd_nxt;
  sb = &tc->sack_sb;
  snd_space = tcp_available_snd_space (tc);

  hole = scoreboard_get_hole (sb, sb->cur_rxt_hole);
  while (hole && snd_space > 0)
    {
      tcp_get_free_buffer_index (tm, &bi);
      b = vlib_get_buffer (vm, bi);

      hole = scoreboard_next_rxt_hole (sb, hole,
				       tcp_fastrecovery_sent_1_smss (tc),
				       &can_rescue, &snd_limited);
      if (!hole)
	{
	  if (!can_rescue || !(seq_lt (sb->rescue_rxt, tc->snd_una)
			       || seq_gt (sb->rescue_rxt,
					  tc->snd_congestion)))
	    break;

	  /* If rescue rxt undefined or less than snd_una then one segment of
	   * up to SMSS octets that MUST include the highest outstanding
	   * unSACKed sequence number SHOULD be returned, and RescueRxt set to
	   * RecoveryPoint. HighRxt MUST NOT be updated.
	   */
	  max_bytes = clib_min (tc->snd_mss, snd_space);
	  offset = tc->snd_congestion - tc->snd_una - max_bytes;
	  sb->rescue_rxt = tc->snd_congestion;
	  tc->snd_nxt = tc->snd_una + offset;
	  tcp_prepare_retransmit_segment (tc, b, offset, max_bytes);
	  tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4);
	  break;
	}

      max_bytes = snd_limited ? tc->snd_mss : hole->end - sb->high_rxt;
      offset = sb->high_rxt - tc->snd_una;
      tc->snd_nxt = tc->snd_una + offset;
      n_written = tcp_prepare_retransmit_segment (tc, b, offset, max_bytes);

      /* Nothing left to retransmit */
      if (n_written == 0)
	{
	  tcp_return_buffer (tm);
	  break;
	}

      sb->high_rxt += n_written;
      tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4);
      snd_space -= n_written;
    }

  /* If window allows, send 1 SMSS of new data */
  tc->snd_nxt = old_snd_nxt;
}

/**
 * Fast retransmit without SACK info
 */
void
tcp_fast_retransmit_no_sack (tcp_connection_t * tc)
{
  tcp_main_t *tm = vnet_get_tcp_main ();
  vlib_main_t *vm = vlib_get_main ();
  u32 n_written = 0, offset = 0, bi, old_snd_nxt;
  int snd_space;
  vlib_buffer_t *b;

  ASSERT (tcp_in_fastrecovery (tc));
  TCP_EVT_DBG (TCP_EVT_CC_EVT, tc, 0);

  /* Start resending from first un-acked segment */
  old_snd_nxt = tc->snd_nxt;
  tc->snd_nxt = tc->snd_una;
  snd_space = tcp_available_snd_space (tc);

  while (snd_space > 0)
    {
      tcp_get_free_buffer_index (tm, &bi);
      b = vlib_get_buffer (vm, bi);

      offset += n_written;
      n_written = tcp_prepare_retransmit_segment (tc, b, offset, snd_space);

      /* Nothing left to retransmit */
      if (n_written == 0)
	{
	  tcp_return_buffer (tm);
	  break;
	}

      tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4);
      snd_space -= n_written;
    }

  /* Restore snd_nxt. If window allows, send 1 SMSS of new data */
  tc->snd_nxt = old_snd_nxt;
}

/**
 * Do fast retransmit
 */
void
tcp_fast_retransmit (tcp_connection_t * tc)
{
  if (tcp_opts_sack_permitted (&tc->rcv_opts)
      && scoreboard_first_hole (&tc->sack_sb))
    tcp_fast_retransmit_sack (tc);
  else
    tcp_fast_retransmit_no_sack (tc);
}

always_inline u32
tcp_session_has_ooo_data (tcp_connection_t * tc)
{
  stream_session_t *s =
    stream_session_get (tc->c_s_index, tc->c_thread_index);
  return svm_fifo_has_ooo_data (s->server_rx_fifo);
}

always_inline uword
tcp46_output_inline (vlib_main_t * vm,
		     vlib_node_runtime_t * node,
		     vlib_frame_t * from_frame, int is_ip4)
{
  u32 n_left_from, next_index, *from, *to_next;
  u32 my_thread_index = vm->thread_index;

  from = vlib_frame_vector_args (from_frame);
  n_left_from = from_frame->n_vectors;

  next_index = node->cached_next_index;

  while (n_left_from > 0)
    {
      u32 n_left_to_next;

      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);

      while (n_left_from > 0 && n_left_to_next > 0)
	{
	  u32 bi0;
	  vlib_buffer_t *b0;
	  tcp_connection_t *tc0;
	  tcp_tx_trace_t *t0;
	  tcp_header_t *th0 = 0;
	  u32 error0 = TCP_ERROR_PKTS_SENT, next0 = TCP_OUTPUT_NEXT_IP_LOOKUP;

	  bi0 = from[0];
	  to_next[0] = bi0;
	  from += 1;
	  to_next += 1;
	  n_left_from -= 1;
	  n_left_to_next -= 1;

	  b0 = vlib_get_buffer (vm, bi0);
	  tc0 = tcp_connection_get (vnet_buffer (b0)->tcp.connection_index,
				    my_thread_index);
	  if (PREDICT_FALSE (tc0 == 0 || tc0->state == TCP_STATE_CLOSED))
	    {
	      error0 = TCP_ERROR_INVALID_CONNECTION;
	      next0 = TCP_OUTPUT_NEXT_DROP;
	      goto done;
	    }

	  th0 = vlib_buffer_get_current (b0);
	  TCP_EVT_DBG (TCP_EVT_OUTPUT, tc0, th0->flags, b0->current_length);

	  if (is_ip4)
	    {
	      ip4_header_t *ih0;
	      ih0 = vlib_buffer_push_ip4 (vm, b0, &tc0->c_lcl_ip4,
					  &tc0->c_rmt_ip4, IP_PROTOCOL_TCP);
	      th0->checksum = ip4_tcp_udp_compute_checksum (vm, b0, ih0);
	    }
	  else
	    {
	      ip6_header_t *ih0;
	      int bogus = ~0;

	      ih0 = vlib_buffer_push_ip6 (vm, b0, &tc0->c_lcl_ip6,
					  &tc0->c_rmt_ip6, IP_PROTOCOL_TCP);
	      th0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b0, ih0,
								 &bogus);
	      ASSERT (!bogus);
	    }

	  /* Filter out DUPACKs if there are no OOO segments left */
	  if (PREDICT_FALSE
	      (vnet_buffer (b0)->tcp.flags & TCP_BUF_FLAG_DUPACK))
	    {
	      if (!tcp_session_has_ooo_data (tc0))
		{
		  error0 = TCP_ERROR_FILTERED_DUPACKS;
		  next0 = TCP_OUTPUT_NEXT_DROP;
		  goto done;
		}
	    }

	  /* Stop DELACK timer and fix flags */
	  tc0->flags &= ~(TCP_CONN_SNDACK);
	  tcp_timer_reset (tc0, TCP_TIMER_DELACK);

	  /* If not retransmitting
	   * 1) update snd_una_max (SYN, SYNACK, FIN)
	   * 2) If we're not tracking an ACK, start tracking */
	  if (seq_lt (tc0->snd_una_max, tc0->snd_nxt))
	    {
	      tc0->snd_una_max = tc0->snd_nxt;
	      if (tc0->rtt_ts == 0)
		{
		  tc0->rtt_ts = tcp_time_now ();
		  tc0->rtt_seq = tc0->snd_nxt;
		}
	    }

	  /* Set the retransmit timer if not set already and not
	   * doing a pure ACK */
	  if (!tcp_timer_is_active (tc0, TCP_TIMER_RETRANSMIT)
	      && tc0->snd_nxt != tc0->snd_una)
	    {
	      tcp_retransmit_timer_set (tc0);
	      tc0->rto_boff = 0;
	    }

	  /* set fib index to default and lookup node */
	  /* XXX network virtualization (vrf/vni) */
	  vnet_buffer (b0)->sw_if_index[VLIB_RX] = 0;
	  vnet_buffer (b0)->sw_if_index[VLIB_TX] = (u32) ~ 0;

	  b0->flags |= VNET_BUFFER_LOCALLY_ORIGINATED;
	done:
	  b0->error = node->errors[error0];
	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
	    {
	      t0 = vlib_add_trace (vm, node, b0, sizeof (*t0));
	      if (th0)
		{
		  clib_memcpy (&t0->tcp_header, th0, sizeof (t0->tcp_header));
		}
	      else
		{
		  memset (&t0->tcp_header, 0, sizeof (t0->tcp_header));
		}
	      clib_memcpy (&t0->tcp_connection, tc0,
			   sizeof (t0->tcp_connection));
	    }

	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
					   n_left_to_next, bi0, next0);
	}

      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
    }

  return from_frame->n_vectors;
}

static uword
tcp4_output (vlib_main_t * vm, vlib_node_runtime_t * node,
	     vlib_frame_t * from_frame)
{
  return tcp46_output_inline (vm, node, from_frame, 1 /* is_ip4 */ );
}

static uword
tcp6_output (vlib_main_t * vm, vlib_node_runtime_t * node,
	     vlib_frame_t * from_frame)
{
  return tcp46_output_inline (vm, node, from_frame, 0 /* is_ip4 */ );
}

/* *INDENT-OFF* */
VLIB_REGISTER_NODE (tcp4_output_node) =
{
  .function = tcp4_output,.name = "tcp4-output",
    /* Takes a vector of packets. */
    .vector_size = sizeof (u32),
    .n_errors = TCP_N_ERROR,
    .error_strings = tcp_error_strings,
    .n_next_nodes = TCP_OUTPUT_N_NEXT,
    .next_nodes = {
#define _(s,n) [TCP_OUTPUT_NEXT_##s] = n,
    foreach_tcp4_output_next
#undef _
    },
    .format_buffer = format_tcp_header,
    .format_trace = format_tcp_tx_trace,
};
/* *INDENT-ON* */

VLIB_NODE_FUNCTION_MULTIARCH (tcp4_output_node, tcp4_output);

/* *INDENT-OFF* */
VLIB_REGISTER_NODE (tcp6_output_node) =
{
  .function = tcp6_output,
  .name = "tcp6-output",
    /* Takes a vector of packets. */
  .vector_size = sizeof (u32),
  .n_errors = TCP_N_ERROR,
  .error_strings = tcp_error_strings,
  .n_next_nodes = TCP_OUTPUT_N_NEXT,
  .next_nodes = {
#define _(s,n) [TCP_OUTPUT_NEXT_##s] = n,
    foreach_tcp6_output_next
#undef _
  },
  .format_buffer = format_tcp_header,
  .format_trace = format_tcp_tx_trace,
};
/* *INDENT-ON* */

VLIB_NODE_FUNCTION_MULTIARCH (tcp6_output_node, tcp6_output);

u32
tcp_push_header (transport_connection_t * tconn, vlib_buffer_t * b)
{
  tcp_connection_t *tc;

  tc = (tcp_connection_t *) tconn;
  tcp_push_hdr_i (tc, b, TCP_STATE_ESTABLISHED, 0);

  if (tc->rtt_ts == 0 && !tcp_in_cong_recovery (tc))
    {
      tc->rtt_ts = tcp_time_now ();
      tc->rtt_seq = tc->snd_nxt;
    }
  return 0;
}

typedef enum _tcp_reset_next
{
  TCP_RESET_NEXT_DROP,
  TCP_RESET_NEXT_IP_LOOKUP,
  TCP_RESET_N_NEXT
} tcp_reset_next_t;

#define foreach_tcp4_reset_next        	\
  _(DROP, "error-drop")                 \
  _(IP_LOOKUP, "ip4-lookup")

#define foreach_tcp6_reset_next        	\
  _(DROP, "error-drop")                 \
  _(IP_LOOKUP, "ip6-lookup")

static uword
tcp46_send_reset_inline (vlib_main_t * vm, vlib_node_runtime_t * node,
			 vlib_frame_t * from_frame, u8 is_ip4)
{
  u32 n_left_from, next_index, *from, *to_next;
  u32 my_thread_index = vm->thread_index;

  from = vlib_frame_vector_args (from_frame);
  n_left_from = from_frame->n_vectors;

  next_index = node->cached_next_index;

  while (n_left_from > 0)
    {
      u32 n_left_to_next;

      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);

      while (n_left_from > 0 && n_left_to_next > 0)
	{
	  u32 bi0;
	  vlib_buffer_t *b0;
	  tcp_tx_trace_t *t0;
	  tcp_header_t *th0;
	  u32 error0 = TCP_ERROR_RST_SENT, next0 = TCP_RESET_NEXT_IP_LOOKUP;

	  bi0 = from[0];
	  to_next[0] = bi0;
	  from += 1;
	  to_next += 1;
	  n_left_from -= 1;
	  n_left_to_next -= 1;

	  b0 = vlib_get_buffer (vm, bi0);

	  if (tcp_make_reset_in_place (vm, b0, vnet_buffer (b0)->tcp.flags,
				       my_thread_index, is_ip4))
	    {
	      error0 = TCP_ERROR_LOOKUP_DROPS;
	      next0 = TCP_RESET_NEXT_DROP;
	      goto done;
	    }

	  /* Prepare to send to IP lookup */
	  vnet_buffer (b0)->sw_if_index[VLIB_TX] = 0;
	  next0 = TCP_RESET_NEXT_IP_LOOKUP;

	done:
	  b0->error = node->errors[error0];
	  b0->flags |= VNET_BUFFER_LOCALLY_ORIGINATED;
	  if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
	    {
	      th0 = vlib_buffer_get_current (b0);
	      if (is_ip4)
		th0 = ip4_next_header ((ip4_header_t *) th0);
	      else
		th0 = ip6_next_header ((ip6_header_t *) th0);
	      t0 = vlib_add_trace (vm, node, b0, sizeof (*t0));
	      clib_memcpy (&t0->tcp_header, th0, sizeof (t0->tcp_header));
	    }

	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
					   n_left_to_next, bi0, next0);
	}
      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
    }
  return from_frame->n_vectors;
}

static uword
tcp4_send_reset (vlib_main_t * vm, vlib_node_runtime_t * node,
		 vlib_frame_t * from_frame)
{
  return tcp46_send_reset_inline (vm, node, from_frame, 1);
}

static uword
tcp6_send_reset (vlib_main_t * vm, vlib_node_runtime_t * node,
		 vlib_frame_t * from_frame)
{
  return tcp46_send_reset_inline (vm, node, from_frame, 0);
}

/* *INDENT-OFF* */
VLIB_REGISTER_NODE (tcp4_reset_node) = {
  .function = tcp4_send_reset,
  .name = "tcp4-reset",
  .vector_size = sizeof (u32),
  .n_errors = TCP_N_ERROR,
  .error_strings = tcp_error_strings,
  .n_next_nodes = TCP_RESET_N_NEXT,
  .next_nodes = {
#define _(s,n) [TCP_RESET_NEXT_##s] = n,
    foreach_tcp4_reset_next
#undef _
  },
  .format_trace = format_tcp_tx_trace,
};
/* *INDENT-ON* */

VLIB_NODE_FUNCTION_MULTIARCH (tcp4_reset_node, tcp4_send_reset);

/* *INDENT-OFF* */
VLIB_REGISTER_NODE (tcp6_reset_node) = {
  .function = tcp6_send_reset,
  .name = "tcp6-reset",
  .vector_size = sizeof (u32),
  .n_errors = TCP_N_ERROR,
  .error_strings = tcp_error_strings,
  .n_next_nodes = TCP_RESET_N_NEXT,
  .next_nodes = {
#define _(s,n) [TCP_RESET_NEXT_##s] = n,
    foreach_tcp6_reset_next
#undef _
  },
  .format_trace = format_tcp_tx_trace,
};
/* *INDENT-ON* */

VLIB_NODE_FUNCTION_MULTIARCH (tcp6_reset_node, tcp6_send_reset);

/*
 * fd.io coding-style-patch-verification: ON
 *
 * Local Variables:
 * eval: (c-set-style "gnu")
 * End:
 */