summaryrefslogtreecommitdiffstats
path: root/src/vnet/ipip/ipip.api
blob: 4fbfa1d584eaa4d4a62733efd52b9daa79e88d82 (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
/* Hey Emacs use -*- mode: C -*- */
/*
 * Copyright (c) 2018 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.
 */

/**
 * The IPIP module implements IP{v4,v6} over IP{v4,v6} tunnelling as
 * described in RFC2473 and to some extent the largely historical
 * RFC1853.  The module also supports an IPv4 over IPv6 automatic
 * tunnelling mechanism called 6RD (RFC5969).
 *
 * The IPIP API module supports a CRD model for adding, deleting and
 * listing tunnels. A tunnel is represented as an interface in
 * VPP. The "handle" representing a tunnel is the sw_if_index.  As any
 * interface, the user must configure an IPv4 and/or IPv6 address on
 * the interface. This is the inner or payload protocol.
 *
 * Tunnel MTU: The tunnel MTU (the payload MTU) is configurable per
 * protocol. If a tunnel MTU is larger than the path MTU, the outer
 * packet will be fragmented. Fragmentation support is configurable,
 * as it can have severe performance issues, and might be used as an
 * attack vector (the remote side must reassemble.)
 *
 * Traffic class / TOS field can either be configured to a fixed
 * value, or can be copied from the inner to the outer header.
 * (For now we have stolen ~0 to indicate copy).
 *
 * Note:
 *
 * - The Tunnel encapsulation limit described in RFC2473 is not
 *   implemented.
 *
 * - ICMP proxying, as in a tunnel head-end receiving ICMP erors on
 *   the outer packet is currently not relayed to the original source
 *   of the packet.
 *
 * - PMTUD / MTU probing and tunnel keepalives are not yet implemented.
 *
 */

option version = "2.0.1";

import "vnet/interface_types.api";
import "vnet/ip/ip_types.api";
import "vnet/tunnel/tunnel_types.api";

/**
 * An IP{v4,v6} over IP{v4,v6} tunnel.
 */
typedef ipip_tunnel
{
  u32 instance; /* If non-~0, specifies a custom dev instance */
  vl_api_address_t src;
  vl_api_address_t dst;
  vl_api_interface_index_t sw_if_index; /* ignored on create, set in
					   details/dump */
  u32 table_id;
  vl_api_tunnel_encap_decap_flags_t flags;
  vl_api_tunnel_mode_t mode;
  vl_api_ip_dscp_t dscp; /* DSCP value for the tunnel encap,
                            ignored if ECNAP_COPY_DSCP flag is set */
};

/**
 * Create an IP{v4,v6} over IP{v4,v6} tunnel.
 */
define ipip_add_tunnel
{
  u32 client_index;
  u32 context;
  vl_api_ipip_tunnel_t tunnel;
};

define ipip_add_tunnel_reply
{
  u32 context;
  i32 retval;
  vl_api_interface_index_t sw_if_index;
};

/**
 * Delete an IP{v4,v6} over IP{v4,v6} tunnel.
 */
autoreply define ipip_del_tunnel
{
  u32 client_index;
  u32 context;
  vl_api_interface_index_t sw_if_index;
};

/**
 * Create an IPv4 over IPv6 automatic tunnel (6RD)
 */
define ipip_6rd_add_tunnel
{
  u32 client_index;
  u32 context;
  u32 ip6_table_id;
  u32 ip4_table_id;
  vl_api_ip6_prefix_t ip6_prefix;
  vl_api_ip4_prefix_t ip4_prefix;
  vl_api_ip4_address_t ip4_src;
  bool security_check;
  u8 tc_tos; /* If ~0, the TOS/TC value is copied from
                inner packet, otherwise set to value */
};

define ipip_6rd_add_tunnel_reply
{
  u32 context;
  i32 retval;
  vl_api_interface_index_t sw_if_index;
};

/**
 * Delete an IPv4 over IPv6 automatic tunnel (6RD)
 */
autoreply define ipip_6rd_del_tunnel
{
  u32 client_index;
  u32 context;
  vl_api_interface_index_t sw_if_index;
};

/**
 * List all IPIP tunnels
 */
define ipip_tunnel_dump
{
  u32 client_index;
  u32 context;
  vl_api_interface_index_t sw_if_index;
};

define ipip_tunnel_details
{
  u32 context;
  vl_api_ipip_tunnel_t tunnel;
};
='n559' href='#n559'>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
/*
 * Copyright (c) 2017 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.
 */
/**
 * @file
 * @brief Session and session manager
 */

#include <vnet/session/session.h>
#include <vnet/session/session_debug.h>
#include <vnet/session/application.h>
#include <vlibmemory/api.h>
#include <vnet/dpo/load_balance.h>
#include <vnet/fib/ip4_fib.h>

session_manager_main_t session_manager_main;
extern transport_proto_vft_t *tp_vfts;

static inline int
session_send_evt_to_thread (void *data, void *args, u32 thread_index,
			    session_evt_type_t evt_type)
{
  session_event_t *evt;
  svm_msg_q_msg_t msg;
  svm_msg_q_t *mq;
  u32 tries = 0, max_tries;

  mq = session_manager_get_vpp_event_queue (thread_index);
  while (svm_msg_q_try_lock (mq))
    {
      max_tries = vlib_get_current_process (vlib_get_main ())? 1e6 : 3;
      if (tries++ == max_tries)
	{
	  SESSION_DBG ("failed to enqueue evt");
	  return -1;
	}
    }
  if (PREDICT_FALSE (svm_msg_q_ring_is_full (mq, SESSION_MQ_IO_EVT_RING)))
    {
      svm_msg_q_unlock (mq);
      return -2;
    }
  msg = svm_msg_q_alloc_msg_w_ring (mq, SESSION_MQ_IO_EVT_RING);
  if (PREDICT_FALSE (svm_msg_q_msg_is_invalid (&msg)))
    {
      svm_msg_q_unlock (mq);
      return -2;
    }
  evt = (session_event_t *) svm_msg_q_msg_data (mq, &msg);
  evt->event_type = evt_type;
  switch (evt_type)
    {
    case FIFO_EVENT_RPC:
      evt->rpc_args.fp = data;
      evt->rpc_args.arg = args;
      break;
    case FIFO_EVENT_APP_TX:
    case FIFO_EVENT_BUILTIN_RX:
      evt->fifo = data;
      break;
    case FIFO_EVENT_BUILTIN_TX:
    case FIFO_EVENT_DISCONNECT:
      evt->session_handle = session_handle ((stream_session_t *) data);
      break;
    default:
      clib_warning ("evt unhandled!");
      svm_msg_q_unlock (mq);
      return -1;
    }

  svm_msg_q_add_and_unlock (mq, &msg);
  return 0;
}

int
session_send_io_evt_to_thread (svm_fifo_t * f, session_evt_type_t evt_type)
{
  return session_send_evt_to_thread (f, 0, f->master_thread_index, evt_type);
}

int
session_send_io_evt_to_thread_custom (void *data, u32 thread_index,
				      session_evt_type_t evt_type)
{
  return session_send_evt_to_thread (data, 0, thread_index, evt_type);
}

int
session_send_ctrl_evt_to_thread (stream_session_t * s,
				 session_evt_type_t evt_type)
{
  /* only event supported for now is disconnect */
  ASSERT (evt_type == FIFO_EVENT_DISCONNECT);
  return session_send_evt_to_thread (s, 0, s->thread_index,
				     FIFO_EVENT_DISCONNECT);
}

void
session_send_rpc_evt_to_thread (u32 thread_index, void *fp, void *rpc_args)
{
  if (thread_index != vlib_get_thread_index ())
    session_send_evt_to_thread (fp, rpc_args, thread_index, FIFO_EVENT_RPC);
  else
    {
      void (*fnp) (void *) = fp;
      fnp (rpc_args);
    }
}

stream_session_t *
session_alloc (u32 thread_index)
{
  session_manager_worker_t *wrk = &session_manager_main.wrk[thread_index];
  stream_session_t *s;
  u8 will_expand = 0;
  pool_get_aligned_will_expand (wrk->sessions, will_expand,
				CLIB_CACHE_LINE_BYTES);
  /* If we have peekers, let them finish */
  if (PREDICT_FALSE (will_expand && vlib_num_workers ()))
    {
      clib_rwlock_writer_lock (&wrk->peekers_rw_locks);
      pool_get_aligned (wrk->sessions, s, CLIB_CACHE_LINE_BYTES);
      clib_rwlock_writer_unlock (&wrk->peekers_rw_locks);
    }
  else
    {
      pool_get_aligned (wrk->sessions, s, CLIB_CACHE_LINE_BYTES);
    }
  clib_memset (s, 0, sizeof (*s));
  s->session_index = s - wrk->sessions;
  s->thread_index = thread_index;
  return s;
}

void
session_free (stream_session_t * s)
{
  pool_put (session_manager_main.wrk[s->thread_index].sessions, s);
  if (CLIB_DEBUG)
    clib_memset (s, 0xFA, sizeof (*s));
}

void
session_free_w_fifos (stream_session_t * s)
{
  segment_manager_dealloc_fifos (s->svm_segment_index, s->server_rx_fifo,
				 s->server_tx_fifo);
  session_free (s);
}

int
session_alloc_fifos (segment_manager_t * sm, stream_session_t * s)
{
  svm_fifo_t *server_rx_fifo = 0, *server_tx_fifo = 0;
  u32 fifo_segment_index;
  int rv;

  if ((rv = segment_manager_alloc_session_fifos (sm, &server_rx_fifo,
						 &server_tx_fifo,
						 &fifo_segment_index)))
    return rv;
  /* Initialize backpointers */
  server_rx_fifo->master_session_index = s->session_index;
  server_rx_fifo->master_thread_index = s->thread_index;

  server_tx_fifo->master_session_index = s->session_index;
  server_tx_fifo->master_thread_index = s->thread_index;

  s->server_rx_fifo = server_rx_fifo;
  s->server_tx_fifo = server_tx_fifo;
  s->svm_segment_index = fifo_segment_index;
  return 0;
}

static stream_session_t *
session_alloc_for_connection (transport_connection_t * tc)
{
  stream_session_t *s;
  u32 thread_index = tc->thread_index;

  ASSERT (thread_index == vlib_get_thread_index ()
	  || transport_protocol_is_cl (tc->proto));

  s = session_alloc (thread_index);
  s->session_type = session_type_from_proto_and_ip (tc->proto, tc->is_ip4);
  s->session_state = SESSION_STATE_CONNECTING;
  s->enqueue_epoch = (u64) ~ 0;

  /* Attach transport to session and vice versa */
  s->connection_index = tc->c_index;
  tc->s_index = s->session_index;
  return s;
}

static int
session_alloc_and_init (segment_manager_t * sm, transport_connection_t * tc,
			u8 alloc_fifos, stream_session_t ** ret_s)
{
  stream_session_t *s;
  int rv;

  s = session_alloc_for_connection (tc);
  if (alloc_fifos && (rv = session_alloc_fifos (sm, s)))
    {
      session_free (s);
      *ret_s = 0;
      return rv;
    }

  /* Add to the main lookup table */
  session_lookup_add_connection (tc, session_handle (s));

  *ret_s = s;
  return 0;
}

/**
 * Discards bytes from buffer chain
 *
 * It discards n_bytes_to_drop starting at first buffer after chain_b
 */
always_inline void
session_enqueue_discard_chain_bytes (vlib_main_t * vm, vlib_buffer_t * b,
				     vlib_buffer_t ** chain_b,
				     u32 n_bytes_to_drop)
{
  vlib_buffer_t *next = *chain_b;
  u32 to_drop = n_bytes_to_drop;
  ASSERT (b->flags & VLIB_BUFFER_NEXT_PRESENT);
  while (to_drop && (next->flags & VLIB_BUFFER_NEXT_PRESENT))
    {
      next = vlib_get_buffer (vm, next->next_buffer);
      if (next->current_length > to_drop)
	{
	  vlib_buffer_advance (next, to_drop);
	  to_drop = 0;
	}
      else
	{
	  to_drop -= next->current_length;
	  next->current_length = 0;
	}
    }
  *chain_b = next;

  if (to_drop == 0)
    b->total_length_not_including_first_buffer -= n_bytes_to_drop;
}

/**
 * Enqueue buffer chain tail
 */
always_inline int
session_enqueue_chain_tail (stream_session_t * s, vlib_buffer_t * b,
			    u32 offset, u8 is_in_order)
{
  vlib_buffer_t *chain_b;
  u32 chain_bi, len, diff;
  vlib_main_t *vm = vlib_get_main ();
  u8 *data;
  u32 written = 0;
  int rv = 0;

  if (is_in_order && offset)
    {
      diff = offset - b->current_length;
      if (diff > b->total_length_not_including_first_buffer)
	return 0;
      chain_b = b;
      session_enqueue_discard_chain_bytes (vm, b, &chain_b, diff);
      chain_bi = vlib_get_buffer_index (vm, chain_b);
    }
  else
    chain_bi = b->next_buffer;

  do
    {
      chain_b = vlib_get_buffer (vm, chain_bi);
      data = vlib_buffer_get_current (chain_b);
      len = chain_b->current_length;
      if (!len)
	continue;
      if (is_in_order)
	{
	  rv = svm_fifo_enqueue_nowait (s->server_rx_fifo, len, data);
	  if (rv == len)
	    {
	      written += rv;
	    }
	  else if (rv < len)
	    {
	      return (rv > 0) ? (written + rv) : written;
	    }
	  else if (rv > len)
	    {
	      written += rv;

	      /* written more than what was left in chain */
	      if (written > b->total_length_not_including_first_buffer)
		return written;

	      /* drop the bytes that have already been delivered */
	      session_enqueue_discard_chain_bytes (vm, b, &chain_b, rv - len);
	    }
	}
      else
	{
	  rv = svm_fifo_enqueue_with_offset (s->server_rx_fifo, offset, len,
					     data);
	  if (rv)
	    {
	      clib_warning ("failed to enqueue multi-buffer seg");
	      return -1;
	    }
	  offset += len;
	}
    }
  while ((chain_bi = (chain_b->flags & VLIB_BUFFER_NEXT_PRESENT)
	  ? chain_b->next_buffer : 0));

  if (is_in_order)
    return written;

  return 0;
}

/*
 * Enqueue data for delivery to session peer. Does not notify peer of enqueue
 * event but on request can queue notification events for later delivery by
 * calling stream_server_flush_enqueue_events().
 *
 * @param tc Transport connection which is to be enqueued data
 * @param b Buffer to be enqueued
 * @param offset Offset at which to start enqueueing if out-of-order
 * @param queue_event Flag to indicate if peer is to be notified or if event
 *                    is to be queued. The former is useful when more data is
 *                    enqueued and only one event is to be generated.
 * @param is_in_order Flag to indicate if data is in order
 * @return Number of bytes enqueued or a negative value if enqueueing failed.
 */
int
session_enqueue_stream_connection (transport_connection_t * tc,
				   vlib_buffer_t * b, u32 offset,
				   u8 queue_event, u8 is_in_order)
{
  stream_session_t *s;
  int enqueued = 0, rv, in_order_off;

  s = session_get (tc->s_index, tc->thread_index);

  if (is_in_order)
    {
      enqueued = svm_fifo_enqueue_nowait (s->server_rx_fifo,
					  b->current_length,
					  vlib_buffer_get_current (b));
      if (PREDICT_FALSE ((b->flags & VLIB_BUFFER_NEXT_PRESENT)
			 && enqueued >= 0))
	{
	  in_order_off = enqueued > b->current_length ? enqueued : 0;
	  rv = session_enqueue_chain_tail (s, b, in_order_off, 1);
	  if (rv > 0)
	    enqueued += rv;
	}
    }
  else
    {
      rv = svm_fifo_enqueue_with_offset (s->server_rx_fifo, offset,
					 b->current_length,
					 vlib_buffer_get_current (b));
      if (PREDICT_FALSE ((b->flags & VLIB_BUFFER_NEXT_PRESENT) && !rv))
	session_enqueue_chain_tail (s, b, offset + b->current_length, 0);
      /* if something was enqueued, report even this as success for ooo
       * segment handling */
      return rv;
    }

  if (queue_event)
    {
      /* Queue RX event on this fifo. Eventually these will need to be flushed
       * by calling stream_server_flush_enqueue_events () */
      session_manager_worker_t *wrk;

      wrk = session_manager_get_worker (s->thread_index);
      if (s->enqueue_epoch != wrk->current_enqueue_epoch[tc->proto])
	{
	  s->enqueue_epoch = wrk->current_enqueue_epoch[tc->proto];
	  vec_add1 (wrk->session_to_enqueue[tc->proto], s->session_index);
	}
    }

  return enqueued;
}

int
session_enqueue_dgram_connection (stream_session_t * s,
				  session_dgram_hdr_t * hdr,
				  vlib_buffer_t * b, u8 proto, u8 queue_event)
{
  int enqueued = 0, rv, in_order_off;

  ASSERT (svm_fifo_max_enqueue (s->server_rx_fifo)
	  >= b->current_length + sizeof (*hdr));

  svm_fifo_enqueue_nowait (s->server_rx_fifo, sizeof (session_dgram_hdr_t),
			   (u8 *) hdr);
  enqueued = svm_fifo_enqueue_nowait (s->server_rx_fifo, b->current_length,
				      vlib_buffer_get_current (b));
  if (PREDICT_FALSE ((b->flags & VLIB_BUFFER_NEXT_PRESENT) && enqueued >= 0))
    {
      in_order_off = enqueued > b->current_length ? enqueued : 0;
      rv = session_enqueue_chain_tail (s, b, in_order_off, 1);
      if (rv > 0)
	enqueued += rv;
    }
  if (queue_event)
    {
      /* Queue RX event on this fifo. Eventually these will need to be flushed
       * by calling stream_server_flush_enqueue_events () */
      session_manager_worker_t *wrk;

      wrk = session_manager_get_worker (s->thread_index);
      if (s->enqueue_epoch != wrk->current_enqueue_epoch[proto])
	{
	  s->enqueue_epoch = wrk->current_enqueue_epoch[proto];
	  vec_add1 (wrk->session_to_enqueue[proto], s->session_index);
	}
    }
  return enqueued;
}

/** Check if we have space in rx fifo to push more bytes */
u8
stream_session_no_space (transport_connection_t * tc, u32 thread_index,
			 u16 data_len)
{
  stream_session_t *s = session_get (tc->s_index, thread_index);

  if (PREDICT_FALSE (s->session_state != SESSION_STATE_READY))
    return 1;

  if (data_len > svm_fifo_max_enqueue (s->server_rx_fifo))
    return 1;

  return 0;
}

u32
session_tx_fifo_max_dequeue (transport_connection_t * tc)
{
  stream_session_t *s = session_get (tc->s_index, tc->thread_index);
  if (!s->server_tx_fifo)
    return 0;
  return svm_fifo_max_dequeue (s->server_tx_fifo);
}

int
stream_session_peek_bytes (transport_connection_t * tc, u8 * buffer,
			   u32 offset, u32 max_bytes)
{
  stream_session_t *s = session_get (tc->s_index, tc->thread_index);
  return svm_fifo_peek (s->server_tx_fifo, offset, max_bytes, buffer);
}

u32
stream_session_dequeue_drop (transport_connection_t * tc, u32 max_bytes)
{
  stream_session_t *s = session_get (tc->s_index, tc->thread_index);
  return svm_fifo_dequeue_drop (s->server_tx_fifo, max_bytes);
}

/**
 * Notify session peer that new data has been enqueued.
 *
 * @param s 	Stream session for which the event is to be generated.
 * @param lock 	Flag to indicate if call should lock message queue.
 *
 * @return 0 on success or negative number if failed to send notification.
 */
static inline int
session_enqueue_notify (stream_session_t * s)
{
  app_worker_t *app;

  app = app_worker_get_if_valid (s->app_wrk_index);
  if (PREDICT_FALSE (!app))
    {
      SESSION_DBG ("invalid s->app_index = %d", s->app_wrk_index);
      return 0;
    }

  /* *INDENT-OFF* */
  SESSION_EVT_DBG(SESSION_EVT_ENQ, s, ({
      ed->data[0] = FIFO_EVENT_APP_RX;
      ed->data[1] = svm_fifo_max_dequeue (s->server_rx_fifo);
  }));
  /* *INDENT-ON* */

  return app_worker_lock_and_send_event (app, s, FIFO_EVENT_APP_RX);
}

int
session_dequeue_notify (stream_session_t * s)
{
  app_worker_t *app;

  app = app_worker_get_if_valid (s->app_wrk_index);
  if (PREDICT_FALSE (!app))
    return -1;

  return app_worker_lock_and_send_event (app, s, FIFO_EVENT_APP_TX);
}

/**
 * Flushes queue of sessions that are to be notified of new data
 * enqueued events.
 *
 * @param thread_index Thread index for which the flush is to be performed.
 * @return 0 on success or a positive number indicating the number of
 *         failures due to API queue being full.
 */
int
session_manager_flush_enqueue_events (u8 transport_proto, u32 thread_index)
{
  session_manager_worker_t *wrk = session_manager_get_worker (thread_index);
  stream_session_t *s;
  int i, errors = 0;
  u32 *indices;

  indices = wrk->session_to_enqueue[transport_proto];

  for (i = 0; i < vec_len (indices); i++)
    {
      s = session_get_if_valid (indices[i], thread_index);
      if (PREDICT_FALSE (!s))
	{
	  errors++;
	  continue;
	}
      if (PREDICT_FALSE (session_enqueue_notify (s)))
	errors++;
    }

  vec_reset_length (indices);
  wrk->session_to_enqueue[transport_proto] = indices;
  wrk->current_enqueue_epoch[transport_proto]++;

  return errors;
}

int
session_manager_flush_all_enqueue_events (u8 transport_proto)
{
  vlib_thread_main_t *vtm = vlib_get_thread_main ();
  int i, errors = 0;
  for (i = 0; i < 1 + vtm->n_threads; i++)
    errors += session_manager_flush_enqueue_events (transport_proto, i);
  return errors;
}

/**
 * Init fifo tail and head pointers
 *
 * Useful if transport uses absolute offsets for tracking ooo segments.
 */
void
stream_session_init_fifos_pointers (transport_connection_t * tc,
				    u32 rx_pointer, u32 tx_pointer)
{
  stream_session_t *s;
  s = session_get (tc->s_index, tc->thread_index);
  svm_fifo_init_pointers (s->server_rx_fifo, rx_pointer);
  svm_fifo_init_pointers (s->server_tx_fifo, tx_pointer);
}

int
session_stream_connect_notify (transport_connection_t * tc, u8 is_fail)
{
  u32 opaque = 0, new_ti, new_si;
  stream_session_t *new_s = 0;
  segment_manager_t *sm;
  app_worker_t *app_wrk;
  application_t *app;
  u8 alloc_fifos;
  int error = 0;
  u64 handle;

  /*
   * Find connection handle and cleanup half-open table
   */
  handle = session_lookup_half_open_handle (tc);
  if (handle == HALF_OPEN_LOOKUP_INVALID_VALUE)
    {
      SESSION_DBG ("half-open was removed!");
      return -1;
    }
  session_lookup_del_half_open (tc);

  /* Get the app's index from the handle we stored when opening connection
   * and the opaque (api_context for external apps) from transport session
   * index */
  app_wrk = app_worker_get_if_valid (handle >> 32);
  if (!app_wrk)
    return -1;
  opaque = tc->s_index;
  app = application_get (app_wrk->app_index);

  /*
   * Allocate new session with fifos (svm segments are allocated if needed)
   */
  if (!is_fail)
    {
      sm = app_worker_get_connect_segment_manager (app_wrk);
      alloc_fifos = !application_is_builtin_proxy (app);
      if (session_alloc_and_init (sm, tc, alloc_fifos, &new_s))
	{
	  is_fail = 1;
	  error = -1;
	}
      else
	{
	  new_s->app_wrk_index = app_wrk->wrk_index;
	  new_si = new_s->session_index;
	  new_ti = new_s->thread_index;
	}
    }

  /*
   * Notify client application
   */
  if (app->cb_fns.session_connected_callback (app_wrk->wrk_index, opaque,
					      new_s, is_fail))
    {
      SESSION_DBG ("failed to notify app");
      if (!is_fail)
	{
	  new_s = session_get (new_si, new_ti);
	  stream_session_disconnect_transport (new_s);
	}
    }
  else
    {
      if (!is_fail)
	{
	  new_s = session_get (new_si, new_ti);
	  new_s->session_state = SESSION_STATE_READY;
	}
    }

  return error;
}

typedef struct _session_switch_pool_args
{
  u32 session_index;
  u32 thread_index;
  u32 new_thread_index;
  u32 new_session_index;
} session_switch_pool_args_t;

static void
session_switch_pool (void *cb_args)
{
  session_switch_pool_args_t *args = (session_switch_pool_args_t *) cb_args;
  transport_proto_t tp;
  stream_session_t *s;
  ASSERT (args->thread_index == vlib_get_thread_index ());
  s = session_get (args->session_index, args->thread_index);
  s->server_tx_fifo->master_session_index = args->new_session_index;
  s->server_tx_fifo->master_thread_index = args->new_thread_index;
  tp = session_get_transport_proto (s);
  tp_vfts[tp].cleanup (s->connection_index, s->thread_index);
  session_free (s);
  clib_mem_free (cb_args);
}

/**
 * Move dgram session to the right thread
 */
int
session_dgram_connect_notify (transport_connection_t * tc,
			      u32 old_thread_index,
			      stream_session_t ** new_session)
{
  stream_session_t *new_s;
  session_switch_pool_args_t *rpc_args;

  /*
   * Clone half-open session to the right thread.
   */
  new_s = session_clone_safe (tc->s_index, old_thread_index);
  new_s->connection_index = tc->c_index;
  new_s->server_rx_fifo->master_session_index = new_s->session_index;
  new_s->server_rx_fifo->master_thread_index = new_s->thread_index;
  new_s->session_state = SESSION_STATE_READY;
  session_lookup_add_connection (tc, session_handle (new_s));

  /*
   * Ask thread owning the old session to clean it up and make us the tx
   * fifo owner
   */
  rpc_args = clib_mem_alloc (sizeof (*rpc_args));
  rpc_args->new_session_index = new_s->session_index;
  rpc_args->new_thread_index = new_s->thread_index;
  rpc_args->session_index = tc->s_index;
  rpc_args->thread_index = old_thread_index;
  session_send_rpc_evt_to_thread (rpc_args->thread_index, session_switch_pool,
				  rpc_args);

  tc->s_index = new_s->session_index;
  new_s->connection_index = tc->c_index;
  *new_session = new_s;
  return 0;
}

void
stream_session_accept_notify (transport_connection_t * tc)
{
  app_worker_t *app_wrk;
  application_t *app;
  stream_session_t *s;

  s = session_get (tc->s_index, tc->thread_index);
  app_wrk = app_worker_get_if_valid (s->app_wrk_index);
  if (!app_wrk)
    return;
  app = application_get (app_wrk->app_index);
  app->cb_fns.session_accept_callback (s);
}

/**
 * Notification from transport that connection is being closed.
 *
 * A disconnect is sent to application but state is not removed. Once
 * disconnect is acknowledged by application, session disconnect is called.
 * Ultimately this leads to close being called on transport (passive close).
 */
void
stream_session_disconnect_notify (transport_connection_t * tc)
{
  app_worker_t *app_wrk;
  application_t *app;
  stream_session_t *s;

  s = session_get (tc->s_index, tc->thread_index);
  if (s->session_state >= SESSION_STATE_TRANSPORT_CLOSING)
    return;
  s->session_state = SESSION_STATE_TRANSPORT_CLOSING;
  app_wrk = app_worker_get_if_valid (s->app_wrk_index);
  if (!app_wrk)
    return;
  app = application_get (app_wrk->app_index);
  app->cb_fns.session_disconnect_callback (s);
}

/**
 * Cleans up session and lookup table.
 *
 * Transport connection must still be valid.
 */
void
stream_session_delete (stream_session_t * s)
{
  int rv;

  /* Delete from the main lookup table. */
  if ((rv = session_lookup_del_session (s)))
    clib_warning ("hash delete error, rv %d", rv);

  session_free_w_fifos (s);
}

/**
 * Notification from transport that connection is being deleted
 *
 * This removes the session if it is still valid. It should be called only on
 * previously fully established sessions. For instance failed connects should
 * call stream_session_connect_notify and indicate that the connect has
 * failed.
 */
void
stream_session_delete_notify (transport_connection_t * tc)
{
  stream_session_t *s;

  /* App might've been removed already */
  if (!(s = session_get_if_valid (tc->s_index, tc->thread_index)))
    return;

  /* Make sure we don't try to send anything more */
  svm_fifo_dequeue_drop_all (s->server_tx_fifo);

  switch (s->session_state)
    {
    case SESSION_STATE_TRANSPORT_CLOSING:
      /* If transport finishes or times out before we get a reply
       * from the app, do the whole disconnect since we might still
       * have lingering events */
      stream_session_disconnect (s);
      s->session_state = SESSION_STATE_CLOSED;
      break;
    case SESSION_STATE_CLOSING:
      /* Cleanup lookup table. Transport needs to still be valid */
      session_lookup_del_session (s);
      s->session_state = SESSION_STATE_CLOSED;
      break;
    case SESSION_STATE_CLOSED:
    case SESSION_STATE_ACCEPTING:
      stream_session_delete (s);
      break;
    default:
      /* Assume connection was not yet added the lookup table */
      session_free_w_fifos (s);
      break;
    }
}

/**
 * Notify application that connection has been reset.
 */
void
stream_session_reset_notify (transport_connection_t * tc)
{
  stream_session_t *s;
  app_worker_t *app_wrk;
  application_t *app;
  s = session_get (tc->s_index, tc->thread_index);
  s->session_state = SESSION_STATE_CLOSED;
  app_wrk = app_worker_get (s->app_wrk_index);
  app = application_get (app_wrk->app_index);
  app->cb_fns.session_reset_callback (s);
}

/**
 * Accept a stream session. Optionally ping the server by callback.
 */
int
stream_session_accept (transport_connection_t * tc, u32 listener_index,
		       u8 notify)
{
  stream_session_t *s, *listener;
  app_worker_t *app_wrk;
  segment_manager_t *sm;
  int rv;

  /* Find the server */
  listener = listen_session_get (listener_index);
  app_wrk = application_listener_select_worker (listener, 0);

  sm = app_worker_get_listen_segment_manager (app_wrk, listener);
  if ((rv = session_alloc_and_init (sm, tc, 1, &s)))
    return rv;

  s->app_wrk_index = app_wrk->wrk_index;
  s->listener_index = listener_index;
  s->session_state = SESSION_STATE_ACCEPTING;

  /* Shoulder-tap the server */
  if (notify)
    {
      application_t *app = application_get (app_wrk->app_index);
      app->cb_fns.session_accept_callback (s);
    }

  return 0;
}

int
session_open_cl (u32 app_wrk_index, session_endpoint_t * rmt, u32 opaque)
{
  transport_connection_t *tc;
  transport_endpoint_cfg_t *tep;
  segment_manager_t *sm;
  app_worker_t *app_wrk;
  stream_session_t *s;
  application_t *app;
  int rv;

  tep = session_endpoint_to_transport_cfg (rmt);
  rv = tp_vfts[rmt->transport_proto].open (tep);
  if (rv < 0)
    {
      SESSION_DBG ("Transport failed to open connection.");
      return VNET_API_ERROR_SESSION_CONNECT;
    }

  tc = tp_vfts[rmt->transport_proto].get_half_open ((u32) rv);

  /* For dgram type of service, allocate session and fifos now.
   */
  app_wrk = app_worker_get (app_wrk_index);
  sm = app_worker_get_connect_segment_manager (app_wrk);

  if (session_alloc_and_init (sm, tc, 1, &s))
    return -1;
  s->app_wrk_index = app_wrk->wrk_index;
  s->session_state = SESSION_STATE_OPENED;

  /* Tell the app about the new event fifo for this session */
  app = application_get (app_wrk->app_index);
  app->cb_fns.session_connected_callback (app_wrk->wrk_index, opaque, s, 0);

  return 0;
}

int
session_open_vc (u32 app_wrk_index, session_endpoint_t * rmt, u32 opaque)
{
  transport_connection_t *tc;
  transport_endpoint_cfg_t *tep;
  u64 handle;
  int rv;

  tep = session_endpoint_to_transport_cfg (rmt);
  rv = tp_vfts[rmt->transport_proto].open (tep);
  if (rv < 0)
    {
      SESSION_DBG ("Transport failed to open connection.");
      return VNET_API_ERROR_SESSION_CONNECT;
    }

  tc = tp_vfts[rmt->transport_proto].get_half_open ((u32) rv);

  /* If transport offers a stream service, only allocate session once the
   * connection has been established.
   * Add connection to half-open table and save app and tc index. The
   * latter is needed to help establish the connection while the former
   * is needed when the connect notify comes and we have to notify the
   * external app
   */
  handle = (((u64) app_wrk_index) << 32) | (u64) tc->c_index;
  session_lookup_add_half_open (tc, handle);

  /* Store api_context (opaque) for when the reply comes. Not the nicest
   * thing but better than allocating a separate half-open pool.
   */
  tc->s_index = opaque;
  return 0;
}

int
session_open_app (u32 app_wrk_index, session_endpoint_t * rmt, u32 opaque)
{
  session_endpoint_cfg_t *sep = (session_endpoint_cfg_t *) rmt;
  transport_endpoint_cfg_t *tep_cfg = session_endpoint_to_transport_cfg (sep);

  sep->app_wrk_index = app_wrk_index;
  sep->opaque = opaque;

  return tp_vfts[rmt->transport_proto].open (tep_cfg);
}

typedef int (*session_open_service_fn) (u32, session_endpoint_t *, u32);

/* *INDENT-OFF* */
static session_open_service_fn session_open_srv_fns[TRANSPORT_N_SERVICES] = {
  session_open_vc,
  session_open_cl,
  session_open_app,
};
/* *INDENT-ON* */

/**
 * Ask transport to open connection to remote transport endpoint.
 *
 * Stores handle for matching request with reply since the call can be
 * asynchronous. For instance, for TCP the 3-way handshake must complete
 * before reply comes. Session is only created once connection is established.
 *
 * @param app_index Index of the application requesting the connect
 * @param st Session type requested.
 * @param tep Remote transport endpoint
 * @param opaque Opaque data (typically, api_context) the application expects
 * 		 on open completion.
 */
int
session_open (u32 app_wrk_index, session_endpoint_t * rmt, u32 opaque)
{
  transport_service_type_t tst = tp_vfts[rmt->transport_proto].service_type;
  return session_open_srv_fns[tst] (app_wrk_index, rmt, opaque);
}

/**
 * Ask transport to listen on session endpoint.
 *
 * @param s Session for which listen will be called. Note that unlike
 * 	    established sessions, listen sessions are not associated to a
 * 	    thread.
 * @param sep Local endpoint to be listened on.
 */
int
session_listen (stream_session_t * ls, session_endpoint_cfg_t * sep)
{
  transport_connection_t *tc;
  transport_endpoint_t *tep;
  u32 tc_index, s_index;

  /* Transport bind/listen */
  tep = session_endpoint_to_transport (sep);
  s_index = ls->session_index;
  tc_index = tp_vfts[sep->transport_proto].bind (s_index, tep);

  if (tc_index == (u32) ~ 0)
    return -1;

  /* Attach transport to session */
  ls = listen_session_get (s_index);
  ls->connection_index = tc_index;

  /* Add to the main lookup table after transport was initialized */
  tc = tp_vfts[sep->transport_proto].get_listener (tc_index);
  session_lookup_add_connection (tc, s_index);
  return 0;
}

/**
 * Ask transport to stop listening on local transport endpoint.
 *
 * @param s Session to stop listening on. It must be in state LISTENING.
 */
int
session_stop_listen (stream_session_t * s)
{
  transport_proto_t tp = session_get_transport_proto (s);
  transport_connection_t *tc;
  if (s->session_state != SESSION_STATE_LISTENING)
    {
      clib_warning ("not a listening session");
      return -1;
    }

  tc = tp_vfts[tp].get_listener (s->connection_index);
  if (!tc)
    {
      clib_warning ("no transport");
      return VNET_API_ERROR_ADDRESS_NOT_IN_USE;
    }

  session_lookup_del_connection (tc);
  tp_vfts[tp].unbind (s->connection_index);
  return 0;
}

/**
 * Initialize session disconnect.
 *
 * Request is always sent to session node to ensure that all outstanding
 * requests are served before transport is notified.
 */
void
stream_session_disconnect (stream_session_t * s)
{
  u32 thread_index = vlib_get_thread_index ();
  session_manager_worker_t *wrk;
  session_event_t *evt;

  if (!s)
    return;

  if (s->session_state >= SESSION_STATE_CLOSING)
    {
      /* Session already closed. Clear the tx fifo */
      if (s->session_state == SESSION_STATE_CLOSED)
	svm_fifo_dequeue_drop_all (s->server_tx_fifo);
      return;
    }

  s->session_state = SESSION_STATE_CLOSING;

  /* If we are in the handler thread, or being called with the worker barrier
   * held, just append a new event to pending disconnects vector. */
  if (vlib_thread_is_main_w_barrier () || thread_index == s->thread_index)
    {
      wrk = session_manager_get_worker (s->thread_index);
      vec_add2 (wrk->pending_disconnects, evt, 1);
      clib_memset (evt, 0, sizeof (*evt));
      evt->session_handle = session_handle (s);
      evt->event_type = FIFO_EVENT_DISCONNECT;
    }
  else
    session_send_ctrl_evt_to_thread (s, FIFO_EVENT_DISCONNECT);
}

/**
 * Notify transport the session can be disconnected. This should eventually
 * result in a delete notification that allows us to cleanup session state.
 * Called for both active/passive disconnects.
 *
 * Must be called from the session's thread.
 */
void
stream_session_disconnect_transport (stream_session_t * s)
{
  /* If transport is already closed, just free the session */
  if (s->session_state == SESSION_STATE_CLOSED)
    {
      session_free_w_fifos (s);
      return;
    }
  s->session_state = SESSION_STATE_CLOSED;
  tp_vfts[session_get_transport_proto (s)].close (s->connection_index,
						  s->thread_index);
}

/**
 * Cleanup transport and session state.
 *
 * Notify transport of the cleanup and free the session. This should
 * be called only if transport reported some error and is already
 * closed.
 */
void
stream_session_cleanup (stream_session_t * s)
{
  s->session_state = SESSION_STATE_CLOSED;

  /* Delete from main lookup table before we axe the the transport */
  session_lookup_del_session (s);
  tp_vfts[session_get_transport_proto (s)].cleanup (s->connection_index,
						    s->thread_index);
  /* Since we called cleanup, no delete notification will come. So, make
   * sure the session is properly freed. */
  session_free_w_fifos (s);
}

transport_service_type_t
session_transport_service_type (stream_session_t * s)
{
  transport_proto_t tp;
  tp = session_get_transport_proto (s);
  return transport_protocol_service_type (tp);
}

transport_tx_fn_type_t
session_transport_tx_fn_type (stream_session_t * s)
{
  transport_proto_t tp;
  tp = session_get_transport_proto (s);
  return transport_protocol_tx_fn_type (tp);
}

u8
session_tx_is_dgram (stream_session_t * s)
{
  return (session_transport_tx_fn_type (s) == TRANSPORT_TX_DGRAM);
}

/**
 * Allocate event queues in the shared-memory segment
 *
 * That can either be a newly created memfd segment, that will need to be
 * mapped by all stack users, or the binary api's svm region. The latter is
 * assumed to be already mapped. NOTE that this assumption DOES NOT hold if
 * api clients bootstrap shm api over sockets (i.e. use memfd segments) and
 * vpp uses api svm region for event queues.
 */
void
session_vpp_event_queues_allocate (session_manager_main_t * smm)
{
  u32 evt_q_length = 2048, evt_size = sizeof (session_event_t);
  ssvm_private_t *eqs = &smm->evt_qs_segment;
  api_main_t *am = &api_main;
  u64 eqs_size = 64 << 20;
  pid_t vpp_pid = getpid ();
  void *oldheap;
  int i;

  if (smm->configured_event_queue_length)
    evt_q_length = smm->configured_event_queue_length;

  if (smm->evt_qs_use_memfd_seg)
    {
      if (smm->evt_qs_segment_size)
	eqs_size = smm->evt_qs_segment_size;

      eqs->ssvm_size = eqs_size;
      eqs->i_am_master = 1;
      eqs->my_pid = vpp_pid;
      eqs->name = format (0, "%s%c", "evt-qs-segment", 0);
      eqs->requested_va = smm->session_baseva;

      if (ssvm_master_init (eqs, SSVM_SEGMENT_MEMFD))
	{
	  clib_warning ("failed to initialize queue segment");
	  return;
	}
    }

  if (smm->evt_qs_use_memfd_seg)
    oldheap = ssvm_push_heap (eqs->sh);
  else
    oldheap = svm_push_data_heap (am->vlib_rp);

  for (i = 0; i < vec_len (smm->wrk); i++)
    {
      svm_msg_q_cfg_t _cfg, *cfg = &_cfg;
      u32 notif_q_size = clib_max (16, evt_q_length >> 4);
      svm_msg_q_ring_cfg_t rc[SESSION_MQ_N_RINGS] = {
	{evt_q_length, evt_size, 0}
	,
	{notif_q_size, 256, 0}
      };
      cfg->consumer_pid = 0;
      cfg->n_rings = 2;
      cfg->q_nitems = evt_q_length;
      cfg->ring_cfgs = rc;
      smm->wrk[i].vpp_event_queue = svm_msg_q_alloc (cfg);
      if (smm->evt_qs_use_memfd_seg)
	{
	  if (svm_msg_q_alloc_consumer_eventfd (smm->wrk[i].vpp_event_queue))
	    clib_warning ("eventfd returned");
	}
    }

  if (smm->evt_qs_use_memfd_seg)
    ssvm_pop_heap (oldheap);
  else
    svm_pop_heap (oldheap);
}

ssvm_private_t *
session_manager_get_evt_q_segment (void)
{
  session_manager_main_t *smm = &session_manager_main;
  if (smm->evt_qs_use_memfd_seg)
    return &smm->evt_qs_segment;
  return 0;
}

/* *INDENT-OFF* */
static session_fifo_rx_fn *session_tx_fns[TRANSPORT_TX_N_FNS] = {
    session_tx_fifo_peek_and_snd,
    session_tx_fifo_dequeue_and_snd,
    session_tx_fifo_dequeue_internal,
    session_tx_fifo_dequeue_and_snd
};
/* *INDENT-ON* */

/**
 * Initialize session layer for given transport proto and ip version
 *
 * Allocates per session type (transport proto + ip version) data structures
 * and adds arc from session queue node to session type output node.
 */
void
session_register_transport (transport_proto_t transport_proto,
			    const transport_proto_vft_t * vft, u8 is_ip4,
			    u32 output_node)
{
  session_manager_main_t *smm = &session_manager_main;
  session_type_t session_type;
  u32 next_index = ~0;

  session_type = session_type_from_proto_and_ip (transport_proto, is_ip4);

  vec_validate (smm->session_type_to_next, session_type);
  vec_validate (smm->session_tx_fns, session_type);

  /* *INDENT-OFF* */
  if (output_node != ~0)
    {
      foreach_vlib_main (({
          next_index = vlib_node_add_next (this_vlib_main,
                                           session_queue_node.index,
                                           output_node);
      }));
    }
  /* *INDENT-ON* */

  smm->session_type_to_next[session_type] = next_index;
  smm->session_tx_fns[session_type] = session_tx_fns[vft->tx_type];
}

transport_connection_t *
session_get_transport (stream_session_t * s)
{
  transport_proto_t tp;
  if (s->session_state != SESSION_STATE_LISTENING)
    {
      tp = session_get_transport_proto (s);
      return tp_vfts[tp].get_connection (s->connection_index,
					 s->thread_index);
    }
  return 0;
}

transport_connection_t *
listen_session_get_transport (stream_session_t * s)
{
  transport_proto_t tp = session_get_transport_proto (s);
  return tp_vfts[tp].get_listener (s->connection_index);
}

int
listen_session_get_local_session_endpoint (stream_session_t * listener,
					   session_endpoint_t * sep)
{
  transport_proto_t tp = session_get_transport_proto (listener);
  transport_connection_t *tc;
  tc = tp_vfts[tp].get_listener (listener->connection_index);
  if (!tc)
    {
      clib_warning ("no transport");
      return -1;
    }

  /* N.B. The ip should not be copied because this is the local endpoint */
  sep->port = tc->lcl_port;
  sep->transport_proto = tc->proto;
  sep->is_ip4 = tc->is_ip4;
  return 0;
}

void
session_flush_frames_main_thread (vlib_main_t * vm)
{
  ASSERT (vlib_get_thread_index () == 0);
  vlib_process_signal_event_mt (vm, session_queue_process_node.index,
				SESSION_Q_PROCESS_FLUSH_FRAMES, 0);
}

static clib_error_t *
session_manager_main_enable (vlib_main_t * vm)
{
  segment_manager_main_init_args_t _sm_args = { 0 }, *sm_args = &_sm_args;
  session_manager_main_t *smm = &session_manager_main;
  vlib_thread_main_t *vtm = vlib_get_thread_main ();
  u32 num_threads, preallocated_sessions_per_worker;
  session_manager_worker_t *wrk;
  int i, j;

  num_threads = 1 /* main thread */  + vtm->n_threads;

  if (num_threads < 1)
    return clib_error_return (0, "n_thread_stacks not set");

  /* Allocate cache line aligned worker contexts */
  vec_validate_aligned (smm->wrk, num_threads - 1, CLIB_CACHE_LINE_BYTES);

  for (i = 0; i < TRANSPORT_N_PROTO; i++)
    {
      for (j = 0; j < num_threads; j++)
	smm->wrk[j].current_enqueue_epoch[i] = 1;
    }

  for (i = 0; i < num_threads; i++)
    {
      wrk = &smm->wrk[i];
      vec_validate (wrk->free_event_vector, 128);
      _vec_len (wrk->free_event_vector) = 0;
      vec_validate (wrk->pending_event_vector, 128);
      _vec_len (wrk->pending_event_vector) = 0;
      vec_validate (wrk->pending_disconnects, 128);
      _vec_len (wrk->pending_disconnects) = 0;
      vec_validate (wrk->postponed_event_vector, 128);
      _vec_len (wrk->postponed_event_vector) = 0;

      wrk->last_vlib_time = vlib_time_now (vlib_mains[i]);
      wrk->dispatch_period = 500e-6;

      if (num_threads > 1)
	clib_rwlock_init (&smm->wrk[i].peekers_rw_locks);
    }

#if SESSION_DEBUG
  vec_validate (smm->last_event_poll_by_thread, num_threads - 1);
#endif

  /* Allocate vpp event queues segment and queue */
  session_vpp_event_queues_allocate (smm);

  /* Initialize fifo segment main baseva and timeout */
  sm_args->baseva = smm->session_baseva + smm->evt_qs_segment_size;
  sm_args->size = smm->session_va_space_size;
  segment_manager_main_init (sm_args);

  /* Preallocate sessions */
  if (smm->preallocated_sessions)
    {
      if (num_threads == 1)
	{
	  pool_init_fixed (smm->wrk[0].sessions, smm->preallocated_sessions);
	}
      else
	{
	  int j;
	  preallocated_sessions_per_worker =
	    (1.1 * (f64) smm->preallocated_sessions /
	     (f64) (num_threads - 1));

	  for (j = 1; j < num_threads; j++)
	    {
	      pool_init_fixed (smm->wrk[j].sessions,
			       preallocated_sessions_per_worker);
	    }
	}
    }

  session_lookup_init ();
  app_namespaces_init ();
  transport_init ();

  smm->is_enabled = 1;

  /* Enable transports */
  transport_enable_disable (vm, 1);
  transport_init_tx_pacers_period ();
  return 0;
}

void
session_node_enable_disable (u8 is_en)
{
  u8 state = is_en ? VLIB_NODE_STATE_POLLING : VLIB_NODE_STATE_DISABLED;
  vlib_thread_main_t *vtm = vlib_get_thread_main ();
  u8 have_workers = vtm->n_threads != 0;

  /* *INDENT-OFF* */
  foreach_vlib_main (({
    if (have_workers && ii == 0)
      {
	vlib_node_set_state (this_vlib_main, session_queue_process_node.index,
	                     state);
	if (is_en)
	  {
	    vlib_node_t *n = vlib_get_node (this_vlib_main,
	                                    session_queue_process_node.index);
	    vlib_start_process (this_vlib_main, n->runtime_index);
	  }
	else
	  {
	    vlib_process_signal_event_mt (this_vlib_main,
	                                  session_queue_process_node.index,
	                                  SESSION_Q_PROCESS_STOP, 0);
	  }

	continue;
      }
    vlib_node_set_state (this_vlib_main, session_queue_node.index,
                         state);
  }));
  /* *INDENT-ON* */
}

clib_error_t *
vnet_session_enable_disable (vlib_main_t * vm, u8 is_en)
{
  clib_error_t *error = 0;
  if (is_en)
    {
      if (session_manager_main.is_enabled)
	return 0;

      session_node_enable_disable (is_en);
      error = session_manager_main_enable (vm);
    }
  else
    {
      session_manager_main.is_enabled = 0;
      session_node_enable_disable (is_en);
    }

  return error;
}

clib_error_t *
session_manager_main_init (vlib_main_t * vm)
{
  session_manager_main_t *smm = &session_manager_main;
  smm->session_baseva = 0x200000000ULL;
  smm->session_va_space_size = (u64) 128 << 30;
  smm->evt_qs_segment_size = 64 << 20;
  smm->is_enabled = 0;
  return 0;
}

VLIB_INIT_FUNCTION (session_manager_main_init);

static clib_error_t *
session_config_fn (vlib_main_t * vm, unformat_input_t * input)
{
  session_manager_main_t *smm = &session_manager_main;
  u32 nitems;
  uword tmp;

  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
    {
      if (unformat (input, "event-queue-length %d", &nitems))
	{
	  if (nitems >= 2048)
	    smm->configured_event_queue_length = nitems;
	  else
	    clib_warning ("event queue length %d too small, ignored", nitems);
	}
      else if (unformat (input, "preallocated-sessions %d",
			 &smm->preallocated_sessions))
	;
      else if (unformat (input, "v4-session-table-buckets %d",
			 &smm->configured_v4_session_table_buckets))
	;
      else if (unformat (input, "v4-halfopen-table-buckets %d",
			 &smm->configured_v4_halfopen_table_buckets))
	;
      else if (unformat (input, "v6-session-table-buckets %d",
			 &smm->configured_v6_session_table_buckets))
	;
      else if (unformat (input, "v6-halfopen-table-buckets %d",
			 &smm->configured_v6_halfopen_table_buckets))
	;
      else if (unformat (input, "v4-session-table-memory %U",
			 unformat_memory_size, &tmp))
	{
	  if (tmp >= 0x100000000)
	    return clib_error_return (0, "memory size %llx (%lld) too large",
				      tmp, tmp);
	  smm->configured_v4_session_table_memory = tmp;
	}
      else if (unformat (input, "v4-halfopen-table-memory %U",
			 unformat_memory_size, &tmp))
	{
	  if (tmp >= 0x100000000)
	    return clib_error_return (0, "memory size %llx (%lld) too large",
				      tmp, tmp);
	  smm->configured_v4_halfopen_table_memory = tmp;
	}
      else if (unformat (input, "v6-session-table-memory %U",
			 unformat_memory_size, &tmp))
	{
	  if (tmp >= 0x100000000)
	    return clib_error_return (0, "memory size %llx (%lld) too large",
				      tmp, tmp);
	  smm->configured_v6_session_table_memory = tmp;
	}
      else if (unformat (input, "v6-halfopen-table-memory %U",
			 unformat_memory_size, &tmp))
	{
	  if (tmp >= 0x100000000)
	    return clib_error_return (0, "memory size %llx (%lld) too large",
				      tmp, tmp);
	  smm->configured_v6_halfopen_table_memory = tmp;
	}
      else if (unformat (input, "local-endpoints-table-memory %U",
			 unformat_memory_size, &tmp))
	{
	  if (tmp >= 0x100000000)
	    return clib_error_return (0, "memory size %llx (%lld) too large",
				      tmp, tmp);
	  smm->local_endpoints_table_memory = tmp;
	}
      else if (unformat (input, "local-endpoints-table-buckets %d",
			 &smm->local_endpoints_table_buckets))
	;
      else if (unformat (input, "evt_qs_memfd_seg"))
	smm->evt_qs_use_memfd_seg = 1;
      else
	return clib_error_return (0, "unknown input `%U'",
				  format_unformat_error, input);
    }
  return 0;
}

VLIB_CONFIG_FUNCTION (session_config_fn, "session");

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