FD.io VPP  v18.04-17-g3a0d853
Vector Packet Processing
tcp.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2016 Cisco and/or its affiliates.
3  * Licensed under the Apache License, Version 2.0 (the "License");
4  * you may not use this file except in compliance with the License.
5  * You may obtain a copy of the License at:
6  *
7  * http://www.apache.org/licenses/LICENSE-2.0
8  *
9  * Unless required by applicable law or agreed to in writing, software
10  * distributed under the License is distributed on an "AS IS" BASIS,
11  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12  * See the License for the specific language governing permissions and
13  * limitations under the License.
14  */
15 
16 /**
17  * @file
18  * @brief TCP host stack utilities
19  */
20 
21 #include <vnet/tcp/tcp.h>
22 #include <vnet/session/session.h>
23 #include <vnet/fib/fib.h>
24 #include <vnet/dpo/load_balance.h>
25 #include <vnet/dpo/receive_dpo.h>
26 #include <vnet/ip/ip6_neighbor.h>
27 #include <math.h>
28 
30 
31 static u32
33 {
34  tcp_main_t *tm = &tcp_main;
35  tcp_connection_t *listener;
36  void *iface_ip;
37 
38  pool_get (tm->listener_pool, listener);
39  memset (listener, 0, sizeof (*listener));
40 
41  listener->c_c_index = listener - tm->listener_pool;
42  listener->c_lcl_port = lcl->port;
43 
44  /* If we are provided a sw_if_index, bind using one of its ips */
45  if (ip_is_zero (&lcl->ip, 1) && lcl->sw_if_index != ENDPOINT_INVALID_INDEX)
46  {
47  if ((iface_ip = ip_interface_get_first_ip (lcl->sw_if_index,
48  lcl->is_ip4)))
49  ip_set (&lcl->ip, iface_ip, lcl->is_ip4);
50  }
51  ip_copy (&listener->c_lcl_ip, &lcl->ip, lcl->is_ip4);
52  listener->c_is_ip4 = lcl->is_ip4;
53  listener->c_proto = TRANSPORT_PROTO_TCP;
54  listener->c_s_index = session_index;
55  listener->c_fib_index = lcl->fib_index;
56  listener->state = TCP_STATE_LISTEN;
57 
58  tcp_connection_timers_init (listener);
59 
60  TCP_EVT_DBG (TCP_EVT_BIND, listener);
61 
62  return listener->c_c_index;
63 }
64 
65 u32
67 {
68  return tcp_connection_bind (session_index, tep);
69 }
70 
71 static void
72 tcp_connection_unbind (u32 listener_index)
73 {
75  tcp_connection_t *tc;
76 
77  tc = pool_elt_at_index (tm->listener_pool, listener_index);
78 
79  TCP_EVT_DBG (TCP_EVT_UNBIND, tc);
80 
81  /* Poison the entry */
82  if (CLIB_DEBUG > 0)
83  memset (tc, 0xFA, sizeof (*tc));
84 
85  pool_put_index (tm->listener_pool, listener_index);
86 }
87 
88 u32
89 tcp_session_unbind (u32 listener_index)
90 {
91  tcp_connection_unbind (listener_index);
92  return 0;
93 }
94 
96 tcp_session_get_listener (u32 listener_index)
97 {
99  tcp_connection_t *tc;
100  tc = pool_elt_at_index (tm->listener_pool, listener_index);
101  return &tc->connection;
102 }
103 
104 /**
105  * Cleanup half-open connection
106  *
107  */
108 void
110 {
111  tcp_main_t *tm = vnet_get_tcp_main ();
112  clib_spinlock_lock_if_init (&tm->half_open_lock);
113  pool_put_index (tm->half_open_connections, tc->c_c_index);
114  if (CLIB_DEBUG)
115  memset (tc, 0xFA, sizeof (*tc));
116  clib_spinlock_unlock_if_init (&tm->half_open_lock);
117 }
118 
119 /**
120  * Try to cleanup half-open connection
121  *
122  * If called from a thread that doesn't own tc, the call won't have any
123  * effect.
124  *
125  * @param tc - connection to be cleaned up
126  * @return non-zero if cleanup failed.
127  */
128 int
130 {
131  /* Make sure this is the owning thread */
132  if (tc->c_thread_index != vlib_get_thread_index ())
133  return 1;
134  tcp_timer_reset (tc, TCP_TIMER_ESTABLISH);
135  tcp_timer_reset (tc, TCP_TIMER_RETRANSMIT_SYN);
137  return 0;
138 }
139 
142 {
143  tcp_main_t *tm = vnet_get_tcp_main ();
144  tcp_connection_t *tc = 0;
145  ASSERT (vlib_get_thread_index () == 0);
146  pool_get (tm->half_open_connections, tc);
147  memset (tc, 0, sizeof (*tc));
148  tc->c_c_index = tc - tm->half_open_connections;
149  return tc;
150 }
151 
152 /**
153  * Cleans up connection state.
154  *
155  * No notifications.
156  */
157 void
159 {
160  tcp_main_t *tm = &tcp_main;
161 
162  /* Cleanup local endpoint if this was an active connect */
164  tc->c_lcl_port);
165 
166  /* Check if connection is not yet fully established */
167  if (tc->state == TCP_STATE_SYN_SENT)
168  {
169  /* Try to remove the half-open connection. If this is not the owning
170  * thread, tc won't be removed. Retransmit or establish timers will
171  * eventually expire and call again cleanup on the right thread. */
173  }
174  else
175  {
176  int thread_index = tc->c_thread_index;
177 
178  /* Make sure all timers are cleared */
180 
181  /* Poison the entry */
182  if (CLIB_DEBUG > 0)
183  memset (tc, 0xFA, sizeof (*tc));
184  pool_put (tm->connections[thread_index], tc);
185  }
186 }
187 
188 /**
189  * Connection removal.
190  *
191  * This should be called only once connection enters CLOSED state. Note
192  * that it notifies the session of the removal event, so if the goal is to
193  * just remove the connection, call tcp_connection_cleanup instead.
194  */
195 void
197 {
198  TCP_EVT_DBG (TCP_EVT_DELETE, tc);
199  stream_session_delete_notify (&tc->connection);
201 }
202 
204 tcp_connection_new (u8 thread_index)
205 {
206  tcp_main_t *tm = vnet_get_tcp_main ();
207  tcp_connection_t *tc;
208 
209  pool_get (tm->connections[thread_index], tc);
210  memset (tc, 0, sizeof (*tc));
211  tc->c_c_index = tc - tm->connections[thread_index];
212  tc->c_thread_index = thread_index;
213  return tc;
214 }
215 
216 /** Notify session that connection has been reset.
217  *
218  * Switch state to closed and wait for session to call cleanup.
219  */
220 void
222 {
223  TCP_EVT_DBG (TCP_EVT_RST_RCVD, tc);
224  switch (tc->state)
225  {
226  case TCP_STATE_SYN_RCVD:
227  /* Cleanup everything. App wasn't notified yet */
228  stream_session_delete_notify (&tc->connection);
230  break;
231  case TCP_STATE_SYN_SENT:
232  session_stream_connect_notify (&tc->connection, 1 /* fail */ );
234  break;
235  case TCP_STATE_ESTABLISHED:
236  stream_session_reset_notify (&tc->connection);
237  /* fall through */
238  case TCP_STATE_CLOSE_WAIT:
239  case TCP_STATE_FIN_WAIT_1:
240  case TCP_STATE_FIN_WAIT_2:
241  case TCP_STATE_CLOSING:
242  tc->state = TCP_STATE_CLOSED;
243  TCP_EVT_DBG (TCP_EVT_STATE_CHANGE, tc);
244 
245  /* Make sure all timers are cleared */
247 
248  /* Wait for cleanup from session layer but not forever */
249  tcp_timer_update (tc, TCP_TIMER_WAITCLOSE, TCP_CLEANUP_TIME);
250  break;
251  case TCP_STATE_CLOSED:
252  return;
253  }
254 }
255 
256 /**
257  * Begin connection closing procedure.
258  *
259  * If at the end the connection is not in CLOSED state, it is not removed.
260  * Instead, we rely on on TCP to advance through state machine to either
261  * 1) LAST_ACK (passive close) whereby when the last ACK is received
262  * tcp_connection_del is called. This notifies session of the delete and
263  * calls cleanup.
264  * 2) TIME_WAIT (active close) whereby after 2MSL the 2MSL timer triggers
265  * and cleanup is called.
266  *
267  * N.B. Half-close connections are not supported
268  */
269 void
271 {
272  TCP_EVT_DBG (TCP_EVT_CLOSE, tc);
273 
274  /* Send/Program FIN if needed and switch state */
275  switch (tc->state)
276  {
277  case TCP_STATE_SYN_SENT:
278  tc->state = TCP_STATE_CLOSED;
279  break;
280  case TCP_STATE_SYN_RCVD:
281  tcp_send_fin (tc);
282  tc->state = TCP_STATE_FIN_WAIT_1;
283  break;
284  case TCP_STATE_ESTABLISHED:
285  if (!stream_session_tx_fifo_max_dequeue (&tc->connection))
286  tcp_send_fin (tc);
287  else
288  tc->flags |= TCP_CONN_FINPNDG;
289  tc->state = TCP_STATE_FIN_WAIT_1;
290  break;
291  case TCP_STATE_CLOSE_WAIT:
293  tcp_send_fin (tc);
294  tc->state = TCP_STATE_LAST_ACK;
295  tcp_timer_update (tc, TCP_TIMER_WAITCLOSE, TCP_2MSL_TIME);
296  break;
297  case TCP_STATE_FIN_WAIT_1:
298  tcp_timer_update (tc, TCP_TIMER_WAITCLOSE, TCP_2MSL_TIME);
299  break;
300  default:
301  TCP_DBG ("state: %u", tc->state);
302  }
303 
304  TCP_EVT_DBG (TCP_EVT_STATE_CHANGE, tc);
305 
306  /* If in CLOSED and WAITCLOSE timer is not set, delete connection now */
307  if (!tcp_timer_is_active (tc, TCP_TIMER_WAITCLOSE)
308  && tc->state == TCP_STATE_CLOSED)
309  tcp_connection_del (tc);
310 }
311 
312 void
313 tcp_session_close (u32 conn_index, u32 thread_index)
314 {
315  tcp_connection_t *tc;
316  tc = tcp_connection_get (conn_index, thread_index);
318 }
319 
320 void
321 tcp_session_cleanup (u32 conn_index, u32 thread_index)
322 {
323  tcp_connection_t *tc;
324  tc = tcp_connection_get (conn_index, thread_index);
326 
327  /* Wait for the session tx events to clear */
328  tc->state = TCP_STATE_CLOSED;
329  TCP_EVT_DBG (TCP_EVT_STATE_CHANGE, tc);
330  tcp_timer_update (tc, TCP_TIMER_WAITCLOSE, TCP_CLEANUP_TIME);
331 }
332 
333 /**
334  * Initialize all connection timers as invalid
335  */
336 void
338 {
339  int i;
340 
341  /* Set all to invalid */
342  for (i = 0; i < TCP_N_TIMERS; i++)
343  {
344  tc->timers[i] = TCP_TIMER_HANDLE_INVALID;
345  }
346 
347  tc->rto = TCP_RTO_INIT;
348 }
349 
350 /**
351  * Stop all connection timers
352  */
353 void
355 {
356  int i;
357  for (i = 0; i < TCP_N_TIMERS; i++)
358  {
359  tcp_timer_reset (tc, i);
360  }
361 }
362 
363 #if 0
364 typedef struct ip4_tcp_hdr
365 {
366  ip4_header_t ip;
367  tcp_header_t tcp;
368 } ip4_tcp_hdr_t;
369 
370 typedef struct ip6_tcp_hdr
371 {
372  ip6_header_t ip;
373  tcp_header_t tcp;
374 } ip6_tcp_hdr_t;
375 
376 static void
377 tcp_connection_select_lb_bucket (tcp_connection_t * tc, const dpo_id_t * dpo,
378  dpo_id_t * result)
379 {
380  const dpo_id_t *choice;
381  load_balance_t *lb;
382  int hash;
383 
384  lb = load_balance_get (dpo->dpoi_index);
385  if (tc->c_is_ip4)
386  {
387  ip4_tcp_hdr_t hdr;
388  memset (&hdr, 0, sizeof (hdr));
389  hdr.ip.protocol = IP_PROTOCOL_TCP;
390  hdr.ip.address_pair.src.as_u32 = tc->c_lcl_ip.ip4.as_u32;
391  hdr.ip.address_pair.dst.as_u32 = tc->c_rmt_ip.ip4.as_u32;
392  hdr.tcp.src_port = tc->c_lcl_port;
393  hdr.tcp.dst_port = tc->c_rmt_port;
394  hash = ip4_compute_flow_hash (&hdr.ip, lb->lb_hash_config);
395  }
396  else
397  {
398  ip6_tcp_hdr_t hdr;
399  memset (&hdr, 0, sizeof (hdr));
400  hdr.ip.protocol = IP_PROTOCOL_TCP;
401  clib_memcpy (&hdr.ip.src_address, &tc->c_lcl_ip.ip6,
402  sizeof (ip6_address_t));
403  clib_memcpy (&hdr.ip.dst_address, &tc->c_rmt_ip.ip6,
404  sizeof (ip6_address_t));
405  hdr.tcp.src_port = tc->c_lcl_port;
406  hdr.tcp.dst_port = tc->c_rmt_port;
407  hash = ip6_compute_flow_hash (&hdr.ip, lb->lb_hash_config);
408  }
409  choice = load_balance_get_bucket_i (lb, hash & lb->lb_n_buckets_minus_1);
410  dpo_copy (result, choice);
411 }
412 
415 {
416  fib_prefix_t prefix;
417  u32 fib_index;
418 
419  clib_memcpy (&prefix.fp_addr, &tc->c_rmt_ip, sizeof (prefix.fp_addr));
420  prefix.fp_proto = tc->c_is_ip4 ? FIB_PROTOCOL_IP4 : FIB_PROTOCOL_IP6;
421  prefix.fp_len = tc->c_is_ip4 ? 32 : 128;
422  fib_index = fib_table_find (prefix.fp_proto, tc->c_fib_index);
423  return fib_table_lookup (fib_index, &prefix);
424 }
425 
426 static int
427 tcp_connection_stack_on_fib_entry (tcp_connection_t * tc)
428 {
429  dpo_id_t choice = DPO_INVALID;
430  u32 output_node_index;
431  fib_entry_t *fe;
432 
433  fe = fib_entry_get (tc->c_rmt_fei);
434  if (fe->fe_lb.dpoi_type != DPO_LOAD_BALANCE)
435  return -1;
436 
437  tcp_connection_select_lb_bucket (tc, &fe->fe_lb, &choice);
438 
439  output_node_index =
440  tc->c_is_ip4 ? tcp4_output_node.index : tcp6_output_node.index;
441  dpo_stack_from_node (output_node_index, &tc->c_rmt_dpo, &choice);
442  return 0;
443 }
444 
445 /** Stack tcp connection on peer's fib entry.
446  *
447  * This ultimately populates the dpo the connection will use to send packets.
448  */
449 static void
450 tcp_connection_fib_attach (tcp_connection_t * tc)
451 {
452  tc->c_rmt_fei = tcp_lookup_rmt_in_fib (tc);
453 
454  ASSERT (tc->c_rmt_fei != FIB_NODE_INDEX_INVALID);
455 
456  tcp_connection_stack_on_fib_entry (tc);
457 }
458 #endif /* 0 */
459 
460 /**
461  * Initialize connection send variables.
462  */
463 void
465 {
466  u32 time_now;
467 
468  /*
469  * We use the time to randomize iss and for setting up the initial
470  * timestamp. Make sure it's updated otherwise syn and ack in the
471  * handshake may make it look as if time has flown in the opposite
472  * direction for us.
473  */
475  time_now = tcp_time_now ();
476 
477  tc->iss = random_u32 (&time_now);
478  tc->snd_una = tc->iss;
479  tc->snd_nxt = tc->iss + 1;
480  tc->snd_una_max = tc->snd_nxt;
481 }
482 
483 /** Initialize tcp connection variables
484  *
485  * Should be called after having received a msg from the peer, i.e., a SYN or
486  * a SYNACK, such that connection options have already been exchanged. */
487 void
489 {
491  tcp_init_mss (tc);
492  scoreboard_init (&tc->sack_sb);
493  tcp_cc_init (tc);
494  if (tc->state == TCP_STATE_SYN_RCVD)
495  tcp_init_snd_vars (tc);
496 
497  // tcp_connection_fib_attach (tc);
498 }
499 
500 static int
501 tcp_alloc_custom_local_endpoint (tcp_main_t * tm, ip46_address_t * lcl_addr,
502  u16 * lcl_port, u8 is_ip4)
503 {
504  int index, port;
505  if (is_ip4)
506  {
507  index = tm->last_v4_address_rotor++;
508  if (tm->last_v4_address_rotor >= vec_len (tm->ip4_src_addresses))
509  tm->last_v4_address_rotor = 0;
510  lcl_addr->ip4.as_u32 = tm->ip4_src_addresses[index].as_u32;
511  }
512  else
513  {
514  index = tm->last_v6_address_rotor++;
515  if (tm->last_v6_address_rotor >= vec_len (tm->ip6_src_addresses))
516  tm->last_v6_address_rotor = 0;
517  clib_memcpy (&lcl_addr->ip6, &tm->ip6_src_addresses[index],
518  sizeof (ip6_address_t));
519  }
521  if (port < 1)
522  {
523  clib_warning ("Failed to allocate src port");
524  return -1;
525  }
526  *lcl_port = port;
527  return 0;
528 }
529 
530 int
532 {
533  tcp_main_t *tm = vnet_get_tcp_main ();
534  tcp_connection_t *tc;
535  ip46_address_t lcl_addr;
536  u16 lcl_port;
537  int rv;
538 
539  /*
540  * Allocate local endpoint
541  */
542  if ((rmt->is_ip4 && vec_len (tm->ip4_src_addresses))
543  || (!rmt->is_ip4 && vec_len (tm->ip6_src_addresses)))
544  rv = tcp_alloc_custom_local_endpoint (tm, &lcl_addr, &lcl_port,
545  rmt->is_ip4);
546  else
548  rmt, &lcl_addr, &lcl_port);
549 
550  if (rv)
551  return -1;
552 
553  /*
554  * Create connection and send SYN
555  */
556  clib_spinlock_lock_if_init (&tm->half_open_lock);
558  ip_copy (&tc->c_rmt_ip, &rmt->ip, rmt->is_ip4);
559  ip_copy (&tc->c_lcl_ip, &lcl_addr, rmt->is_ip4);
560  tc->c_rmt_port = rmt->port;
561  tc->c_lcl_port = clib_host_to_net_u16 (lcl_port);
562  tc->c_is_ip4 = rmt->is_ip4;
563  tc->c_proto = TRANSPORT_PROTO_TCP;
564  tc->c_fib_index = rmt->fib_index;
565  /* The other connection vars will be initialized after SYN ACK */
567 
568  TCP_EVT_DBG (TCP_EVT_OPEN, tc);
569  tc->state = TCP_STATE_SYN_SENT;
570  tcp_init_snd_vars (tc);
571  tcp_send_syn (tc);
572  clib_spinlock_unlock_if_init (&tm->half_open_lock);
573 
574  return tc->c_c_index;
575 }
576 
577 int
579 {
580  return tcp_connection_open (tep);
581 }
582 
583 const char *tcp_dbg_evt_str[] = {
584 #define _(sym, str) str,
586 #undef _
587 };
588 
589 const char *tcp_fsm_states[] = {
590 #define _(sym, str) str,
592 #undef _
593 };
594 
595 u8 *
596 format_tcp_state (u8 * s, va_list * args)
597 {
598  u32 state = va_arg (*args, u32);
599 
600  if (state < TCP_N_STATES)
601  s = format (s, "%s", tcp_fsm_states[state]);
602  else
603  s = format (s, "UNKNOWN (%d (0x%x))", state, state);
604  return s;
605 }
606 
607 const char *tcp_connection_flags_str[] = {
608 #define _(sym, str) str,
610 #undef _
611 };
612 
613 u8 *
614 format_tcp_connection_flags (u8 * s, va_list * args)
615 {
616  tcp_connection_t *tc = va_arg (*args, tcp_connection_t *);
617  int i, last = -1;
618 
619  for (i = 0; i < TCP_CONN_N_FLAG_BITS; i++)
620  if (tc->flags & (1 << i))
621  last = i;
622  for (i = 0; i < last; i++)
623  {
624  if (tc->flags & (1 << i))
625  s = format (s, "%s, ", tcp_connection_flags_str[i]);
626  }
627  if (last >= 0)
628  s = format (s, "%s", tcp_connection_flags_str[last]);
629  return s;
630 }
631 
632 const char *tcp_conn_timers[] = {
633 #define _(sym, str) str,
635 #undef _
636 };
637 
638 u8 *
639 format_tcp_timers (u8 * s, va_list * args)
640 {
641  tcp_connection_t *tc = va_arg (*args, tcp_connection_t *);
642  int i, last = -1;
643 
644  for (i = 0; i < TCP_N_TIMERS; i++)
645  if (tc->timers[i] != TCP_TIMER_HANDLE_INVALID)
646  last = i;
647 
648  s = format (s, "[");
649  for (i = 0; i < last; i++)
650  {
651  if (tc->timers[i] != TCP_TIMER_HANDLE_INVALID)
652  s = format (s, "%s,", tcp_conn_timers[i]);
653  }
654 
655  if (last >= 0)
656  s = format (s, "%s]", tcp_conn_timers[i]);
657  else
658  s = format (s, "]");
659 
660  return s;
661 }
662 
663 u8 *
664 format_tcp_congestion_status (u8 * s, va_list * args)
665 {
666  tcp_connection_t *tc = va_arg (*args, tcp_connection_t *);
667  if (tcp_in_recovery (tc))
668  s = format (s, "recovery");
669  else if (tcp_in_fastrecovery (tc))
670  s = format (s, "fastrecovery");
671  else
672  s = format (s, "none");
673  return s;
674 }
675 
676 u8 *
677 format_tcp_vars (u8 * s, va_list * args)
678 {
679  tcp_connection_t *tc = va_arg (*args, tcp_connection_t *);
680  s = format (s, " flags: %U timers: %U\n", format_tcp_connection_flags, tc,
681  format_tcp_timers, tc);
682  s = format (s, " snd_una %u snd_nxt %u snd_una_max %u",
683  tc->snd_una - tc->iss, tc->snd_nxt - tc->iss,
684  tc->snd_una_max - tc->iss);
685  s = format (s, " rcv_nxt %u rcv_las %u\n",
686  tc->rcv_nxt - tc->irs, tc->rcv_las - tc->irs);
687  s = format (s, " snd_wnd %u rcv_wnd %u snd_wl1 %u snd_wl2 %u\n",
688  tc->snd_wnd, tc->rcv_wnd, tc->snd_wl1 - tc->irs,
689  tc->snd_wl2 - tc->iss);
690  s = format (s, " flight size %u send space %u rcv_wnd_av %d\n",
692  tcp_rcv_wnd_available (tc));
693  s = format (s, " cong %U ", format_tcp_congestion_status, tc);
694  s = format (s, "cwnd %u ssthresh %u rtx_bytes %u bytes_acked %u\n",
695  tc->cwnd, tc->ssthresh, tc->snd_rxt_bytes, tc->bytes_acked);
696  s = format (s, " prev_ssthresh %u snd_congestion %u dupack %u",
697  tc->prev_ssthresh, tc->snd_congestion - tc->iss,
698  tc->rcv_dupacks);
699  s = format (s, " limited_transmit %u\n", tc->limited_transmit - tc->iss);
700  s = format (s, " tsecr %u tsecr_last_ack %u\n", tc->rcv_opts.tsecr,
701  tc->tsecr_last_ack);
702  s = format (s, " rto %u rto_boff %u srtt %u rttvar %u rtt_ts %u ", tc->rto,
703  tc->rto_boff, tc->srtt, tc->rttvar, tc->rtt_ts);
704  s = format (s, "rtt_seq %u\n", tc->rtt_seq);
705  s = format (s, " tsval_recent %u tsval_recent_age %u\n", tc->tsval_recent,
706  tcp_time_now () - tc->tsval_recent_age);
707  if (tc->state >= TCP_STATE_ESTABLISHED)
708  s = format (s, " scoreboard: %U\n", format_tcp_scoreboard, &tc->sack_sb,
709  tc);
710  if (vec_len (tc->snd_sacks))
711  s = format (s, " sacks tx: %U\n", format_tcp_sacks, tc);
712 
713  return s;
714 }
715 
716 u8 *
717 format_tcp_connection_id (u8 * s, va_list * args)
718 {
719  tcp_connection_t *tc = va_arg (*args, tcp_connection_t *);
720  if (!tc)
721  return s;
722  if (tc->c_is_ip4)
723  {
724  s = format (s, "[#%d][%s] %U:%d->%U:%d", tc->c_thread_index, "T",
725  format_ip4_address, &tc->c_lcl_ip4,
726  clib_net_to_host_u16 (tc->c_lcl_port), format_ip4_address,
727  &tc->c_rmt_ip4, clib_net_to_host_u16 (tc->c_rmt_port));
728  }
729  else
730  {
731  s = format (s, "[#%d][%s] %U:%d->%U:%d", tc->c_thread_index, "T",
732  format_ip6_address, &tc->c_lcl_ip6,
733  clib_net_to_host_u16 (tc->c_lcl_port), format_ip6_address,
734  &tc->c_rmt_ip6, clib_net_to_host_u16 (tc->c_rmt_port));
735  }
736 
737  return s;
738 }
739 
740 u8 *
741 format_tcp_connection (u8 * s, va_list * args)
742 {
743  tcp_connection_t *tc = va_arg (*args, tcp_connection_t *);
744  u32 verbose = va_arg (*args, u32);
745 
746  if (!tc)
747  return s;
748  s = format (s, "%-50U", format_tcp_connection_id, tc);
749  if (verbose)
750  {
751  s = format (s, "%-15U", format_tcp_state, tc->state);
752  if (verbose > 1)
753  s = format (s, "\n%U", format_tcp_vars, tc);
754  }
755 
756  return s;
757 }
758 
759 u8 *
760 format_tcp_session (u8 * s, va_list * args)
761 {
762  u32 tci = va_arg (*args, u32);
763  u32 thread_index = va_arg (*args, u32);
764  u32 verbose = va_arg (*args, u32);
765  tcp_connection_t *tc;
766 
767  tc = tcp_connection_get (tci, thread_index);
768  if (tc)
769  s = format (s, "%U", format_tcp_connection, tc, verbose);
770  else
771  s = format (s, "empty\n");
772  return s;
773 }
774 
775 u8 *
776 format_tcp_listener_session (u8 * s, va_list * args)
777 {
778  u32 tci = va_arg (*args, u32);
780  return format (s, "%U", format_tcp_connection_id, tc);
781 }
782 
783 u8 *
784 format_tcp_half_open_session (u8 * s, va_list * args)
785 {
786  u32 tci = va_arg (*args, u32);
788  return format (s, "%U", format_tcp_connection_id, tc);
789 }
790 
791 u8 *
792 format_tcp_sacks (u8 * s, va_list * args)
793 {
794  tcp_connection_t *tc = va_arg (*args, tcp_connection_t *);
795  sack_block_t *sacks = tc->snd_sacks;
796  sack_block_t *block;
797  int i, len = 0;
798 
799  len = vec_len (sacks);
800  for (i = 0; i < len - 1; i++)
801  {
802  block = &sacks[i];
803  s = format (s, " start %u end %u\n", block->start - tc->irs,
804  block->end - tc->irs);
805  }
806  if (len)
807  {
808  block = &sacks[len - 1];
809  s = format (s, " start %u end %u", block->start - tc->irs,
810  block->end - tc->irs);
811  }
812  return s;
813 }
814 
815 u8 *
816 format_tcp_rcv_sacks (u8 * s, va_list * args)
817 {
818  tcp_connection_t *tc = va_arg (*args, tcp_connection_t *);
819  sack_block_t *sacks = tc->rcv_opts.sacks;
820  sack_block_t *block;
821  int i, len = 0;
822 
823  len = vec_len (sacks);
824  for (i = 0; i < len - 1; i++)
825  {
826  block = &sacks[i];
827  s = format (s, " start %u end %u\n", block->start - tc->iss,
828  block->end - tc->iss);
829  }
830  if (len)
831  {
832  block = &sacks[len - 1];
833  s = format (s, " start %u end %u", block->start - tc->iss,
834  block->end - tc->iss);
835  }
836  return s;
837 }
838 
839 u8 *
840 format_tcp_sack_hole (u8 * s, va_list * args)
841 {
842  sack_scoreboard_hole_t *hole = va_arg (*args, sack_scoreboard_hole_t *);
843  tcp_connection_t *tc = va_arg (*args, tcp_connection_t *);
844  if (tc)
845  s = format (s, " [%u, %u]", hole->start - tc->iss, hole->end - tc->iss);
846  else
847  s = format (s, " [%u, %u]", hole->start, hole->end);
848  return s;
849 }
850 
851 u8 *
852 format_tcp_scoreboard (u8 * s, va_list * args)
853 {
854  sack_scoreboard_t *sb = va_arg (*args, sack_scoreboard_t *);
855  tcp_connection_t *tc = va_arg (*args, tcp_connection_t *);
857  s = format (s, "sacked_bytes %u last_sacked_bytes %u lost_bytes %u\n",
858  sb->sacked_bytes, sb->last_sacked_bytes, sb->lost_bytes);
859  s = format (s, " last_bytes_delivered %u high_sacked %u snd_una_adv %u\n",
860  sb->last_bytes_delivered, sb->high_sacked, sb->snd_una_adv);
861  s = format (s, " cur_rxt_hole %u high_rxt %u rescue_rxt %u",
862  sb->cur_rxt_hole, sb->high_rxt, sb->rescue_rxt);
863 
864  hole = scoreboard_first_hole (sb);
865  if (hole)
866  s = format (s, "\n head %u tail %u holes:\n", sb->head, sb->tail);
867 
868  while (hole)
869  {
870  s = format (s, "%U", format_tcp_sack_hole, hole, tc);
871  hole = scoreboard_next_hole (sb, hole);
872  }
873 
874  return s;
875 }
876 
878 tcp_session_get_transport (u32 conn_index, u32 thread_index)
879 {
880  tcp_connection_t *tc = tcp_connection_get (conn_index, thread_index);
881  return &tc->connection;
882 }
883 
886 {
888  return &tc->connection;
889 }
890 
891 /**
892  * Compute maximum segment size for session layer.
893  *
894  * Since the result needs to be the actual data length, it first computes
895  * the tcp options to be used in the next burst and subtracts their
896  * length from the connection's snd_mss.
897  */
898 u16
900 {
901  tcp_connection_t *tc = (tcp_connection_t *) trans_conn;
902 
903  /* Ensure snd_mss does accurately reflect the amount of data we can push
904  * in a segment. This also makes sure that options are updated according to
905  * the current state of the connection. */
906  tcp_update_snd_mss (tc);
907 
908  return tc->snd_mss;
909 }
910 
913 {
914  if (PREDICT_FALSE (tc->snd_wnd < tc->snd_mss))
915  {
916  return tc->snd_wnd <= snd_space ? tc->snd_wnd : 0;
917  }
918 
919  /* If not snd_wnd constrained and we can't write at least a segment,
920  * don't try at all */
921  if (PREDICT_FALSE (snd_space < tc->snd_mss))
922  return snd_space < tc->cwnd ? 0 : snd_space;
923 
924  /* round down to mss multiple */
925  return snd_space - (snd_space % tc->snd_mss);
926 }
927 
928 /**
929  * Compute tx window session is allowed to fill.
930  *
931  * Takes into account available send space, snd_mss and the congestion
932  * state of the connection. If possible, the value returned is a multiple
933  * of snd_mss.
934  *
935  * @param tc tcp connection
936  * @return number of bytes session is allowed to write
937  */
938 u32
940 {
941  int snd_space, snt_limited;
942 
943  if (PREDICT_TRUE (tcp_in_cong_recovery (tc) == 0))
944  {
945  snd_space = tcp_available_output_snd_space (tc);
946 
947  /* If we haven't gotten dupacks or if we did and have gotten sacked
948  * bytes then we can still send as per Limited Transmit (RFC3042) */
949  if (PREDICT_FALSE (tc->rcv_dupacks != 0
950  && (tcp_opts_sack_permitted (tc)
951  && tc->sack_sb.last_sacked_bytes == 0)))
952  {
953  if (tc->rcv_dupacks == 1 && tc->limited_transmit != tc->snd_nxt)
954  tc->limited_transmit = tc->snd_nxt;
955  ASSERT (seq_leq (tc->limited_transmit, tc->snd_nxt));
956 
957  snt_limited = tc->snd_nxt - tc->limited_transmit;
958  snd_space = clib_max (2 * tc->snd_mss - snt_limited, 0);
959  }
960  return tcp_round_snd_space (tc, snd_space);
961  }
962 
963  if (tcp_in_recovery (tc))
964  {
965  tc->snd_nxt = tc->snd_una_max;
966  snd_space = tcp_available_snd_wnd (tc) - tc->snd_rxt_bytes
967  - (tc->snd_una_max - tc->snd_congestion);
968  if (snd_space <= 0 || (tc->snd_una_max - tc->snd_una) >= tc->snd_wnd)
969  return 0;
970  return tcp_round_snd_space (tc, snd_space);
971  }
972 
973  /* RFC 5681: When previously unsent data is available and the new value of
974  * cwnd and the receiver's advertised window allow, a TCP SHOULD send 1*SMSS
975  * bytes of previously unsent data. */
977  {
978  if (tcp_available_output_snd_space (tc) < tc->snd_mss)
979  return 0;
981  return tc->snd_mss;
982  }
983 
984  return 0;
985 }
986 
987 u32
989 {
990  tcp_connection_t *tc = (tcp_connection_t *) trans_conn;
991  return clib_min (tcp_snd_space (tc),
992  tc->snd_wnd - (tc->snd_nxt - tc->snd_una));
993 }
994 
995 i32
997 {
998  return (i32) tc->rcv_wnd - (tc->rcv_nxt - tc->rcv_las);
999 }
1000 
1001 u32
1003 {
1004  tcp_connection_t *tc = (tcp_connection_t *) trans_conn;
1005 
1006  ASSERT (seq_geq (tc->snd_nxt, tc->snd_una));
1007 
1008  /* This still works if fast retransmit is on */
1009  return (tc->snd_nxt - tc->snd_una);
1010 }
1011 
1012 void
1013 tcp_update_time (f64 now, u8 thread_index)
1014 {
1015  tcp_set_time_now (thread_index);
1016  tw_timer_expire_timers_16t_2w_512sl (&tcp_main.timer_wheels[thread_index],
1017  now);
1018  tcp_flush_frames_to_output (thread_index);
1019 }
1020 
1021 /* *INDENT-OFF* */
1023  .enable = vnet_tcp_enable_disable,
1024  .bind = tcp_session_bind,
1025  .unbind = tcp_session_unbind,
1026  .push_header = tcp_push_header,
1027  .get_connection = tcp_session_get_transport,
1028  .get_listener = tcp_session_get_listener,
1029  .get_half_open = tcp_half_open_session_get_transport,
1030  .open = tcp_session_open,
1031  .close = tcp_session_close,
1032  .cleanup = tcp_session_cleanup,
1033  .send_mss = tcp_session_send_mss,
1034  .send_space = tcp_session_send_space,
1035  .update_time = tcp_update_time,
1036  .tx_fifo_offset = tcp_session_tx_fifo_offset,
1037  .format_connection = format_tcp_session,
1038  .format_listener = format_tcp_listener_session,
1039  .format_half_open = format_tcp_half_open_session,
1040  .tx_type = TRANSPORT_TX_PEEK,
1041  .service_type = TRANSPORT_SERVICE_VC,
1042 };
1043 /* *INDENT-ON* */
1044 
1045 void
1047 {
1048  u32 thread_index = vlib_get_thread_index ();
1049  tcp_connection_t *tc;
1050 
1051  tc = tcp_connection_get (conn_index, thread_index);
1052  tc->timers[TCP_TIMER_KEEP] = TCP_TIMER_HANDLE_INVALID;
1053 
1054  tcp_connection_close (tc);
1055 }
1056 
1057 void
1059 {
1060  tcp_connection_t *tc;
1061 
1062  tc = tcp_half_open_connection_get (conn_index);
1063  if (tc)
1064  {
1065  ASSERT (tc->state == TCP_STATE_SYN_SENT);
1066  session_stream_connect_notify (&tc->connection, 1 /* fail */ );
1067  TCP_DBG ("establish pop: %U", format_tcp_connection, tc, 2);
1068  }
1069  else
1070  {
1071  tc = tcp_connection_get (conn_index, vlib_get_thread_index ());
1072  /* note: the connection may have already disappeared */
1073  if (PREDICT_FALSE (tc == 0))
1074  return;
1075  TCP_DBG ("establish pop: %U", format_tcp_connection, tc, 2);
1076  ASSERT (tc->state == TCP_STATE_SYN_RCVD);
1077  /* Start cleanup. App wasn't notified yet so use delete notify as
1078  * opposed to delete to cleanup session layer state. */
1079  stream_session_delete_notify (&tc->connection);
1080  }
1081  tc->timers[TCP_TIMER_ESTABLISH] = TCP_TIMER_HANDLE_INVALID;
1083 }
1084 
1085 void
1087 {
1088  u32 thread_index = vlib_get_thread_index ();
1089  tcp_connection_t *tc;
1090 
1091  tc = tcp_connection_get (conn_index, thread_index);
1092  if (!tc)
1093  return;
1094  tc->timers[TCP_TIMER_WAITCLOSE] = TCP_TIMER_HANDLE_INVALID;
1095 
1096  /* Session didn't come back with a close(). Send FIN either way
1097  * and switch to LAST_ACK. */
1098  if (tc->state == TCP_STATE_CLOSE_WAIT)
1099  {
1100  if (tc->flags & TCP_CONN_FINSNT)
1101  {
1102  clib_warning ("FIN was sent and still in CLOSE WAIT. Weird!");
1103  }
1104 
1105  tcp_send_fin (tc);
1106  tc->state = TCP_STATE_LAST_ACK;
1107 
1108  /* Make sure we don't wait in LAST ACK forever */
1109  tcp_timer_set (tc, TCP_TIMER_WAITCLOSE, TCP_2MSL_TIME);
1110 
1111  /* Don't delete the connection yet */
1112  return;
1113  }
1114 
1115  tcp_connection_del (tc);
1116 }
1117 
1118 /* *INDENT-OFF* */
1120 {
1128 };
1129 /* *INDENT-ON* */
1130 
1131 static void
1133 {
1134  int i;
1135  u32 connection_index, timer_id;
1136 
1137  for (i = 0; i < vec_len (expired_timers); i++)
1138  {
1139  /* Get session index and timer id */
1140  connection_index = expired_timers[i] & 0x0FFFFFFF;
1141  timer_id = expired_timers[i] >> 28;
1142 
1143  TCP_EVT_DBG (TCP_EVT_TIMER_POP, connection_index, timer_id);
1144 
1145  /* Handle expiration */
1146  (*timer_expiration_handlers[timer_id]) (connection_index);
1147  }
1148 }
1149 
1150 void
1152 {
1153  tw_timer_wheel_16t_2w_512sl_t *tw;
1154  /* *INDENT-OFF* */
1155  foreach_vlib_main (({
1156  tw = &tm->timer_wheels[ii];
1157  tw_timer_wheel_init_16t_2w_512sl (tw, tcp_expired_timers_dispatch,
1158  100e-3 /* timer period 100ms */ , ~0);
1159  tw->last_run_time = vlib_time_now (this_vlib_main);
1160  }));
1161  /* *INDENT-ON* */
1162 }
1163 
1164 clib_error_t *
1166 {
1167  tcp_main_t *tm = vnet_get_tcp_main ();
1169  clib_error_t *error = 0;
1170  u32 num_threads;
1171  int thread;
1172  tcp_connection_t *tc __attribute__ ((unused));
1173  u32 preallocated_connections_per_thread;
1174 
1175  if ((error = vlib_call_init_function (vm, ip_main_init)))
1176  return error;
1177  if ((error = vlib_call_init_function (vm, ip4_lookup_init)))
1178  return error;
1179  if ((error = vlib_call_init_function (vm, ip6_lookup_init)))
1180  return error;
1181 
1182  /*
1183  * Registrations
1184  */
1185 
1186  ip4_register_protocol (IP_PROTOCOL_TCP, tcp4_input_node.index);
1187  ip6_register_protocol (IP_PROTOCOL_TCP, tcp6_input_node.index);
1188 
1189  /*
1190  * Initialize data structures
1191  */
1192 
1193  num_threads = 1 /* main thread */ + vtm->n_threads;
1194  vec_validate (tm->connections, num_threads - 1);
1195 
1196  /*
1197  * Preallocate connections. Assume that thread 0 won't
1198  * use preallocated threads when running multi-core
1199  */
1200  if (num_threads == 1)
1201  {
1202  thread = 0;
1203  preallocated_connections_per_thread = tm->preallocated_connections;
1204  }
1205  else
1206  {
1207  thread = 1;
1208  preallocated_connections_per_thread =
1209  tm->preallocated_connections / (num_threads - 1);
1210  }
1211  for (; thread < num_threads; thread++)
1212  {
1213  if (preallocated_connections_per_thread)
1214  pool_init_fixed (tm->connections[thread],
1215  preallocated_connections_per_thread);
1216  }
1217 
1218  /*
1219  * Use a preallocated half-open connection pool?
1220  */
1221  if (tm->preallocated_half_open_connections)
1222  pool_init_fixed (tm->half_open_connections,
1223  tm->preallocated_half_open_connections);
1224 
1225  /* Initialize per worker thread tx buffers (used for control messages) */
1226  vec_validate (tm->tx_buffers, num_threads - 1);
1227 
1228  /* Initialize timer wheels */
1229  vec_validate (tm->timer_wheels, num_threads - 1);
1231 
1232  /* Initialize clocks per tick for TCP timestamp. Used to compute
1233  * monotonically increasing timestamps. */
1234  tm->tstamp_ticks_per_clock = vm->clib_time.seconds_per_clock
1236 
1237  if (num_threads > 1)
1238  {
1239  clib_spinlock_init (&tm->half_open_lock);
1240  }
1241 
1242  vec_validate (tm->tx_frames[0], num_threads - 1);
1243  vec_validate (tm->tx_frames[1], num_threads - 1);
1244  vec_validate (tm->ip_lookup_tx_frames[0], num_threads - 1);
1245  vec_validate (tm->ip_lookup_tx_frames[1], num_threads - 1);
1246 
1247  tm->bytes_per_buffer = vlib_buffer_free_list_buffer_size
1249 
1250  vec_validate (tm->time_now, num_threads - 1);
1251  return error;
1252 }
1253 
1254 clib_error_t *
1256 {
1257  if (is_en)
1258  {
1259  if (tcp_main.is_enabled)
1260  return 0;
1261 
1262  return tcp_main_enable (vm);
1263  }
1264  else
1265  {
1266  tcp_main.is_enabled = 0;
1267  }
1268 
1269  return 0;
1270 }
1271 
1272 void
1273 tcp_punt_unknown (vlib_main_t * vm, u8 is_ip4, u8 is_add)
1274 {
1275  tcp_main_t *tm = &tcp_main;
1276  if (is_ip4)
1277  tm->punt_unknown4 = is_add;
1278  else
1279  tm->punt_unknown6 = is_add;
1280 }
1281 
1282 clib_error_t *
1284 {
1285  tcp_main_t *tm = vnet_get_tcp_main ();
1286  ip_main_t *im = &ip_main;
1287  ip_protocol_info_t *pi;
1288 
1289  /* Session layer, and by implication tcp, are disabled by default */
1290  tm->is_enabled = 0;
1291 
1292  /* Register with IP for header parsing */
1293  pi = ip_get_protocol_info (im, IP_PROTOCOL_TCP);
1294  if (pi == 0)
1295  return clib_error_return (0, "TCP protocol info AWOL");
1298 
1299  /* Register as transport with session layer */
1304 
1305  tcp_api_reference ();
1306  return 0;
1307 }
1308 
1310 
1311 static clib_error_t *
1313 {
1314  tcp_main_t *tm = vnet_get_tcp_main ();
1315 
1316  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
1317  {
1318  if (unformat
1319  (input, "preallocated-connections %d",
1320  &tm->preallocated_connections))
1321  ;
1322  else if (unformat (input, "preallocated-half-open-connections %d",
1323  &tm->preallocated_half_open_connections))
1324  ;
1325  else if (unformat (input, "buffer-fail-fraction %f",
1326  &tm->buffer_fail_fraction))
1327  ;
1328  else
1329  return clib_error_return (0, "unknown input `%U'",
1330  format_unformat_error, input);
1331  }
1332  return 0;
1333 }
1334 
1336 
1337 
1338 /**
1339  * \brief Configure an ipv4 source address range
1340  * @param vm vlib_main_t pointer
1341  * @param start first ipv4 address in the source address range
1342  * @param end last ipv4 address in the source address range
1343  * @param table_id VRF / table ID, 0 for the default FIB
1344  * @return 0 if all OK, else an error indication from api_errno.h
1345  */
1346 
1347 int
1349  ip4_address_t * start,
1350  ip4_address_t * end, u32 table_id)
1351 {
1352  tcp_main_t *tm = vnet_get_tcp_main ();
1353  vnet_main_t *vnm = vnet_get_main ();
1354  u32 start_host_byte_order, end_host_byte_order;
1355  fib_prefix_t prefix;
1356  vnet_sw_interface_t *si;
1357  fib_node_index_t fei;
1358  u32 fib_index = 0;
1359  u32 sw_if_index;
1360  int rv;
1361  int vnet_proxy_arp_add_del (ip4_address_t * lo_addr,
1362  ip4_address_t * hi_addr, u32 fib_index,
1363  int is_del);
1364 
1365  memset (&prefix, 0, sizeof (prefix));
1366 
1367  fib_index = fib_table_find (FIB_PROTOCOL_IP4, table_id);
1368 
1369  if (fib_index == ~0)
1370  return VNET_API_ERROR_NO_SUCH_FIB;
1371 
1372  start_host_byte_order = clib_net_to_host_u32 (start->as_u32);
1373  end_host_byte_order = clib_net_to_host_u32 (end->as_u32);
1374 
1375  /* sanity check for reversed args or some such */
1376  if ((end_host_byte_order - start_host_byte_order) > (10 << 10))
1377  return VNET_API_ERROR_INVALID_ARGUMENT;
1378 
1379  /* Lookup the last address, to identify the interface involved */
1380  prefix.fp_len = 32;
1381  prefix.fp_proto = FIB_PROTOCOL_IP4;
1382  memcpy (&prefix.fp_addr.ip4, end, sizeof (ip4_address_t));
1383 
1384  fei = fib_table_lookup (fib_index, &prefix);
1385 
1386  /* Couldn't find route to destination. Bail out. */
1387  if (fei == FIB_NODE_INDEX_INVALID)
1388  return VNET_API_ERROR_NEXT_HOP_NOT_IN_FIB;
1389 
1390  sw_if_index = fib_entry_get_resolving_interface (fei);
1391 
1392  /* Enable proxy arp on the interface */
1393  si = vnet_get_sw_interface (vnm, sw_if_index);
1395 
1396  /* Configure proxy arp across the range */
1397  rv = vnet_proxy_arp_add_del (start, end, fib_index, 0 /* is_del */ );
1398 
1399  if (rv)
1400  return rv;
1401 
1402  do
1403  {
1404  dpo_id_t dpo = DPO_INVALID;
1405 
1406  vec_add1 (tm->ip4_src_addresses, start[0]);
1407 
1408  /* Add local adjacencies for the range */
1409 
1410  receive_dpo_add_or_lock (DPO_PROTO_IP4, ~0 /* sw_if_index */ ,
1411  NULL, &dpo);
1412  prefix.fp_len = 32;
1413  prefix.fp_proto = FIB_PROTOCOL_IP4;
1414  prefix.fp_addr.ip4.as_u32 = start->as_u32;
1415 
1417  &prefix,
1419  FIB_ENTRY_FLAG_EXCLUSIVE, &dpo);
1420  dpo_reset (&dpo);
1421 
1422  start_host_byte_order++;
1423  start->as_u32 = clib_host_to_net_u32 (start_host_byte_order);
1424  }
1425  while (start_host_byte_order <= end_host_byte_order);
1426 
1427  return 0;
1428 }
1429 
1430 /**
1431  * \brief Configure an ipv6 source address range
1432  * @param vm vlib_main_t pointer
1433  * @param start first ipv6 address in the source address range
1434  * @param end last ipv6 address in the source address range
1435  * @param table_id VRF / table ID, 0 for the default FIB
1436  * @return 0 if all OK, else an error indication from api_errno.h
1437  */
1438 
1439 int
1441  ip6_address_t * start,
1442  ip6_address_t * end, u32 table_id)
1443 {
1444  tcp_main_t *tm = vnet_get_tcp_main ();
1445  fib_prefix_t prefix;
1446  u32 fib_index = 0;
1447  fib_node_index_t fei;
1448  u32 sw_if_index;
1449 
1450  memset (&prefix, 0, sizeof (prefix));
1451 
1452  fib_index = fib_table_find (FIB_PROTOCOL_IP6, table_id);
1453 
1454  if (fib_index == ~0)
1455  return VNET_API_ERROR_NO_SUCH_FIB;
1456 
1457  while (1)
1458  {
1459  int i;
1460  ip6_address_t tmp;
1461  dpo_id_t dpo = DPO_INVALID;
1462 
1463  /* Remember this address */
1464  vec_add1 (tm->ip6_src_addresses, start[0]);
1465 
1466  /* Lookup the prefix, to identify the interface involved */
1467  prefix.fp_len = 128;
1468  prefix.fp_proto = FIB_PROTOCOL_IP6;
1469  memcpy (&prefix.fp_addr.ip6, start, sizeof (ip6_address_t));
1470 
1471  fei = fib_table_lookup (fib_index, &prefix);
1472 
1473  /* Couldn't find route to destination. Bail out. */
1474  if (fei == FIB_NODE_INDEX_INVALID)
1475  return VNET_API_ERROR_NEXT_HOP_NOT_IN_FIB;
1476 
1477  sw_if_index = fib_entry_get_resolving_interface (fei);
1478 
1479  if (sw_if_index == (u32) ~ 0)
1480  return VNET_API_ERROR_NO_MATCHING_INTERFACE;
1481 
1482  /* Add a proxy neighbor discovery entry for this address */
1483  ip6_neighbor_proxy_add_del (sw_if_index, start, 0 /* is_del */ );
1484 
1485  /* Add a receive adjacency for this address */
1486  receive_dpo_add_or_lock (DPO_PROTO_IP6, ~0 /* sw_if_index */ ,
1487  NULL, &dpo);
1488 
1490  &prefix,
1492  FIB_ENTRY_FLAG_EXCLUSIVE, &dpo);
1493  dpo_reset (&dpo);
1494 
1495  /* Done with the entire range? */
1496  if (!memcmp (start, end, sizeof (start[0])))
1497  break;
1498 
1499  /* Increment the address. DGMS. */
1500  tmp = start[0];
1501  for (i = 15; i >= 0; i--)
1502  {
1503  tmp.as_u8[i] += 1;
1504  if (tmp.as_u8[i] != 0)
1505  break;
1506  }
1507  start[0] = tmp;
1508  }
1509  return 0;
1510 }
1511 
1512 static clib_error_t *
1514  unformat_input_t * input, vlib_cli_command_t * cmd_arg)
1515 {
1516  ip4_address_t v4start, v4end;
1517  ip6_address_t v6start, v6end;
1518  u32 table_id = 0;
1519  int v4set = 0;
1520  int v6set = 0;
1521  int rv;
1522 
1523  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
1524  {
1525  if (unformat (input, "%U - %U", unformat_ip4_address, &v4start,
1526  unformat_ip4_address, &v4end))
1527  v4set = 1;
1528  else if (unformat (input, "%U", unformat_ip4_address, &v4start))
1529  {
1530  memcpy (&v4end, &v4start, sizeof (v4start));
1531  v4set = 1;
1532  }
1533  else if (unformat (input, "%U - %U", unformat_ip6_address, &v6start,
1534  unformat_ip6_address, &v6end))
1535  v6set = 1;
1536  else if (unformat (input, "%U", unformat_ip6_address, &v6start))
1537  {
1538  memcpy (&v6end, &v6start, sizeof (v6start));
1539  v6set = 1;
1540  }
1541  else if (unformat (input, "fib-table %d", &table_id))
1542  ;
1543  else
1544  break;
1545  }
1546 
1547  if (!v4set && !v6set)
1548  return clib_error_return (0, "at least one v4 or v6 address required");
1549 
1550  if (v4set)
1551  {
1552  rv = tcp_configure_v4_source_address_range (vm, &v4start, &v4end,
1553  table_id);
1554  switch (rv)
1555  {
1556  case 0:
1557  break;
1558 
1559  case VNET_API_ERROR_NO_SUCH_FIB:
1560  return clib_error_return (0, "Invalid table-id %d", table_id);
1561 
1562  case VNET_API_ERROR_INVALID_ARGUMENT:
1563  return clib_error_return (0, "Invalid address range %U - %U",
1564  format_ip4_address, &v4start,
1565  format_ip4_address, &v4end);
1566  default:
1567  return clib_error_return (0, "error %d", rv);
1568  break;
1569  }
1570  }
1571  if (v6set)
1572  {
1573  rv = tcp_configure_v6_source_address_range (vm, &v6start, &v6end,
1574  table_id);
1575  switch (rv)
1576  {
1577  case 0:
1578  break;
1579 
1580  case VNET_API_ERROR_NO_SUCH_FIB:
1581  return clib_error_return (0, "Invalid table-id %d", table_id);
1582 
1583  default:
1584  return clib_error_return (0, "error %d", rv);
1585  break;
1586  }
1587  }
1588  return 0;
1589 }
1590 
1591 /* *INDENT-OFF* */
1592 VLIB_CLI_COMMAND (tcp_src_address_command, static) =
1593 {
1594  .path = "tcp src-address",
1595  .short_help = "tcp src-address <ip-addr> [- <ip-addr>] add src address range",
1596  .function = tcp_src_address,
1597 };
1598 /* *INDENT-ON* */
1599 
1600 static u8 *
1602 {
1603 #if TCP_SCOREBOARD_TRACE
1604 
1605  scoreboard_trace_elt_t *block;
1606  int i = 0;
1607 
1608  if (!sb->trace)
1609  return s;
1610 
1611  s = format (s, "scoreboard trace:");
1612  vec_foreach (block, sb->trace)
1613  {
1614  s = format (s, "{%u, %u, %u, %u, %u}, ", block->start, block->end,
1615  block->ack, block->snd_una_max, block->group);
1616  if ((++i % 3) == 0)
1617  s = format (s, "\n");
1618  }
1619  return s;
1620 #else
1621  return 0;
1622 #endif
1623 }
1624 
1625 static clib_error_t *
1627  vlib_cli_command_t * cmd_arg)
1628 {
1629  transport_connection_t *tconn = 0;
1630  tcp_connection_t *tc;
1631  u8 *s = 0;
1632  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
1633  {
1634  if (unformat (input, "%U", unformat_transport_connection, &tconn,
1636  ;
1637  else
1638  return clib_error_return (0, "unknown input `%U'",
1639  format_unformat_error, input);
1640  }
1641 
1642  if (!TCP_SCOREBOARD_TRACE)
1643  {
1644  vlib_cli_output (vm, "scoreboard tracing not enabled");
1645  return 0;
1646  }
1647 
1648  tc = tcp_get_connection_from_transport (tconn);
1649  s = tcp_scoreboard_dump_trace (s, &tc->sack_sb);
1650  vlib_cli_output (vm, "%v", s);
1651  return 0;
1652 }
1653 
1654 /* *INDENT-OFF* */
1655 VLIB_CLI_COMMAND (tcp_show_scoreboard_trace_command, static) =
1656 {
1657  .path = "show tcp scoreboard trace",
1658  .short_help = "show tcp scoreboard trace <connection>",
1659  .function = tcp_show_scoreboard_trace_fn,
1660 };
1661 /* *INDENT-ON* */
1662 
1663 u8 *
1665 {
1666  int i, trace_len;
1668  u32 next_ack, left, group, has_new_ack = 0;
1669  tcp_connection_t _dummy_tc, *dummy_tc = &_dummy_tc;
1670  sack_block_t *block;
1671 
1672  if (!tc)
1673  return s;
1674 
1675  memset (dummy_tc, 0, sizeof (*dummy_tc));
1676  tcp_connection_timers_init (dummy_tc);
1677  scoreboard_init (&dummy_tc->sack_sb);
1678  dummy_tc->rcv_opts.flags |= TCP_OPTS_FLAG_SACK;
1679 
1680 #if TCP_SCOREBOARD_TRACE
1681  trace = tc->sack_sb.trace;
1682  trace_len = vec_len (tc->sack_sb.trace);
1683 #else
1684  trace = 0;
1685  trace_len = 0;
1686 #endif
1687 
1688  for (i = 0; i < trace_len; i++)
1689  {
1690  if (trace[i].ack != 0)
1691  {
1692  dummy_tc->snd_una = trace[i].ack - 1448;
1693  dummy_tc->snd_una_max = trace[i].ack;
1694  }
1695  }
1696 
1697  left = 0;
1698  while (left < trace_len)
1699  {
1700  group = trace[left].group;
1701  vec_reset_length (dummy_tc->rcv_opts.sacks);
1702  has_new_ack = 0;
1703  while (trace[left].group == group)
1704  {
1705  if (trace[left].ack != 0)
1706  {
1707  if (verbose)
1708  s = format (s, "Adding ack %u, snd_una_max %u, segs: ",
1709  trace[left].ack, trace[left].snd_una_max);
1710  dummy_tc->snd_una_max = trace[left].snd_una_max;
1711  next_ack = trace[left].ack;
1712  has_new_ack = 1;
1713  }
1714  else
1715  {
1716  if (verbose)
1717  s = format (s, "[%u, %u], ", trace[left].start,
1718  trace[left].end);
1719  vec_add2 (dummy_tc->rcv_opts.sacks, block, 1);
1720  block->start = trace[left].start;
1721  block->end = trace[left].end;
1722  }
1723  left++;
1724  }
1725 
1726  /* Push segments */
1727  tcp_rcv_sacks (dummy_tc, next_ack);
1728  if (has_new_ack)
1729  dummy_tc->snd_una = next_ack + dummy_tc->sack_sb.snd_una_adv;
1730 
1731  if (verbose)
1732  s = format (s, "result: %U", format_tcp_scoreboard,
1733  &dummy_tc->sack_sb);
1734 
1735  }
1736  s = format (s, "result: %U", format_tcp_scoreboard, &dummy_tc->sack_sb);
1737 
1738  return s;
1739 }
1740 
1741 static clib_error_t *
1743  vlib_cli_command_t * cmd_arg)
1744 {
1745  transport_connection_t *tconn = 0;
1746  tcp_connection_t *tc = 0;
1747  u8 *str = 0;
1748  while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT)
1749  {
1750  if (unformat (input, "%U", unformat_transport_connection, &tconn,
1752  ;
1753  else
1754  return clib_error_return (0, "unknown input `%U'",
1755  format_unformat_error, input);
1756  }
1757 
1758  if (!TCP_SCOREBOARD_TRACE)
1759  {
1760  vlib_cli_output (vm, "scoreboard tracing not enabled");
1761  return 0;
1762  }
1763 
1764  tc = tcp_get_connection_from_transport (tconn);
1765  if (!tc)
1766  {
1767  vlib_cli_output (vm, "connection not found");
1768  return 0;
1769  }
1770  str = tcp_scoreboard_replay (str, tc, 1);
1771  vlib_cli_output (vm, "%v", str);
1772  return 0;
1773 }
1774 
1775 /* *INDENT-OFF* */
1776 VLIB_CLI_COMMAND (tcp_replay_scoreboard_command, static) =
1777 {
1778  .path = "tcp replay scoreboard",
1779  .short_help = "tcp replay scoreboard <connection>",
1780  .function = tcp_scoreboard_trace_fn,
1781 };
1782 /* *INDENT-ON* */
1783 
1784 static clib_error_t *
1786  vlib_cli_command_t * cmd_arg)
1787 {
1788  tcp_main_t *tm = vnet_get_tcp_main ();
1790  return clib_error_return (0, "unknown input `%U'", format_unformat_error,
1791  input);
1792  vlib_cli_output (vm, "IPv4 TCP punt: %s",
1793  tm->punt_unknown4 ? "enabled" : "disabled");
1794  vlib_cli_output (vm, "IPv6 TCP punt: %s",
1795  tm->punt_unknown6 ? "enabled" : "disabled");
1796  return 0;
1797 }
1798 /* *INDENT-OFF* */
1799 VLIB_CLI_COMMAND (show_tcp_punt_command, static) =
1800 {
1801  .path = "show tcp punt",
1802  .short_help = "show tcp punt",
1803  .function = show_tcp_punt_fn,
1804 };
1805 /* *INDENT-ON* */
1806 
1807 /*
1808  * fd.io coding-style-patch-verification: ON
1809  *
1810  * Local Variables:
1811  * eval: (c-set-style "gnu")
1812  * End:
1813  */
#define vec_validate(V, I)
Make sure vector is long enough for given index (no header, unspecified alignment) ...
Definition: vec.h:434
#define tcp_in_cong_recovery(tc)
Definition: tcp.h:330
void dpo_stack_from_node(u32 child_node_index, dpo_id_t *dpo, const dpo_id_t *parent)
Stack one DPO object on another, and thus establish a child parent relationship.
Definition: dpo.c:530
fib_protocol_t fp_proto
protocol type
Definition: fib_types.h:197
#define TCP_2MSL_TIME
Definition: tcp.h:101
static vlib_cli_command_t trace
(constructor) VLIB_CLI_COMMAND (trace)
Definition: vlib_api_cli.c:862
void tcp_timer_keep_handler(u32 conn_index)
Definition: tcp.c:1046
void receive_dpo_add_or_lock(dpo_proto_t proto, u32 sw_if_index, const ip46_address_t *nh_addr, dpo_id_t *dpo)
Definition: receive_dpo.c:56
An entry in a FIB table.
Definition: fib_entry.h:453
#define clib_min(x, y)
Definition: clib.h:340
virtual circuit service
const char * tcp_fsm_states[]
Definition: tcp.c:589
static clib_error_t * tcp_config_fn(vlib_main_t *vm, unformat_input_t *input)
Definition: tcp.c:1312
#define tcp_in_recovery(tc)
Definition: tcp.h:324
u8 * format_tcp_vars(u8 *s, va_list *args)
Definition: tcp.c:677
void ip_copy(ip46_address_t *dst, ip46_address_t *src, u8 is_ip4)
Definition: ip.c:81
#define seq_leq(_s1, _s2)
Definition: tcp.h:531
struct _sack_block sack_block_t
u8 * format_tcp_rcv_sacks(u8 *s, va_list *args)
Definition: tcp.c:816
void ip6_register_protocol(u32 protocol, u32 node_index)
Definition: ip6_forward.c:1392
struct _scoreboard_trace_elt scoreboard_trace_elt_t
u8 * format_tcp_sacks(u8 *s, va_list *args)
Definition: tcp.c:792
vnet_main_t * vnet_get_main(void)
Definition: misc.c:47
struct _transport_connection transport_connection_t
const char * tcp_connection_flags_str[]
Definition: tcp.c:607
vlib_node_registration_t tcp4_output_node
(constructor) VLIB_REGISTER_NODE (tcp4_output_node)
Definition: tcp_output.c:1932
uword unformat_transport_connection(unformat_input_t *input, va_list *args)
Definition: session_cli.c:164
u32 tcp_session_unbind(u32 listener_index)
Definition: tcp.c:89
#define PREDICT_TRUE(x)
Definition: clib.h:106
tcp_connection_t * tcp_connection_new(u8 thread_index)
Definition: tcp.c:204
u8 as_u8[16]
Definition: ip6_packet.h:48
u32 tcp_session_bind(u32 session_index, transport_endpoint_t *tep)
Definition: tcp.c:66
void ip_set(ip46_address_t *dst, void *src, u8 is_ip4)
Definition: ip.c:90
#define NULL
Definition: clib.h:55
static u32 ip4_compute_flow_hash(const ip4_header_t *ip, flow_hash_config_t flow_hash_config)
Definition: ip4.h:287
struct _sack_scoreboard sack_scoreboard_t
static int tcp_alloc_custom_local_endpoint(tcp_main_t *tm, ip46_address_t *lcl_addr, u16 *lcl_port, u8 is_ip4)
Definition: tcp.c:501
static f64 vlib_time_now(vlib_main_t *vm)
Definition: main.h:227
static tcp_connection_t * tcp_half_open_connection_get(u32 conn_index)
Definition: tcp.h:499
flow_hash_config_t lb_hash_config
the hash config to use when selecting a bucket.
Definition: load_balance.h:134
static_always_inline void clib_spinlock_unlock_if_init(clib_spinlock_t *p)
Definition: lock.h:98
struct _tcp_main tcp_main_t
struct _transport_proto_vft transport_proto_vft_t
void tcp_connection_timers_reset(tcp_connection_t *tc)
Stop all connection timers.
Definition: tcp.c:354
vlib_node_registration_t tcp6_output_node
(constructor) VLIB_REGISTER_NODE (tcp6_output_node)
Definition: tcp_output.c:1953
void dpo_copy(dpo_id_t *dst, const dpo_id_t *src)
atomic copy a data-plane object.
Definition: dpo.c:261
#define vec_add1(V, E)
Add 1 element to end of vector (unspecified alignment).
Definition: vec.h:520
timer_expiration_handler tcp_timer_retransmit_handler
static void scoreboard_init(sack_scoreboard_t *sb)
Definition: tcp.h:878
static heap_elt_t * last(heap_header_t *h)
Definition: heap.c:53
#define vec_add2(V, P, N)
Add N elements to end of vector V, return pointer to new elements in P.
Definition: vec.h:559
int i
u8 * format_tcp_connection_flags(u8 *s, va_list *args)
Definition: tcp.c:614
int ip6_neighbor_proxy_add_del(u32 sw_if_index, ip6_address_t *addr, u8 is_del)
static vnet_sw_interface_t * vnet_get_sw_interface(vnet_main_t *vnm, u32 sw_if_index)
void * ip_interface_get_first_ip(u32 sw_if_index, u8 is_ip4)
Definition: ip.c:133
struct _tcp_connection tcp_connection_t
u8 * format(u8 *s, const char *fmt,...)
Definition: format.c:419
clib_time_t clib_time
Definition: main.h:63
static u32 tcp_available_snd_wnd(const tcp_connection_t *tc)
Definition: tcp.h:598
Definition: ip.h:109
transport_connection_t * tcp_half_open_session_get_transport(u32 conn_index)
Definition: tcp.c:885
static tcp_connection_t * tcp_get_connection_from_transport(transport_connection_t *tconn)
Definition: tcp.h:468
void tcp_timer_establish_handler(u32 conn_index)
Definition: tcp.c:1058
void ip4_register_protocol(u32 protocol, u32 node_index)
Definition: ip4_forward.c:1508
#define pool_get(P, E)
Allocate an object E from a pool P (unspecified alignment).
Definition: pool.h:227
struct _tcp_header tcp_header_t
int tcp_half_open_connection_cleanup(tcp_connection_t *tc)
Try to cleanup half-open connection.
Definition: tcp.c:129
struct _sack_scoreboard_hole sack_scoreboard_hole_t
#define vec_reset_length(v)
Reset vector length to zero NULL-pointer tolerant.
pthread_t thread[MAX_CONNS]
Definition: main.c:125
u32 tcp_session_tx_fifo_offset(transport_connection_t *trans_conn)
Definition: tcp.c:1002
#define TCP_RTO_INIT
Definition: tcp.h:111
format_function_t format_ip4_address
Definition: format.h:79
unformat_function_t * unformat_pg_edit
Definition: ip.h:90
#define VNET_SW_INTERFACE_FLAG_PROXY_ARP
Definition: interface.h:593
unformat_function_t unformat_ip4_address
Definition: format.h:76
void tcp_init_snd_vars(tcp_connection_t *tc)
Initialize connection send variables.
Definition: tcp.c:464
u8 * format_tcp_connection_id(u8 *s, va_list *args)
Definition: tcp.c:717
#define VLIB_INIT_FUNCTION(x)
Definition: init.h:111
#define always_inline
Definition: clib.h:92
u16 lb_n_buckets_minus_1
number of buckets in the load-balance - 1.
Definition: load_balance.h:99
void tcp_connection_timers_init(tcp_connection_t *tc)
Initialize all connection timers as invalid.
Definition: tcp.c:337
static u32 tcp_available_output_snd_space(const tcp_connection_t *tc)
Definition: tcp.h:604
tcp_main_t tcp_main
Definition: tcp.c:29
int i32
Definition: types.h:81
static void tcp_connection_unbind(u32 listener_index)
Definition: tcp.c:72
clib_error_t * tcp_main_enable(vlib_main_t *vm)
Definition: tcp.c:1165
Aggregrate type for a prefix.
Definition: fib_types.h:188
#define clib_error_return(e, args...)
Definition: error.h:99
int tcp_connection_open(transport_endpoint_t *rmt)
Definition: tcp.c:531
void stream_session_delete_notify(transport_connection_t *tc)
Notification from transport that connection is being deleted.
Definition: session.c:742
static u8 * tcp_scoreboard_dump_trace(u8 *s, sack_scoreboard_t *sb)
Definition: tcp.c:1601
timer_expiration_handler tcp_timer_retransmit_syn_handler
static u32 tcp_time_now(void)
Definition: tcp.h:656
u32 tcp_snd_space(tcp_connection_t *tc)
Compute tx window session is allowed to fill.
Definition: tcp.c:939
void tcp_half_open_connection_del(tcp_connection_t *tc)
Cleanup half-open connection.
Definition: tcp.c:109
u32 fib_table_find(fib_protocol_t proto, u32 table_id)
Get the index of the FIB for a Table-ID.
Definition: fib_table.c:1037
u16 fp_len
The mask length.
Definition: fib_types.h:192
#define vlib_call_init_function(vm, x)
Definition: init.h:162
tcp_connection_t * tcp_half_open_connection_new(void)
Definition: tcp.c:141
void tcp_api_reference(void)
Definition: tcp_api.c:109
#define TCP_EVT_DBG(_evt, _args...)
Definition: tcp_debug.h:235
fib_node_index_t fib_table_lookup(u32 fib_index, const fib_prefix_t *prefix)
Perfom a longest prefix match in the non-forwarding table.
Definition: fib_table.c:66
static u32 tcp_connection_bind(u32 session_index, transport_endpoint_t *lcl)
Definition: tcp.c:32
static void clib_spinlock_init(clib_spinlock_t *p)
Definition: lock.h:57
static void tcp_timer_set(tcp_connection_t *tc, u8 timer_id, u32 interval)
Definition: tcp.h:689
static timer_expiration_handler * timer_expiration_handlers[TCP_N_TIMERS]
Definition: tcp.c:1119
The identity of a DPO is a combination of its type and its instance number/index of objects of that t...
Definition: dpo.h:168
Definition: fib_entry.h:274
#define pool_elt_at_index(p, i)
Returns pointer to element at given index.
Definition: pool.h:461
transport_connection_t * tcp_session_get_listener(u32 listener_index)
Definition: tcp.c:96
timer_expiration_handler tcp_timer_persist_handler
static ip_protocol_info_t * ip_get_protocol_info(ip_main_t *im, u32 protocol)
Definition: ip.h:136
#define tcp_in_fastrecovery(tc)
Definition: tcp.h:323
ip46_address_t fp_addr
The address type is not deriveable from the fp_addr member.
Definition: fib_types.h:211
dpo_type_t dpoi_type
the type
Definition: dpo.h:172
format_function_t * format_header
Definition: ip.h:81
static const dpo_id_t * load_balance_get_bucket_i(const load_balance_t *lb, u32 bucket)
Definition: load_balance.h:209
void tcp_rcv_sacks(tcp_connection_t *tc, u32 ack)
Definition: tcp_input.c:724
void tcp_session_cleanup(u32 conn_index, u32 thread_index)
Definition: tcp.c:321
void tcp_send_syn(tcp_connection_t *tc)
Send SYN.
Definition: tcp_output.c:970
struct _unformat_input_t unformat_input_t
static sack_scoreboard_hole_t * scoreboard_next_hole(sack_scoreboard_t *sb, sack_scoreboard_hole_t *hole)
Definition: tcp.h:813
load-balancing over a choice of [un]equal cost paths
Definition: dpo.h:102
#define VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX
Definition: buffer.h:446
static u32 ip6_compute_flow_hash(const ip6_header_t *ip, flow_hash_config_t flow_hash_config)
Definition: ip6.h:428
#define pool_put(P, E)
Free an object E in pool P.
Definition: pool.h:273
static void tcp_expired_timers_dispatch(u32 *expired_timers)
Definition: tcp.c:1132
#define TCP_TIMER_HANDLE_INVALID
Definition: tcp.h:93
The FIB DPO provieds;.
Definition: load_balance.h:84
void tcp_timer_waitclose_handler(u32 conn_index)
Definition: tcp.c:1086
#define TCP_CLEANUP_TIME
Definition: tcp.h:104
static u32 tcp_flight_size(const tcp_connection_t *tc)
Our estimate of the number of bytes in flight (pipe size)
Definition: tcp.h:556
#define PREDICT_FALSE(x)
Definition: clib.h:105
u8 * format_tcp_congestion_status(u8 *s, va_list *args)
Definition: tcp.c:664
#define VLIB_CONFIG_FUNCTION(x, n,...)
Definition: init.h:119
reliable transport protos
u8 * format_tcp_scoreboard(u8 *s, va_list *args)
Definition: tcp.c:852
clib_error_t * vnet_tcp_enable_disable(vlib_main_t *vm, u8 is_en)
Definition: tcp.c:1255
f64 seconds_per_clock
Definition: time.h:57
u32 tcp_push_header(transport_connection_t *tconn, vlib_buffer_t *b)
Definition: tcp_output.c:1975
void tcp_session_close(u32 conn_index, u32 thread_index)
Definition: tcp.c:313
void( timer_expiration_handler)(u32 index)
Definition: tcp.h:86
#define foreach_vlib_main(body)
Definition: threads.h:244
#define TCP_DBG(_fmt, _args...)
Definition: tcp_debug.h:86
u32 fib_entry_get_resolving_interface(fib_node_index_t entry_index)
Definition: fib_entry.c:1409
static void tcp_timer_reset(tcp_connection_t *tc, u8 timer_id)
Definition: tcp.h:699
static sack_scoreboard_hole_t * scoreboard_first_hole(sack_scoreboard_t *sb)
Definition: tcp.h:829
#define tcp_fastrecovery_sent_1_smss(tc)
Definition: tcp.h:326
unformat_function_t unformat_ip6_address
Definition: format.h:94
u8 * format_tcp_half_open_session(u8 *s, va_list *args)
Definition: tcp.c:784
clib_error_t * ip_main_init(vlib_main_t *vm)
Definition: ip_init.c:45
fib_node_index_t fib_table_entry_special_dpo_update(u32 fib_index, const fib_prefix_t *prefix, fib_source_t source, fib_entry_flag_t flags, const dpo_id_t *dpo)
Update a &#39;special&#39; entry to the FIB that links to the DPO passed A special entry is an entry that the...
Definition: fib_table.c:338
#define UNFORMAT_END_OF_INPUT
Definition: format.h:143
int tcp_configure_v4_source_address_range(vlib_main_t *vm, ip4_address_t *start, ip4_address_t *end, u32 table_id)
Configure an ipv4 source address range.
Definition: tcp.c:1348
static_always_inline uword vlib_get_thread_index(void)
Definition: threads.h:221
format_function_t format_ip6_address
Definition: format.h:95
vlib_main_t * vm
Definition: buffer.c:294
void tcp_cc_init(tcp_connection_t *tc)
Definition: tcp_input.c:1204
ip_main_t ip_main
Definition: ip_init.c:42
u32 stream_session_tx_fifo_max_dequeue(transport_connection_t *tc)
Definition: session.c:410
static clib_error_t * tcp_src_address(vlib_main_t *vm, unformat_input_t *input, vlib_cli_command_t *cmd_arg)
Definition: tcp.c:1513
void tcp_connection_close(tcp_connection_t *tc)
Begin connection closing procedure.
Definition: tcp.c:270
#define ENDPOINT_INVALID_INDEX
Definition: transport.h:115
#define clib_warning(format, args...)
Definition: error.h:59
#define clib_memcpy(a, b, c)
Definition: string.h:75
u8 * format_tcp_session(u8 *s, va_list *args)
Definition: tcp.c:760
format_function_t format_tcp_header
Definition: format.h:102
const char * tcp_dbg_evt_str[]
Definition: tcp.c:583
u32 fib_node_index_t
A typedef of a node index.
Definition: fib_types.h:30
void tcp_update_time(f64 now, u8 thread_index)
Definition: tcp.c:1013
void tcp_flush_frames_to_output(u8 thread_index)
Flush v4 and v6 tcp and ip-lookup tx frames for thread index.
Definition: tcp_output.c:1039
vlib_node_registration_t tcp6_input_node
(constructor) VLIB_REGISTER_NODE (tcp6_input_node)
Definition: tcp_input.c:3133
u16 tcp_session_send_mss(transport_connection_t *trans_conn)
Compute maximum segment size for session layer.
Definition: tcp.c:899
fib_node_index_t tcp_lookup_rmt_in_fib(tcp_connection_t *tc)
fib_entry_t * fib_entry_get(fib_node_index_t index)
Definition: fib_entry.c:45
int vnet_proxy_arp_add_del(ip4_address_t *lo_addr, ip4_address_t *hi_addr, u32 fib_index, int is_del)
Definition: arp.c:1952
#define pool_init_fixed(pool, max_elts)
initialize a fixed-size, preallocated pool
Definition: pool.h:86
void transport_endpoint_cleanup(u8 proto, ip46_address_t *lcl_ip, u16 port)
Definition: transport.c:207
void transport_register_protocol(transport_proto_t transport_proto, const transport_proto_vft_t *vft, fib_protocol_t fib_proto, u32 output_node)
Register transport virtual function table.
Definition: transport.c:163
#define VLIB_CLI_COMMAND(x,...)
Definition: cli.h:154
static void tcp_timer_update(tcp_connection_t *tc, u8 timer_id, u32 interval)
Definition: tcp.h:711
#define pool_put_index(p, i)
Free pool element with given index.
Definition: pool.h:296
#define ASSERT(truth)
unsigned int u32
Definition: types.h:88
static clib_error_t * tcp_scoreboard_trace_fn(vlib_main_t *vm, unformat_input_t *input, vlib_cli_command_t *cmd_arg)
Definition: tcp.c:1742
unformat_function_t unformat_pg_tcp_header
Definition: format.h:104
static load_balance_t * load_balance_get(index_t lbi)
Definition: load_balance.h:200
#define seq_geq(_s1, _s2)
Definition: tcp.h:533
vhost_vring_state_t state
Definition: vhost-user.h:82
void stream_session_reset_notify(transport_connection_t *tc)
Notify application that connection has been reset.
Definition: session.c:757
From the control plane API.
Definition: fib_entry.h:70
i32 tcp_rcv_wnd_available(tcp_connection_t *tc)
Definition: tcp.c:996
u8 * tcp_scoreboard_replay(u8 *s, tcp_connection_t *tc, u8 verbose)
Definition: tcp.c:1664
clib_error_t * ip4_lookup_init(vlib_main_t *vm)
Definition: ip4_forward.c:821
int tcp_configure_v6_source_address_range(vlib_main_t *vm, ip6_address_t *start, ip6_address_t *end, u32 table_id)
Configure an ipv6 source address range.
Definition: tcp.c:1440
#define clib_max(x, y)
Definition: clib.h:333
u8 * format_tcp_timers(u8 *s, va_list *args)
Definition: tcp.c:639
dpo_id_t fe_lb
The load-balance used for forwarding.
Definition: fib_entry.h:479
u8 ip_is_zero(ip46_address_t *ip46_address, u8 is_ip4)
Definition: ip.c:20
static const transport_proto_vft_t tcp_proto
Definition: tcp.c:1022
void tcp_update_snd_mss(tcp_connection_t *tc)
Update snd_mss to reflect the effective segment size that we can send by taking into account all TCP ...
Definition: tcp_output.c:402
void tcp_punt_unknown(vlib_main_t *vm, u8 is_ip4, u8 is_add)
Definition: tcp.c:1273
#define tcp_fastrecovery_1_smss_on(tc)
Definition: tcp.h:327
int tcp_session_open(transport_endpoint_t *tep)
Definition: tcp.c:578
unsigned short u16
Definition: types.h:57
void tcp_connection_reset(tcp_connection_t *tc)
Notify session that connection has been reset.
Definition: tcp.c:221
vlib_node_registration_t tcp4_input_node
(constructor) VLIB_REGISTER_NODE (tcp4_input_node)
Definition: tcp_input.c:3110
index_t dpoi_index
the index of objects of that type
Definition: dpo.h:184
#define FIB_NODE_INDEX_INVALID
Definition: fib_types.h:31
#define foreach_tcp_fsm_state
TCP FSM state definitions as per RFC793.
Definition: tcp.h:42
void tcp_send_fin(tcp_connection_t *tc)
Send FIN.
Definition: tcp_output.c:1052
#define foreach_tcp_connection_flag
TCP connection flags.
Definition: tcp.h:114
#define foreach_tcp_timer
TCP timers.
Definition: tcp.h:69
#define vec_len(v)
Number of elements in vector (rvalue-only, NULL tolerant)
double f64
Definition: types.h:142
unsigned char u8
Definition: types.h:56
static u32 tcp_set_time_now(u32 thread_index)
Definition: tcp.h:662
static u32 tcp_round_snd_space(tcp_connection_t *tc, u32 snd_space)
Definition: tcp.c:912
struct _transport_endpoint transport_endpoint_t
void tcp_connection_init_vars(tcp_connection_t *tc)
Initialize tcp connection variables.
Definition: tcp.c:488
#define DPO_INVALID
An initialiser for DPOs declared on the stack.
Definition: dpo.h:195
const char * tcp_conn_timers[]
Definition: tcp.c:632
static u32 vlib_buffer_free_list_buffer_size(vlib_main_t *vm, vlib_buffer_free_list_index_t index)
Definition: buffer_funcs.h:533
left
clib_error_t * tcp_init(vlib_main_t *vm)
Definition: tcp.c:1283
u8 * format_tcp_connection(u8 *s, va_list *args)
Definition: tcp.c:741
static tcp_connection_t * tcp_connection_get(u32 conn_index, u32 thread_index)
Definition: tcp.h:449
static u32 random_u32(u32 *seed)
32-bit random number generator
Definition: random.h:69
u8 * format_unformat_error(u8 *s, va_list *va)
Definition: unformat.c:91
#define TCP_SCOREBOARD_TRACE
Definition: tcp.h:161
int session_stream_connect_notify(transport_connection_t *tc, u8 is_fail)
Definition: session.c:549
static vlib_thread_main_t * vlib_get_thread_main()
Definition: global_funcs.h:32
void tcp_connection_cleanup(tcp_connection_t *tc)
Cleans up connection state.
Definition: tcp.c:158
void tcp_connection_del(tcp_connection_t *tc)
Connection removal.
Definition: tcp.c:196
void tcp_init_mss(tcp_connection_t *tc)
Definition: tcp_output.c:415
void dpo_reset(dpo_id_t *dpo)
reset a DPO ID The DPO will be unlocked.
Definition: dpo.c:231
#define vec_foreach(var, vec)
Vector iterator.
int transport_alloc_local_endpoint(u8 proto, transport_endpoint_t *rmt, ip46_address_t *lcl_addr, u16 *lcl_port)
Definition: transport.c:273
u8 * format_tcp_sack_hole(u8 *s, va_list *args)
Definition: tcp.c:840
#define TCP_TSTAMP_RESOLUTION
Time stamp resolution.
Definition: tcp.h:29
#define tcp_opts_sack_permitted(_to)
Definition: tcp_packet.h:160
int transport_alloc_local_port(u8 proto, ip46_address_t *ip)
Allocate local port and add if successful add entry to local endpoint table to mark the pair as used...
Definition: transport.c:228
u8 * format_tcp_state(u8 *s, va_list *args)
Definition: tcp.c:596
u32 tcp_session_send_space(transport_connection_t *trans_conn)
Definition: tcp.c:988
static clib_error_t * ip6_lookup_init(vlib_main_t *vm)
Definition: ip6_forward.c:2753
static clib_error_t * show_tcp_punt_fn(vlib_main_t *vm, unformat_input_t *input, vlib_cli_command_t *cmd_arg)
Definition: tcp.c:1785
static tcp_main_t * vnet_get_tcp_main()
Definition: tcp.h:422
timer_expiration_handler tcp_timer_delack_handler
void vlib_cli_output(vlib_main_t *vm, char *fmt,...)
Definition: cli.c:680
static_always_inline void clib_spinlock_lock_if_init(clib_spinlock_t *p)
Definition: lock.h:82
transport_connection_t * tcp_session_get_transport(u32 conn_index, u32 thread_index)
Definition: tcp.c:878
static clib_error_t * tcp_show_scoreboard_trace_fn(vlib_main_t *vm, unformat_input_t *input, vlib_cli_command_t *cmd_arg)
Definition: tcp.c:1626
u8 * format_tcp_listener_session(u8 *s, va_list *args)
Definition: tcp.c:776
void tcp_initialize_timer_wheels(tcp_main_t *tm)
Definition: tcp.c:1151
uword unformat(unformat_input_t *i, const char *fmt,...)
Definition: unformat.c:972
static u8 tcp_timer_is_active(tcp_connection_t *tc, tcp_timers_e timer)
Definition: tcp.h:781
static tcp_connection_t * tcp_listener_get(u32 tli)
Definition: tcp.h:493
static uword unformat_check_input(unformat_input_t *i)
Definition: format.h:169