summaryrefslogtreecommitdiffstats
path: root/src/vnet/ip/ping.c
blob: a2b96ab62896c72a76fd45c17c4f702d79f9b52b (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
/*
 * Copyright (c) 2016 Cisco and/or its affiliates.
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at:
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

#include <stddef.h>
#include <vnet/ip/ping.h>
#include <vnet/fib/ip6_fib.h>
#include <vnet/fib/ip4_fib.h>
#include <vnet/fib/fib_entry.h>

/**
 * @file
 * @brief IPv4 and IPv6 ICMP Ping.
 *
 * This file contains code to suppport IPv4 or IPv6 ICMP ECHO_REQUEST to
 * network hosts.
 *
 */


u8 *
format_icmp_echo_trace (u8 * s, va_list * va)
{
  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*va, vlib_main_t *);
  CLIB_UNUSED (vlib_node_t * node) = va_arg (*va, vlib_node_t *);
  icmp_echo_trace_t *t = va_arg (*va, icmp_echo_trace_t *);

  s = format (s, "ICMP echo id %d seq %d%s",
	      clib_net_to_host_u16 (t->id),
	      clib_net_to_host_u16 (t->seq), t->bound ? "" : " (unknown)");

  return s;
}

/*
 * If we can find the ping run by an ICMP ID, then we send the signal
 * to the CLI process referenced by that ping run, alongside with
 * a freshly made copy of the packet.
 * I opted for a packet copy to keep the main packet processing path
 * the same as for all the other nodes.
 *
 */

static int
signal_ip46_icmp_reply_event (vlib_main_t * vm,
			      u8 event_type, vlib_buffer_t * b0)
{
  ping_main_t *pm = &ping_main;
  u16 net_icmp_id = 0;
  u32 bi0_copy = 0;

  switch (event_type)
    {
    case PING_RESPONSE_IP4:
      {
	icmp4_echo_request_header_t *h0 = vlib_buffer_get_current (b0);
	net_icmp_id = h0->icmp_echo.id;
      }
      break;
    case PING_RESPONSE_IP6:
      {
	icmp6_echo_request_header_t *h0 = vlib_buffer_get_current (b0);
	net_icmp_id = h0->icmp_echo.id;
      }
      break;
    default:
      return 0;
    }

  uword *p = hash_get (pm->ping_run_by_icmp_id,
		       clib_net_to_host_u16 (net_icmp_id));
  if (!p)
    return 0;

  ping_run_t *pr = vec_elt_at_index (pm->ping_runs, p[0]);
  if (vlib_buffer_alloc (vm, &bi0_copy, 1) == 1)
    {
      void *dst = vlib_buffer_get_current (vlib_get_buffer (vm, bi0_copy));
      clib_memcpy (dst, vlib_buffer_get_current (b0), b0->current_length);
    }
  /* If buffer_alloc failed, bi0_copy == 0 - just signaling an event. */

  vlib_process_signal_event (vm, pr->cli_process_id, event_type, bi0_copy);
  return 1;
}

/*
 * Process ICMPv6 echo replies
 */
static uword
ip6_icmp_echo_reply_node_fn (vlib_main_t * vm,
			     vlib_node_runtime_t * node, vlib_frame_t * frame)
{
  u32 n_left_from, *from;

  from = vlib_frame_vector_args (frame);	/* array of buffer indices */
  n_left_from = frame->n_vectors;	/* number of buffer indices */

  while (n_left_from > 0)
    {
      u32 bi0;
      vlib_buffer_t *b0;
      u32 next0;

      bi0 = from[0];
      b0 = vlib_get_buffer (vm, bi0);

      next0 = signal_ip46_icmp_reply_event (vm, PING_RESPONSE_IP6, b0) ?
	ICMP6_ECHO_REPLY_NEXT_DROP : ICMP6_ECHO_REPLY_NEXT_PUNT;

      if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
	{
	  icmp6_echo_request_header_t *h0 = vlib_buffer_get_current (b0);
	  icmp_echo_trace_t *tr = vlib_add_trace (vm, node, b0, sizeof (*tr));
	  tr->id = h0->icmp_echo.id;
	  tr->seq = h0->icmp_echo.seq;
	  tr->bound = (next0 == ICMP6_ECHO_REPLY_NEXT_DROP);
	}

      /* push this pkt to the next graph node */
      vlib_set_next_frame_buffer (vm, node, next0, bi0);

      from += 1;
      n_left_from -= 1;
    }

  return frame->n_vectors;
}

/* *INDENT-OFF* */
VLIB_REGISTER_NODE (ip6_icmp_echo_reply_node, static) =
{
  .function = ip6_icmp_echo_reply_node_fn,
  .name = "ip6-icmp-echo-reply",
  .vector_size = sizeof (u32),
  .format_trace = format_icmp_echo_trace,
  .n_next_nodes = ICMP6_ECHO_REPLY_N_NEXT,
  .next_nodes = {
    [ICMP6_ECHO_REPLY_NEXT_DROP] = "error-drop",
    [ICMP6_ECHO_REPLY_NEXT_PUNT] = "error-punt",
  },
};
/* *INDENT-ON* */

/*
 * Process ICMPv4 echo replies
 */
static uword
ip4_icmp_echo_reply_node_fn (vlib_main_t * vm,
			     vlib_node_runtime_t * node, vlib_frame_t * frame)
{
  u32 n_left_from, *from;

  from = vlib_frame_vector_args (frame);	/* array of buffer indices */
  n_left_from = frame->n_vectors;	/* number of buffer indices */

  while (n_left_from > 0)
    {
      u32 bi0;
      vlib_buffer_t *b0;
      u32 next0;

      bi0 = from[0];
      b0 = vlib_get_buffer (vm, bi0);

      next0 = signal_ip46_icmp_reply_event (vm, PING_RESPONSE_IP4, b0) ?
	ICMP4_ECHO_REPLY_NEXT_DROP : ICMP4_ECHO_REPLY_NEXT_PUNT;

      if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
	{
	  icmp4_echo_request_header_t *h0 = vlib_buffer_get_current (b0);
	  icmp_echo_trace_t *tr = vlib_add_trace (vm, node, b0, sizeof (*tr));
	  tr->id = h0->icmp_echo.id;
	  tr->seq = h0->icmp_echo.seq;
	  tr->bound = (next0 == ICMP4_ECHO_REPLY_NEXT_DROP);
	}

      /* push this pkt to the next graph node */
      vlib_set_next_frame_buffer (vm, node, next0, bi0);

      from += 1;
      n_left_from -= 1;
    }

  return frame->n_vectors;
}

/* *INDENT-OFF* */
VLIB_REGISTER_NODE (ip4_icmp_echo_reply_node, static) =
{
  .function = ip4_icmp_echo_reply_node_fn,
  .name = "ip4-icmp-echo-reply",
  .vector_size = sizeof (u32),
  .format_trace = format_icmp_echo_trace,
  .n_next_nodes = ICMP4_ECHO_REPLY_N_NEXT,
  .next_nodes = {
    [ICMP4_ECHO_REPLY_NEXT_DROP] = "error-drop",
    [ICMP4_ECHO_REPLY_NEXT_PUNT] = "error-punt",
  },
};
/* *INDENT-ON* */

char *ip6_lookup_next_nodes[] = IP6_LOOKUP_NEXT_NODES;
char *ip4_lookup_next_nodes[] = IP4_LOOKUP_NEXT_NODES;

/* Fill in the ICMP ECHO structure, return the safety-checked and possibly shrunk data_len */
static u16
init_icmp46_echo_request (icmp46_echo_request_t * icmp46_echo,
			  u16 seq_host, u16 id_host, u16 data_len)
{
  int i;
  icmp46_echo->seq = clib_host_to_net_u16 (seq_host);
  icmp46_echo->id = clib_host_to_net_u16 (id_host);

  if (data_len > PING_MAXIMUM_DATA_SIZE)
    data_len = PING_MAXIMUM_DATA_SIZE;
  for (i = 0; i < data_len; i++)
    icmp46_echo->data[i] = i % 256;
  return data_len;
}

static send_ip46_ping_result_t
send_ip6_ping (vlib_main_t * vm, ip6_main_t * im,
	       u32 table_id, ip6_address_t * pa6,
	       u32 sw_if_index, u16 seq_host, u16 id_host, u16 data_len,
	       u8 verbose)
{
  icmp6_echo_request_header_t *h0;
  u32 bi0 = 0;
  int bogus_length = 0;
  vlib_buffer_t *p0;
  vlib_frame_t *f;
  u32 *to_next;
  vlib_buffer_free_list_t *fl;

  if (vlib_buffer_alloc (vm, &bi0, 1) != 1)
    return SEND_PING_ALLOC_FAIL;

  p0 = vlib_get_buffer (vm, bi0);
  fl = vlib_buffer_get_free_list (vm, VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX);
  vlib_buffer_init_for_free_list (p0, fl);
  VLIB_BUFFER_TRACE_TRAJECTORY_INIT (p0);

  /*
   * if the user did not provide a source interface, use the any interface
   * that the destination resolves via.
   */
  if (~0 == sw_if_index)
    {
      fib_node_index_t fib_entry_index;
      u32 fib_index;

      fib_index = ip6_fib_index_from_table_id (table_id);

      if (~0 == fib_index)
	{
	  vlib_buffer_free (vm, &bi0, 1);
	  return SEND_PING_NO_TABLE;
	}

      fib_entry_index = ip6_fib_table_lookup (fib_index, pa6, 128);
      sw_if_index = fib_entry_get_resolving_interface (fib_entry_index);
      /*
       * Set the TX interface to force ip-lookup to use its table ID
       */
      vnet_buffer (p0)->sw_if_index[VLIB_TX] = fib_index;
    }
  else
    {
      /*
       * force an IP lookup in the table bound to the user's chosen
       * source interface.
       */
      vnet_buffer (p0)->sw_if_index[VLIB_TX] =
	ip6_fib_table_get_index_for_sw_if_index (sw_if_index);
    }

  if (~0 == sw_if_index)
    {
      vlib_buffer_free (vm, &bi0, 1);
      return SEND_PING_NO_INTERFACE;
    }

  vnet_buffer (p0)->sw_if_index[VLIB_RX] = sw_if_index;

  h0 = vlib_buffer_get_current (p0);

  /* Fill in ip6 header fields */
  h0->ip6.ip_version_traffic_class_and_flow_label =
    clib_host_to_net_u32 (0x6 << 28);
  h0->ip6.payload_length = 0;	/* Set below */
  h0->ip6.protocol = IP_PROTOCOL_ICMP6;
  h0->ip6.hop_limit = 255;
  h0->ip6.dst_address = *pa6;
  h0->ip6.src_address = *pa6;

  /* Fill in the correct source now */
  ip6_address_t *a = ip6_interface_first_address (im, sw_if_index);
  if (!a)
    {
      vlib_buffer_free (vm, &bi0, 1);
      return SEND_PING_NO_SRC_ADDRESS;
    }
  h0->ip6.src_address = a[0];

  /* Fill in icmp fields */
  h0->icmp.type = ICMP6_echo_request;
  h0->icmp.code = 0;
  h0->icmp.checksum = 0;

  data_len =
    init_icmp46_echo_request (&h0->icmp_echo, seq_host, id_host, data_len);
  h0->icmp_echo.time_sent = vlib_time_now (vm);

  /* Fix up the lengths */
  h0->ip6.payload_length =
    clib_host_to_net_u16 (data_len + sizeof (icmp46_header_t));

  p0->current_length = clib_net_to_host_u16 (h0->ip6.payload_length) +
    STRUCT_OFFSET_OF (icmp6_echo_request_header_t, icmp);

  /* Calculate the ICMP checksum */
  h0->icmp.checksum = 0;
  h0->icmp.checksum =
    ip6_tcp_udp_icmp_compute_checksum (vm, 0, &h0->ip6, &bogus_length);

  /* Enqueue the packet right now */
  f = vlib_get_frame_to_node (vm, ip6_lookup_node.index);
  to_next = vlib_frame_vector_args (f);
  to_next[0] = bi0;
  f->n_vectors = 1;
  vlib_put_frame_to_node (vm, ip6_lookup_node.index, f);

  return SEND_PING_OK;
}

static send_ip46_ping_result_t
send_ip4_ping (vlib_main_t * vm,
	       ip4_main_t * im,
	       u32 table_id,
	       ip4_address_t * pa4,
	       u32 sw_if_index,
	       u16 seq_host, u16 id_host, u16 data_len, u8 verbose)
{
  icmp4_echo_request_header_t *h0;
  u32 bi0 = 0;
  ip_lookup_main_t *lm = &im->lookup_main;
  vlib_buffer_t *p0;
  vlib_frame_t *f;
  u32 *to_next;
  u32 if_add_index0;
  vlib_buffer_free_list_t *fl;

  if (vlib_buffer_alloc (vm, &bi0, 1) != 1)
    return SEND_PING_ALLOC_FAIL;

  p0 = vlib_get_buffer (vm, bi0);
  fl = vlib_buffer_get_free_list (vm, VLIB_BUFFER_DEFAULT_FREE_LIST_INDEX);
  vlib_buffer_init_for_free_list (p0, fl);
  VLIB_BUFFER_TRACE_TRAJECTORY_INIT (p0);

  /*
   * if the user did not provide a source interface, use the any interface
   * that the destination resolves via.
   */
  if (~0 == sw_if_index)
    {
      fib_node_index_t fib_entry_index;
      u32 fib_index;

      fib_index = ip4_fib_index_from_table_id (table_id);

      if (~0 == fib_index)
	{
	  vlib_buffer_free (vm, &bi0, 1);
	  return SEND_PING_NO_TABLE;
	}

      fib_entry_index =
	ip4_fib_table_lookup (ip4_fib_get (fib_index), pa4, 32);
      sw_if_index = fib_entry_get_resolving_interface (fib_entry_index);
      /*
       * Set the TX interface to force ip-lookup to use the user's table ID
       */
      vnet_buffer (p0)->sw_if_index[VLIB_TX] = fib_index;
    }
  else
    {
      /*
       * force an IP lookup in the table bound to the user's chosen
       * source interface.
       */
      vnet_buffer (p0)->sw_if_index[VLIB_TX] =
	ip4_fib_table_get_index_for_sw_if_index (sw_if_index);
    }

  if (~0 == sw_if_index)
    {
      vlib_buffer_free (vm, &bi0, 1);
      return SEND_PING_NO_INTERFACE;
    }

  vnet_buffer (p0)->sw_if_index[VLIB_RX] = sw_if_index;

  h0 = vlib_buffer_get_current (p0);

  /* Fill in ip4 header fields */
  h0->ip4.checksum = 0;
  h0->ip4.ip_version_and_header_length = 0x45;
  h0->ip4.tos = 0;
  h0->ip4.length = 0;		/* Set below */
  h0->ip4.fragment_id = 0;
  h0->ip4.flags_and_fragment_offset = 0;
  h0->ip4.ttl = 0xff;
  h0->ip4.protocol = IP_PROTOCOL_ICMP;
  h0->ip4.dst_address = *pa4;
  h0->ip4.src_address = *pa4;

  /* Fill in the correct source now */
  if_add_index0 = lm->if_address_pool_index_by_sw_if_index[sw_if_index];
  if (PREDICT_TRUE (if_add_index0 != ~0))
    {
      ip_interface_address_t *if_add =
	pool_elt_at_index (lm->if_address_pool, if_add_index0);
      ip4_address_t *if_ip = ip_interface_address_get_address (lm, if_add);
      h0->ip4.src_address = *if_ip;
      if (verbose)
	{
	  vlib_cli_output (vm, "Source address: %U",
			   format_ip4_address, &h0->ip4.src_address);
	}
    }

  /* Fill in icmp fields */
  h0->icmp.type = ICMP4_echo_request;
  h0->icmp.code = 0;
  h0->icmp.checksum = 0;

  data_len =
    init_icmp46_echo_request (&h0->icmp_echo, seq_host, id_host, data_len);
  h0->icmp_echo.time_sent = vlib_time_now (vm);

  /* Fix up the lengths */
  h0->ip4.length =
    clib_host_to_net_u16 (data_len + sizeof (icmp46_header_t) +
			  sizeof (ip4_header_t));

  p0->current_length = clib_net_to_host_u16 (h0->ip4.length);

  /* Calculate the IP and ICMP checksums */
  h0->ip4.checksum = ip4_header_checksum (&(h0->ip4));
  h0->icmp.checksum =
    ~ip_csum_fold (ip_incremental_checksum (0, &(h0->icmp),
					    p0->current_length -
					    sizeof (ip4_header_t)));

  /* Enqueue the packet right now */
  f = vlib_get_frame_to_node (vm, ip4_lookup_node.index);
  to_next = vlib_frame_vector_args (f);
  to_next[0] = bi0;
  f->n_vectors = 1;
  vlib_put_frame_to_node (vm, ip4_lookup_node.index, f);

  return SEND_PING_OK;
}


static void
print_ip6_icmp_reply (vlib_main_t * vm, u32 bi0)
{
  vlib_buffer_t *b0 = vlib_get_buffer (vm, bi0);
  icmp6_echo_request_header_t *h0 = vlib_buffer_get_current (b0);
  f64 rtt = vlib_time_now (vm) - h0->icmp_echo.time_sent;

  vlib_cli_output (vm,
		   "%d bytes from %U: icmp_seq=%d ttl=%d time=%.4f ms",
		   clib_host_to_net_u16 (h0->ip6.payload_length),
		   format_ip6_address,
		   &h0->ip6.src_address,
		   clib_host_to_net_u16 (h0->icmp_echo.seq),
		   h0->ip6.hop_limit, rtt * 1000.0);
}

static void
print_ip4_icmp_reply (vlib_main_t * vm, u32 bi0)
{
  vlib_buffer_t *b0 = vlib_get_buffer (vm, bi0);
  icmp4_echo_request_header_t *h0 = vlib_buffer_get_current (b0);
  f64 rtt = vlib_time_now (vm) - h0->icmp_echo.time_sent;
  u32 rcvd_icmp_len =
    clib_host_to_net_u16 (h0->ip4.length) -
    (4 * (0xF & h0->ip4.ip_version_and_header_length));

  vlib_cli_output (vm,
		   "%d bytes from %U: icmp_seq=%d ttl=%d time=%.4f ms",
		   rcvd_icmp_len,
		   format_ip4_address,
		   &h0->ip4.src_address,
		   clib_host_to_net_u16 (h0->icmp_echo.seq),
		   h0->ip4.ttl, rtt * 1000.0);
}


/*
 * Perform the ping run with the given parameters in the current CLI process.
 * Depending on whether pa4 or pa6 is set, runs IPv4 or IPv6 ping.
 * The amusing side effect is of course if both are set, then both pings are sent.
 * This behavior can be used to ping a dualstack host over IPv4 and IPv6 at once.
 */

static void
run_ping_ip46_address (vlib_main_t * vm, u32 table_id, ip4_address_t * pa4,
		       ip6_address_t * pa6, u32 sw_if_index,
		       f64 ping_interval, u32 ping_repeat, u32 data_len,
		       u32 verbose)
{
  int i;
  ping_main_t *pm = &ping_main;
  uword curr_proc = vlib_current_process (vm);
  u32 n_replies = 0;
  u32 n_requests = 0;
  ping_run_t *pr = 0;
  u32 ping_run_index = 0;
  u16 icmp_id;

  static u32 rand_seed = 0;

  if (PREDICT_FALSE (!rand_seed))
    rand_seed = random_default_seed ();

  icmp_id = random_u32 (&rand_seed) & 0xffff;

  while (hash_get (pm->ping_run_by_icmp_id, icmp_id))
    {
      vlib_cli_output (vm, "ICMP ID collision at %d, incrementing", icmp_id);
      icmp_id++;
    }
  pool_get (pm->ping_runs, pr);
  ping_run_index = pr - pm->ping_runs;
  pr->cli_process_id = curr_proc;
  pr->icmp_id = icmp_id;
  hash_set (pm->ping_run_by_icmp_id, icmp_id, ping_run_index);
  for (i = 1; i <= ping_repeat; i++)
    {
      f64 sleep_interval;
      f64 time_ping_sent = vlib_time_now (vm);
      /* Reset pr: running ping in other process could have changed pm->ping_runs */
      pr = vec_elt_at_index (pm->ping_runs, ping_run_index);
      pr->curr_seq = i;
      if (pa6 &&
	  (SEND_PING_OK ==
	   send_ip6_ping (vm, ping_main.ip6_main, table_id, pa6, sw_if_index,
			  i, icmp_id, data_len, verbose)))
	{
	  n_requests++;
	}
      if (pa4 &&
	  (SEND_PING_OK ==
	   send_ip4_ping (vm, ping_main.ip4_main, table_id, pa4, sw_if_index,
			  i, icmp_id, data_len, verbose)))
	{
	  n_requests++;
	}
      while ((i <= ping_repeat)
	     &&
	     ((sleep_interval =
	       time_ping_sent + ping_interval - vlib_time_now (vm)) > 0.0))
	{
	  uword event_type, *event_data = 0;
	  vlib_process_wait_for_event_or_clock (vm, sleep_interval);
	  event_type = vlib_process_get_events (vm, &event_data);
	  switch (event_type)
	    {
	    case ~0:		/* no events => timeout */
	      break;
	    case PING_RESPONSE_IP6:
	      {
		int i;
		for (i = 0; i < vec_len (event_data); i++)
		  {
		    u32 bi0 = event_data[i];
		    print_ip6_icmp_reply (vm, bi0);
		    n_replies++;
		    if (0 != bi0)
		      {
			vlib_buffer_free (vm, &bi0, 1);
		      }
		  }
	      }
	      break;
	    case PING_RESPONSE_IP4:
	      {
		int i;
		for (i = 0; i < vec_len (event_data); i++)
		  {
		    u32 bi0 = event_data[i];
		    print_ip4_icmp_reply (vm, bi0);
		    n_replies++;
		    if (0 != bi0)
		      {
			vlib_buffer_free (vm, &bi0, 1);
		      }
		  }
	      }
	      break;
	    default:
	      /* someone pressed a key, abort */
	      vlib_cli_output (vm, "Aborted due to a keypress.");
	      i = 1 + ping_repeat;
	      break;
	    }
	}
    }
  vlib_cli_output (vm, "\n");
  {
    float loss =
      (0 ==
       n_requests) ? 0 : 100.0 * ((float) n_requests -
				  (float) n_replies) / (float) n_requests;
    vlib_cli_output (vm,
		     "Statistics: %u sent, %u received, %f%% packet loss\n",
		     n_requests, n_replies, loss);
    /* Reset pr: running ping in other process could have changed pm->ping_runs */
    pr = vec_elt_at_index (pm->ping_runs, ping_run_index);
    hash_unset (pm->ping_run_by_icmp_id, icmp_id);
    pool_put (pm->ping_runs, pr);
  }
}





static clib_error_t *
ping_ip_address (vlib_main_t * vm,
		 unformat_input_t * input, vlib_cli_command_t * cmd)
{
  ip4_address_t a4;
  ip6_address_t a6;
  clib_error_t *error = 0;
  u32 ping_repeat = 5;
  u8 ping_ip4, ping_ip6;
  vnet_main_t *vnm = vnet_get_main ();
  u32 data_len = PING_DEFAULT_DATA_LEN;
  u32 verbose = 0;
  f64 ping_interval = PING_DEFAULT_INTERVAL;
  u32 sw_if_index, table_id;

  table_id = 0;
  ping_ip4 = ping_ip6 = 0;
  sw_if_index = ~0;

  if (unformat (input, "%U", unformat_ip4_address, &a4))
    {
      ping_ip4 = 1;
    }
  else if (unformat (input, "%U", unformat_ip6_address, &a6))
    {
      ping_ip6 = 1;
    }
  else if (unformat (input, "ipv4"))
    {
      if (unformat (input, "%U", unformat_ip4_address, &a4))
	{
	  ping_ip4 = 1;
	}
      else
	{
	  error =
	    clib_error_return (0,
			       "expecting IPv4 address but got `%U'",
			       format_unformat_error, input);
	}
    }
  else if (unformat (input, "ipv6"))
    {
      if (unformat (input, "%U", unformat_ip6_address, &a6))
	{
	  ping_ip6 = 1;
	}
      else
	{
	  error =
	    clib_error_return (0,
			       "expecting IPv6 address but got `%U'",
			       format_unformat_error, input);
	}
    }
  else
    {
      error =
	clib_error_return (0,
			   "expecting IP4/IP6 address `%U'. Usage: ping <addr> [source <intf>] [size <datasz>] [repeat <count>] [verbose]",
			   format_unformat_error, input);
      goto done;
    }

  /* allow for the second AF in the same ping */
  if (!ping_ip4 && (unformat (input, "ipv4")))
    {
      if (unformat (input, "%U", unformat_ip4_address, &a4))
	{
	  ping_ip4 = 1;
	}
    }
  else if (!ping_ip6 && (unformat (input, "ipv6")))
    {
      if (unformat (input, "%U", unformat_ip6_address, &a6))
	{
	  ping_ip6 = 1;
	}
    }

  /* parse the rest of the parameters  in a cycle */
  while (!unformat_eof (input, NULL))
    {
      if (unformat (input, "source"))
	{
	  if (!unformat_user
	      (input, unformat_vnet_sw_interface, vnm, &sw_if_index))
	    {
	      error =
		clib_error_return (0,
				   "unknown interface `%U'",
				   format_unformat_error, input);
	      goto done;
	    }
	}
      else if (unformat (input, "size"))
	{
	  if (!unformat (input, "%u", &data_len))
	    {
	      error =
		clib_error_return (0,
				   "expecting size but got `%U'",
				   format_unformat_error, input);
	      goto done;
	    }
	  if (data_len > PING_MAXIMUM_DATA_SIZE)
	    {
	      error =
		clib_error_return (0,
				   "%d is bigger than maximum allowed payload size %d",
				   data_len, PING_MAXIMUM_DATA_SIZE);
	      goto done;
	    }
	}
      else if (unformat (input, "table-id"))
	{
	  if (!unformat (input, "du", &table_id))
	    {
	      error =
		clib_error_return (0,
				   "expecting table-id but got `%U'",
				   format_unformat_error, input);
	      goto done;
	    }
	}
      else if (unformat (input, "interval"))
	{
	  if (!unformat (input, "%f", &ping_interval))
	    {
	      error =
		clib_error_return (0,
				   "expecting interval (floating point number) got `%U'",
				   format_unformat_error, input);
	      goto done;
	    }
	}
      else if (unformat (input, "repeat"))
	{
	  if (!unformat (input, "%u", &ping_repeat))
	    {
	      error =
		clib_error_return (0,
				   "expecting repeat count but got `%U'",
				   format_unformat_error, input);
	      goto done;
	    }
	}
      else if (unformat (input, "verbose"))
	{
	  verbose = 1;
	}
      else
	{
	  error = clib_error_return (0, "unknown input `%U'",
				     format_unformat_error, input);
	  goto done;
	}
    }

  run_ping_ip46_address (vm, table_id, ping_ip4 ? &a4 : NULL,
			 ping_ip6 ? &a6 : NULL, sw_if_index, ping_interval,
			 ping_repeat, data_len, verbose);
done:
  return error;
}

/*?
 * This command sends an ICMP ECHO_REQUEST to network hosts. The address
 * can be an IPv4 or IPv6 address (or both at the same time).
 *
 * @cliexpar
 * @parblock
 * Example of how ping an IPv4 address:
 * @cliexstart{ping 172.16.1.2 source GigabitEthernet2/0/0 repeat 2}
 * 64 bytes from 172.16.1.2: icmp_seq=1 ttl=64 time=.1090 ms
 * 64 bytes from 172.16.1.2: icmp_seq=2 ttl=64 time=.0914 ms
 *
 * Statistics: 2 sent, 2 received, 0% packet loss
 * @cliexend
 *
 * Example of how ping both an IPv4 address and IPv6 address at the same time:
 * @cliexstart{ping 172.16.1.2 ipv6 fe80::24a5:f6ff:fe9c:3a36 source GigabitEthernet2/0/0 repeat 2 verbose}
 * Adjacency index: 10, sw_if_index: 1
 * Adj: ip6-discover-neighbor
 * Adj Interface: 0
 * Forced set interface: 1
 * Adjacency index: 0, sw_if_index: 4294967295
 * Adj: ip4-miss
 * Adj Interface: 0
 * Forced set interface: 1
 * Source address: 172.16.1.1
 * 64 bytes from 172.16.1.2: icmp_seq=1 ttl=64 time=.1899 ms
 * Adjacency index: 10, sw_if_index: 1
 * Adj: ip6-discover-neighbor
 * Adj Interface: 0
 * Forced set interface: 1
 * Adjacency index: 0, sw_if_index: 4294967295
 * Adj: ip4-miss
 * Adj Interface: 0
 * Forced set interface: 1
 * Source address: 172.16.1.1
 * 64 bytes from 172.16.1.2: icmp_seq=2 ttl=64 time=.0910 ms
 *
 * Statistics: 4 sent, 2 received, 50% packet loss
 * @cliexend
 * @endparblock
?*/
/* *INDENT-OFF* */
VLIB_CLI_COMMAND (ping_command, static) =
{
  .path = "ping",
  .function = ping_ip_address,
  .short_help = "ping {<ip-addr> | ipv4 <ip4-addr> | ipv6 <ip6-addr>}"
  " [ipv4 <ip4-addr> | ipv6 <ip6-addr>] [source <interface>]"
  " [size <pktsize>] [interval <sec>] [repeat <cnt>] [table-id <id>]"
  " [verbose]",
};
/* *INDENT-ON* */

static clib_error_t *
ping_cli_init (vlib_main_t * vm)
{
  ping_main_t *pm = &ping_main;
  pm->ip6_main = &ip6_main;
  pm->ip4_main = &ip4_main;
  icmp6_register_type (vm, ICMP6_echo_reply, ip6_icmp_echo_reply_node.index);
  ip4_icmp_register_type (vm, ICMP4_echo_reply,
			  ip4_icmp_echo_reply_node.index);
  return 0;
}

VLIB_INIT_FUNCTION (ping_cli_init);

/*
 * fd.io coding-style-patch-verification: ON
 *
 * Local Variables:
 * eval: (c-set-style "gnu")
 * End:
 */
x and trace flag */ b->flags &= VLIB_BUFFER_NEXT_PRESENT - 1; b->current_data = 0; b->current_length = 0; b->total_length_not_including_first_buffer = 0; vnet_buffer (b)->tcp.flags = 0; /* Leave enough space for headers */ return vlib_buffer_make_headroom (b, MAX_HDRS_LEN); } static void * tcp_init_buffer (vlib_main_t * vm, vlib_buffer_t * b) { ASSERT ((b->flags & VLIB_BUFFER_NEXT_PRESENT) == 0); b->flags &= VLIB_BUFFER_NON_DEFAULT_FREELIST; b->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED; b->total_length_not_including_first_buffer = 0; b->current_data = 0; vnet_buffer (b)->tcp.flags = 0; VLIB_BUFFER_TRACE_TRAJECTORY_INIT (b); /* Leave enough space for headers */ return vlib_buffer_make_headroom (b, MAX_HDRS_LEN); } /** * Prepare ACK */ static void tcp_make_ack_i (tcp_connection_t * tc, vlib_buffer_t * b, tcp_state_t state, u8 flags) { tcp_options_t _snd_opts, *snd_opts = &_snd_opts; u8 tcp_opts_len, tcp_hdr_opts_len; tcp_header_t *th; u16 wnd; wnd = tcp_window_to_advertise (tc, state); /* Make and write options */ tcp_opts_len = tcp_make_established_options (tc, snd_opts); tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t); th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->snd_nxt, tc->rcv_nxt, tcp_hdr_opts_len, flags, wnd); tcp_options_write ((u8 *) (th + 1), snd_opts); vnet_buffer (b)->tcp.connection_index = tc->c_c_index; } /** * Convert buffer to ACK */ void tcp_make_ack (tcp_connection_t * tc, vlib_buffer_t * b) { vlib_main_t *vm = vlib_get_main (); tcp_reuse_buffer (vm, b); tcp_make_ack_i (tc, b, TCP_STATE_ESTABLISHED, TCP_FLAG_ACK); TCP_EVT_DBG (TCP_EVT_ACK_SENT, tc); vnet_buffer (b)->tcp.flags = TCP_BUF_FLAG_ACK; tc->rcv_las = tc->rcv_nxt; } /** * Convert buffer to FIN-ACK */ void tcp_make_fin (tcp_connection_t * tc, vlib_buffer_t * b) { vlib_main_t *vm = vlib_get_main (); u8 flags = 0; tcp_reuse_buffer (vm, b); flags = TCP_FLAG_FIN | TCP_FLAG_ACK; tcp_make_ack_i (tc, b, TCP_STATE_ESTABLISHED, flags); /* Reset flags, make sure ack is sent */ vnet_buffer (b)->tcp.flags &= ~TCP_BUF_FLAG_DUPACK; } /** * Convert buffer to SYN */ void tcp_make_syn (tcp_connection_t * tc, vlib_buffer_t * b) { u8 tcp_hdr_opts_len, tcp_opts_len; tcp_header_t *th; u16 initial_wnd; tcp_options_t snd_opts; initial_wnd = tcp_initial_window_to_advertise (tc); /* Make and write options */ memset (&snd_opts, 0, sizeof (snd_opts)); tcp_opts_len = tcp_make_syn_options (&snd_opts, tc->rcv_wscale); tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t); th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->iss, tc->rcv_nxt, tcp_hdr_opts_len, TCP_FLAG_SYN, initial_wnd); vnet_buffer (b)->tcp.connection_index = tc->c_c_index; tcp_options_write ((u8 *) (th + 1), &snd_opts); } /** * Convert buffer to SYN-ACK */ void tcp_make_synack (tcp_connection_t * tc, vlib_buffer_t * b) { vlib_main_t *vm = vlib_get_main (); tcp_options_t _snd_opts, *snd_opts = &_snd_opts; u8 tcp_opts_len, tcp_hdr_opts_len; tcp_header_t *th; u16 initial_wnd; memset (snd_opts, 0, sizeof (*snd_opts)); tcp_reuse_buffer (vm, b); initial_wnd = tcp_initial_window_to_advertise (tc); tcp_opts_len = tcp_make_synack_options (tc, snd_opts); tcp_hdr_opts_len = tcp_opts_len + sizeof (tcp_header_t); th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->iss, tc->rcv_nxt, tcp_hdr_opts_len, TCP_FLAG_SYN | TCP_FLAG_ACK, initial_wnd); tcp_options_write ((u8 *) (th + 1), snd_opts); vnet_buffer (b)->tcp.connection_index = tc->c_c_index; vnet_buffer (b)->tcp.flags = TCP_BUF_FLAG_ACK; /* Init retransmit timer. Use update instead of set because of * retransmissions */ tcp_retransmit_timer_force_update (tc); TCP_EVT_DBG (TCP_EVT_SYNACK_SENT, tc); } always_inline void tcp_enqueue_to_ip_lookup_i (vlib_main_t * vm, vlib_buffer_t * b, u32 bi, u8 is_ip4, u32 fib_index, u8 flush) { tcp_main_t *tm = vnet_get_tcp_main (); u32 thread_index = vlib_get_thread_index (); u32 *to_next, next_index; vlib_frame_t *f; b->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED; b->error = 0; vnet_buffer (b)->sw_if_index[VLIB_TX] = fib_index; vnet_buffer (b)->sw_if_index[VLIB_RX] = 0; /* Send to IP lookup */ next_index = is_ip4 ? ip4_lookup_node.index : ip6_lookup_node.index; tcp_trajectory_add_start (b, 1); f = tm->wrk_ctx[thread_index].ip_lookup_tx_frames[!is_ip4]; if (!f) { f = vlib_get_frame_to_node (vm, next_index); ASSERT (f); tm->wrk_ctx[thread_index].ip_lookup_tx_frames[!is_ip4] = f; } to_next = vlib_frame_vector_args (f); to_next[f->n_vectors] = bi; f->n_vectors += 1; if (flush || f->n_vectors == VLIB_FRAME_SIZE) { vlib_put_frame_to_node (vm, next_index, f); tm->wrk_ctx[thread_index].ip_lookup_tx_frames[!is_ip4] = 0; } } static void tcp_enqueue_to_ip_lookup_now (vlib_main_t * vm, vlib_buffer_t * b, u32 bi, u8 is_ip4, u32 fib_index) { tcp_enqueue_to_ip_lookup_i (vm, b, bi, is_ip4, fib_index, 1); } static void tcp_enqueue_to_ip_lookup (vlib_main_t * vm, vlib_buffer_t * b, u32 bi, u8 is_ip4, u32 fib_index) { tcp_enqueue_to_ip_lookup_i (vm, b, bi, is_ip4, fib_index, 0); if (vm->thread_index == 0 && vlib_num_workers ()) session_flush_frames_main_thread (vm); } always_inline void tcp_enqueue_to_output_i (vlib_main_t * vm, vlib_buffer_t * b, u32 bi, u8 is_ip4, u8 flush) { tcp_main_t *tm = vnet_get_tcp_main (); u32 thread_index = vlib_get_thread_index (); u32 *to_next, next_index; vlib_frame_t *f; b->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED; b->error = 0; /* Decide where to send the packet */ next_index = is_ip4 ? tcp4_output_node.index : tcp6_output_node.index; tcp_trajectory_add_start (b, 2); /* Get frame to v4/6 output node */ f = tm->wrk_ctx[thread_index].tx_frames[!is_ip4]; if (!f) { f = vlib_get_frame_to_node (vm, next_index); ASSERT (f); tm->wrk_ctx[thread_index].tx_frames[!is_ip4] = f; } to_next = vlib_frame_vector_args (f); to_next[f->n_vectors] = bi; f->n_vectors += 1; if (flush || f->n_vectors == VLIB_FRAME_SIZE) { vlib_put_frame_to_node (vm, next_index, f); tm->wrk_ctx[thread_index].tx_frames[!is_ip4] = 0; } } static void tcp_enqueue_to_output (vlib_main_t * vm, vlib_buffer_t * b, u32 bi, u8 is_ip4) { tcp_enqueue_to_output_i (vm, b, bi, is_ip4, 0); } static void tcp_enqueue_to_output_now (vlib_main_t * vm, vlib_buffer_t * b, u32 bi, u8 is_ip4) { tcp_enqueue_to_output_i (vm, b, bi, is_ip4, 1); } static int tcp_make_reset_in_place (vlib_main_t * vm, vlib_buffer_t * b0, tcp_state_t state, u8 thread_index, u8 is_ip4) { ip4_header_t *ih4; ip6_header_t *ih6; tcp_header_t *th0; ip4_address_t src_ip40, dst_ip40; ip6_address_t src_ip60, dst_ip60; u16 src_port, dst_port; u32 tmp; u32 seq, ack; u8 flags; /* Find IP and TCP headers */ th0 = tcp_buffer_hdr (b0); /* Save src and dst ip */ if (is_ip4) { ih4 = vlib_buffer_get_current (b0); ASSERT ((ih4->ip_version_and_header_length & 0xF0) == 0x40); src_ip40.as_u32 = ih4->src_address.as_u32; dst_ip40.as_u32 = ih4->dst_address.as_u32; } else { ih6 = vlib_buffer_get_current (b0); ASSERT ((ih6->ip_version_traffic_class_and_flow_label & 0xF0) == 0x60); clib_memcpy (&src_ip60, &ih6->src_address, sizeof (ip6_address_t)); clib_memcpy (&dst_ip60, &ih6->dst_address, sizeof (ip6_address_t)); } src_port = th0->src_port; dst_port = th0->dst_port; /* Try to determine what/why we're actually resetting */ if (state == TCP_STATE_CLOSED) { if (!tcp_syn (th0)) return -1; tmp = clib_net_to_host_u32 (th0->seq_number); /* Got a SYN for no listener. */ flags = TCP_FLAG_RST | TCP_FLAG_ACK; ack = clib_host_to_net_u32 (tmp + 1); seq = 0; } else { flags = TCP_FLAG_RST; seq = th0->ack_number; ack = 0; } tcp_reuse_buffer (vm, b0); tcp_trajectory_add_start (b0, 4); th0 = vlib_buffer_push_tcp_net_order (b0, dst_port, src_port, seq, ack, sizeof (tcp_header_t), flags, 0); if (is_ip4) { ih4 = vlib_buffer_push_ip4 (vm, b0, &dst_ip40, &src_ip40, IP_PROTOCOL_TCP, 1); th0->checksum = ip4_tcp_udp_compute_checksum (vm, b0, ih4); } else { int bogus = ~0; ih6 = vlib_buffer_push_ip6 (vm, b0, &dst_ip60, &src_ip60, IP_PROTOCOL_TCP); th0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b0, ih6, &bogus); ASSERT (!bogus); } return 0; } /** * Send reset without reusing existing buffer * * It extracts connection info out of original packet */ void tcp_send_reset_w_pkt (tcp_connection_t * tc, vlib_buffer_t * pkt, u8 is_ip4) { vlib_buffer_t *b; u32 bi, sw_if_index, fib_index; tcp_main_t *tm = vnet_get_tcp_main (); vlib_main_t *vm = vlib_get_main (); u8 tcp_hdr_len, flags = 0; tcp_header_t *th, *pkt_th; u32 seq, ack; ip4_header_t *ih4, *pkt_ih4; ip6_header_t *ih6, *pkt_ih6; fib_protocol_t fib_proto; if (PREDICT_FALSE (tcp_get_free_buffer_index (tm, &bi))) return; b = vlib_get_buffer (vm, bi); sw_if_index = vnet_buffer (pkt)->sw_if_index[VLIB_RX]; fib_proto = is_ip4 ? FIB_PROTOCOL_IP4 : FIB_PROTOCOL_IP6; fib_index = fib_table_get_index_for_sw_if_index (fib_proto, sw_if_index); tcp_init_buffer (vm, b); /* Make and write options */ tcp_hdr_len = sizeof (tcp_header_t); if (is_ip4) { pkt_ih4 = vlib_buffer_get_current (pkt); pkt_th = ip4_next_header (pkt_ih4); } else { pkt_ih6 = vlib_buffer_get_current (pkt); pkt_th = ip6_next_header (pkt_ih6); } if (tcp_ack (pkt_th)) { flags = TCP_FLAG_RST; seq = pkt_th->ack_number; ack = (tc && tc->state >= TCP_STATE_SYN_RCVD) ? tc->rcv_nxt : 0; } else { flags = TCP_FLAG_RST | TCP_FLAG_ACK; seq = 0; ack = clib_host_to_net_u32 (vnet_buffer (pkt)->tcp.seq_end); } th = vlib_buffer_push_tcp_net_order (b, pkt_th->dst_port, pkt_th->src_port, seq, ack, tcp_hdr_len, flags, 0); /* Swap src and dst ip */ if (is_ip4) { ASSERT ((pkt_ih4->ip_version_and_header_length & 0xF0) == 0x40); ih4 = vlib_buffer_push_ip4 (vm, b, &pkt_ih4->dst_address, &pkt_ih4->src_address, IP_PROTOCOL_TCP, 1); th->checksum = ip4_tcp_udp_compute_checksum (vm, b, ih4); } else { int bogus = ~0; ASSERT ((pkt_ih6->ip_version_traffic_class_and_flow_label & 0xF0) == 0x60); ih6 = vlib_buffer_push_ip6 (vm, b, &pkt_ih6->dst_address, &pkt_ih6->src_address, IP_PROTOCOL_TCP); th->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b, ih6, &bogus); ASSERT (!bogus); } tcp_enqueue_to_ip_lookup_now (vm, b, bi, is_ip4, fib_index); TCP_EVT_DBG (TCP_EVT_RST_SENT, tc); } /** * Build and set reset packet for connection */ void tcp_send_reset (tcp_connection_t * tc) { vlib_main_t *vm = vlib_get_main (); tcp_main_t *tm = vnet_get_tcp_main (); vlib_buffer_t *b; u32 bi; tcp_header_t *th; u16 tcp_hdr_opts_len, advertise_wnd, opts_write_len; u8 flags; if (PREDICT_FALSE (tcp_get_free_buffer_index (tm, &bi))) return; b = vlib_get_buffer (vm, bi); tcp_init_buffer (vm, b); tc->snd_opts_len = tcp_make_options (tc, &tc->snd_opts, tc->state); tcp_hdr_opts_len = tc->snd_opts_len + sizeof (tcp_header_t); advertise_wnd = tcp_window_to_advertise (tc, TCP_STATE_ESTABLISHED); flags = TCP_FLAG_RST; th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->snd_nxt, tc->rcv_nxt, tcp_hdr_opts_len, flags, advertise_wnd); opts_write_len = tcp_options_write ((u8 *) (th + 1), &tc->snd_opts); ASSERT (opts_write_len == tc->snd_opts_len); vnet_buffer (b)->tcp.connection_index = tc->c_c_index; if (tc->c_is_ip4) { ip4_header_t *ih4; ih4 = vlib_buffer_push_ip4 (vm, b, &tc->c_lcl_ip.ip4, &tc->c_rmt_ip.ip4, IP_PROTOCOL_TCP, 0); th->checksum = ip4_tcp_udp_compute_checksum (vm, b, ih4); } else { int bogus = ~0; ip6_header_t *ih6; ih6 = vlib_buffer_push_ip6 (vm, b, &tc->c_lcl_ip.ip6, &tc->c_rmt_ip.ip6, IP_PROTOCOL_TCP); th->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b, ih6, &bogus); ASSERT (!bogus); } tcp_enqueue_to_ip_lookup_now (vm, b, bi, tc->c_is_ip4, tc->c_fib_index); TCP_EVT_DBG (TCP_EVT_RST_SENT, tc); } static void tcp_push_ip_hdr (tcp_main_t * tm, tcp_connection_t * tc, vlib_buffer_t * b) { tcp_header_t *th = vlib_buffer_get_current (b); vlib_main_t *vm = vlib_get_main (); if (tc->c_is_ip4) { ip4_header_t *ih; ih = vlib_buffer_push_ip4 (vm, b, &tc->c_lcl_ip4, &tc->c_rmt_ip4, IP_PROTOCOL_TCP, 1); th->checksum = ip4_tcp_udp_compute_checksum (vm, b, ih); } else { ip6_header_t *ih; int bogus = ~0; ih = vlib_buffer_push_ip6 (vm, b, &tc->c_lcl_ip6, &tc->c_rmt_ip6, IP_PROTOCOL_TCP); th->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b, ih, &bogus); ASSERT (!bogus); } } /** * Send SYN * * Builds a SYN packet for a half-open connection and sends it to ipx_lookup. * The packet is not forwarded through tcpx_output to avoid doing lookups * in the half_open pool. */ void tcp_send_syn (tcp_connection_t * tc) { vlib_buffer_t *b; u32 bi; tcp_main_t *tm = vnet_get_tcp_main (); vlib_main_t *vm = vlib_get_main (); /* * Setup retransmit and establish timers before requesting buffer * such that we can return if we've ran out. */ tcp_timer_set (tc, TCP_TIMER_ESTABLISH, TCP_ESTABLISH_TIME); tcp_timer_update (tc, TCP_TIMER_RETRANSMIT_SYN, tc->rto * TCP_TO_TIMER_TICK); if (PREDICT_FALSE (tcp_get_free_buffer_index (tm, &bi))) return; b = vlib_get_buffer (vm, bi); tcp_init_buffer (vm, b); tcp_make_syn (tc, b); /* Measure RTT with this */ tc->rtt_ts = tcp_time_now (); tc->rtt_seq = tc->snd_nxt; tc->rto_boff = 0; tcp_push_ip_hdr (tm, tc, b); tcp_enqueue_to_ip_lookup (vm, b, bi, tc->c_is_ip4, tc->c_fib_index); TCP_EVT_DBG (TCP_EVT_SYN_SENT, tc); } /** * Flush tx frame populated by retransmits and timer pops */ void tcp_flush_frame_to_output (vlib_main_t * vm, u8 thread_index, u8 is_ip4) { if (tcp_main.wrk_ctx[thread_index].tx_frames[!is_ip4]) { u32 next_index; next_index = is_ip4 ? tcp4_output_node.index : tcp6_output_node.index; vlib_put_frame_to_node (vm, next_index, tcp_main. wrk_ctx[thread_index].tx_frames[!is_ip4]); tcp_main.wrk_ctx[thread_index].tx_frames[!is_ip4] = 0; } } /** * Flush ip lookup tx frames populated by timer pops */ static void tcp_flush_frame_to_ip_lookup (vlib_main_t * vm, u8 thread_index, u8 is_ip4) { if (tcp_main.wrk_ctx[thread_index].ip_lookup_tx_frames[!is_ip4]) { u32 next_index; next_index = is_ip4 ? ip4_lookup_node.index : ip6_lookup_node.index; vlib_put_frame_to_node (vm, next_index, tcp_main. wrk_ctx[thread_index].ip_lookup_tx_frames [!is_ip4]); tcp_main.wrk_ctx[thread_index].ip_lookup_tx_frames[!is_ip4] = 0; } } /** * Flush v4 and v6 tcp and ip-lookup tx frames for thread index */ void tcp_flush_frames_to_output (u8 thread_index) { vlib_main_t *vm = vlib_get_main (); tcp_flush_frame_to_output (vm, thread_index, 1); tcp_flush_frame_to_output (vm, thread_index, 0); tcp_flush_frame_to_ip_lookup (vm, thread_index, 1); tcp_flush_frame_to_ip_lookup (vm, thread_index, 0); } /** * Send FIN */ void tcp_send_fin (tcp_connection_t * tc) { tcp_main_t *tm = vnet_get_tcp_main (); vlib_main_t *vm = vlib_get_main (); vlib_buffer_t *b; u32 bi; u8 fin_snt = 0; tcp_retransmit_timer_force_update (tc); if (PREDICT_FALSE (tcp_get_free_buffer_index (tm, &bi))) return; b = vlib_get_buffer (vm, bi); tcp_init_buffer (vm, b); fin_snt = tc->flags & TCP_CONN_FINSNT; if (fin_snt) tc->snd_nxt = tc->snd_una; tcp_make_fin (tc, b); tcp_enqueue_to_output_now (vm, b, bi, tc->c_is_ip4); if (!fin_snt) { tc->flags |= TCP_CONN_FINSNT; tc->flags &= ~TCP_CONN_FINPNDG; /* Account for the FIN */ tc->snd_una_max += 1; tc->snd_nxt = tc->snd_una_max; } else { tc->snd_nxt = tc->snd_una_max; } TCP_EVT_DBG (TCP_EVT_FIN_SENT, tc); } always_inline u8 tcp_make_state_flags (tcp_connection_t * tc, tcp_state_t next_state) { switch (next_state) { case TCP_STATE_ESTABLISHED: case TCP_STATE_CLOSE_WAIT: return TCP_FLAG_ACK; case TCP_STATE_SYN_RCVD: return TCP_FLAG_SYN | TCP_FLAG_ACK; case TCP_STATE_SYN_SENT: return TCP_FLAG_SYN; case TCP_STATE_LAST_ACK: case TCP_STATE_FIN_WAIT_1: if (tc->snd_nxt + 1 < tc->snd_una_max) return TCP_FLAG_ACK; else return TCP_FLAG_FIN; default: clib_warning ("Shouldn't be here!"); } return 0; } /** * Push TCP header and update connection variables */ always_inline void tcp_push_hdr_i (tcp_connection_t * tc, vlib_buffer_t * b, tcp_state_t next_state, u8 compute_opts, u8 maybe_burst) { u32 advertise_wnd, data_len; u8 tcp_hdr_opts_len, flags; tcp_main_t *tm = &tcp_main; tcp_header_t *th; data_len = b->current_length; if (PREDICT_FALSE (b->flags & VLIB_BUFFER_NEXT_PRESENT)) data_len += b->total_length_not_including_first_buffer; vnet_buffer (b)->tcp.flags = 0; vnet_buffer (b)->tcp.connection_index = tc->c_c_index; if (compute_opts) tc->snd_opts_len = tcp_make_options (tc, &tc->snd_opts, tc->state); tcp_hdr_opts_len = tc->snd_opts_len + sizeof (tcp_header_t); if (maybe_burst) advertise_wnd = tc->rcv_wnd >> tc->rcv_wscale; else advertise_wnd = tcp_window_to_advertise (tc, next_state); flags = tcp_make_state_flags (tc, next_state); th = vlib_buffer_push_tcp (b, tc->c_lcl_port, tc->c_rmt_port, tc->snd_nxt, tc->rcv_nxt, tcp_hdr_opts_len, flags, advertise_wnd); if (maybe_burst) { clib_memcpy ((u8 *) (th + 1), tm->wrk_ctx[tc->c_thread_index].cached_opts, tc->snd_opts_len); } else { u8 len = tcp_options_write ((u8 *) (th + 1), &tc->snd_opts); ASSERT (len == tc->snd_opts_len); } /* * Update connection variables */ tc->snd_nxt += data_len; tc->rcv_las = tc->rcv_nxt; TCP_EVT_DBG (TCP_EVT_PKTIZE, tc); } u32 tcp_push_header (tcp_connection_t * tc, vlib_buffer_t * b) { tcp_push_hdr_i (tc, b, TCP_STATE_ESTABLISHED, /* compute opts */ 0, /* burst */ 1); tc->snd_una_max = tc->snd_nxt; ASSERT (seq_leq (tc->snd_una_max, tc->snd_una + tc->snd_wnd + tcp_fastrecovery_sent_1_smss (tc) * tc->snd_mss)); tcp_validate_txf_size (tc, tc->snd_una_max - tc->snd_una); /* If not tracking an ACK, start tracking */ if (tc->rtt_ts == 0 && !tcp_in_cong_recovery (tc)) { tc->rtt_ts = tcp_time_now (); tc->rtt_seq = tc->snd_nxt; } if (PREDICT_FALSE (!tcp_timer_is_active (tc, TCP_TIMER_RETRANSMIT))) { tcp_retransmit_timer_set (tc); tc->rto_boff = 0; } tcp_trajectory_add_start (b, 3); return 0; } void tcp_send_ack (tcp_connection_t * tc) { tcp_main_t *tm = vnet_get_tcp_main (); vlib_main_t *vm = vlib_get_main (); vlib_buffer_t *b; u32 bi; /* Get buffer */ if (PREDICT_FALSE (tcp_get_free_buffer_index (tm, &bi))) return; b = vlib_get_buffer (vm, bi); tcp_init_buffer (vm, b); /* Fill in the ACK */ tcp_make_ack (tc, b); tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4); } /** * Delayed ack timer handler * * Sends delayed ACK when timer expires */ void tcp_timer_delack_handler (u32 index) { u32 thread_index = vlib_get_thread_index (); tcp_connection_t *tc; tc = tcp_connection_get (index, thread_index); tc->timers[TCP_TIMER_DELACK] = TCP_TIMER_HANDLE_INVALID; tcp_send_ack (tc); } /** * Build a retransmit segment * * @return the number of bytes in the segment or 0 if there's nothing to * retransmit */ static u32 tcp_prepare_retransmit_segment (tcp_connection_t * tc, u32 offset, u32 max_deq_bytes, vlib_buffer_t ** b) { tcp_main_t *tm = vnet_get_tcp_main (); vlib_main_t *vm = vlib_get_main (); int n_bytes = 0; u32 start, bi, available_bytes, seg_size; u8 *data; ASSERT (tc->state >= TCP_STATE_ESTABLISHED); ASSERT (max_deq_bytes != 0); /* * Make sure we can retransmit something */ available_bytes = session_tx_fifo_max_dequeue (&tc->connection); ASSERT (available_bytes >= offset); available_bytes -= offset; if (!available_bytes) return 0; max_deq_bytes = clib_min (tc->snd_mss, max_deq_bytes); max_deq_bytes = clib_min (available_bytes, max_deq_bytes); /* Start is beyond snd_congestion */ start = tc->snd_una + offset; if (seq_geq (start, tc->snd_congestion)) goto done; /* Don't overshoot snd_congestion */ if (seq_gt (start + max_deq_bytes, tc->snd_congestion)) { max_deq_bytes = tc->snd_congestion - start; if (max_deq_bytes == 0) goto done; } seg_size = max_deq_bytes + MAX_HDRS_LEN; /* * Prepare options */ tc->snd_opts_len = tcp_make_options (tc, &tc->snd_opts, tc->state); /* * Allocate and fill in buffer(s) */ /* Easy case, buffer size greater than mss */ if (PREDICT_TRUE (seg_size <= tm->bytes_per_buffer)) { if (PREDICT_FALSE (tcp_get_free_buffer_index (tm, &bi))) return 0; *b = vlib_get_buffer (vm, bi); data = tcp_init_buffer (vm, *b); n_bytes = stream_session_peek_bytes (&tc->connection, data, offset, max_deq_bytes); ASSERT (n_bytes == max_deq_bytes); b[0]->current_length = n_bytes; tcp_push_hdr_i (tc, *b, tc->state, /* compute opts */ 0, /* burst */ 0); if (seq_gt (tc->snd_nxt, tc->snd_una_max)) tc->snd_una_max = tc->snd_nxt; } /* Split mss into multiple buffers */ else { u32 chain_bi = ~0, n_bufs_per_seg; u32 thread_index = vlib_get_thread_index (); u16 n_peeked, len_to_deq, available_bufs; vlib_buffer_t *chain_b, *prev_b; int i; /* Make sure we have enough buffers */ n_bufs_per_seg = ceil ((double) seg_size / tm->bytes_per_buffer); available_bufs = vec_len (tm->wrk_ctx[thread_index].tx_buffers); if (n_bufs_per_seg > available_bufs) { tcp_alloc_tx_buffers (tm, thread_index, &available_bufs, VLIB_FRAME_SIZE); if (n_bufs_per_seg > available_bufs) { *b = 0; return 0; } } tcp_get_free_buffer_index (tm, &bi); ASSERT (bi != (u32) ~ 0); *b = vlib_get_buffer (vm, bi); data = tcp_init_buffer (vm, *b); n_bytes = stream_session_peek_bytes (&tc->connection, data, offset, tm->bytes_per_buffer - MAX_HDRS_LEN); b[0]->current_length = n_bytes; b[0]->flags |= VLIB_BUFFER_TOTAL_LENGTH_VALID; b[0]->total_length_not_including_first_buffer = 0; max_deq_bytes -= n_bytes; chain_b = *b; for (i = 1; i < n_bufs_per_seg; i++) { prev_b = chain_b; len_to_deq = clib_min (max_deq_bytes, tm->bytes_per_buffer); tcp_get_free_buffer_index (tm, &chain_bi); ASSERT (chain_bi != (u32) ~ 0); chain_b = vlib_get_buffer (vm, chain_bi); chain_b->current_data = 0; data = vlib_buffer_get_current (chain_b); n_peeked = stream_session_peek_bytes (&tc->connection, data, offset + n_bytes, len_to_deq); ASSERT (n_peeked == len_to_deq); n_bytes += n_peeked; chain_b->current_length = n_peeked; chain_b->next_buffer = 0; /* update previous buffer */ prev_b->next_buffer = chain_bi; prev_b->flags |= VLIB_BUFFER_NEXT_PRESENT; max_deq_bytes -= n_peeked; b[0]->total_length_not_including_first_buffer += n_peeked; } tcp_push_hdr_i (tc, *b, tc->state, /* compute opts */ 0, /* burst */ 0); if (seq_gt (tc->snd_nxt, tc->snd_una_max)) tc->snd_una_max = tc->snd_nxt; } ASSERT (n_bytes > 0); ASSERT (((*b)->current_data + (*b)->current_length) <= tm->bytes_per_buffer); if (tcp_in_fastrecovery (tc)) tc->snd_rxt_bytes += n_bytes; done: TCP_EVT_DBG (TCP_EVT_CC_RTX, tc, offset, n_bytes); return n_bytes; } /** * Reset congestion control, switch cwnd to loss window and try again. */ static void tcp_rxt_timeout_cc (tcp_connection_t * tc) { TCP_EVT_DBG (TCP_EVT_CC_EVT, tc, 6); tc->prev_ssthresh = tc->ssthresh; tc->prev_cwnd = tc->cwnd; /* Cleanly recover cc (also clears up fast retransmit) */ if (tcp_in_fastrecovery (tc)) tcp_cc_fastrecovery_exit (tc); /* Start again from the beginning */ tc->cc_algo->congestion (tc); tc->cwnd = tcp_loss_wnd (tc); tc->snd_congestion = tc->snd_una_max; tc->rtt_ts = 0; tc->cwnd_acc_bytes = 0; tcp_recovery_on (tc); } static inline void tcp_timer_retransmit_handler_i (u32 index, u8 is_syn) { tcp_main_t *tm = vnet_get_tcp_main (); vlib_main_t *vm = vlib_get_main (); u32 thread_index = vlib_get_thread_index (); tcp_connection_t *tc; vlib_buffer_t *b = 0; u32 bi, n_bytes; if (is_syn) { tc = tcp_half_open_connection_get (index); /* Note: the connection may have transitioned to ESTABLISHED... */ if (PREDICT_FALSE (tc == 0)) return; tc->timers[TCP_TIMER_RETRANSMIT_SYN] = TCP_TIMER_HANDLE_INVALID; } else { tc = tcp_connection_get (index, thread_index); /* Note: the connection may have been closed and pool_put */ if (PREDICT_FALSE (tc == 0)) return; tc->timers[TCP_TIMER_RETRANSMIT] = TCP_TIMER_HANDLE_INVALID; } TCP_EVT_DBG (TCP_EVT_CC_EVT, tc, 1); if (tc->state >= TCP_STATE_ESTABLISHED) { /* Lost FIN, retransmit and return */ if (tcp_is_lost_fin (tc)) { tcp_send_fin (tc); tc->rto_boff += 1; tc->rto = clib_min (tc->rto << 1, TCP_RTO_MAX); return; } /* Shouldn't be here. This condition is tricky because it has to take * into account boff > 0 due to persist timeout. */ if ((tc->rto_boff == 0 && tc->snd_una == tc->snd_una_max) || (tc->rto_boff > 0 && seq_geq (tc->snd_una, tc->snd_congestion) && !tcp_flight_size (tc))) { ASSERT (!tcp_in_recovery (tc)); tc->rto_boff = 0; return; } /* We're not in recovery so make sure rto_boff is 0. Can be non 0 due * to persist timer timeout */ if (!tcp_in_recovery (tc) && tc->rto_boff > 0) { tc->rto_boff = 0; tcp_update_rto (tc); } /* Increment RTO backoff (also equal to number of retries) and go back * to first un-acked byte */ tc->rto_boff += 1; /* First retransmit timeout */ if (tc->rto_boff == 1) tcp_rxt_timeout_cc (tc); /* If we've sent beyond snd_congestion, update it */ if (seq_gt (tc->snd_una_max, tc->snd_congestion)) tc->snd_congestion = tc->snd_una_max; tc->snd_una_max = tc->snd_nxt = tc->snd_una; tc->rto = clib_min (tc->rto << 1, TCP_RTO_MAX); /* Send one segment. Note that n_bytes may be zero due to buffer * shortfall */ n_bytes = tcp_prepare_retransmit_segment (tc, 0, tc->snd_mss, &b); /* TODO be less aggressive about this */ scoreboard_clear (&tc->sack_sb); if (n_bytes == 0) { tcp_retransmit_timer_force_update (tc); return; } bi = vlib_get_buffer_index (vm, b); /* For first retransmit, record timestamp (Eifel detection RFC3522) */ if (tc->rto_boff == 1) tc->snd_rxt_ts = tcp_time_now (); tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4); tcp_retransmit_timer_force_update (tc); } /* Retransmit for SYN */ else if (tc->state == TCP_STATE_SYN_SENT) { /* Half-open connection actually moved to established but we were * waiting for syn retransmit to pop to call cleanup from the right * thread. */ if (tc->flags & TCP_CONN_HALF_OPEN_DONE) { if (tcp_half_open_connection_cleanup (tc)) { clib_warning ("could not remove half-open connection"); ASSERT (0); } return; } /* Try without increasing RTO a number of times. If this fails, * start growing RTO exponentially */ tc->rto_boff += 1; if (tc->rto_boff > TCP_RTO_SYN_RETRIES) tc->rto = clib_min (tc->rto << 1, TCP_RTO_MAX); tcp_timer_update (tc, TCP_TIMER_RETRANSMIT_SYN, tc->rto * TCP_TO_TIMER_TICK); if (PREDICT_FALSE (tcp_get_free_buffer_index (tm, &bi))) return; b = vlib_get_buffer (vm, bi); tcp_init_buffer (vm, b); tcp_make_syn (tc, b); tc->rtt_ts = 0; TCP_EVT_DBG (TCP_EVT_SYN_RXT, tc, 0); /* This goes straight to ipx_lookup. Retransmit timer set already */ tcp_push_ip_hdr (tm, tc, b); tcp_enqueue_to_ip_lookup (vm, b, bi, tc->c_is_ip4, tc->c_fib_index); } /* Retransmit SYN-ACK */ else if (tc->state == TCP_STATE_SYN_RCVD) { tc->rto_boff += 1; if (tc->rto_boff > TCP_RTO_SYN_RETRIES) tc->rto = clib_min (tc->rto << 1, TCP_RTO_MAX); tc->rtt_ts = 0; if (PREDICT_FALSE (tcp_get_free_buffer_index (tm, &bi))) { tcp_retransmit_timer_force_update (tc); return; } b = vlib_get_buffer (vm, bi); tcp_init_buffer (vm, b); tcp_make_synack (tc, b); TCP_EVT_DBG (TCP_EVT_SYN_RXT, tc, 1); /* Retransmit timer already updated, just enqueue to output */ tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4); } else { ASSERT (tc->state == TCP_STATE_CLOSED); return; } } void tcp_timer_retransmit_handler (u32 index) { tcp_timer_retransmit_handler_i (index, 0); } void tcp_timer_retransmit_syn_handler (u32 index) { tcp_timer_retransmit_handler_i (index, 1); } /** * Got 0 snd_wnd from peer, try to do something about it. * */ void tcp_timer_persist_handler (u32 index) { tcp_main_t *tm = vnet_get_tcp_main (); vlib_main_t *vm = vlib_get_main (); u32 thread_index = vlib_get_thread_index (); tcp_connection_t *tc; vlib_buffer_t *b; u32 bi, max_snd_bytes, available_bytes, offset; int n_bytes = 0; u8 *data; tc = tcp_connection_get_if_valid (index, thread_index); if (!tc) return; /* Make sure timer handle is set to invalid */ tc->timers[TCP_TIMER_PERSIST] = TCP_TIMER_HANDLE_INVALID; /* Problem already solved or worse */ if (tc->state == TCP_STATE_CLOSED || tc->state > TCP_STATE_ESTABLISHED || tc->snd_wnd > tc->snd_mss || tcp_in_recovery (tc)) return; available_bytes = session_tx_fifo_max_dequeue (&tc->connection); offset = tc->snd_una_max - tc->snd_una; /* Reprogram persist if no new bytes available to send. We may have data * next time */ if (!available_bytes) { tcp_persist_timer_set (tc); return; } if (available_bytes <= offset) { ASSERT (tcp_timer_is_active (tc, TCP_TIMER_RETRANSMIT)); return; } /* Increment RTO backoff */ tc->rto_boff += 1; tc->rto = clib_min (tc->rto << 1, TCP_RTO_MAX); /* * Try to force the first unsent segment (or buffer) */ if (PREDICT_FALSE (tcp_get_free_buffer_index (tm, &bi))) { tcp_persist_timer_set (tc); return; } b = vlib_get_buffer (vm, bi); data = tcp_init_buffer (vm, b); tcp_validate_txf_size (tc, offset); tc->snd_opts_len = tcp_make_options (tc, &tc->snd_opts, tc->state); max_snd_bytes = clib_min (tc->snd_mss, tm->bytes_per_buffer - MAX_HDRS_LEN); n_bytes = stream_session_peek_bytes (&tc->connection, data, offset, max_snd_bytes); b->current_length = n_bytes; ASSERT (n_bytes != 0 && (tcp_timer_is_active (tc, TCP_TIMER_RETRANSMIT) || tc->snd_nxt == tc->snd_una_max || tc->rto_boff > 1)); tcp_push_hdr_i (tc, b, tc->state, /* compute opts */ 0, /* burst */ 0); tc->snd_una_max = tc->snd_nxt; tcp_validate_txf_size (tc, tc->snd_una_max - tc->snd_una); tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4); /* Just sent new data, enable retransmit */ tcp_retransmit_timer_update (tc); } /** * Retransmit first unacked segment */ void tcp_retransmit_first_unacked (tcp_connection_t * tc) { vlib_main_t *vm = vlib_get_main (); vlib_buffer_t *b; u32 bi, old_snd_nxt, n_bytes; old_snd_nxt = tc->snd_nxt; tc->snd_nxt = tc->snd_una; TCP_EVT_DBG (TCP_EVT_CC_EVT, tc, 2); n_bytes = tcp_prepare_retransmit_segment (tc, 0, tc->snd_mss, &b); if (!n_bytes) return; bi = vlib_get_buffer_index (vm, b); tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4); tc->snd_nxt = old_snd_nxt; } /** * Do fast retransmit with SACKs */ void tcp_fast_retransmit_sack (tcp_connection_t * tc) { vlib_main_t *vm = vlib_get_main (); u32 n_written = 0, offset, max_bytes, n_segs = 0; vlib_buffer_t *b = 0; sack_scoreboard_hole_t *hole; sack_scoreboard_t *sb; u32 bi, old_snd_nxt; int snd_space; u8 snd_limited = 0, can_rescue = 0; ASSERT (tcp_in_fastrecovery (tc)); old_snd_nxt = tc->snd_nxt; sb = &tc->sack_sb; snd_space = tcp_available_cc_snd_space (tc); if (snd_space < tc->snd_mss) goto done; TCP_EVT_DBG (TCP_EVT_CC_EVT, tc, 0); hole = scoreboard_get_hole (sb, sb->cur_rxt_hole); while (hole && snd_space > 0 && n_segs++ < VLIB_FRAME_SIZE) { hole = scoreboard_next_rxt_hole (sb, hole, tcp_fastrecovery_sent_1_smss (tc), &can_rescue, &snd_limited); if (!hole) { if (!can_rescue || !(seq_lt (sb->rescue_rxt, tc->snd_una) || seq_gt (sb->rescue_rxt, tc->snd_congestion))) break; /* If rescue rxt undefined or less than snd_una then one segment of * up to SMSS octets that MUST include the highest outstanding * unSACKed sequence number SHOULD be returned, and RescueRxt set to * RecoveryPoint. HighRxt MUST NOT be updated. */ max_bytes = clib_min (tc->snd_mss, tc->snd_congestion - tc->snd_una); max_bytes = clib_min (max_bytes, snd_space); offset = tc->snd_congestion - tc->snd_una - max_bytes; sb->rescue_rxt = tc->snd_congestion; tc->snd_nxt = tc->snd_una + offset; n_written = tcp_prepare_retransmit_segment (tc, offset, max_bytes, &b); if (!n_written) goto done; bi = vlib_get_buffer_index (vm, b); tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4); break; } max_bytes = clib_min (hole->end - sb->high_rxt, snd_space); max_bytes = snd_limited ? clib_min (max_bytes, tc->snd_mss) : max_bytes; if (max_bytes == 0) break; offset = sb->high_rxt - tc->snd_una; tc->snd_nxt = sb->high_rxt; n_written = tcp_prepare_retransmit_segment (tc, offset, max_bytes, &b); /* Nothing left to retransmit */ if (n_written == 0) break; bi = vlib_get_buffer_index (vm, b); sb->high_rxt += n_written; tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4); ASSERT (n_written <= snd_space); snd_space -= n_written; } done: /* If window allows, send 1 SMSS of new data */ tc->snd_nxt = old_snd_nxt; } /** * Fast retransmit without SACK info */ void tcp_fast_retransmit_no_sack (tcp_connection_t * tc) { vlib_main_t *vm = vlib_get_main (); u32 n_written = 0, offset = 0, bi, old_snd_nxt; int snd_space; vlib_buffer_t *b; ASSERT (tcp_in_fastrecovery (tc)); TCP_EVT_DBG (TCP_EVT_CC_EVT, tc, 0); /* Start resending from first un-acked segment */ old_snd_nxt = tc->snd_nxt; tc->snd_nxt = tc->snd_una; snd_space = tcp_available_cc_snd_space (tc); while (snd_space > 0) { offset += n_written; n_written = tcp_prepare_retransmit_segment (tc, offset, snd_space, &b); /* Nothing left to retransmit */ if (n_written == 0) break; bi = vlib_get_buffer_index (vm, b); tcp_enqueue_to_output (vm, b, bi, tc->c_is_ip4); snd_space -= n_written; } /* Restore snd_nxt. If window allows, send 1 SMSS of new data */ tc->snd_nxt = old_snd_nxt; } /** * Do fast retransmit */ void tcp_fast_retransmit (tcp_connection_t * tc) { if (tcp_opts_sack_permitted (&tc->rcv_opts)) tcp_fast_retransmit_sack (tc); else tcp_fast_retransmit_no_sack (tc); } static u32 tcp_session_has_ooo_data (tcp_connection_t * tc) { stream_session_t *s = session_get (tc->c_s_index, tc->c_thread_index); return svm_fifo_has_ooo_data (s->server_rx_fifo); } static void tcp_output_handle_link_local (tcp_connection_t * tc0, vlib_buffer_t * b0, u16 * next0, u32 * error0) { ip_adjacency_t *adj; adj_index_t ai; /* Not thread safe but as long as the connection exists the adj should * not be removed */ ai = adj_nbr_find (FIB_PROTOCOL_IP6, VNET_LINK_IP6, &tc0->c_rmt_ip, tc0->sw_if_index); if (ai == ADJ_INDEX_INVALID) { vnet_buffer (b0)->sw_if_index[VLIB_TX] = ~0; *next0 = TCP_OUTPUT_NEXT_DROP; *error0 = TCP_ERROR_LINK_LOCAL_RW; return; } adj = adj_get (ai); if (PREDICT_TRUE (adj->lookup_next_index == IP_LOOKUP_NEXT_REWRITE)) *next0 = TCP_OUTPUT_NEXT_IP_REWRITE; else if (adj->lookup_next_index == IP_LOOKUP_NEXT_ARP) *next0 = TCP_OUTPUT_NEXT_IP_ARP; else { *next0 = TCP_OUTPUT_NEXT_DROP; *error0 = TCP_ERROR_LINK_LOCAL_RW; } vnet_buffer (b0)->ip.adj_index[VLIB_TX] = ai; } static void tcp46_output_trace_frame (vlib_main_t * vm, vlib_node_runtime_t * node, u32 * to_next, u32 n_bufs) { u32 n_trace = vlib_get_trace_count (vm, node); tcp_connection_t *tc; tcp_tx_trace_t *t; vlib_buffer_t *b; tcp_header_t *th; int i; for (i = 0; i < clib_min (n_trace, n_bufs); i++) { b = vlib_get_buffer (vm, to_next[i]); th = vlib_buffer_get_current (b); tc = tcp_connection_get (vnet_buffer (b)->tcp.connection_index, vm->thread_index); t = vlib_add_trace (vm, node, b, sizeof (*t)); clib_memcpy (&t->tcp_header, th, sizeof (t->tcp_header)); clib_memcpy (&t->tcp_connection, tc, sizeof (t->tcp_connection)); } } always_inline void tcp_output_push_ip (vlib_main_t * vm, vlib_buffer_t * b0, tcp_connection_t * tc0, u8 is_ip4) { tcp_header_t *th0 = 0; th0 = vlib_buffer_get_current (b0); TCP_EVT_DBG (TCP_EVT_OUTPUT, tc0, th0->flags, b0->current_length); if (is_ip4) { vlib_buffer_push_ip4 (vm, b0, &tc0->c_lcl_ip4, &tc0->c_rmt_ip4, IP_PROTOCOL_TCP, 1); b0->flags |= VNET_BUFFER_F_OFFLOAD_TCP_CKSUM; vnet_buffer (b0)->l4_hdr_offset = (u8 *) th0 - b0->data; th0->checksum = 0; } else { ip6_header_t *ih0; ih0 = vlib_buffer_push_ip6 (vm, b0, &tc0->c_lcl_ip6, &tc0->c_rmt_ip6, IP_PROTOCOL_TCP); b0->flags |= VNET_BUFFER_F_OFFLOAD_TCP_CKSUM; vnet_buffer (b0)->l3_hdr_offset = (u8 *) ih0 - b0->data; vnet_buffer (b0)->l4_hdr_offset = (u8 *) th0 - b0->data; th0->checksum = 0; } } always_inline void tcp_output_handle_packet (tcp_connection_t * tc0, vlib_buffer_t * b0, u32 * error0, u16 * next0, u8 is_ip4) { if (PREDICT_FALSE (tc0->state == TCP_STATE_CLOSED)) { *error0 = TCP_ERROR_INVALID_CONNECTION; *next0 = TCP_OUTPUT_NEXT_DROP; return; } vnet_buffer (b0)->sw_if_index[VLIB_TX] = tc0->c_fib_index; vnet_buffer (b0)->sw_if_index[VLIB_RX] = 0; if (!is_ip4) { if (PREDICT_FALSE (ip6_address_is_link_local_unicast (&tc0->c_rmt_ip6))) tcp_output_handle_link_local (tc0, b0, next0, error0); } /* Filter out DUPACKs if there are no OOO segments left */ if (PREDICT_FALSE (vnet_buffer (b0)->tcp.flags & TCP_BUF_FLAG_DUPACK)) { /* N.B. Should not filter burst of dupacks. Two issues: * 1) dupacks open cwnd on remote peer when congested * 2) acks leaving should have the latest rcv_wnd since the * burst may have eaten up all of it, so only the old ones * could be filtered. */ if (!tcp_session_has_ooo_data (tc0)) { *error0 = TCP_ERROR_FILTERED_DUPACKS; *next0 = TCP_OUTPUT_NEXT_DROP; return; } } /* Stop DELACK timer and fix flags */ tc0->flags &= ~(TCP_CONN_SNDACK); if (!TCP_ALWAYS_ACK) tcp_timer_reset (tc0, TCP_TIMER_DELACK); } always_inline uword tcp46_output_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * frame, int is_ip4) { u32 n_left_from, *from, thread_index = vm->thread_index; vlib_buffer_t *bufs[VLIB_FRAME_SIZE], **b; u16 nexts[VLIB_FRAME_SIZE], *next; from = vlib_frame_vector_args (frame); n_left_from = frame->n_vectors; tcp_set_time_now (thread_index); if (PREDICT_FALSE (node->flags & VLIB_NODE_FLAG_TRACE)) tcp46_output_trace_frame (vm, node, from, n_left_from); vlib_get_buffers (vm, from, bufs, n_left_from); b = bufs; next = nexts; while (n_left_from >= 4) { u32 error0 = TCP_ERROR_PKTS_SENT, error1 = TCP_ERROR_PKTS_SENT; tcp_connection_t *tc0, *tc1; { vlib_prefetch_buffer_header (b[2], STORE); CLIB_PREFETCH (b[2]->data, 2 * CLIB_CACHE_LINE_BYTES, STORE); vlib_prefetch_buffer_header (b[3], STORE); CLIB_PREFETCH (b[3]->data, 2 * CLIB_CACHE_LINE_BYTES, STORE); } next[0] = next[1] = TCP_OUTPUT_NEXT_IP_LOOKUP; tc0 = tcp_connection_get (vnet_buffer (b[0])->tcp.connection_index, thread_index); tc1 = tcp_connection_get (vnet_buffer (b[1])->tcp.connection_index, thread_index); tcp_output_push_ip (vm, b[0], tc0, is_ip4); tcp_output_push_ip (vm, b[1], tc1, is_ip4); tcp_output_handle_packet (tc0, b[0], &error0, &next[0], is_ip4); tcp_output_handle_packet (tc1, b[1], &error1, &next[1], is_ip4); b += 2; next += 2; n_left_from -= 2; } while (n_left_from > 0) { u32 error0 = TCP_ERROR_PKTS_SENT; tcp_connection_t *tc0; if (n_left_from > 1) { vlib_prefetch_buffer_header (b[1], STORE); CLIB_PREFETCH (b[1]->data, 2 * CLIB_CACHE_LINE_BYTES, STORE); } next[0] = TCP_OUTPUT_NEXT_IP_LOOKUP; tc0 = tcp_connection_get (vnet_buffer (b[0])->tcp.connection_index, thread_index); tcp_output_push_ip (vm, b[0], tc0, is_ip4); tcp_output_handle_packet (tc0, b[0], &error0, &next[0], is_ip4); b += 1; next += 1; n_left_from -= 1; } vlib_buffer_enqueue_to_next (vm, node, from, nexts, frame->n_vectors); return frame->n_vectors; } static uword tcp4_output (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame) { return tcp46_output_inline (vm, node, from_frame, 1 /* is_ip4 */ ); } static uword tcp6_output (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame) { return tcp46_output_inline (vm, node, from_frame, 0 /* is_ip4 */ ); } /* *INDENT-OFF* */ VLIB_REGISTER_NODE (tcp4_output_node) = { .function = tcp4_output,.name = "tcp4-output", /* Takes a vector of packets. */ .vector_size = sizeof (u32), .n_errors = TCP_N_ERROR, .error_strings = tcp_error_strings, .n_next_nodes = TCP_OUTPUT_N_NEXT, .next_nodes = { #define _(s,n) [TCP_OUTPUT_NEXT_##s] = n, foreach_tcp4_output_next #undef _ }, .format_buffer = format_tcp_header, .format_trace = format_tcp_tx_trace, }; /* *INDENT-ON* */ VLIB_NODE_FUNCTION_MULTIARCH (tcp4_output_node, tcp4_output); /* *INDENT-OFF* */ VLIB_REGISTER_NODE (tcp6_output_node) = { .function = tcp6_output, .name = "tcp6-output", /* Takes a vector of packets. */ .vector_size = sizeof (u32), .n_errors = TCP_N_ERROR, .error_strings = tcp_error_strings, .n_next_nodes = TCP_OUTPUT_N_NEXT, .next_nodes = { #define _(s,n) [TCP_OUTPUT_NEXT_##s] = n, foreach_tcp6_output_next #undef _ }, .format_buffer = format_tcp_header, .format_trace = format_tcp_tx_trace, }; /* *INDENT-ON* */ VLIB_NODE_FUNCTION_MULTIARCH (tcp6_output_node, tcp6_output); typedef enum _tcp_reset_next { TCP_RESET_NEXT_DROP, TCP_RESET_NEXT_IP_LOOKUP, TCP_RESET_N_NEXT } tcp_reset_next_t; #define foreach_tcp4_reset_next \ _(DROP, "error-drop") \ _(IP_LOOKUP, "ip4-lookup") #define foreach_tcp6_reset_next \ _(DROP, "error-drop") \ _(IP_LOOKUP, "ip6-lookup") static uword tcp46_send_reset_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame, u8 is_ip4) { u32 n_left_from, next_index, *from, *to_next; u32 my_thread_index = vm->thread_index; from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; next_index = node->cached_next_index; while (n_left_from > 0) { u32 n_left_to_next; vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); while (n_left_from > 0 && n_left_to_next > 0) { u32 bi0; vlib_buffer_t *b0; tcp_tx_trace_t *t0; tcp_header_t *th0; u32 error0 = TCP_ERROR_RST_SENT, next0 = TCP_RESET_NEXT_IP_LOOKUP; bi0 = from[0]; to_next[0] = bi0; from += 1; to_next += 1; n_left_from -= 1; n_left_to_next -= 1; b0 = vlib_get_buffer (vm, bi0); if (tcp_make_reset_in_place (vm, b0, vnet_buffer (b0)->tcp.flags, my_thread_index, is_ip4)) { error0 = TCP_ERROR_LOOKUP_DROPS; next0 = TCP_RESET_NEXT_DROP; goto done; } /* Prepare to send to IP lookup */ vnet_buffer (b0)->sw_if_index[VLIB_TX] = ~0; next0 = TCP_RESET_NEXT_IP_LOOKUP; done: b0->error = node->errors[error0]; b0->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED; if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) { th0 = vlib_buffer_get_current (b0); if (is_ip4) th0 = ip4_next_header ((ip4_header_t *) th0); else th0 = ip6_next_header ((ip6_header_t *) th0); t0 = vlib_add_trace (vm, node, b0, sizeof (*t0)); clib_memcpy (&t0->tcp_header, th0, sizeof (t0->tcp_header)); } vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, n_left_to_next, bi0, next0); } vlib_put_next_frame (vm, node, next_index, n_left_to_next); } return from_frame->n_vectors; } static uword tcp4_send_reset (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame) { return tcp46_send_reset_inline (vm, node, from_frame, 1); } static uword tcp6_send_reset (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame) { return tcp46_send_reset_inline (vm, node, from_frame, 0); } /* *INDENT-OFF* */ VLIB_REGISTER_NODE (tcp4_reset_node) = { .function = tcp4_send_reset, .name = "tcp4-reset", .vector_size = sizeof (u32), .n_errors = TCP_N_ERROR, .error_strings = tcp_error_strings, .n_next_nodes = TCP_RESET_N_NEXT, .next_nodes = { #define _(s,n) [TCP_RESET_NEXT_##s] = n, foreach_tcp4_reset_next #undef _ }, .format_trace = format_tcp_tx_trace, }; /* *INDENT-ON* */ VLIB_NODE_FUNCTION_MULTIARCH (tcp4_reset_node, tcp4_send_reset); /* *INDENT-OFF* */ VLIB_REGISTER_NODE (tcp6_reset_node) = { .function = tcp6_send_reset, .name = "tcp6-reset", .vector_size = sizeof (u32), .n_errors = TCP_N_ERROR, .error_strings = tcp_error_strings, .n_next_nodes = TCP_RESET_N_NEXT, .next_nodes = { #define _(s,n) [TCP_RESET_NEXT_##s] = n, foreach_tcp6_reset_next #undef _ }, .format_trace = format_tcp_tx_trace, }; /* *INDENT-ON* */ VLIB_NODE_FUNCTION_MULTIARCH (tcp6_reset_node, tcp6_send_reset); /* * fd.io coding-style-patch-verification: ON * * Local Variables: * eval: (c-set-style "gnu") * End: */