summaryrefslogtreecommitdiffstats
path: root/src/vnet/session/application_local.c
blob: 88536dc27c696f193cf47a1fa6ae8b5b088a103b (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
/*
 * Copyright (c) 2019 Cisco and/or its affiliates.
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at:
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

#include <vnet/session/application_local.h>
#include <vnet/session/session.h>

typedef enum ct_segment_flags_
{
  CT_SEGMENT_F_CLIENT_DETACHED = 1 << 0,
  CT_SEGMENT_F_SERVER_DETACHED = 1 << 1,
} ct_segment_flags_t;

typedef struct ct_segment_
{
  u32 client_n_sessions;
  u32 server_n_sessions;
  u32 seg_ctx_index;
  u32 ct_seg_index;
  u32 segment_index;
  ct_segment_flags_t flags;
} ct_segment_t;

typedef struct ct_segments_
{
  u32 sm_index;
  u32 server_wrk;
  u32 client_wrk;
  u32 fifo_pair_bytes;
  ct_segment_t *segments;
} ct_segments_ctx_t;

typedef struct ct_cleanup_req_
{
  u32 ct_index;
} ct_cleanup_req_t;

typedef struct ct_main_
{
  ct_connection_t **connections;	/**< Per-worker connection pools */
  ct_cleanup_req_t **pending_cleanups;
  u8 *heave_cleanups;
  u32 n_workers;			/**< Number of vpp workers */
  u32 n_sessions;			/**< Cumulative sessions counter */
  u32 *ho_reusable;			/**< Vector of reusable ho indices */
  clib_spinlock_t ho_reuseable_lock;	/**< Lock for reusable ho indices */
  clib_rwlock_t app_segs_lock;		/**< RW lock for seg contexts */
  uword *app_segs_ctxs_table;		/**< App handle to segment pool map */
  ct_segments_ctx_t *app_seg_ctxs;	/**< Pool of ct segment contexts */
} ct_main_t;

static ct_main_t ct_main;

static ct_connection_t *
ct_connection_alloc (u32 thread_index)
{
  ct_connection_t *ct;

  pool_get_zero (ct_main.connections[thread_index], ct);
  ct->c_c_index = ct - ct_main.connections[thread_index];
  ct->c_thread_index = thread_index;
  ct->client_wrk = ~0;
  ct->server_wrk = ~0;
  ct->seg_ctx_index = ~0;
  ct->ct_seg_index = ~0;
  return ct;
}

static ct_connection_t *
ct_connection_get (u32 ct_index, u32 thread_index)
{
  if (pool_is_free_index (ct_main.connections[thread_index], ct_index))
    return 0;
  return pool_elt_at_index (ct_main.connections[thread_index], ct_index);
}

static void
ct_connection_free (ct_connection_t * ct)
{
  if (CLIB_DEBUG)
    {
      u32 thread_index = ct->c_thread_index;
      memset (ct, 0xfc, sizeof (*ct));
      pool_put (ct_main.connections[thread_index], ct);
      return;
    }
  pool_put (ct_main.connections[ct->c_thread_index], ct);
}

static ct_connection_t *
ct_half_open_alloc (void)
{
  ct_main_t *cm = &ct_main;
  u32 *hip;

  clib_spinlock_lock (&cm->ho_reuseable_lock);
  vec_foreach (hip, cm->ho_reusable)
    pool_put_index (cm->connections[0], *hip);
  vec_reset_length (cm->ho_reusable);
  clib_spinlock_unlock (&cm->ho_reuseable_lock);

  return ct_connection_alloc (0);
}

void
ct_half_open_add_reusable (u32 ho_index)
{
  ct_main_t *cm = &ct_main;

  clib_spinlock_lock (&cm->ho_reuseable_lock);
  vec_add1 (cm->ho_reusable, ho_index);
  clib_spinlock_unlock (&cm->ho_reuseable_lock);
}

session_t *
ct_session_get_peer (session_t * s)
{
  ct_connection_t *ct, *peer_ct;
  ct = ct_connection_get (s->connection_index, s->thread_index);
  peer_ct = ct_connection_get (ct->peer_index, s->thread_index);
  return session_get (peer_ct->c_s_index, s->thread_index);
}

void
ct_session_endpoint (session_t * ll, session_endpoint_t * sep)
{
  ct_connection_t *ct;
  ct = (ct_connection_t *) session_get_transport (ll);
  sep->transport_proto = ct->actual_tp;
  sep->port = ct->c_lcl_port;
  sep->is_ip4 = ct->c_is_ip4;
  ip_copy (&sep->ip, &ct->c_lcl_ip, ct->c_is_ip4);
}

static void
ct_session_dealloc_fifos (ct_connection_t *ct, svm_fifo_t *rx_fifo,
			  svm_fifo_t *tx_fifo)
{
  ct_segments_ctx_t *seg_ctx;
  ct_main_t *cm = &ct_main;
  segment_manager_t *sm;
  app_worker_t *app_wrk;
  ct_segment_t *ct_seg;
  fifo_segment_t *fs;
  u8 del_segment = 0;
  u32 seg_index;
  int cnt;

  /*
   * Cleanup fifos
   */

  sm = segment_manager_get (rx_fifo->segment_manager);
  seg_index = rx_fifo->segment_index;

  fs = segment_manager_get_segment_w_lock (sm, seg_index);
  fifo_segment_free_fifo (fs, rx_fifo);
  fifo_segment_free_fifo (fs, tx_fifo);
  segment_manager_segment_reader_unlock (sm);

  /*
   * Atomically update segment context with readers lock
   */

  clib_rwlock_reader_lock (&cm->app_segs_lock);

  seg_ctx = pool_elt_at_index (cm->app_seg_ctxs, ct->seg_ctx_index);
  ct_seg = pool_elt_at_index (seg_ctx->segments, ct->ct_seg_index);

  if (ct->flags & CT_CONN_F_CLIENT)
    {
      cnt =
	__atomic_sub_fetch (&ct_seg->client_n_sessions, 1, __ATOMIC_RELAXED);
    }
  else
    {
      cnt =
	__atomic_sub_fetch (&ct_seg->server_n_sessions, 1, __ATOMIC_RELAXED);
    }

  clib_rwlock_reader_unlock (&cm->app_segs_lock);

  /*
   * No need to do any app updates, return
   */
  ASSERT (cnt >= 0);
  if (cnt)
    return;

  /*
   * Grab exclusive lock and update flags unless some other thread
   * added more sessions
   */
  clib_rwlock_writer_lock (&cm->app_segs_lock);

  seg_ctx = pool_elt_at_index (cm->app_seg_ctxs, ct->seg_ctx_index);
  ct_seg = pool_elt_at_index (seg_ctx->segments, ct->ct_seg_index);
  if (ct->flags & CT_CONN_F_CLIENT)
    {
      cnt = ct_seg->client_n_sessions;
      if (!cnt)
	ct_seg->flags |= CT_SEGMENT_F_CLIENT_DETACHED;
    }
  else
    {
      cnt = ct_seg->server_n_sessions;
      if (!cnt)
	ct_seg->flags |= CT_SEGMENT_F_SERVER_DETACHED;
    }

  /*
   * Remove segment context because both client and server detached
   */

  if (!cnt && (ct_seg->flags & CT_SEGMENT_F_CLIENT_DETACHED) &&
      (ct_seg->flags & CT_SEGMENT_F_SERVER_DETACHED))
    {
      pool_put_index (seg_ctx->segments, ct->ct_seg_index);

      /*
       * No more segment indices left, remove the segments context
       */
      if (!pool_elts (seg_ctx->segments))
	{
	  u64 table_handle = seg_ctx->client_wrk << 16 | seg_ctx->server_wrk;
	  table_handle = (u64) seg_ctx->sm_index << 32 | table_handle;
	  hash_unset (cm->app_segs_ctxs_table, table_handle);
	  pool_free (seg_ctx->segments);
	  pool_put_index (cm->app_seg_ctxs, ct->seg_ctx_index);
	}
      del_segment = 1;
    }

  clib_rwlock_writer_unlock (&cm->app_segs_lock);

  /*
   * Session counter went to zero, notify the app that detached
   */
  if (cnt)
    return;

  if (ct->flags & CT_CONN_F_CLIENT)
    {
      app_wrk = app_worker_get_if_valid (ct->client_wrk);
      /* Determine if client app still needs notification, i.e., if it is
       * still attached. If client detached and this is the last ct session
       * on this segment, then its connects segment manager should also be
       * detached, so do not send notification */
      if (app_wrk)
	{
	  segment_manager_t *csm;
	  csm = app_worker_get_connect_segment_manager (app_wrk);
	  if (!segment_manager_app_detached (csm))
	    app_worker_del_segment_notify (app_wrk, ct->segment_handle);
	}
    }
  else if (!segment_manager_app_detached (sm))
    {
      app_wrk = app_worker_get (ct->server_wrk);
      app_worker_del_segment_notify (app_wrk, ct->segment_handle);
    }

  if (!del_segment)
    return;

  segment_manager_lock_and_del_segment (sm, seg_index);

  /* Cleanup segment manager if needed. If server detaches there's a chance
   * the client's sessions will hold up segment removal */
  if (segment_manager_app_detached (sm) && !segment_manager_has_fifos (sm))
    segment_manager_free_safe (sm);
}

int
ct_session_connect_notify (session_t *ss, session_error_t err)
{
  u32 ss_index, opaque, thread_index;
  ct_connection_t *sct, *cct;
  app_worker_t *client_wrk;
  session_t *cs;

  ss_index = ss->session_index;
  thread_index = ss->thread_index;
  sct = (ct_connection_t *) session_get_transport (ss);
  client_wrk = app_worker_get (sct->client_wrk);
  opaque = sct->client_opaque;

  cct = ct_connection_get (sct->peer_index, thread_index);

  /* Client closed while waiting for reply from server */
  if (PREDICT_FALSE (!cct))
    {
      session_transport_closing_notify (&sct->connection);
      session_transport_delete_notify (&sct->connection);
      ct_connection_free (sct);
      return 0;
    }

  session_half_open_delete_notify (&cct->connection);
  cct->flags &= ~CT_CONN_F_HALF_OPEN;

  if (PREDICT_FALSE (err))
    goto connect_error;

  /*
   * Alloc client session
   */

  cs = session_alloc (thread_index);
  ss = session_get (ss_index, thread_index);
  cs->session_type = ss->session_type;
  cs->listener_handle = SESSION_INVALID_HANDLE;
  cs->session_state = SESSION_STATE_CONNECTING;
  cs->app_wrk_index = client_wrk->wrk_index;
  cs->connection_index = cct->c_c_index;
  cct->c_s_index = cs->session_index;

  /* This will allocate fifos for the session. They won't be used for
   * exchanging data but they will be used to close the connection if
   * the segment manager/worker is freed */
  if ((err = app_worker_init_connected (client_wrk, cs)))
    {
      session_free (cs);
      session_close (ss);
      err = SESSION_E_ALLOC;
      goto connect_error;
    }

  cs->session_state = SESSION_STATE_CONNECTING;

  if (app_worker_connect_notify (client_wrk, cs, 0, opaque))
    {
      segment_manager_dealloc_fifos (cs->rx_fifo, cs->tx_fifo);
      session_free (cs);
      session_close (ss);
      goto cleanup_client;
    }

  cs = session_get (cct->c_s_index, cct->c_thread_index);
  cs->session_state = SESSION_STATE_READY;

  return 0;

connect_error:

  app_worker_connect_notify (client_wrk, 0, err, cct->client_opaque);

cleanup_client:

  if (cct->client_rx_fifo)
    ct_session_dealloc_fifos (cct, cct->client_rx_fifo, cct->client_tx_fifo);
  ct_connection_free (cct);
  return -1;
}

static inline ct_segment_t *
ct_lookup_free_segment (ct_main_t *cm, segment_manager_t *sm,
			u32 seg_ctx_index)
{
  uword free_bytes, max_free_bytes;
  ct_segment_t *ct_seg, *res = 0;
  ct_segments_ctx_t *seg_ctx;
  fifo_segment_t *fs;
  u32 max_fifos;

  seg_ctx = pool_elt_at_index (cm->app_seg_ctxs, seg_ctx_index);
  max_free_bytes = seg_ctx->fifo_pair_bytes;

  pool_foreach (ct_seg, seg_ctx->segments)
    {
      /* Client or server has detached so segment cannot be used */
      if ((ct_seg->flags & CT_SEGMENT_F_SERVER_DETACHED) ||
	  (ct_seg->flags & CT_SEGMENT_F_CLIENT_DETACHED))
	continue;
      fs = segment_manager_get_segment (sm, ct_seg->segment_index);
      free_bytes = fifo_segment_available_bytes (fs);
      max_fifos = fifo_segment_size (fs) / seg_ctx->fifo_pair_bytes;
      if (free_bytes > max_free_bytes &&
	  fifo_segment_num_fifos (fs) / 2 < max_fifos)
	{
	  max_free_bytes = free_bytes;
	  res = ct_seg;
	}
    }

  return res;
}

static ct_segment_t *
ct_alloc_segment (ct_main_t *cm, app_worker_t *server_wrk, u64 table_handle,
		  segment_manager_t *sm, u32 client_wrk_index)
{
  u32 seg_ctx_index = ~0, sm_index, pair_bytes;
  segment_manager_props_t *props;
  const u32 margin = 16 << 10;
  ct_segments_ctx_t *seg_ctx;
  app_worker_t *client_wrk;
  u64 seg_size, seg_handle;
  application_t *server;
  ct_segment_t *ct_seg;
  uword *spp;
  int fs_index;

  server = application_get (server_wrk->app_index);
  props = application_segment_manager_properties (server);
  sm_index = segment_manager_index (sm);
  pair_bytes = props->rx_fifo_size + props->tx_fifo_size + margin;

  /*
   * Make sure another thread did not alloc a segment while acquiring the lock
   */

  spp = hash_get (cm->app_segs_ctxs_table, table_handle);
  if (spp)
    {
      seg_ctx_index = *spp;
      ct_seg = ct_lookup_free_segment (cm, sm, seg_ctx_index);
      if (ct_seg)
	return ct_seg;
    }

  /*
   * No segment, try to alloc one and notify the server and the client.
   * Make sure the segment is not used for other fifos
   */
  seg_size = clib_max (props->segment_size, 128 << 20);
  fs_index =
    segment_manager_add_segment2 (sm, seg_size, FIFO_SEGMENT_F_CUSTOM_USE);
  if (fs_index < 0)
    return 0;

  if (seg_ctx_index == ~0)
    {
      pool_get_zero (cm->app_seg_ctxs, seg_ctx);
      seg_ctx_index = seg_ctx - cm->app_seg_ctxs;
      hash_set (cm->app_segs_ctxs_table, table_handle, seg_ctx_index);
      seg_ctx->server_wrk = server_wrk->wrk_index;
      seg_ctx->client_wrk = client_wrk_index;
      seg_ctx->sm_index = sm_index;
      seg_ctx->fifo_pair_bytes = pair_bytes;
    }
  else
    {
      seg_ctx = pool_elt_at_index (cm->app_seg_ctxs, seg_ctx_index);
    }

  pool_get_zero (seg_ctx->segments, ct_seg);
  ct_seg->segment_index = fs_index;
  ct_seg->server_n_sessions = 0;
  ct_seg->client_n_sessions = 0;
  ct_seg->ct_seg_index = ct_seg - seg_ctx->segments;
  ct_seg->seg_ctx_index = seg_ctx_index;

  /* New segment, notify the server and client */
  seg_handle = segment_manager_make_segment_handle (sm_index, fs_index);
  if (app_worker_add_segment_notify (server_wrk, seg_handle))
    goto error;

  client_wrk = app_worker_get (client_wrk_index);
  if (app_worker_add_segment_notify (client_wrk, seg_handle))
    {
      app_worker_del_segment_notify (server_wrk, seg_handle);
      goto error;
    }

  return ct_seg;

error:

  segment_manager_lock_and_del_segment (sm, fs_index);
  pool_put_index (seg_ctx->segments, ct_seg->seg_ctx_index);
  return 0;
}

static int
ct_init_accepted_session (app_worker_t *server_wrk, ct_connection_t *ct,
			  session_t *ls, session_t *ll)
{
  segment_manager_props_t *props;
  u64 seg_handle, table_handle;
  u32 sm_index, fs_index = ~0;
  ct_segments_ctx_t *seg_ctx;
  ct_main_t *cm = &ct_main;
  application_t *server;
  segment_manager_t *sm;
  ct_segment_t *ct_seg;
  fifo_segment_t *fs;
  uword *spp;
  int rv;

  sm = app_worker_get_listen_segment_manager (server_wrk, ll);
  sm_index = segment_manager_index (sm);
  server = application_get (server_wrk->app_index);
  props = application_segment_manager_properties (server);

  table_handle = ct->client_wrk << 16 | server_wrk->wrk_index;
  table_handle = (u64) sm_index << 32 | table_handle;

  /*
   * Check if we already have a segment that can hold the fifos
   */

  clib_rwlock_reader_lock (&cm->app_segs_lock);

  spp = hash_get (cm->app_segs_ctxs_table, table_handle);
  if (spp)
    {
      ct_seg = ct_lookup_free_segment (cm, sm, *spp);
      if (ct_seg)
	{
	  ct->seg_ctx_index = ct_seg->seg_ctx_index;
	  ct->ct_seg_index = ct_seg->ct_seg_index;
	  fs_index = ct_seg->segment_index;
	  __atomic_add_fetch (&ct_seg->server_n_sessions, 1, __ATOMIC_RELAXED);
	  __atomic_add_fetch (&ct_seg->client_n_sessions, 1, __ATOMIC_RELAXED);
	}
    }

  clib_rwlock_reader_unlock (&cm->app_segs_lock);

  /*
   * If not, grab exclusive lock and allocate segment
   */
  if (fs_index == ~0)
    {
      clib_rwlock_writer_lock (&cm->app_segs_lock);

      ct_seg =
	ct_alloc_segment (cm, server_wrk, table_handle, sm, ct->client_wrk);
      if (!ct_seg)
	{
	  clib_rwlock_writer_unlock (&cm->app_segs_lock);
	  return -1;
	}

      ct->seg_ctx_index = ct_seg->seg_ctx_index;
      ct->ct_seg_index = ct_seg->ct_seg_index;
      ct_seg->server_n_sessions += 1;
      ct_seg->client_n_sessions += 1;
      fs_index = ct_seg->segment_index;

      clib_rwlock_writer_unlock (&cm->app_segs_lock);
    }

  /*
   * Allocate and initialize the fifos
   */
  fs = segment_manager_get_segment_w_lock (sm, fs_index);
  rv = segment_manager_try_alloc_fifos (
    fs, ls->thread_index, props->rx_fifo_size, props->tx_fifo_size,
    &ls->rx_fifo, &ls->tx_fifo);
  if (rv)
    {
      segment_manager_segment_reader_unlock (sm);

      clib_rwlock_reader_lock (&cm->app_segs_lock);

      seg_ctx = pool_elt_at_index (cm->app_seg_ctxs, ct->seg_ctx_index);
      ct_seg = pool_elt_at_index (seg_ctx->segments, ct->ct_seg_index);
      __atomic_sub_fetch (&ct_seg->server_n_sessions, 1, __ATOMIC_RELAXED);
      __atomic_sub_fetch (&ct_seg->client_n_sessions, 1, __ATOMIC_RELAXED);

      clib_rwlock_reader_unlock (&cm->app_segs_lock);

      return rv;
    }

  ls->rx_fifo->shr->master_session_index = ls->session_index;
  ls->tx_fifo->shr->master_session_index = ls->session_index;
  ls->rx_fifo->master_thread_index = ls->thread_index;
  ls->tx_fifo->master_thread_index = ls->thread_index;
  ls->rx_fifo->segment_manager = sm_index;
  ls->tx_fifo->segment_manager = sm_index;
  ls->rx_fifo->segment_index = fs_index;
  ls->tx_fifo->segment_index = fs_index;

  seg_handle = segment_manager_segment_handle (sm, fs);
  segment_manager_segment_reader_unlock (sm);

  ct->segment_handle = seg_handle;

  return 0;
}

static void
ct_accept_rpc_wrk_handler (void *accept_args)
{
  u32 cct_index, ho_index, thread_index, ll_index;
  ct_connection_t *sct, *cct, *ho;
  transport_connection_t *ll_ct;
  app_worker_t *server_wrk;
  session_t *ss, *ll;

  /*
   * Alloc client ct and initialize from ho
   */
  thread_index = vlib_get_thread_index ();
  cct = ct_connection_alloc (thread_index);
  cct_index = cct->c_c_index;

  ho_index = pointer_to_uword (accept_args);
  ho = ct_connection_get (ho_index, 0);

  /* Unlikely but half-open session and transport could have been freed */
  if (PREDICT_FALSE (!ho))
    {
      ct_connection_free (cct);
      return;
    }

  clib_memcpy (cct, ho, sizeof (*ho));
  cct->c_c_index = cct_index;
  cct->c_thread_index = thread_index;
  cct->flags |= CT_CONN_F_HALF_OPEN;

  /* Notify session layer that half-open is on a different thread
   * and mark ho connection index reusable. Avoids another rpc
   */
  session_half_open_migrate_notify (&cct->connection);
  session_half_open_migrated_notify (&cct->connection);
  ct_half_open_add_reusable (ho_index);

  /*
   * Alloc and init server transport
   */

  ll_index = cct->peer_index;
  ll = listen_session_get (ll_index);
  sct = ct_connection_alloc (thread_index);
  /* Transport not necessarily ct but it might, so grab after sct alloc */
  ll_ct = listen_session_get_transport (ll);

  /* Make sure cct is valid after sct alloc */
  cct = ct_connection_get (cct_index, thread_index);

  sct->c_rmt_port = 0;
  sct->c_lcl_port = ll_ct->lcl_port;
  sct->c_is_ip4 = cct->c_is_ip4;
  clib_memcpy (&sct->c_lcl_ip, &cct->c_rmt_ip, sizeof (cct->c_rmt_ip));
  sct->client_wrk = cct->client_wrk;
  sct->c_proto = TRANSPORT_PROTO_NONE;
  sct->client_opaque = cct->client_opaque;
  sct->actual_tp = cct->actual_tp;

  sct->peer_index = cct->c_c_index;
  cct->peer_index = sct->c_c_index;

  /*
   * Accept server session. Client session is created only after
   * server confirms accept.
   */
  ss = session_alloc (thread_index);
  ll = listen_session_get (ll_index);
  ss->session_type = session_type_from_proto_and_ip (TRANSPORT_PROTO_NONE,
						     sct->c_is_ip4);
  ss->connection_index = sct->c_c_index;
  ss->listener_handle = listen_session_get_handle (ll);
  ss->session_state = SESSION_STATE_CREATED;

  server_wrk = application_listener_select_worker (ll);
  ss->app_wrk_index = server_wrk->wrk_index;

  sct->c_s_index = ss->session_index;
  sct->server_wrk = ss->app_wrk_index;

  if (ct_init_accepted_session (server_wrk, sct, ss, ll))
    {
      ct_session_connect_notify (ss, SESSION_E_ALLOC);
      ct_connection_free (sct);
      session_free (ss);
      return;
    }

  cct->seg_ctx_index = sct->seg_ctx_index;
  cct->ct_seg_index = sct->ct_seg_index;
  cct->client_rx_fifo = ss->tx_fifo;
  cct->client_tx_fifo = ss->rx_fifo;
  cct->client_rx_fifo->refcnt++;
  cct->client_tx_fifo->refcnt++;
  cct->segment_handle = sct->segment_handle;

  ss->session_state = SESSION_STATE_ACCEPTING;
  if (app_worker_accept_notify (server_wrk, ss))
    {
      ct_session_connect_notify (ss, SESSION_E_REFUSED);
      ct_session_dealloc_fifos (sct, ss->rx_fifo, ss->tx_fifo);
      ct_connection_free (sct);
      session_free (ss);
    }
}

static int
ct_connect (app_worker_t * client_wrk, session_t * ll,
	    session_endpoint_cfg_t * sep)
{
  u32 thread_index, ho_index;
  ct_main_t *cm = &ct_main;
  ct_connection_t *ho;

  /* Simple round-robin policy for spreading sessions over workers. We skip
   * thread index 0, i.e., offset the index by 1, when we have workers as it
   * is the one dedicated to main thread. Note that n_workers does not include
   * main thread */
  cm->n_sessions += 1;
  thread_index = cm->n_workers ? (cm->n_sessions % cm->n_workers) + 1 : 0;

  /*
   * Alloc and init client half-open transport
   */

  ho = ct_half_open_alloc ();
  ho_index = ho->c_c_index;
  ho->c_rmt_port = sep->port;
  ho->c_lcl_port = 0;
  ho->c_is_ip4 = sep->is_ip4;
  ho->client_opaque = sep->opaque;
  ho->client_wrk = client_wrk->wrk_index;
  ho->peer_index = ll->session_index;
  ho->c_proto = TRANSPORT_PROTO_NONE;
  ho->c_flags |= TRANSPORT_CONNECTION_F_NO_LOOKUP;
  clib_memcpy (&ho->c_rmt_ip, &sep->ip, sizeof (sep->ip));
  ho->flags |= CT_CONN_F_CLIENT;
  ho->c_s_index = ~0;
  ho->actual_tp = sep->transport_proto;

  /*
   * Accept connection on thread selected above. Connected reply comes
   * after server accepts the connection.
   */

  session_send_rpc_evt_to_thread_force (thread_index,
					ct_accept_rpc_wrk_handler,
					uword_to_pointer (ho_index, void *));

  return ho_index;
}

static u32
ct_start_listen (u32 app_listener_index, transport_endpoint_t * tep)
{
  session_endpoint_cfg_t *sep;
  ct_connection_t *ct;

  sep = (session_endpoint_cfg_t *) tep;
  ct = ct_connection_alloc (0);
  ct->server_wrk = sep->app_wrk_index;
  ct->c_is_ip4 = sep->is_ip4;
  clib_memcpy (&ct->c_lcl_ip, &sep->ip, sizeof (sep->ip));
  ct->c_lcl_port = sep->port;
  ct->c_s_index = app_listener_index;
  ct->actual_tp = sep->transport_proto;
  return ct->c_c_index;
}

static u32
ct_stop_listen (u32 ct_index)
{
  ct_connection_t *ct;
  ct = ct_connection_get (ct_index, 0);
  ct_connection_free (ct);
  return 0;
}

static transport_connection_t *
ct_listener_get (u32 ct_index)
{
  return (transport_connection_t *) ct_connection_get (ct_index, 0);
}

static transport_connection_t *
ct_half_open_get (u32 ct_index)
{
  return (transport_connection_t *) ct_connection_get (ct_index, 0);
}

static void
ct_session_cleanup (u32 conn_index, u32 thread_index)
{
  ct_connection_t *ct, *peer_ct;

  ct = ct_connection_get (conn_index, thread_index);
  if (!ct)
    return;

  peer_ct = ct_connection_get (ct->peer_index, thread_index);
  if (peer_ct)
    peer_ct->peer_index = ~0;

  ct_connection_free (ct);
}

static void
ct_cleanup_ho (u32 ho_index)
{
  ct_connection_free (ct_connection_get (ho_index, 0));
}

static int
ct_session_connect (transport_endpoint_cfg_t * tep)
{
  session_endpoint_cfg_t *sep_ext;
  session_endpoint_t _sep, *sep = &_sep;
  app_worker_t *app_wrk;
  session_handle_t lh;
  application_t *app;
  app_listener_t *al;
  u32 table_index;
  session_t *ll;
  u8 fib_proto;

  sep_ext = (session_endpoint_cfg_t *) tep;
  _sep = *(session_endpoint_t *) tep;
  app_wrk = app_worker_get (sep_ext->app_wrk_index);
  app = application_get (app_wrk->app_index);

  sep->transport_proto = sep_ext->original_tp;
  table_index = application_local_session_table (app);
  lh = session_lookup_local_endpoint (table_index, sep);
  if (lh == SESSION_DROP_HANDLE)
    return SESSION_E_FILTERED;

  if (lh == SESSION_INVALID_HANDLE)
    goto global_scope;

  ll = listen_session_get_from_handle (lh);
  al = app_listener_get_w_session (ll);

  /*
   * Break loop if rule in local table points to connecting app. This
   * can happen if client is a generic proxy. Route connect through
   * global table instead.
   */
  if (al->app_index == app->app_index)
    goto global_scope;

  return ct_connect (app_wrk, ll, sep_ext);

  /*
   * If nothing found, check the global scope for locally attached
   * destinations. Make sure first that we're allowed to.
   */

global_scope:
  if (session_endpoint_is_local (sep))
    return SESSION_E_NOROUTE;

  if (!application_has_global_scope (app))
    return SESSION_E_SCOPE;

  fib_proto = session_endpoint_fib_proto (sep);
  table_index = session_lookup_get_index_for_fib (fib_proto, sep->fib_index);
  ll = session_lookup_listener_wildcard (table_index, sep);

  /* Avoid connecting app to own listener */
  if (ll && ll->app_index != app->app_index)
    return ct_connect (app_wrk, ll, sep_ext);

  /* Failed to connect but no error */
  return SESSION_E_LOCAL_CONNECT;
}

static void
ct_session_postponed_cleanup (ct_connection_t *ct)
{
  app_worker_t *app_wrk;
  session_t *s;

  s = session_get (ct->c_s_index, ct->c_thread_index);
  app_wrk = app_worker_get_if_valid (s->app_wrk_index);

  if (ct->flags & CT_CONN_F_CLIENT)
    {
      if (app_wrk)
	app_worker_cleanup_notify (app_wrk, s, SESSION_CLEANUP_TRANSPORT);

      /* Normal free for client session as the fifos are allocated through
       * the connects segment manager in a segment that's not shared with
       * the server */
      session_free_w_fifos (s);
      ct_session_dealloc_fifos (ct, ct->client_rx_fifo, ct->client_tx_fifo);
    }
  else
    {
      /* Manual session and fifo segment cleanup to avoid implicit
       * segment manager cleanups and notifications */
      app_wrk = app_worker_get_if_valid (s->app_wrk_index);
      if (app_wrk)
	{
	  app_worker_cleanup_notify (app_wrk, s, SESSION_CLEANUP_TRANSPORT);
	  app_worker_cleanup_notify (app_wrk, s, SESSION_CLEANUP_SESSION);
	}

      ct_session_dealloc_fifos (ct, s->rx_fifo, s->tx_fifo);
      session_free (s);
    }

  ct_connection_free (ct);
}

static void
ct_handle_cleanups (void *args)
{
  uword thread_index = pointer_to_uword (args);
  const u32 max_cleanups = 100;
  ct_main_t *cm = &ct_main;
  ct_cleanup_req_t *req;
  ct_connection_t *ct;
  u32 n_to_handle = 0;
  session_t *s;

  cm->heave_cleanups[thread_index] = 0;
  n_to_handle = clib_fifo_elts (cm->pending_cleanups[thread_index]);
  n_to_handle = clib_min (n_to_handle, max_cleanups);

  while (n_to_handle)
    {
      clib_fifo_sub2 (cm->pending_cleanups[thread_index], req);
      ct = ct_connection_get (req->ct_index, thread_index);
      s = session_get (ct->c_s_index, ct->c_thread_index);
      if (!svm_fifo_has_event (s->tx_fifo))
	ct_session_postponed_cleanup (ct);
      else
	clib_fifo_add1 (cm->pending_cleanups[thread_index], *req);
      n_to_handle -= 1;
    }

  if (clib_fifo_elts (cm->pending_cleanups[thread_index]))
    {
      cm->heave_cleanups[thread_index] = 1;
      session_send_rpc_evt_to_thread_force (
	thread_index, ct_handle_cleanups,
	uword_to_pointer (thread_index, void *));
    }
}

static void
ct_program_cleanup (ct_connection_t *ct)
{
  ct_main_t *cm = &ct_main;
  ct_cleanup_req_t *req;
  uword thread_index;

  thread_index = ct->c_thread_index;
  clib_fifo_add2 (cm->pending_cleanups[thread_index], req);
  req->ct_index = ct->c_c_index;

  if (cm->heave_cleanups[thread_index])
    return;

  cm->heave_cleanups[thread_index] = 1;
  session_send_rpc_evt_to_thread_force (
    thread_index, ct_handle_cleanups, uword_to_pointer (thread_index, void *));
}

static void
ct_session_close (u32 ct_index, u32 thread_index)
{
  ct_connection_t *ct, *peer_ct;
  session_t *s;

  ct = ct_connection_get (ct_index, thread_index);
  s = session_get (ct->c_s_index, ct->c_thread_index);
  peer_ct = ct_connection_get (ct->peer_index, thread_index);
  if (peer_ct)
    {
      peer_ct->peer_index = ~0;
      /* Make sure session was allocated */
      if (peer_ct->flags & CT_CONN_F_HALF_OPEN)
	{
	  ct_session_connect_notify (s, SESSION_E_REFUSED);
	}
      else if (peer_ct->c_s_index != ~0)
	session_transport_closing_notify (&peer_ct->connection);
      else
	{
	  /* should not happen */
	  clib_warning ("ct peer without session");
	  ct_connection_free (peer_ct);
	}
    }

  /* Do not send closed notify to make sure pending tx events are
   * still delivered and program cleanup */
  ct_program_cleanup (ct);
}

static transport_connection_t *
ct_session_get (u32 ct_index, u32 thread_index)
{
  return (transport_connection_t *) ct_connection_get (ct_index,
						       thread_index);
}

static u8 *
format_ct_connection_id (u8 * s, va_list * args)
{
  ct_connection_t *ct = va_arg (*args, ct_connection_t *);
  if (!ct)
    return s;
  if (ct->c_is_ip4)
    {
      s = format (s, "[%d:%d][CT:%U] %U:%d->%U:%d", ct->c_thread_index,
		  ct->c_s_index, format_transport_proto_short, ct->actual_tp,
		  format_ip4_address, &ct->c_lcl_ip4,
		  clib_net_to_host_u16 (ct->c_lcl_port), format_ip4_address,
		  &ct->c_rmt_ip4, clib_net_to_host_u16 (ct->c_rmt_port));
    }
  else
    {
      s = format (s, "[%d:%d][CT:%U] %U:%d->%U:%d", ct->c_thread_index,
		  ct->c_s_index, format_transport_proto_short, ct->actual_tp,
		  format_ip6_address, &ct->c_lcl_ip6,
		  clib_net_to_host_u16 (ct->c_lcl_port), format_ip6_address,
		  &ct->c_rmt_ip6, clib_net_to_host_u16 (ct->c_rmt_port));
    }

  return s;
}

static int
ct_custom_tx (void *session, transport_send_params_t * sp)
{
  session_t *s = (session_t *) session;
  if (session_has_transport (s))
    return 0;
  /* If event enqueued towards peer, remove from scheduler and remove
   * session tx flag, i.e., accept new tx events. Unset fifo flag now to
   * avoid missing events if peer did not clear fifo flag yet, which is
   * interpreted as successful notification and session is descheduled. */
  svm_fifo_unset_event (s->tx_fifo);
  if (!ct_session_tx (s))
    sp->flags = TRANSPORT_SND_F_DESCHED;

  /* The scheduler uses packet count as a means of upper bounding the amount
   * of work done per dispatch. So make it look like we have sent something */
  return 1;
}

static int
ct_app_rx_evt (transport_connection_t * tc)
{
  ct_connection_t *ct = (ct_connection_t *) tc, *peer_ct;
  session_t *ps;

  peer_ct = ct_connection_get (ct->peer_index, tc->thread_index);
  if (!peer_ct)
    return -1;
  ps = session_get (peer_ct->c_s_index, peer_ct->c_thread_index);
  return session_dequeue_notify (ps);
}

static u8 *
format_ct_listener (u8 * s, va_list * args)
{
  u32 tc_index = va_arg (*args, u32);
  u32 __clib_unused thread_index = va_arg (*args, u32);
  u32 __clib_unused verbose = va_arg (*args, u32);
  ct_connection_t *ct = ct_connection_get (tc_index, 0);
  s = format (s, "%-" SESSION_CLI_ID_LEN "U", format_ct_connection_id, ct);
  if (verbose)
    s = format (s, "%-" SESSION_CLI_STATE_LEN "s", "LISTEN");
  return s;
}

static u8 *
format_ct_half_open (u8 *s, va_list *args)
{
  u32 ho_index = va_arg (*args, u32);
  u32 verbose = va_arg (*args, u32);
  ct_connection_t *ct = ct_connection_get (ho_index, 0);
  s = format (s, "%-" SESSION_CLI_ID_LEN "U", format_ct_connection_id, ct);
  if (verbose)
    s = format (s, "%-" SESSION_CLI_STATE_LEN "s", "HALF-OPEN");
  return s;
}

static u8 *
format_ct_connection (u8 * s, va_list * args)
{
  ct_connection_t *ct = va_arg (*args, ct_connection_t *);
  u32 verbose = va_arg (*args, u32);

  if (!ct)
    return s;
  s = format (s, "%-" SESSION_CLI_ID_LEN "U", format_ct_connection_id, ct);
  if (verbose)
    {
      s = format (s, "%-" SESSION_CLI_STATE_LEN "s", "ESTABLISHED");
      if (verbose > 1)
	{
	  s = format (s, "\n");
	}
    }
  return s;
}

static u8 *
format_ct_session (u8 * s, va_list * args)
{
  u32 ct_index = va_arg (*args, u32);
  u32 thread_index = va_arg (*args, u32);
  u32 verbose = va_arg (*args, u32);
  ct_connection_t *ct;

  ct = ct_connection_get (ct_index, thread_index);
  if (!ct)
    {
      s = format (s, "empty\n");
      return s;
    }

  s = format (s, "%U", format_ct_connection, ct, verbose);
  return s;
}

clib_error_t *
ct_enable_disable (vlib_main_t * vm, u8 is_en)
{
  ct_main_t *cm = &ct_main;

  cm->n_workers = vlib_num_workers ();
  vec_validate (cm->connections, cm->n_workers);
  vec_validate (cm->pending_cleanups, cm->n_workers);
  vec_validate (cm->heave_cleanups, cm->n_workers);
  clib_spinlock_init (&cm->ho_reuseable_lock);
  clib_rwlock_init (&cm->app_segs_lock);
  return 0;
}

/* *INDENT-OFF* */
static const transport_proto_vft_t cut_thru_proto = {
  .enable = ct_enable_disable,
  .start_listen = ct_start_listen,
  .stop_listen = ct_stop_listen,
  .get_connection = ct_session_get,
  .get_listener = ct_listener_get,
  .get_half_open = ct_half_open_get,
  .cleanup = ct_session_cleanup,
  .cleanup_ho = ct_cleanup_ho,
  .connect = ct_session_connect,
  .close = ct_session_close,
  .custom_tx = ct_custom_tx,
  .app_rx_evt = ct_app_rx_evt,
  .format_listener = format_ct_listener,
  .format_half_open = format_ct_half_open,
  .format_connection = format_ct_session,
  .transport_options = {
    .name = "ct",
    .short_name = "C",
    .tx_type = TRANSPORT_TX_INTERNAL,
    .service_type = TRANSPORT_SERVICE_VC,
  },
};
/* *INDENT-ON* */

int
ct_session_tx (session_t * s)
{
  ct_connection_t *ct, *peer_ct;
  session_t *peer_s;

  ct = (ct_connection_t *) session_get_transport (s);
  peer_ct = ct_connection_get (ct->peer_index, ct->c_thread_index);
  if (!peer_ct)
    return 0;
  peer_s = session_get (peer_ct->c_s_index, peer_ct->c_thread_index);
  if (peer_s->session_state >= SESSION_STATE_TRANSPORT_CLOSING)
    return 0;
  return session_enqueue_notify (peer_s);
}

static clib_error_t *
ct_transport_init (vlib_main_t * vm)
{
  transport_register_protocol (TRANSPORT_PROTO_NONE, &cut_thru_proto,
			       FIB_PROTOCOL_IP4, ~0);
  transport_register_protocol (TRANSPORT_PROTO_NONE, &cut_thru_proto,
			       FIB_PROTOCOL_IP6, ~0);
  return 0;
}

VLIB_INIT_FUNCTION (ct_transport_init);

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