summaryrefslogtreecommitdiff
path: root/proto/bfd/bfd.c
blob: c88c1cb27f4265c9ceaff03c0b1528388bf71fc5 (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
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
/*
 *	BIRD -- Bidirectional Forwarding Detection (BFD)
 *
 *	Can be freely distributed and used under the terms of the GNU GPL.
 */

/**
 * DOC: Bidirectional Forwarding Detection
 *
 * The BFD protocol is implemented in three files: |bfd.c| containing the
 * protocol logic and the protocol glue with BIRD core, |packets.c| handling BFD
 * packet processing, RX, TX and protocol sockets. |io.c| then contains generic
 * code for the event loop, threads and event sources (sockets, microsecond
 * timers). This generic code will be merged to the main BIRD I/O code in the
 * future.
 *
 * The BFD implementation uses a separate thread with an internal event loop for
 * handling the protocol logic, which requires high-res and low-latency timing,
 * so it is not affected by the rest of BIRD, which has several low-granularity
 * hooks in the main loop, uses second-based timers and cannot offer good
 * latency. The core of BFD protocol (the code related to BFD sessions,
 * interfaces and packets) runs in the BFD thread, while the rest (the code
 * related to BFD requests, BFD neighbors and the protocol glue) runs in the
 * main thread.
 *
 * BFD sessions are represented by structure &bfd_session that contains a state
 * related to the session and two timers (TX timer for periodic packets and hold
 * timer for session timeout). These sessions are allocated from @session_slab
 * and are accessible by two hash tables, @session_hash_id (by session ID) and
 * @session_hash_ip (by IP addresses of neighbors and associated interfaces).
 * Slab and both hashes are in the main protocol structure &bfd_proto. The
 * protocol logic related to BFD sessions is implemented in internal functions
 * bfd_session_*(), which are expected to be called from the context of BFD
 * thread, and external functions bfd_add_session(), bfd_remove_session() and
 * bfd_reconfigure_session(), which form an interface to the BFD core for the
 * rest and are expected to be called from the context of main thread.
 *
 * Each BFD session has an associated BFD interface, represented by structure
 * &bfd_iface. A BFD interface contains a socket used for TX (the one for RX is
 * shared in &bfd_proto), an interface configuration and reference counter.
 * Compared to interface structures of other protocols, these structures are not
 * created and removed based on interface notification events, but according to
 * the needs of BFD sessions. When a new session is created, it requests a
 * proper BFD interface by function bfd_get_iface(), which either finds an
 * existing one in &iface_list (from &bfd_proto) or allocates a new one. When a
 * session is removed, an associated iface is discharged by bfd_free_iface().
 *
 * BFD requests are the external API for the other protocols. When a protocol
 * wants a BFD session, it calls bfd_request_session(), which creates a
 * structure &bfd_request containing approprite information and an notify hook.
 * This structure is a resource associated with the caller's resource pool. When
 * a BFD protocol is available, a BFD request is submitted to the protocol, an
 * appropriate BFD session is found or created and the request is attached to
 * the session. When a session changes state, all attached requests (and related
 * protocols) are notified. Note that BFD requests do not depend on BFD protocol
 * running. When the BFD protocol is stopped or removed (or not available from
 * beginning), related BFD requests are stored in @bfd_wait_list, where waits
 * for a new protocol.
 *
 * BFD neighbors are just a way to statically configure BFD sessions without
 * requests from other protocol. Structures &bfd_neighbor are part of BFD
 * configuration (like static routes in the static protocol). BFD neighbors are
 * handled by BFD protocol like it is a BFD client -- when a BFD neighbor is
 * ready, the protocol just creates a BFD request like any other protocol.
 *
 * The protocol uses a new generic event loop (structure &birdloop) from |io.c|,
 * which supports sockets, timers and events like the main loop. A birdloop is
 * associated with a thread (field @thread) in which event hooks are executed.
 * Most functions for setting event sources (like sk_start() or tm_start()) must
 * be called from the context of that thread. Birdloop allows to temporarily
 * acquire the context of that thread for the main thread by calling
 * birdloop_enter() and then birdloop_leave(), which also ensures mutual
 * exclusion with all event hooks. Note that resources associated with a
 * birdloop (like timers) should be attached to the independent resource pool,
 * detached from the main resource tree.
 *
 * There are two kinds of interaction between the BFD core (running in the BFD
 * thread) and the rest of BFD (running in the main thread). The first kind are
 * configuration calls from main thread to the BFD thread (like bfd_add_session()).
 * These calls are synchronous and use birdloop_enter() mechanism for mutual
 * exclusion. The second kind is a notification about session changes from the
 * BFD thread to the main thread. This is done in an asynchronous way, sesions
 * with pending notifications are linked (in the BFD thread) to @notify_list in
 * &bfd_proto, and then bfd_notify_hook() in the main thread is activated using
 * a standard event sending code. The hook then processes scheduled sessions and
 * calls hooks from associated BFD requests. This @notify_list (and state fields
 * in structure &bfd_session) is protected by a spinlock in &bfd_proto and
 * functions bfd_lock_sessions() / bfd_unlock_sessions().
 *
 * There are few data races (accessing @p->p.debug from TRACE() from the BFD
 * thread and accessing some some private fields of %bfd_session from
 * bfd_show_sessions() from the main thread, but these are harmless (i hope).
 *
 * TODO: document functions and access restrictions for fields in BFD structures.
 *
 * Supported standards:
 * - RFC 5880 - main BFD standard
 * - RFC 5881 - BFD for IP links
 * - RFC 5882 - generic application of BFD
 * - RFC 5883 - BFD for multihop paths
 */

#include "bfd.h"


#define HASH_ID_KEY(n)		n->loc_id
#define HASH_ID_NEXT(n)		n->next_id
#define HASH_ID_EQ(a,b)		a == b
#define HASH_ID_FN(k)		k

#define HASH_IP_KEY(n)		n->addr, n->ifindex
#define HASH_IP_NEXT(n)		n->next_ip
#define HASH_IP_EQ(a1,n1,a2,n2)	ipa_equal(a1, a2) && n1 == n2
#define HASH_IP_FN(a,n)		ipa_hash(a) ^ u32_hash(n)

#define BFD_LOCK	LOCK_DOMAIN(rtable, bfd_global.lock)
#define BFD_UNLOCK	UNLOCK_DOMAIN(rtable, bfd_global.lock)

static struct {
  DOMAIN(rtable) lock;
  list wait_list;
  list pickup_list;
  list proto_list;
  uint pickup_reload;
} bfd_global;

const char *bfd_state_names[] = { "AdminDown", "Down", "Init", "Up" };

static void bfd_session_set_min_tx(struct bfd_session *s, u32 val);
static struct bfd_iface *bfd_get_iface(struct bfd_proto *p, ip_addr local, struct iface *iface);
static void bfd_free_iface(struct bfd_iface *ifa);


/*
 *	BFD sessions
 */

static inline struct bfd_session_config
bfd_merge_options(const struct bfd_iface_config *cf, const struct bfd_options *opts)
{
  return (struct bfd_session_config) {
    .min_rx_int = opts->min_rx_int ?: cf->min_rx_int,
    .min_tx_int = opts->min_tx_int ?: cf->min_tx_int,
    .idle_tx_int = opts->idle_tx_int ?: cf->idle_tx_int,
    .multiplier = opts->multiplier ?: cf->multiplier,
    .passive = opts->passive_set ? opts->passive : cf->passive
  };
}

static void
bfd_session_update_state(struct bfd_session *s, uint state, uint diag)
{
  struct bfd_proto *p = s->ifa->bfd;
  uint old_state = s->loc_state;
  int notify;

  if (state == old_state)
    return;

  TRACE(D_EVENTS, "Session to %I changed state from %s to %s",
	s->addr, bfd_state_names[old_state], bfd_state_names[state]);

  bfd_lock_sessions(p);
  s->loc_state = state;
  s->loc_diag = diag;
  s->last_state_change = current_time();

  notify = !NODE_VALID(&s->n);
  if (notify)
    add_tail(&p->notify_list, &s->n);
  bfd_unlock_sessions(p);

  if (state == BFD_STATE_UP)
    bfd_session_set_min_tx(s, s->cf.min_tx_int);

  if (old_state == BFD_STATE_UP)
    bfd_session_set_min_tx(s, s->cf.idle_tx_int);

  if (notify)
    ev_send(&global_event_list, &p->notify_event);
}

static void
bfd_session_update_tx_interval(struct bfd_session *s)
{
  u32 tx_int = MAX(s->des_min_tx_int, s->rem_min_rx_int);
  u32 tx_int_l = tx_int - (tx_int / 4);	 // 75 %
  u32 tx_int_h = tx_int - (tx_int / 10); // 90 %

  s->tx_timer->recurrent = tx_int_l;
  s->tx_timer->randomize = tx_int_h - tx_int_l;

  /* Do not set timer if no previous event */
  if (!s->last_tx)
    return;

  /* Set timer relative to last tx_timer event */
  tm_set_in(s->tx_timer, s->last_tx + tx_int_l, s->ifa->bfd->p.loop);
}

static void
bfd_session_update_detection_time(struct bfd_session *s, int kick)
{
  btime timeout = (btime) MAX(s->req_min_rx_int, s->rem_min_tx_int) * s->rem_detect_mult;

  if (kick)
    s->last_rx = current_time();

  if (!s->last_rx)
    return;

  tm_set_in(s->hold_timer, s->last_rx + timeout, s->ifa->bfd->p.loop);
}

static void
bfd_session_control_tx_timer(struct bfd_session *s, int reset)
{
  // if (!s->opened) goto stop;

  if (s->passive && (s->rem_id == 0))
    goto stop;

  if (s->rem_demand_mode &&
      !s->poll_active &&
      (s->loc_state == BFD_STATE_UP) &&
      (s->rem_state == BFD_STATE_UP))
    goto stop;

  if (s->rem_min_rx_int == 0)
    goto stop;

  /* So TX timer should run */
  if (reset || !tm_active(s->tx_timer))
  {
    s->last_tx = 0;
    tm_start_in(s->tx_timer, 0, s->ifa->bfd->p.loop);
  }

  return;

 stop:
  tm_stop(s->tx_timer);
  s->last_tx = 0;
}

static void
bfd_session_request_poll(struct bfd_session *s, u8 request)
{
  /* Not sure about this, but doing poll in this case does not make sense */
  if (s->rem_id == 0)
    return;

  s->poll_scheduled |= request;

  if (s->poll_active)
    return;

  s->poll_active = s->poll_scheduled;
  s->poll_scheduled = 0;

  bfd_session_control_tx_timer(s, 1);
}

static void
bfd_session_terminate_poll(struct bfd_session *s)
{
  u8 poll_done = s->poll_active & ~s->poll_scheduled;

  if (poll_done & BFD_POLL_TX)
    s->des_min_tx_int = s->des_min_tx_new;

  if (poll_done & BFD_POLL_RX)
    s->req_min_rx_int = s->req_min_rx_new;

  s->poll_active = s->poll_scheduled;
  s->poll_scheduled = 0;

  /* Timers are updated by caller - bfd_session_process_ctl() */
}

void
bfd_session_process_ctl(struct bfd_session *s, u8 flags, u32 old_tx_int, u32 old_rx_int)
{
  if (s->poll_active && (flags & BFD_FLAG_FINAL))
    bfd_session_terminate_poll(s);

  if ((s->des_min_tx_int != old_tx_int) || (s->rem_min_rx_int != old_rx_int))
    bfd_session_update_tx_interval(s);

  bfd_session_update_detection_time(s, 1);

  /* Update session state */
  int next_state = 0;
  int diag = BFD_DIAG_NOTHING;

  switch (s->loc_state)
  {
  case BFD_STATE_ADMIN_DOWN:
    return;

  case BFD_STATE_DOWN:
    if (s->rem_state == BFD_STATE_DOWN)		next_state = BFD_STATE_INIT;
    else if (s->rem_state == BFD_STATE_INIT)	next_state = BFD_STATE_UP;
    break;

  case BFD_STATE_INIT:
    if (s->rem_state == BFD_STATE_ADMIN_DOWN)	next_state = BFD_STATE_DOWN, diag = BFD_DIAG_NEIGHBOR_DOWN;
    else if (s->rem_state >= BFD_STATE_INIT)	next_state = BFD_STATE_UP;
    break;

  case BFD_STATE_UP:
    if (s->rem_state <= BFD_STATE_DOWN)		next_state = BFD_STATE_DOWN, diag = BFD_DIAG_NEIGHBOR_DOWN;
    break;
  }

  if (next_state)
    bfd_session_update_state(s, next_state, diag);

  bfd_session_control_tx_timer(s, 0);

  if (flags & BFD_FLAG_POLL)
    bfd_send_ctl(s->ifa->bfd, s, 1);
}

static void
bfd_session_timeout(struct bfd_session *s)
{
  struct bfd_proto *p = s->ifa->bfd;

  TRACE(D_EVENTS, "Session to %I expired", s->addr);

  s->rem_state = BFD_STATE_DOWN;
  s->rem_id = 0;
  s->rem_min_tx_int = 0;
  s->rem_min_rx_int = 1;
  s->rem_demand_mode = 0;
  s->rem_detect_mult = 0;
  s->rx_csn_known = 0;

  s->poll_active = 0;
  s->poll_scheduled = 0;

  bfd_session_update_state(s, BFD_STATE_DOWN, BFD_DIAG_TIMEOUT);

  bfd_session_control_tx_timer(s, 1);
}

static void
bfd_session_set_min_tx(struct bfd_session *s, u32 val)
{
  /* Note that des_min_tx_int <= des_min_tx_new */

  if (val == s->des_min_tx_new)
    return;

  s->des_min_tx_new = val;

  /* Postpone timer update if des_min_tx_int increases and the session is up */
  if ((s->loc_state != BFD_STATE_UP) || (val < s->des_min_tx_int))
  {
    s->des_min_tx_int = val;
    bfd_session_update_tx_interval(s);
  }

  bfd_session_request_poll(s, BFD_POLL_TX);
}

static void
bfd_session_set_min_rx(struct bfd_session *s, u32 val)
{
  /* Note that req_min_rx_int >= req_min_rx_new */

  if (val == s->req_min_rx_new)
    return;

  s->req_min_rx_new = val;

  /* Postpone timer update if req_min_rx_int decreases and the session is up */
  if ((s->loc_state != BFD_STATE_UP) || (val > s->req_min_rx_int))
  {
    s->req_min_rx_int = val;
    bfd_session_update_detection_time(s, 0);
  }

  bfd_session_request_poll(s, BFD_POLL_RX);
}

struct bfd_session *
bfd_find_session_by_id(struct bfd_proto *p, u32 id)
{
  return HASH_FIND(p->session_hash_id, HASH_ID, id);
}

struct bfd_session *
bfd_find_session_by_addr(struct bfd_proto *p, ip_addr addr, uint ifindex)
{
  return HASH_FIND(p->session_hash_ip, HASH_IP, addr, ifindex);
}

static void
bfd_tx_timer_hook(timer *t)
{
  struct bfd_session *s = t->data;

  s->last_tx = current_time();
  bfd_send_ctl(s->ifa->bfd, s, 0);
}

static void
bfd_hold_timer_hook(timer *t)
{
  bfd_session_timeout(t->data);
}

static u32
bfd_get_free_id(struct bfd_proto *p)
{
  u32 id;
  for (id = random_u32(); 1; id++)
    if (id && !bfd_find_session_by_id(p, id))
      break;

  return id;
}

static struct bfd_session *
bfd_add_session(struct bfd_proto *p, ip_addr addr, ip_addr local, struct iface *iface, struct bfd_options *opts)
{
  ASSERT_DIE(birdloop_inside(p->p.loop));

  struct bfd_iface *ifa = bfd_get_iface(p, local, iface);

  struct bfd_session *s = sl_allocz(p->session_slab);
  s->addr = addr;
  s->ifa = ifa;
  s->ifindex = iface ? iface->index : 0;
  s->loc_id = bfd_get_free_id(p);

  HASH_INSERT(p->session_hash_id, HASH_ID, s);
  HASH_INSERT(p->session_hash_ip, HASH_IP, s);

  s->cf = bfd_merge_options(ifa->cf, opts);

  /* Initialization of state variables - see RFC 5880 6.8.1 */
  s->loc_state = BFD_STATE_DOWN;
  s->rem_state = BFD_STATE_DOWN;
  s->des_min_tx_int = s->des_min_tx_new = s->cf.idle_tx_int;
  s->req_min_rx_int = s->req_min_rx_new = s->cf.min_rx_int;
  s->rem_min_rx_int = 1;
  s->detect_mult = s->cf.multiplier;
  s->passive = s->cf.passive;
  s->tx_csn = random_u32();

  s->tx_timer = tm_new_init(p->tpool, bfd_tx_timer_hook, s, 0, 0);
  s->hold_timer = tm_new_init(p->tpool, bfd_hold_timer_hook, s, 0, 0);
  bfd_session_update_tx_interval(s);
  bfd_session_control_tx_timer(s, 1);

  init_list(&s->request_list);
  s->last_state_change = current_time();

  TRACE(D_EVENTS, "Session to %I added", s->addr);

  return s;
}

/*
static void
bfd_open_session(struct bfd_proto *p, struct bfd_session *s, ip_addr local, struct iface *ifa)
{
  birdloop_enter(p->p.loop);

  s->opened = 1;

  bfd_session_control_tx_timer(s);

  birdloop_leave(p->p.loop);
}

static void
bfd_close_session(struct bfd_proto *p, struct bfd_session *s)
{
  birdloop_enter(p->p.loop);

  s->opened = 0;

  bfd_session_update_state(s, BFD_STATE_DOWN, BFD_DIAG_PATH_DOWN);
  bfd_session_control_tx_timer(s);

  birdloop_leave(p->p.loop);
}
*/

static void
bfd_remove_session_locked(struct bfd_proto *p, struct bfd_session *s)
{
  /* Caller should ensure that request list is empty */

  /* Remove session from notify list if scheduled for notification */
  /* No need for bfd_lock_sessions(), we are already protected by birdloop_enter() */
  if (NODE_VALID(&s->n))
    rem_node(&s->n);

  bfd_free_iface(s->ifa);

  rfree(s->tx_timer);
  rfree(s->hold_timer);

  HASH_REMOVE(p->session_hash_id, HASH_ID, s);
  HASH_REMOVE(p->session_hash_ip, HASH_IP, s);

  TRACE(D_EVENTS, "Session to %I removed", s->addr);

  sl_free(s);
}

static void
bfd_remove_session(struct bfd_proto *p, struct bfd_session *s)
{
  birdloop_enter(p->p.loop);
  bfd_remove_session_locked(p, s);
  birdloop_leave(p->p.loop);
}

static void
bfd_reconfigure_session(struct bfd_proto *p, struct bfd_session *s)
{
  if (EMPTY_LIST(s->request_list))
    return;

  birdloop_enter(p->p.loop);

  struct bfd_request *req = SKIP_BACK(struct bfd_request, n, HEAD(s->request_list));
  s->cf = bfd_merge_options(s->ifa->cf, &req->opts);

  u32 tx = (s->loc_state == BFD_STATE_UP) ? s->cf.min_tx_int : s->cf.idle_tx_int;
  bfd_session_set_min_tx(s, tx);
  bfd_session_set_min_rx(s, s->cf.min_rx_int);
  s->detect_mult = s->cf.multiplier;
  s->passive = s->cf.passive;

  bfd_session_control_tx_timer(s, 0);

  birdloop_leave(p->p.loop);

  TRACE(D_EVENTS, "Session to %I reconfigured", s->addr);
}


/*
 *	BFD interfaces
 */

static struct bfd_iface_config bfd_default_iface = {
  .min_rx_int = BFD_DEFAULT_MIN_RX_INT,
  .min_tx_int = BFD_DEFAULT_MIN_TX_INT,
  .idle_tx_int = BFD_DEFAULT_IDLE_TX_INT,
  .multiplier = BFD_DEFAULT_MULTIPLIER
};

static inline struct bfd_iface_config *
bfd_find_iface_config(struct bfd_config *cf, struct iface *iface)
{
  struct bfd_iface_config *ic;

  ic = iface ? (void *) iface_patt_find(&cf->patt_list, iface, NULL) : cf->multihop;

  return ic ? ic : &bfd_default_iface;
}

static struct bfd_iface *
bfd_get_iface(struct bfd_proto *p, ip_addr local, struct iface *iface)
{
  struct bfd_iface *ifa;

  WALK_LIST(ifa, p->iface_list)
    if (ipa_equal(ifa->local, local) && (ifa->iface == iface))
      return ifa->uc++, ifa;

  struct bfd_config *cf = (struct bfd_config *) (p->p.cf);
  struct bfd_iface_config *ic = bfd_find_iface_config(cf, iface);

  ifa = mb_allocz(p->tpool, sizeof(struct bfd_iface));
  ifa->local = local;
  ifa->iface = iface;
  ifa->cf = ic;
  ifa->bfd = p;

  ifa->sk = bfd_open_tx_sk(p, local, iface);
  ifa->uc = 1;

  if (cf->strict_bind)
    ifa->rx = bfd_open_rx_sk_bound(p, local, iface);

  add_tail(&p->iface_list, &ifa->n);

  return ifa;
}

static void
bfd_free_iface(struct bfd_iface *ifa)
{
  if (!ifa || --ifa->uc)
    return;

  if (ifa->sk)
    rfree(ifa->sk);

  if (ifa->rx)
    rfree(ifa->rx);

  rem_node(&ifa->n);
  mb_free(ifa);
}

static void
bfd_reconfigure_iface(struct bfd_proto *p, struct bfd_iface *ifa, struct bfd_config *nc)
{
  struct bfd_iface_config *new = bfd_find_iface_config(nc, ifa->iface);
  struct bfd_iface_config *old = ifa->cf;

  /* Check options that are handled in bfd_reconfigure_session() */
  ifa->changed =
    (new->min_rx_int != old->min_rx_int) ||
    (new->min_tx_int != old->min_tx_int) ||
    (new->idle_tx_int != old->idle_tx_int) ||
    (new->multiplier != old->multiplier) ||
    (new->passive != old->passive);

  /* This should be probably changed to not access ifa->cf from the BFD thread */
  birdloop_enter(p->p.loop);
  ifa->cf = new;
  birdloop_leave(p->p.loop);
}


/*
 *	BFD requests
 */

static void
bfd_request_notify(struct bfd_request *req, u8 state, u8 diag)
{
  u8 old_state = req->state;

  if (state == old_state)
    return;

  req->state = state;
  req->diag = diag;
  req->old_state = old_state;
  req->down = (old_state == BFD_STATE_UP) && (state == BFD_STATE_DOWN);

  if (req->hook)
  {
    struct birdloop *target = !birdloop_inside(req->target) ? req->target : NULL;

    if (target)
      birdloop_enter(target);

    req->hook(req);

    if (target)
      birdloop_leave(target);
  }
}

static int
bfd_add_request(struct bfd_proto *p, struct bfd_request *req)
{
  struct bfd_config *cf = (struct bfd_config *) (p->p.cf);

  if (p->p.vrf && (p->p.vrf != req->vrf))
  {
    TRACE(D_EVENTS, "Not accepting request to %I with different VRF", req->addr);
    return 0;
  }

  if (ipa_is_ip4(req->addr) ? !cf->accept_ipv4 : !cf->accept_ipv6)
  {
    TRACE(D_EVENTS, "Not accepting request to %I (AF limit)", req->addr);
    return 0;
  }

  if (req->iface ? !cf->accept_direct : !cf->accept_multihop)
  {
    TRACE(D_EVENTS, "Not accepting %s request to %I", req->iface ? "direct" : "multihop", req->addr);
    return 0;
  }

  uint ifindex = req->iface ? req->iface->index : 0;
  struct bfd_session *s = bfd_find_session_by_addr(p, req->addr, ifindex);

  if (s)
    TRACE(D_EVENTS, "Session to %I reused", s->addr);
  else
    s = bfd_add_session(p, req->addr, req->local, req->iface, &req->opts);

  rem_node(&req->n);
  add_tail(&s->request_list, &req->n);
  req->session = s;

  bfd_lock_sessions(p);

  int notify = !NODE_VALID(&s->n);
  if (notify)
    add_tail(&p->notify_list, &s->n);

  bfd_unlock_sessions(p);

  if (notify)
    ev_send(&global_event_list, &p->notify_event);

  return 1;
}

static void
bfd_pickup_requests(void *_data UNUSED)
{
  /* NOTE TO MY FUTURE SELF
   *
   * Functions bfd_take_requests() and bfd_drop_requests() need to have
   * consistent &bfd_global.wait_list and this is ensured only by having these
   * functions called from bfd_start() and bfd_shutdown() which are both called
   * in PROTO_LOCKED_FROM_MAIN context, i.e. always from &main_birdloop.
   *
   * This pickup event is also called in &main_birdloop, therefore we can
   * freely do BFD_LOCK/BFD_UNLOCK while processing all the requests. All BFD
   * protocols capable of bfd_add_request() are either started before this code
   * happens or after that.
   *
   * If BFD protocols could start in parallel with this routine, they might
   * miss some of the waiting requests, thus if anybody tries to start
   * protocols or run this pickup event outside &main_birdloop in future, they
   * shall ensure that this race condition is mitigated somehow.
   *
   * Thank you, my future self, for understanding. Have a nice day!
   */

  DBG("BFD pickup loop starting");

  BFD_LOCK;
  do {
    bfd_global.pickup_reload = 0;
    BFD_UNLOCK;

    node *n;
    WALK_LIST(n, bfd_global.proto_list)
    {
      struct bfd_proto *p = SKIP_BACK(struct bfd_proto, bfd_node, n);
      birdloop_enter(p->p.loop);
      BFD_LOCK;

      TRACE(D_EVENTS, "Picking up new requests (%d available)", list_length(&bfd_global.pickup_list));

      node *rn, *rnxt;
      WALK_LIST_DELSAFE(rn, rnxt, bfd_global.pickup_list)
	bfd_add_request(p, SKIP_BACK(struct bfd_request, n, rn));

      BFD_UNLOCK;

      /* Remove sessions with no requests */
      HASH_WALK_DELSAFE(p->session_hash_id, next_id, s)
      {
	if (EMPTY_LIST(s->request_list))
	  bfd_remove_session_locked(p, s);
      }
      HASH_WALK_END;

      birdloop_leave(p->p.loop);
    }

    BFD_LOCK;
  } while (bfd_global.pickup_reload);

  list tmp_list;
  init_list(&tmp_list);
  add_tail_list(&tmp_list, &bfd_global.pickup_list);

  init_list(&bfd_global.pickup_list);
  BFD_UNLOCK;

  log(L_TRACE "No protocol for %d BFD requests", list_length(&tmp_list));

  node *n;
  WALK_LIST(n, tmp_list)
    bfd_request_notify(SKIP_BACK(struct bfd_request, n, n), BFD_STATE_ADMIN_DOWN, 0);

  BFD_LOCK;
  add_tail_list(&bfd_global.wait_list, &tmp_list);
  BFD_UNLOCK;
}

static event bfd_pickup_event = { .hook = bfd_pickup_requests };

static void
bfd_take_requests(struct bfd_proto *p)
{
  node *n, *nn;
  BFD_LOCK;
  WALK_LIST_DELSAFE(n, nn, bfd_global.wait_list)
    bfd_add_request(p, SKIP_BACK(struct bfd_request, n, n));
  BFD_UNLOCK;
}

static void
bfd_drop_requests(struct bfd_proto *p)
{
  node *n;
  BFD_LOCK;
  HASH_WALK_DELSAFE(p->session_hash_id, next_id, s)
  {
    WALK_LIST_FIRST(n, s->request_list)
    {
      struct bfd_request *req = SKIP_BACK(struct bfd_request, n, n);
      rem_node(&req->n);
      add_tail(&bfd_global.pickup_list, &req->n);
      req->session = NULL;
    }

    ev_send(&global_event_list, &bfd_pickup_event);

    bfd_remove_session_locked(p, s);
  }
  HASH_WALK_END;
  BFD_UNLOCK;
}

static struct resclass bfd_request_class;

struct bfd_request *
bfd_request_session(pool *p, ip_addr addr, ip_addr local,
		    struct iface *iface, struct iface *vrf,
		    void (*hook)(struct bfd_request *), void *data,
		    struct birdloop *target,
		    const struct bfd_options *opts)
{
  struct bfd_request *req = ralloc(p, &bfd_request_class);

  req->addr = addr;
  req->local = local;
  req->iface = iface;
  req->vrf = vrf;

  if (opts)
    req->opts = *opts;

  ASSERT_DIE(target || !hook);
  req->hook = hook;
  req->data = data;
  req->target = target;

  req->session = NULL;

  BFD_LOCK;
  bfd_global.pickup_reload++;
  add_tail(&bfd_global.pickup_list, &req->n);
  ev_send(&global_event_list, &bfd_pickup_event);
  DBG("New BFD request enlisted.\n");
  BFD_UNLOCK;

  return req;
}

void
bfd_update_request(struct bfd_request *req, const struct bfd_options *opts)
{
  struct bfd_session *s = req->session;

  if (!memcmp(opts, &req->opts, sizeof(const struct bfd_options)))
    return;

  req->opts = *opts;

  if (s)
    bfd_reconfigure_session(s->ifa->bfd, s);
}

static void
bfd_request_free(resource *r)
{
  struct bfd_request *req = (struct bfd_request *) r;

  BFD_LOCK;
  rem_node(&req->n);
  BFD_UNLOCK;

  ev_send(&global_event_list, &bfd_pickup_event);
}

static void
bfd_request_dump(resource *r, unsigned indent UNUSED)
{
  struct bfd_request *req = (struct bfd_request *) r;

  debug("(code %p, data %p)\n", req->hook, req->data);
}

static struct resclass bfd_request_class = {
  "BFD request",
  sizeof(struct bfd_request),
  bfd_request_free,
  bfd_request_dump,
  NULL,
  NULL
};


/*
 *	BFD neighbors
 */

static void
bfd_neigh_notify(struct neighbor *nb)
{
  struct bfd_proto *p = (struct bfd_proto *) nb->proto;
  struct bfd_neighbor *n = nb->data;

  if (!n)
    return;

  if ((nb->scope > 0) && !n->req)
  {
    ip_addr local = ipa_nonzero(n->local) ? n->local : nb->ifa->ip;
    n->req = bfd_request_session(p->p.pool, n->addr, local, nb->iface, p->p.vrf, NULL, NULL, NULL, NULL);
  }

  if ((nb->scope <= 0) && n->req)
  {
    rfree(n->req);
    n->req = NULL;
  }
}

static void
bfd_start_neighbor(struct bfd_proto *p, struct bfd_neighbor *n)
{
  n->active = 1;

  if (n->multihop)
  {
    n->req = bfd_request_session(p->p.pool, n->addr, n->local, NULL, p->p.vrf, NULL, NULL, NULL, NULL);
    return;
  }

  struct neighbor *nb = neigh_find(&p->p, n->addr, n->iface, NEF_STICKY);
  if (!nb)
  {
    log(L_ERR "%s: Invalid remote address %I%J", p->p.name, n->addr, n->iface);
    return;
  }

  if (nb->data)
  {
    log(L_ERR "%s: Duplicate neighbor %I", p->p.name, n->addr);
    return;
  }

  n->neigh = nb;
  nb->data = n;

  if (nb->scope > 0)
    bfd_neigh_notify(nb);
  else
    TRACE(D_EVENTS, "Waiting for %I%J to become my neighbor", n->addr, n->iface);
}

static void
bfd_stop_neighbor(struct bfd_proto *p UNUSED, struct bfd_neighbor *n)
{
  if (n->neigh)
    n->neigh->data = NULL;
  n->neigh = NULL;

  rfree(n->req);
  n->req = NULL;
}

static inline int
bfd_same_neighbor(struct bfd_neighbor *x, struct bfd_neighbor *y)
{
  return ipa_equal(x->addr, y->addr) && ipa_equal(x->local, y->local) &&
    (x->iface == y->iface) && (x->multihop == y->multihop);
}

static void
bfd_reconfigure_neighbors(struct bfd_proto *p, struct bfd_config *new)
{
  struct bfd_config *old = (struct bfd_config *) (p->p.cf);
  struct bfd_neighbor *on, *nn;

  WALK_LIST(on, old->neigh_list)
  {
    WALK_LIST(nn, new->neigh_list)
      if (bfd_same_neighbor(nn, on))
      {
	nn->neigh = on->neigh;
	if (nn->neigh)
	  nn->neigh->data = nn;

	nn->req = on->req;
	nn->active = 1;
	goto next;
      }

    bfd_stop_neighbor(p, on);
  next:;
  }

  WALK_LIST(nn, new->neigh_list)
    if (!nn->active)
      bfd_start_neighbor(p, nn);
}


/*
 *	BFD notify socket
 */

/* This core notify code should be replaced after main loop transition to birdloop */

static void
bfd_notify_hook(void *data)
{
  struct bfd_proto *p = data;
  struct bfd_session *s;
  list tmp_list;
  u8 state, diag;
  node *n, *nn;

  bfd_lock_sessions(p);
  init_list(&tmp_list);
  add_tail_list(&tmp_list, &p->notify_list);
  init_list(&p->notify_list);
  bfd_unlock_sessions(p);

  WALK_LIST_FIRST(s, tmp_list)
  {
    bfd_lock_sessions(p);
    rem_node(&s->n);
    state = s->loc_state;
    diag = s->loc_diag;
    bfd_unlock_sessions(p);

    WALK_LIST_DELSAFE(n, nn, s->request_list)
      bfd_request_notify(SKIP_BACK(struct bfd_request, n, n), state, diag);

    /* Remove the session if all requests were removed in notify hooks */
    if (EMPTY_LIST(s->request_list))
      bfd_remove_session(p, s);
  }
}

/*
 *	BFD protocol glue
 */

static struct proto *
bfd_init(struct proto_config *c)
{
  struct proto *p = proto_new(c);

  p->iface_sub.neigh_notify = bfd_neigh_notify;

  return p;
}

static int
bfd_start(struct proto *P)
{
  struct bfd_proto *p = (struct bfd_proto *) P;
  struct bfd_config *cf = (struct bfd_config *) (P->cf);

  pthread_spin_init(&p->lock, PTHREAD_PROCESS_PRIVATE);

  p->tpool = rp_new(P->pool, "BFD loop pool");

  p->session_slab = sl_new(P->pool, sizeof(struct bfd_session));
  HASH_INIT(p->session_hash_id, P->pool, 8);
  HASH_INIT(p->session_hash_ip, P->pool, 8);

  init_list(&p->iface_list);

  init_list(&p->notify_list);
  p->notify_event = (event) {
    .hook = bfd_notify_hook,
    .data = p,
  };

  add_tail(&bfd_global.proto_list, &p->bfd_node);

  if (!cf->strict_bind)
  {
    if (cf->accept_ipv4 && cf->accept_direct)
      p->rx4_1 = bfd_open_rx_sk(p, 0, SK_IPV4);

    if (cf->accept_ipv4 && cf->accept_multihop)
      p->rx4_m = bfd_open_rx_sk(p, 1, SK_IPV4);

    if (cf->accept_ipv6 && cf->accept_direct)
      p->rx6_1 = bfd_open_rx_sk(p, 0, SK_IPV6);

    if (cf->accept_ipv6 && cf->accept_multihop)
      p->rx6_m = bfd_open_rx_sk(p, 1, SK_IPV6);
  }

  bfd_take_requests(p);

  struct bfd_neighbor *n;
  WALK_LIST(n, cf->neigh_list)
    bfd_start_neighbor(p, n);

  return PS_UP;
}

static int
bfd_shutdown(struct proto *P)
{
  struct bfd_proto *p = (struct bfd_proto *) P;
  struct bfd_config *cf = (struct bfd_config *) (p->p.cf);

  rem_node(&p->bfd_node);

  struct bfd_neighbor *bn;
  WALK_LIST(bn, cf->neigh_list)
    bfd_stop_neighbor(p, bn);

  bfd_drop_requests(p);

  return PS_DOWN;
}

static int
bfd_reconfigure(struct proto *P, struct proto_config *c)
{
  struct bfd_proto *p = (struct bfd_proto *) P;
  struct bfd_config *old = (struct bfd_config *) (P->cf);
  struct bfd_config *new = (struct bfd_config *) c;
  struct bfd_iface *ifa;

  /* TODO: Improve accept reconfiguration */
  if ((new->accept_ipv4 != old->accept_ipv4) ||
      (new->accept_ipv6 != old->accept_ipv6) ||
      (new->accept_direct != old->accept_direct) ||
      (new->accept_multihop != old->accept_multihop) ||
      (new->strict_bind != old->strict_bind))
    return 0;

  birdloop_mask_wakeups(p->p.loop);

  WALK_LIST(ifa, p->iface_list)
    bfd_reconfigure_iface(p, ifa, new);

  HASH_WALK(p->session_hash_id, next_id, s)
  {
    if (s->ifa->changed)
      bfd_reconfigure_session(p, s);
  }
  HASH_WALK_END;

  bfd_reconfigure_neighbors(p, new);

  birdloop_unmask_wakeups(p->p.loop);

  return 1;
}

static void
bfd_copy_config(struct proto_config *dest, struct proto_config *src UNUSED)
{
  struct bfd_config *d = (struct bfd_config *) dest;
  // struct bfd_config *s = (struct bfd_config *) src;

  /* We clean up patt_list and neigh_list, neighbors and ifaces are non-sharable */
  init_list(&d->patt_list);
  init_list(&d->neigh_list);
}

void
bfd_show_sessions(struct proto *P)
{
  byte tbuf[TM_DATETIME_BUFFER_SIZE];
  struct bfd_proto *p = (struct bfd_proto *) P;
  uint state, diag UNUSED;
  btime tx_int, timeout;
  const char *ifname;

  if (p->p.proto_state != PS_UP)
  {
    cli_msg(-1020, "%s: is not up", p->p.name);
    return;
  }

  cli_msg(-1020, "%s:", p->p.name);
  cli_msg(-1020, "%-25s %-10s %-10s %-12s  %8s %8s",
	  "IP address", "Interface", "State", "Since", "Interval", "Timeout");


  HASH_WALK(p->session_hash_id, next_id, s)
  {
    /* FIXME: this is thread-unsafe, but perhaps harmless */
    state = s->loc_state;
    diag = s->loc_diag;
    ifname = (s->ifa && s->ifa->iface) ? s->ifa->iface->name : "---";
    tx_int = s->last_tx ? MAX(s->des_min_tx_int, s->rem_min_rx_int) : 0;
    timeout = (btime) MAX(s->req_min_rx_int, s->rem_min_tx_int) * s->rem_detect_mult;

    state = (state < 4) ? state : 0;
    tm_format_time(tbuf, &config->tf_proto, s->last_state_change);

    cli_msg(-1020, "%-25I %-10s %-10s %-12s  %7t  %7t",
	    s->addr, ifname, bfd_state_names[state], tbuf, tx_int, timeout);
  }
  HASH_WALK_END;
}


struct protocol proto_bfd = {
  .name =		"BFD",
  .template =		"bfd%d",
  .proto_size =		sizeof(struct bfd_proto),
  .config_size =	sizeof(struct bfd_config),
  .init =		bfd_init,
  .start =		bfd_start,
  .shutdown =		bfd_shutdown,
  .reconfigure =	bfd_reconfigure,
  .copy_config =	bfd_copy_config,
};

void
bfd_build(void)
{
  proto_build(&proto_bfd);

  bfd_global.lock = DOMAIN_NEW(rtable, "BFD Global");
  init_list(&bfd_global.wait_list);
  init_list(&bfd_global.pickup_list);
  init_list(&bfd_global.proto_list);
}