summaryrefslogtreecommitdiffstats
path: root/clientlog.c
blob: adf0c59ba223c1603bfdd925ec62969dad727e1a (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
/*
  chronyd/chronyc - Programs for keeping computer clocks accurate.

 **********************************************************************
 * Copyright (C) Richard P. Curnow  1997-2003
 * Copyright (C) Miroslav Lichvar  2009, 2015-2017, 2021
 * 
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of version 2 of the GNU General Public License as
 * published by the Free Software Foundation.
 * 
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 * 
 * You should have received a copy of the GNU General Public License along
 * with this program; if not, write to the Free Software Foundation, Inc.,
 * 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
 * 
 **********************************************************************

  =======================================================================

  This module keeps a count of the number of successful accesses by
  clients, and the times of the last accesses.

  This can be used for status reporting, and (in the case of a
  server), if it needs to know which clients have made use of its data
  recently.

  */

#include "config.h"

#include "sysincl.h"

#include "array.h"
#include "clientlog.h"
#include "conf.h"
#include "local.h"
#include "memory.h"
#include "ntp.h"
#include "reports.h"
#include "util.h"
#include "logging.h"

#define MAX_SERVICES 3

typedef struct {
  IPAddr ip_addr;
  uint32_t last_hit[MAX_SERVICES];
  uint32_t hits[MAX_SERVICES];
  uint16_t drops[MAX_SERVICES];
  uint16_t tokens[MAX_SERVICES];
  int8_t rate[MAX_SERVICES];
  int8_t ntp_timeout_rate;
  uint8_t drop_flags;
} Record;

/* Hash table of records, there is a fixed number of records per slot */
static ARR_Instance records;

#define SLOT_BITS 4

/* Number of records in one slot of the hash table */
#define SLOT_SIZE (1U << SLOT_BITS)

/* Minimum number of slots */
#define MIN_SLOTS 1

/* Maximum number of slots, this is a hard limit */
#define MAX_SLOTS (1U << (24 - SLOT_BITS))

/* Number of slots in the hash table */
static unsigned int slots;

/* Maximum number of slots given memory allocation limit */
static unsigned int max_slots;

/* Times of last hits are saved as 32-bit fixed point values */
#define TS_FRAC 4
#define INVALID_TS 0

/* Static offset included in conversion to the fixed-point timestamps to
   randomise their alignment */
static uint32_t ts_offset;

/* Request rates are saved in the record as 8-bit scaled log2 values */
#define RATE_SCALE 4
#define MIN_RATE (-14 * RATE_SCALE)
#define INVALID_RATE -128

/* Response rates are controlled by token buckets.  The capacity and
   number of tokens spent on response are determined from configured
   minimum inverval between responses (in log2) and burst length. */

#define MIN_LIMIT_INTERVAL (-15 - TS_FRAC)
#define MAX_LIMIT_INTERVAL 12
#define MIN_LIMIT_BURST 1
#define MAX_LIMIT_BURST 255

static uint16_t max_tokens[MAX_SERVICES];
static uint16_t tokens_per_hit[MAX_SERVICES];

/* Reduction of token rates to avoid overflow of 16-bit counters.  Negative
   shift is used for coarse limiting with intervals shorter than -TS_FRAC. */
static int token_shift[MAX_SERVICES];

/* Rates at which responses are randomly allowed (in log2) when the
   buckets don't have enough tokens.  This is necessary in order to
   prevent an attacker sending requests with spoofed source address
   from blocking responses to the address completely. */

#define MIN_LEAK_RATE 1
#define MAX_LEAK_RATE 4

static int leak_rate[MAX_SERVICES];

/* Limit intervals in log2 */
static int limit_interval[MAX_SERVICES];

/* Flag indicating whether facility is turned on or not */
static int active;

/* RX and TX timestamp saved for clients using interleaved mode */
typedef struct {
  uint64_t rx_ts;
  uint16_t flags;
  uint16_t slew_epoch;
  int32_t tx_ts_offset;
} NtpTimestamps;

/* Flags for NTP timestamps */
#define NTPTS_DISABLED 1
#define NTPTS_VALID_TX 2

/* RX->TX map using a circular buffer with ordered timestamps */
typedef struct {
  ARR_Instance timestamps;
  uint32_t first;
  uint32_t size;
  uint32_t max_size;
  uint32_t cached_index;
  uint64_t cached_rx_ts;
  uint16_t slew_epoch;
  double slew_offset;
} NtpTimestampMap;

static NtpTimestampMap ntp_ts_map;

/* Maximum interval of NTP timestamps in future after a backward step */
#define NTPTS_FUTURE_LIMIT (1LL << 32) /* 1 second */

/* Maximum number of timestamps moved in the array to insert a new timestamp */
#define NTPTS_INSERT_LIMIT 64

/* Global statistics */
static uint32_t total_hits[MAX_SERVICES];
static uint32_t total_drops[MAX_SERVICES];
static uint32_t total_ntp_auth_hits;
static uint32_t total_ntp_interleaved_hits;
static uint32_t total_record_drops;

#define NSEC_PER_SEC 1000000000U

/* ================================================== */

static int expand_hashtable(void);
static void handle_slew(struct timespec *raw, struct timespec *cooked, double dfreq,
                        double doffset, LCL_ChangeType change_type, void *anything);

/* ================================================== */

static int
compare_ts(uint32_t x, uint32_t y)
{
  if (x == y)
    return 0;
  if (y == INVALID_TS)
    return 1;
  return (int32_t)(x - y) > 0 ? 1 : -1;
}

/* ================================================== */

static int
compare_total_hits(Record *x, Record *y)
{
  uint32_t x_hits, y_hits;
  int i;

  for (i = 0, x_hits = y_hits = 0; i < MAX_SERVICES; i++) {
    x_hits += x->hits[i];
    y_hits += y->hits[i];
  }

  return x_hits > y_hits ? 1 : -1;
}

/* ================================================== */

static Record *
get_record(IPAddr *ip)
{
  uint32_t last_hit = 0, oldest_hit = 0;
  Record *record, *oldest_record;
  unsigned int first, i, j;

  if (!active || (ip->family != IPADDR_INET4 && ip->family != IPADDR_INET6))
    return NULL;

  while (1) {
    /* Get index of the first record in the slot */
    first = UTI_IPToHash(ip) % slots * SLOT_SIZE;

    for (i = 0, oldest_record = NULL; i < SLOT_SIZE; i++) {
      record = ARR_GetElement(records, first + i);

      if (!UTI_CompareIPs(ip, &record->ip_addr, NULL))
        return record;

      if (record->ip_addr.family == IPADDR_UNSPEC)
        break;

      for (j = 0; j < MAX_SERVICES; j++) {
        if (j == 0 || compare_ts(last_hit, record->last_hit[j]) < 0)
          last_hit = record->last_hit[j];
      }

      if (!oldest_record || compare_ts(oldest_hit, last_hit) > 0 ||
          (oldest_hit == last_hit && compare_total_hits(oldest_record, record) > 0)) {
        oldest_record = record;
        oldest_hit = last_hit;
      }
    }

    /* If the slot still has an empty record, use it */
    if (record->ip_addr.family == IPADDR_UNSPEC)
      break;

    /* Resize the table if possible and try again as the new slot may
       have some empty records */
    if (expand_hashtable())
      continue;

    /* There is no other option, replace the oldest record */
    record = oldest_record;
    total_record_drops++;
    break;
  }

  record->ip_addr = *ip;
  for (i = 0; i < MAX_SERVICES; i++)
    record->last_hit[i] = INVALID_TS;
  for (i = 0; i < MAX_SERVICES; i++)
    record->hits[i] = 0;
  for (i = 0; i < MAX_SERVICES; i++)
    record->drops[i] = 0;
  for (i = 0; i < MAX_SERVICES; i++)
    record->tokens[i] = max_tokens[i];
  for (i = 0; i < MAX_SERVICES; i++)
    record->rate[i] = INVALID_RATE;
  record->ntp_timeout_rate = INVALID_RATE;
  record->drop_flags = 0;

  return record;
}

/* ================================================== */

static int
expand_hashtable(void)
{
  ARR_Instance old_records;
  Record *old_record, *new_record;
  unsigned int i;

  old_records = records;

  if (2 * slots > max_slots)
    return 0;

  records = ARR_CreateInstance(sizeof (Record));

  slots = MAX(MIN_SLOTS, 2 * slots);
  assert(slots <= max_slots);

  ARR_SetSize(records, slots * SLOT_SIZE);

  /* Mark all new records as empty */
  for (i = 0; i < slots * SLOT_SIZE; i++) {
    new_record = ARR_GetElement(records, i);
    new_record->ip_addr.family = IPADDR_UNSPEC;
  }

  if (!old_records)
    return 1;

  /* Copy old records to the new hash table */
  for (i = 0; i < ARR_GetSize(old_records); i++) {
    old_record = ARR_GetElement(old_records, i);
    if (old_record->ip_addr.family == IPADDR_UNSPEC)
      continue;

    new_record = get_record(&old_record->ip_addr);

    assert(new_record);
    *new_record = *old_record;
  }

  ARR_DestroyInstance(old_records);

  return 1;
}

/* ================================================== */

static void
set_bucket_params(int interval, int burst, uint16_t *max_tokens,
                  uint16_t *tokens_per_packet, int *token_shift)
{
  interval = CLAMP(MIN_LIMIT_INTERVAL, interval, MAX_LIMIT_INTERVAL);
  burst = CLAMP(MIN_LIMIT_BURST, burst, MAX_LIMIT_BURST);

  if (interval >= -TS_FRAC) {
    /* Find the smallest shift with which the maximum number fits in 16 bits */
    for (*token_shift = 0; *token_shift < interval + TS_FRAC; (*token_shift)++) {
      if (burst << (TS_FRAC + interval - *token_shift) < 1U << 16)
        break;
    }
  } else {
    /* Coarse rate limiting */
    *token_shift = interval + TS_FRAC;
    *tokens_per_packet = 1;
    burst = MAX(1U << -*token_shift, burst);
  }

  *tokens_per_packet = 1U << (TS_FRAC + interval - *token_shift);
  *max_tokens = *tokens_per_packet * burst;

  DEBUG_LOG("Tokens max %d packet %d shift %d",
            *max_tokens, *tokens_per_packet, *token_shift);
}

/* ================================================== */

void
CLG_Initialise(void)
{
  int i, interval, burst, lrate, slots2;

  for (i = 0; i < MAX_SERVICES; i++) {
    max_tokens[i] = 0;
    tokens_per_hit[i] = 0;
    token_shift[i] = 0;
    leak_rate[i] = 0;
    limit_interval[i] = MIN_LIMIT_INTERVAL;

    switch (i) {
      case CLG_NTP:
        if (!CNF_GetNTPRateLimit(&interval, &burst, &lrate))
          continue;
        break;
      case CLG_NTSKE:
        if (!CNF_GetNtsRateLimit(&interval, &burst, &lrate))
          continue;
        break;
      case CLG_CMDMON:
        if (!CNF_GetCommandRateLimit(&interval, &burst, &lrate))
          continue;
        break;
      default:
        assert(0);
    }

    set_bucket_params(interval, burst, &max_tokens[i], &tokens_per_hit[i], &token_shift[i]);
    leak_rate[i] = CLAMP(MIN_LEAK_RATE, lrate, MAX_LEAK_RATE);
    limit_interval[i] = CLAMP(MIN_LIMIT_INTERVAL, interval, MAX_LIMIT_INTERVAL);
  }

  active = !CNF_GetNoClientLog();
  if (!active) {
    for (i = 0; i < MAX_SERVICES; i++) {
      if (leak_rate[i] != 0)
        LOG_FATAL("Rate limiting cannot be enabled with noclientlog");
    }
    return;
  }

  /* Calculate the maximum number of slots that can be allocated in the
     configured memory limit.  Take into account expanding of the hash
     table where two copies exist at the same time. */
  max_slots = CNF_GetClientLogLimit() /
              ((sizeof (Record) + sizeof (NtpTimestamps)) * SLOT_SIZE * 3 / 2);
  max_slots = CLAMP(MIN_SLOTS, max_slots, MAX_SLOTS);
  for (slots2 = 0; 1U << (slots2 + 1) <= max_slots; slots2++)
    ;

  DEBUG_LOG("Max records %u", 1U << (slots2 + SLOT_BITS));

  slots = 0;
  records = NULL;

  expand_hashtable();

  UTI_GetRandomBytes(&ts_offset, sizeof (ts_offset));
  ts_offset %= NSEC_PER_SEC / (1U << TS_FRAC);

  ntp_ts_map.timestamps = NULL;
  ntp_ts_map.first = 0;
  ntp_ts_map.size = 0;
  ntp_ts_map.max_size = 1U << (slots2 + SLOT_BITS);
  ntp_ts_map.cached_index = 0;
  ntp_ts_map.cached_rx_ts = 0ULL;
  ntp_ts_map.slew_epoch = 0;
  ntp_ts_map.slew_offset = 0.0;

  LCL_AddParameterChangeHandler(handle_slew, NULL);
}

/* ================================================== */

void
CLG_Finalise(void)
{
  if (!active)
    return;

  ARR_DestroyInstance(records);
  if (ntp_ts_map.timestamps)
    ARR_DestroyInstance(ntp_ts_map.timestamps);

  LCL_RemoveParameterChangeHandler(handle_slew, NULL);
}

/* ================================================== */

static uint32_t
get_ts_from_timespec(struct timespec *ts)
{
  uint32_t sec = ts->tv_sec, nsec = ts->tv_nsec;

  nsec += ts_offset;
  if (nsec >= NSEC_PER_SEC) {
    nsec -= NSEC_PER_SEC;
    sec++;
  }

  /* This is fast and accurate enough */
  return sec << TS_FRAC | (140740U * (nsec >> 15)) >> (32 - TS_FRAC);
}

/* ================================================== */

static void
update_record(CLG_Service service, Record *record, struct timespec *now)
{
  uint32_t interval, now_ts, prev_hit, tokens;
  int interval2, tshift, mtokens;
  int8_t *rate;

  now_ts = get_ts_from_timespec(now);

  prev_hit = record->last_hit[service];
  record->last_hit[service] = now_ts;
  record->hits[service]++;

  interval = now_ts - prev_hit;

  if (prev_hit == INVALID_TS || (int32_t)interval < 0)
    return;

  tshift = token_shift[service];
  mtokens = max_tokens[service];

  if (tshift >= 0)
    tokens = (now_ts >> tshift) - (prev_hit >> tshift);
  else if (now_ts - prev_hit > mtokens)
    tokens = mtokens;
  else
    tokens = (now_ts - prev_hit) << -tshift;
  record->tokens[service] = MIN(record->tokens[service] + tokens, mtokens);

  /* Convert the interval to scaled and rounded log2 */
  if (interval) {
    interval += interval >> 1;
    for (interval2 = -RATE_SCALE * TS_FRAC; interval2 < -MIN_RATE;
         interval2 += RATE_SCALE) {
      if (interval <= 1)
        break;
      interval >>= 1;
    }
  } else {
    interval2 = -RATE_SCALE * (TS_FRAC + 1);
  }

  /* For the NTP service, update one of the two rates depending on whether
     the previous request of the client had a reply or it timed out */
  rate = service == CLG_NTP && record->drop_flags & (1U << service) ?
           &record->ntp_timeout_rate : &record->rate[service];

  /* Update the rate in a rough approximation of exponential moving average */
  if (*rate == INVALID_RATE) {
    *rate = -interval2;
  } else {
    if (*rate < -interval2) {
      (*rate)++;
    } else if (*rate > -interval2) {
      if (*rate > RATE_SCALE * 5 / 2 - interval2)
        *rate = RATE_SCALE * 5 / 2 - interval2;
      else
        *rate = (*rate - interval2 - 1) / 2;
    }
  }
}

/* ================================================== */

static int
get_index(Record *record)
{
  return record - (Record *)ARR_GetElements(records);
}

/* ================================================== */

int
CLG_GetClientIndex(IPAddr *client)
{
  Record *record;

  record = get_record(client);
  if (record == NULL)
    return -1;

  return get_index(record);
}

/* ================================================== */

static void
check_service_number(CLG_Service service)
{
  assert(service >= 0 && service <= MAX_SERVICES);
}

/* ================================================== */

int
CLG_LogServiceAccess(CLG_Service service, IPAddr *client, struct timespec *now)
{
  Record *record;

  check_service_number(service);

  total_hits[service]++;

  record = get_record(client);
  if (record == NULL)
    return -1;

  update_record(service, record, now);

  DEBUG_LOG("service %d hits %"PRIu32" rate %d trate %d tokens %d",
            (int)service, record->hits[service], record->rate[service],
            service == CLG_NTP ? record->ntp_timeout_rate : INVALID_RATE,
            record->tokens[service]);

  return get_index(record);
}

/* ================================================== */

static int
limit_response_random(int leak_rate)
{
  static uint32_t rnd;
  static int bits_left = 0;
  int r;

  if (bits_left < leak_rate) {
    UTI_GetRandomBytes(&rnd, sizeof (rnd));
    bits_left = 8 * sizeof (rnd);
  }

  /* Return zero on average once per 2^leak_rate */
  r = rnd % (1U << leak_rate) ? 1 : 0;
  rnd >>= leak_rate;
  bits_left -= leak_rate;

  return r;
}

/* ================================================== */

int
CLG_LimitServiceRate(CLG_Service service, int index)
{
  Record *record;
  int drop;

  check_service_number(service);

  if (tokens_per_hit[service] == 0)
    return 0;

  record = ARR_GetElement(records, index);
  record->drop_flags &= ~(1U << service);

  if (record->tokens[service] >= tokens_per_hit[service]) {
    record->tokens[service] -= tokens_per_hit[service];
    return 0;
  }

  drop = limit_response_random(leak_rate[service]);

  /* Poorly implemented NTP clients can send requests at a higher rate
     when they are not getting replies.  If the request rate seems to be more
     than twice as much as when replies are sent, give up on rate limiting to
     reduce the amount of traffic.  Invert the sense of the leak to respond to
     most of the requests, but still keep the estimated rate updated. */
  if (service == CLG_NTP && record->ntp_timeout_rate != INVALID_RATE &&
      record->ntp_timeout_rate > record->rate[service] + RATE_SCALE)
    drop = !drop;

  if (!drop) {
    record->tokens[service] = 0;
    return 0;
  }

  record->drop_flags |= 1U << service;
  record->drops[service]++;
  total_drops[service]++;

  return 1;
}

/* ================================================== */

void
CLG_LogAuthNtpRequest(void)
{
  total_ntp_auth_hits++;
}

/* ================================================== */

int
CLG_GetNtpMinPoll(void)
{
  return limit_interval[CLG_NTP];
}

/* ================================================== */

static NtpTimestamps *
get_ntp_tss(uint32_t index)
{
  return ARR_GetElement(ntp_ts_map.timestamps,
                        (ntp_ts_map.first + index) & (ntp_ts_map.max_size - 1));
}

/* ================================================== */

static int
find_ntp_rx_ts(uint64_t rx_ts, uint32_t *index)
{
  uint64_t rx_x, rx_lo, rx_hi, step;
  uint32_t i, x, lo, hi;

  if (ntp_ts_map.cached_rx_ts == rx_ts && rx_ts != 0ULL) {
    *index = ntp_ts_map.cached_index;
    return 1;
  }

  if (ntp_ts_map.size == 0) {
    *index = 0;
    return 0;
  }

  lo = 0;
  hi = ntp_ts_map.size - 1;
  rx_lo = get_ntp_tss(lo)->rx_ts;
  rx_hi = get_ntp_tss(hi)->rx_ts;

  /* Check for ts < lo before ts > hi to trim timestamps from "future" later
     if both conditions are true to not break the order of the endpoints.
     Compare timestamps by their difference to allow adjacent NTP eras. */
  if ((int64_t)(rx_ts - rx_lo) < 0) {
    *index = 0;
    return 0;
  } else if ((int64_t)(rx_ts - rx_hi) > 0) {
    *index = ntp_ts_map.size;
    return 0;
  }

  /* Perform a combined linear interpolation and binary search */

  for (i = 0; ; i++) {
    if (rx_ts == rx_hi) {
      *index = ntp_ts_map.cached_index = hi;
      ntp_ts_map.cached_rx_ts = rx_ts;
      return 1;
    } else if (rx_ts == rx_lo) {
      *index = ntp_ts_map.cached_index = lo;
      ntp_ts_map.cached_rx_ts = rx_ts;
      return 1;
    } else if (lo + 1 == hi) {
      *index = hi;
      return 0;
    }

    if (hi - lo > 3 && i % 2 == 0) {
      step = (rx_hi - rx_lo) / (hi - lo);
      if (step == 0)
        step = 1;
      x = lo + (rx_ts - rx_lo) / step;
    } else {
      x = lo + (hi - lo) / 2;
    }

    if (x <= lo)
      x = lo + 1;
    else if (x >= hi)
      x = hi - 1;

    rx_x = get_ntp_tss(x)->rx_ts;

    if ((int64_t)(rx_x - rx_ts) <= 0) {
      lo = x;
      rx_lo = rx_x;
    } else {
      hi = x;
      rx_hi = rx_x;
    }
  }
}

/* ================================================== */

static uint64_t
ntp64_to_int64(NTP_int64 *ts)
{
  return (uint64_t)ntohl(ts->hi) << 32 | ntohl(ts->lo);
}

/* ================================================== */

static void
int64_to_ntp64(uint64_t ts, NTP_int64 *ntp_ts)
{
  ntp_ts->hi = htonl(ts >> 32);
  ntp_ts->lo = htonl(ts);
}

/* ================================================== */

static uint32_t
push_ntp_tss(uint32_t index)
{
  if (ntp_ts_map.size < ntp_ts_map.max_size) {
    ntp_ts_map.size++;
  } else {
    ntp_ts_map.first = (ntp_ts_map.first + 1) % (ntp_ts_map.max_size);
    if (index > 0)
      index--;
  }

  return index;
}

/* ================================================== */

static void
set_ntp_tx_offset(NtpTimestamps *tss, NTP_int64 *rx_ts, struct timespec *tx_ts)
{
  struct timespec ts;

  if (!tx_ts) {
    tss->flags &= ~NTPTS_VALID_TX;
    return;
  }

  UTI_Ntp64ToTimespec(rx_ts, &ts);
  UTI_DiffTimespecs(&ts, tx_ts, &ts);

  if (ts.tv_sec < -2 || ts.tv_sec > 1) {
    tss->flags &= ~NTPTS_VALID_TX;
    return;
  }

  tss->tx_ts_offset = (int32_t)ts.tv_nsec + (int32_t)ts.tv_sec * (int32_t)NSEC_PER_SEC;
  tss->flags |= NTPTS_VALID_TX;
}

/* ================================================== */

static void
get_ntp_tx(NtpTimestamps *tss, struct timespec *tx_ts)
{
  int32_t offset = tss->tx_ts_offset;
  NTP_int64 ntp_ts;

  if (tss->flags & NTPTS_VALID_TX) {
    int64_to_ntp64(tss->rx_ts, &ntp_ts);
    UTI_Ntp64ToTimespec(&ntp_ts, tx_ts);
    if (offset >= (int32_t)NSEC_PER_SEC) {
      offset -= NSEC_PER_SEC;
      tx_ts->tv_sec++;
    }
    tx_ts->tv_nsec += offset;
    UTI_NormaliseTimespec(tx_ts);
  } else {
    UTI_ZeroTimespec(tx_ts);
  }
}

/* ================================================== */

void
CLG_SaveNtpTimestamps(NTP_int64 *rx_ts, struct timespec *tx_ts)
{
  NtpTimestamps *tss;
  uint32_t i, index;
  uint64_t rx;

  if (!active)
    return;

  /* Allocate the array on first use */
  if (!ntp_ts_map.timestamps) {
    ntp_ts_map.timestamps = ARR_CreateInstance(sizeof (NtpTimestamps));
    ARR_SetSize(ntp_ts_map.timestamps, ntp_ts_map.max_size);
  }

  rx = ntp64_to_int64(rx_ts);

  if (rx == 0ULL)
    return;

  /* Disable the RX timestamp if it already exists to avoid responding
     with a wrong TX timestamp */
  if (find_ntp_rx_ts(rx, &index)) {
    get_ntp_tss(index)->flags |= NTPTS_DISABLED;
    return;
  }

  assert(index <= ntp_ts_map.size);

  if (index == ntp_ts_map.size) {
    /* Increase the size or drop the oldest timestamp to make room for
       the new timestamp */
    index = push_ntp_tss(index);
  } else {
    /* Trim timestamps in distant future after backward step */
    while (index < ntp_ts_map.size &&
           get_ntp_tss(ntp_ts_map.size - 1)->rx_ts - rx > NTPTS_FUTURE_LIMIT)
      ntp_ts_map.size--;

    /* Insert the timestamp if it is close to the latest timestamp.
       Otherwise, replace the closest older or the oldest timestamp. */
    if (index + NTPTS_INSERT_LIMIT >= ntp_ts_map.size) {
      index = push_ntp_tss(index);
      for (i = ntp_ts_map.size - 1; i > index; i--)
        *get_ntp_tss(i) = *get_ntp_tss(i - 1);
    } else {
      if (index > 0)
        index--;
    }
  }

  ntp_ts_map.cached_index = index;
  ntp_ts_map.cached_rx_ts = rx;

  tss = get_ntp_tss(index);
  tss->rx_ts = rx;
  tss->flags = 0;
  tss->slew_epoch = ntp_ts_map.slew_epoch;
  set_ntp_tx_offset(tss, rx_ts, tx_ts);

  DEBUG_LOG("Saved RX+TX index=%"PRIu32" first=%"PRIu32" size=%"PRIu32,
            index, ntp_ts_map.first, ntp_ts_map.size);
}

/* ================================================== */

static void
handle_slew(struct timespec *raw, struct timespec *cooked, double dfreq,
            double doffset, LCL_ChangeType change_type, void *anything)
{
  /* Drop all timestamps on unknown step */
  if (change_type == LCL_ChangeUnknownStep) {
    ntp_ts_map.size = 0;
    ntp_ts_map.cached_rx_ts = 0ULL;
  }

  ntp_ts_map.slew_epoch++;
  ntp_ts_map.slew_offset = doffset;
}

/* ================================================== */

void
CLG_UndoNtpTxTimestampSlew(NTP_int64 *rx_ts, struct timespec *tx_ts)
{
  uint32_t index;

  if (!ntp_ts_map.timestamps)
    return;

  if (!find_ntp_rx_ts(ntp64_to_int64(rx_ts), &index))
    return;

  /* If the RX timestamp was captured before the last correction of the clock,
     remove the adjustment from the TX timestamp */
  if ((uint16_t)(get_ntp_tss(index)->slew_epoch + 1U) == ntp_ts_map.slew_epoch)
    UTI_AddDoubleToTimespec(tx_ts, ntp_ts_map.slew_offset, tx_ts);
}

/* ================================================== */

void
CLG_UpdateNtpTxTimestamp(NTP_int64 *rx_ts, struct timespec *tx_ts)
{
  uint32_t index;

  if (!ntp_ts_map.timestamps)
    return;

  if (!find_ntp_rx_ts(ntp64_to_int64(rx_ts), &index))
    return;

  set_ntp_tx_offset(get_ntp_tss(index), rx_ts, tx_ts);
}

/* ================================================== */

int
CLG_GetNtpTxTimestamp(NTP_int64 *rx_ts, struct timespec *tx_ts)
{
  NtpTimestamps *tss;
  uint32_t index;

  if (!ntp_ts_map.timestamps)
    return 0;

  if (!find_ntp_rx_ts(ntp64_to_int64(rx_ts), &index))
    return 0;

  tss = get_ntp_tss(index);

  if (tss->flags & NTPTS_DISABLED)
    return 0;

  get_ntp_tx(tss, tx_ts);

  return 1;
}

/* ================================================== */

void
CLG_DisableNtpTimestamps(NTP_int64 *rx_ts)
{
  uint32_t index;

  if (!ntp_ts_map.timestamps)
    return;

  if (find_ntp_rx_ts(ntp64_to_int64(rx_ts), &index))
    get_ntp_tss(index)->flags |= NTPTS_DISABLED;

  /* This assumes the function is called only to prevent multiple
     interleaved responses to the same timestamp */
  total_ntp_interleaved_hits++;
}

/* ================================================== */

int
CLG_GetNumberOfIndices(void)
{
  if (!active)
    return -1;

  return ARR_GetSize(records);
}

/* ================================================== */

static int get_interval(int rate)
{
  if (rate == INVALID_RATE)
    return 127;

  rate += rate > 0 ? RATE_SCALE / 2 : -RATE_SCALE / 2;

  return rate / -RATE_SCALE;
}

/* ================================================== */

static uint32_t get_last_ago(uint32_t x, uint32_t y)
{
  if (y == INVALID_TS || (int32_t)(x - y) < 0)
    return -1;

  return (x - y) >> TS_FRAC;
}

/* ================================================== */

int
CLG_GetClientAccessReportByIndex(int index, int reset, uint32_t min_hits,
                                 RPT_ClientAccessByIndex_Report *report, struct timespec *now)
{
  Record *record;
  uint32_t now_ts;
  int i, r;

  if (!active || index < 0 || index >= ARR_GetSize(records))
    return 0;

  record = ARR_GetElement(records, index);

  if (record->ip_addr.family == IPADDR_UNSPEC)
    return 0;

  if (min_hits == 0) {
    r = 1;
  } else {
    for (i = r = 0; i < MAX_SERVICES; i++) {
      if (record->hits[i] >= min_hits) {
        r = 1;
        break;
      }
    }
  }

  if (r) {
    now_ts = get_ts_from_timespec(now);

    report->ip_addr = record->ip_addr;
    report->ntp_hits = record->hits[CLG_NTP];
    report->nke_hits = record->hits[CLG_NTSKE];
    report->cmd_hits = record->hits[CLG_CMDMON];
    report->ntp_drops = record->drops[CLG_NTP];
    report->nke_drops = record->drops[CLG_NTSKE];
    report->cmd_drops = record->drops[CLG_CMDMON];
    report->ntp_interval = get_interval(record->rate[CLG_NTP]);
    report->nke_interval = get_interval(record->rate[CLG_NTSKE]);
    report->cmd_interval = get_interval(record->rate[CLG_CMDMON]);
    report->ntp_timeout_interval = get_interval(record->ntp_timeout_rate);
    report->last_ntp_hit_ago = get_last_ago(now_ts, record->last_hit[CLG_NTP]);
    report->last_nke_hit_ago = get_last_ago(now_ts, record->last_hit[CLG_NTSKE]);
    report->last_cmd_hit_ago = get_last_ago(now_ts, record->last_hit[CLG_CMDMON]);
  }

  if (reset) {
    for (i = 0; i < MAX_SERVICES; i++) {
      record->hits[i] = 0;
      record->drops[i] = 0;
    }
  }

  return r;
}

/* ================================================== */

void
CLG_GetServerStatsReport(RPT_ServerStatsReport *report)
{
  report->ntp_hits = total_hits[CLG_NTP];
  report->nke_hits = total_hits[CLG_NTSKE];
  report->cmd_hits = total_hits[CLG_CMDMON];
  report->ntp_drops = total_drops[CLG_NTP];
  report->nke_drops = total_drops[CLG_NTSKE];
  report->cmd_drops = total_drops[CLG_CMDMON];
  report->log_drops = total_record_drops;
  report->ntp_auth_hits = total_ntp_auth_hits;
  report->ntp_interleaved_hits = total_ntp_interleaved_hits;
  report->ntp_timestamps = ntp_ts_map.size;
  report->ntp_span_seconds = ntp_ts_map.size > 1 ?
                             (get_ntp_tss(ntp_ts_map.size - 1)->rx_ts -
                              get_ntp_tss(0)->rx_ts) >> 32 : 0;
}