summaryrefslogtreecommitdiff
path: root/ares_process.c
blob: 25eb24c0bc700d4775f2ee92fa7083d6727293a8 (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
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311

/* Copyright 1998 by the Massachusetts Institute of Technology.
 * Copyright (C) 2004-2012 by Daniel Stenberg
 *
 * Permission to use, copy, modify, and distribute this
 * software and its documentation for any purpose and without
 * fee is hereby granted, provided that the above copyright
 * notice appear in all copies and that both that copyright
 * notice and this permission notice appear in supporting
 * documentation, and that the name of M.I.T. not be used in
 * advertising or publicity pertaining to distribution of the
 * software without specific, written prior permission.
 * M.I.T. makes no representations about the suitability of
 * this software for any purpose.  It is provided "as is"
 * without express or implied warranty.
 */

#include "ares_setup.h"

#ifdef HAVE_SYS_SOCKET_H
#  include <sys/socket.h>
#endif
#ifdef HAVE_SYS_UIO_H
#  include <sys/uio.h>
#endif
#ifdef HAVE_NETINET_IN_H
#  include <netinet/in.h>
#endif
#ifdef HAVE_NETINET_TCP_H
#  include <netinet/tcp.h>
#endif
#ifdef HAVE_NETDB_H
#  include <netdb.h>
#endif
#ifdef HAVE_ARPA_NAMESER_H
#  include <arpa/nameser.h>
#else
#  include "nameser.h"
#endif
#ifdef HAVE_ARPA_NAMESER_COMPAT_H
#  include <arpa/nameser_compat.h>
#endif

#ifdef HAVE_STRINGS_H
#  include <strings.h>
#endif
#ifdef HAVE_UNISTD_H
#  include <unistd.h>
#endif
#ifdef HAVE_SYS_IOCTL_H
#  include <sys/ioctl.h>
#endif
#ifdef NETWARE
#  include <sys/filio.h>
#endif

#include <assert.h>
#include <fcntl.h>

#include "ares.h"
#include "ares_dns.h"
#include "ares_nowarn.h"
#include "ares_private.h"


static int try_again(int errnum);
static void write_tcp_data(ares_channel channel, fd_set *write_fds,
                           ares_socket_t write_fd, struct timeval *now);
static void read_tcp_data(ares_channel channel, fd_set *read_fds,
                          ares_socket_t read_fd, struct timeval *now);
static void read_udp_packets(ares_channel channel, fd_set *read_fds,
                             ares_socket_t read_fd, struct timeval *now);
static void advance_tcp_send_queue(ares_channel channel, int whichserver,
                                   ssize_t num_bytes);
static void process_timeouts(ares_channel channel, struct timeval *now);
static void process_broken_connections(ares_channel channel,
                                       struct timeval *now);
static void process_answer(ares_channel channel, unsigned char *abuf,
                           int alen, int whichserver, int tcp,
                           struct timeval *now);
static void handle_error(ares_channel channel, int whichserver,
                         struct timeval *now);
static void skip_server(ares_channel channel, struct query *query,
                        int whichserver);
static void next_server(ares_channel channel, struct query *query,
                        struct timeval *now);
static int open_tcp_socket(ares_channel channel, struct server_state *server);
static int open_udp_socket(ares_channel channel, struct server_state *server);
static int same_questions(const unsigned char *qbuf, int qlen,
                          const unsigned char *abuf, int alen);
static int same_address(struct sockaddr *sa, struct ares_addr *aa);
static void end_query(ares_channel channel, struct query *query, int status,
                      unsigned char *abuf, int alen);

/* return true if now is exactly check time or later */
int ares__timedout(struct timeval *now,
                   struct timeval *check)
{
  long secs = (now->tv_sec - check->tv_sec);

  if(secs > 0)
    return 1; /* yes, timed out */
  if(secs < 0)
    return 0; /* nope, not timed out */

  /* if the full seconds were identical, check the sub second parts */
  return (now->tv_usec - check->tv_usec >= 0);
}

/* add the specific number of milliseconds to the time in the first argument */
int ares__timeadd(struct timeval *now,
                  int millisecs)
{
  now->tv_sec += millisecs/1000;
  now->tv_usec += (millisecs%1000)*1000;

  if(now->tv_usec >= 1000000) {
    ++(now->tv_sec);
    now->tv_usec -= 1000000;
  }

  return 0;
}

/* return time offset between now and (future) check, in milliseconds */
long ares__timeoffset(struct timeval *now,
                      struct timeval *check)
{
  return (check->tv_sec - now->tv_sec)*1000 +
         (check->tv_usec - now->tv_usec)/1000;
}


/*
 * generic process function
 */
static void processfds(ares_channel channel,
                       fd_set *read_fds, ares_socket_t read_fd,
                       fd_set *write_fds, ares_socket_t write_fd)
{
  struct timeval now = ares__tvnow();

  write_tcp_data(channel, write_fds, write_fd, &now);
  read_tcp_data(channel, read_fds, read_fd, &now);
  read_udp_packets(channel, read_fds, read_fd, &now);
  process_timeouts(channel, &now);
  process_broken_connections(channel, &now);
}

/* Something interesting happened on the wire, or there was a timeout.
 * See what's up and respond accordingly.
 */
void ares_process(ares_channel channel, fd_set *read_fds, fd_set *write_fds)
{
  processfds(channel, read_fds, ARES_SOCKET_BAD, write_fds, ARES_SOCKET_BAD);
}

/* Something interesting happened on the wire, or there was a timeout.
 * See what's up and respond accordingly.
 */
void ares_process_fd(ares_channel channel,
                     ares_socket_t read_fd, /* use ARES_SOCKET_BAD or valid
                                               file descriptors */
                     ares_socket_t write_fd)
{
  processfds(channel, NULL, read_fd, NULL, write_fd);
}


/* Return 1 if the specified error number describes a readiness error, or 0
 * otherwise. This is mostly for HP-UX, which could return EAGAIN or
 * EWOULDBLOCK. See this man page
 *
 * http://devrsrc1.external.hp.com/STKS/cgi-bin/man2html?
 *     manpage=/usr/share/man/man2.Z/send.2
 */
static int try_again(int errnum)
{
#if !defined EWOULDBLOCK && !defined EAGAIN
#error "Neither EWOULDBLOCK nor EAGAIN defined"
#endif
  switch (errnum)
    {
#ifdef EWOULDBLOCK
    case EWOULDBLOCK:
      return 1;
#endif
#if defined EAGAIN && EAGAIN != EWOULDBLOCK
    case EAGAIN:
      return 1;
#endif
    }
  return 0;
}

/* If any TCP sockets select true for writing, write out queued data
 * we have for them.
 */
static void write_tcp_data(ares_channel channel,
                           fd_set *write_fds,
                           ares_socket_t write_fd,
                           struct timeval *now)
{
  struct server_state *server;
  struct send_request *sendreq;
  struct iovec *vec;
  int i;
  ssize_t scount;
  ssize_t wcount;
  size_t n;

  if(!write_fds && (write_fd == ARES_SOCKET_BAD))
    /* no possible action */
    return;

  for (i = 0; i < channel->nservers; i++)
    {
      /* Make sure server has data to send and is selected in write_fds or
         write_fd. */
      server = &channel->servers[i];
      if (!server->qhead || server->tcp_socket == ARES_SOCKET_BAD ||
          server->is_broken)
        continue;

      if(write_fds) {
        if(!FD_ISSET(server->tcp_socket, write_fds))
          continue;
      }
      else {
        if(server->tcp_socket != write_fd)
          continue;
      }

      if(write_fds)
        /* If there's an error and we close this socket, then open
         * another with the same fd to talk to another server, then we
         * don't want to think that it was the new socket that was
         * ready. This is not disastrous, but is likely to result in
         * extra system calls and confusion. */
        FD_CLR(server->tcp_socket, write_fds);

      /* Count the number of send queue items. */
      n = 0;
      for (sendreq = server->qhead; sendreq; sendreq = sendreq->next)
        n++;

      /* Allocate iovecs so we can send all our data at once. */
      vec = malloc(n * sizeof(struct iovec));
      if (vec)
        {
          /* Fill in the iovecs and send. */
          n = 0;
          for (sendreq = server->qhead; sendreq; sendreq = sendreq->next)
            {
              vec[n].iov_base = (char *) sendreq->data;
              vec[n].iov_len = sendreq->len;
              n++;
            }
          wcount = (ssize_t)writev(server->tcp_socket, vec, (int)n);
          free(vec);
          if (wcount < 0)
            {
              if (!try_again(SOCKERRNO))
                handle_error(channel, i, now);
              continue;
            }

          /* Advance the send queue by as many bytes as we sent. */
          advance_tcp_send_queue(channel, i, wcount);
        }
      else
        {
          /* Can't allocate iovecs; just send the first request. */
          sendreq = server->qhead;

          scount = swrite(server->tcp_socket, sendreq->data, sendreq->len);
          if (scount < 0)
            {
              if (!try_again(SOCKERRNO))
                handle_error(channel, i, now);
              continue;
            }

          /* Advance the send queue by as many bytes as we sent. */
          advance_tcp_send_queue(channel, i, scount);
        }
    }
}

/* Consume the given number of bytes from the head of the TCP send queue. */
static void advance_tcp_send_queue(ares_channel channel, int whichserver,
                                   ssize_t num_bytes)
{
  struct send_request *sendreq;
  struct server_state *server = &channel->servers[whichserver];
  while (num_bytes > 0) {
    sendreq = server->qhead;
    if ((size_t)num_bytes >= sendreq->len) {
      num_bytes -= sendreq->len;
      server->qhead = sendreq->next;
      if (sendreq->data_storage)
        free(sendreq->data_storage);
      free(sendreq);
      if (server->qhead == NULL) {
        SOCK_STATE_CALLBACK(channel, server->tcp_socket, 1, 0);
        server->qtail = NULL;

        /* qhead is NULL so we cannot continue this loop */
        break;
      }
    }
    else {
      sendreq->data += num_bytes;
      sendreq->len -= num_bytes;
      num_bytes = 0;
    }
  }
}

/* If any TCP socket selects true for reading, read some data,
 * allocate a buffer if we finish reading the length word, and process
 * a packet if we finish reading one.
 */
static void read_tcp_data(ares_channel channel, fd_set *read_fds,
                          ares_socket_t read_fd, struct timeval *now)
{
  struct server_state *server;
  int i;
  ssize_t count;

  if(!read_fds && (read_fd == ARES_SOCKET_BAD))
    /* no possible action */
    return;

  for (i = 0; i < channel->nservers; i++)
    {
      /* Make sure the server has a socket and is selected in read_fds. */
      server = &channel->servers[i];
      if (server->tcp_socket == ARES_SOCKET_BAD || server->is_broken)
        continue;

      if(read_fds) {
        if(!FD_ISSET(server->tcp_socket, read_fds))
          continue;
      }
      else {
        if(server->tcp_socket != read_fd)
          continue;
      }

      if(read_fds)
        /* If there's an error and we close this socket, then open another
         * with the same fd to talk to another server, then we don't want to
         * think that it was the new socket that was ready. This is not
         * disastrous, but is likely to result in extra system calls and
         * confusion. */
        FD_CLR(server->tcp_socket, read_fds);

      if (server->tcp_lenbuf_pos != 2)
        {
          /* We haven't yet read a length word, so read that (or
           * what's left to read of it).
           */
          count = sread(server->tcp_socket,
                        server->tcp_lenbuf + server->tcp_lenbuf_pos,
                        2 - server->tcp_lenbuf_pos);
          if (count <= 0)
            {
              if (!(count == -1 && try_again(SOCKERRNO)))
                handle_error(channel, i, now);
              continue;
            }

          server->tcp_lenbuf_pos += (int)count;
          if (server->tcp_lenbuf_pos == 2)
            {
              /* We finished reading the length word.  Decode the
               * length and allocate a buffer for the data.
               */
              server->tcp_length = server->tcp_lenbuf[0] << 8
                | server->tcp_lenbuf[1];
              server->tcp_buffer = malloc(server->tcp_length);
              if (!server->tcp_buffer)
                handle_error(channel, i, now);
              server->tcp_buffer_pos = 0;
            }
        }
      else
        {
          /* Read data into the allocated buffer. */
          count = sread(server->tcp_socket,
                        server->tcp_buffer + server->tcp_buffer_pos,
                        server->tcp_length - server->tcp_buffer_pos);
          if (count <= 0)
            {
              if (!(count == -1 && try_again(SOCKERRNO)))
                handle_error(channel, i, now);
              continue;
            }

          server->tcp_buffer_pos += (int)count;
          if (server->tcp_buffer_pos == server->tcp_length)
            {
              /* We finished reading this answer; process it and
               * prepare to read another length word.
               */
              process_answer(channel, server->tcp_buffer, server->tcp_length,
                             i, 1, now);
          if (server->tcp_buffer)
                        free(server->tcp_buffer);
              server->tcp_buffer = NULL;
              server->tcp_lenbuf_pos = 0;
              server->tcp_buffer_pos = 0;
            }
        }
    }
}

/* If any UDP sockets select true for reading, process them. */
static void read_udp_packets(ares_channel channel, fd_set *read_fds,
                             ares_socket_t read_fd, struct timeval *now)
{
  struct server_state *server;
  int i;
  ssize_t count;
  unsigned char buf[MAXENDSSZ + 1];
#ifdef HAVE_RECVFROM
  ares_socklen_t fromlen;
  union {
    struct sockaddr     sa;
    struct sockaddr_in  sa4;
    struct sockaddr_in6 sa6;
  } from;
#endif

  if(!read_fds && (read_fd == ARES_SOCKET_BAD))
    /* no possible action */
    return;

  for (i = 0; i < channel->nservers; i++)
    {
      /* Make sure the server has a socket and is selected in read_fds. */
      server = &channel->servers[i];

      if (server->udp_socket == ARES_SOCKET_BAD || server->is_broken)
        continue;

      if(read_fds) {
        if(!FD_ISSET(server->udp_socket, read_fds))
          continue;
      }
      else {
        if(server->udp_socket != read_fd)
          continue;
      }

      if(read_fds)
        /* If there's an error and we close this socket, then open
         * another with the same fd to talk to another server, then we
         * don't want to think that it was the new socket that was
         * ready. This is not disastrous, but is likely to result in
         * extra system calls and confusion. */
        FD_CLR(server->udp_socket, read_fds);

      /* To reduce event loop overhead, read and process as many
       * packets as we can. */
      do {
        if (server->udp_socket == ARES_SOCKET_BAD)
          count = 0;

        else {
#ifdef HAVE_RECVFROM
          if (server->addr.family == AF_INET)
            fromlen = sizeof(from.sa4);
          else
            fromlen = sizeof(from.sa6);
          count = (ssize_t)recvfrom(server->udp_socket, (void *)buf,
                                    sizeof(buf), 0, &from.sa, &fromlen);
#else
          count = sread(server->udp_socket, buf, sizeof(buf));
#endif
        }

        if (count == -1 && try_again(SOCKERRNO))
          continue;
        else if (count <= 0)
          handle_error(channel, i, now);
#ifdef HAVE_RECVFROM
        else if (!same_address(&from.sa, &server->addr))
          /* The address the response comes from does not match the address we
           * sent the request to. Someone may be attempting to perform a cache
           * poisoning attack. */
          break;
#endif
        else
          process_answer(channel, buf, (int)count, i, 0, now);
       } while (count > 0);
    }
}

/* If any queries have timed out, note the timeout and move them on. */
static void process_timeouts(ares_channel channel, struct timeval *now)
{
  time_t t;  /* the time of the timeouts we're processing */
  struct query *query;
  struct list_node* list_head;
  struct list_node* list_node;

  /* Process all the timeouts that have fired since the last time we processed
   * timeouts. If things are going well, then we'll have hundreds/thousands of
   * queries that fall into future buckets, and only a handful of requests
   * that fall into the "now" bucket, so this should be quite quick.
   */
  for (t = channel->last_timeout_processed; t <= now->tv_sec; t++)
    {
      list_head = &(channel->queries_by_timeout[t % ARES_TIMEOUT_TABLE_SIZE]);
      for (list_node = list_head->next; list_node != list_head; )
        {
          query = list_node->data;
          list_node = list_node->next;  /* in case the query gets deleted */
          if (query->timeout.tv_sec && ares__timedout(now, &query->timeout))
            {
              query->error_status = ARES_ETIMEOUT;
              ++query->timeouts;
              next_server(channel, query, now);
            }
        }
     }
  channel->last_timeout_processed = now->tv_sec;
}

/* Handle an answer from a server. */
static void process_answer(ares_channel channel, unsigned char *abuf,
                           int alen, int whichserver, int tcp,
                           struct timeval *now)
{
  int tc, rcode, packetsz;
  unsigned short id;
  struct query *query;
  struct list_node* list_head;
  struct list_node* list_node;

  /* If there's no room in the answer for a header, we can't do much
   * with it. */
  if (alen < HFIXEDSZ)
    return;

  /* Grab the query ID, truncate bit, and response code from the packet. */
  id = DNS_HEADER_QID(abuf);
  tc = DNS_HEADER_TC(abuf);
  rcode = DNS_HEADER_RCODE(abuf);

  /* Find the query corresponding to this packet. The queries are
   * hashed/bucketed by query id, so this lookup should be quick.  Note that
   * both the query id and the questions must be the same; when the query id
   * wraps around we can have multiple outstanding queries with the same query
   * id, so we need to check both the id and question.
   */
  query = NULL;
  list_head = &(channel->queries_by_qid[id % ARES_QID_TABLE_SIZE]);
  for (list_node = list_head->next; list_node != list_head;
       list_node = list_node->next)
    {
      struct query *q = list_node->data;
      if ((q->qid == id) && same_questions(q->qbuf, q->qlen, abuf, alen))
        {
          query = q;
          break;
        }
    }
  if (!query)
    return;

  packetsz = PACKETSZ;
  /* If we use EDNS and server answers with one of these RCODES, the protocol
   * extension is not understood by the responder. We must retry the query
   * without EDNS enabled.
   */
  if (channel->flags & ARES_FLAG_EDNS)
  {
      packetsz = channel->ednspsz;
      if (rcode == NOTIMP || rcode == FORMERR || rcode == SERVFAIL)
      {
          int qlen = alen - EDNSFIXEDSZ;
          channel->flags ^= ARES_FLAG_EDNS;
          query->tcplen -= EDNSFIXEDSZ;
          query->qlen -= EDNSFIXEDSZ;
          query->tcpbuf[0] = (unsigned char)((qlen >> 8) & 0xff);
          query->tcpbuf[1] = (unsigned char)(qlen & 0xff);
          DNS_HEADER_SET_ARCOUNT(query->tcpbuf + 2, 0);
          query->tcpbuf = realloc(query->tcpbuf, query->tcplen);
          ares__send_query(channel, query, now);
          return;
      }
  }

  /* If we got a truncated UDP packet and are not ignoring truncation,
   * don't accept the packet, and switch the query to TCP if we hadn't
   * done so already.
   */
  if ((tc || alen > packetsz) && !tcp && !(channel->flags & ARES_FLAG_IGNTC))
    {
      if (!query->using_tcp)
        {
          query->using_tcp = 1;
          ares__send_query(channel, query, now);
        }
      return;
    }

  /* Limit alen to PACKETSZ if we aren't using TCP (only relevant if we
   * are ignoring truncation.
   */
  if (alen > packetsz && !tcp)
      alen = packetsz;

  /* If we aren't passing through all error packets, discard packets
   * with SERVFAIL, NOTIMP, or REFUSED response codes.
   */
  if (!(channel->flags & ARES_FLAG_NOCHECKRESP))
    {
      if (rcode == SERVFAIL || rcode == NOTIMP || rcode == REFUSED)
        {
          skip_server(channel, query, whichserver);
          if (query->server == whichserver)
            next_server(channel, query, now);
          return;
        }
    }

  end_query(channel, query, ARES_SUCCESS, abuf, alen);
}

/* Close all the connections that are no longer usable. */
static void process_broken_connections(ares_channel channel,
                                       struct timeval *now)
{
  int i;
  for (i = 0; i < channel->nservers; i++)
    {
      struct server_state *server = &channel->servers[i];
      if (server->is_broken)
        {
          handle_error(channel, i, now);
        }
    }
}

static void handle_error(ares_channel channel, int whichserver,
                         struct timeval *now)
{
  struct server_state *server;
  struct query *query;
  struct list_node list_head;
  struct list_node* list_node;

  server = &channel->servers[whichserver];

  /* Reset communications with this server. */
  ares__close_sockets(channel, server);

  /* Tell all queries talking to this server to move on and not try this
   * server again. We steal the current list of queries that were in-flight to
   * this server, since when we call next_server this can cause the queries to
   * be re-sent to this server, which will re-insert these queries in that
   * same server->queries_to_server list.
   */
  ares__init_list_head(&list_head);
  ares__swap_lists(&list_head, &(server->queries_to_server));
  for (list_node = list_head.next; list_node != &list_head; )
    {
      query = list_node->data;
      list_node = list_node->next;  /* in case the query gets deleted */
      assert(query->server == whichserver);
      skip_server(channel, query, whichserver);
      next_server(channel, query, now);
    }
  /* Each query should have removed itself from our temporary list as
   * it re-sent itself or finished up...
   */
  assert(ares__is_list_empty(&list_head));
}

static void skip_server(ares_channel channel, struct query *query,
                        int whichserver)
{
  /* The given server gave us problems with this query, so if we have the
   * luxury of using other servers, then let's skip the potentially broken
   * server and just use the others. If we only have one server and we need to
   * retry then we should just go ahead and re-use that server, since it's our
   * only hope; perhaps we just got unlucky, and retrying will work (eg, the
   * server timed out our TCP connection just as we were sending another
   * request).
   */
  if (channel->nservers > 1)
    {
      query->server_info[whichserver].skip_server = 1;
    }
}

static void next_server(ares_channel channel, struct query *query,
                        struct timeval *now)
{
  /* We need to try each server channel->tries times. We have channel->nservers
   * servers to try. In total, we need to do channel->nservers * channel->tries
   * attempts. Use query->try to remember how many times we already attempted
   * this query. Use modular arithmetic to find the next server to try. */
  while (++(query->try_count) < (channel->nservers * channel->tries))
    {
      struct server_state *server;

      /* Move on to the next server. */
      query->server = (query->server + 1) % channel->nservers;
      server = &channel->servers[query->server];

      /* We don't want to use this server if (1) we decided this connection is
       * broken, and thus about to be closed, (2) we've decided to skip this
       * server because of earlier errors we encountered, or (3) we already
       * sent this query over this exact connection.
       */
      if (!server->is_broken &&
           !query->server_info[query->server].skip_server &&
           !(query->using_tcp &&
             (query->server_info[query->server].tcp_connection_generation ==
              server->tcp_connection_generation)))
        {
           ares__send_query(channel, query, now);
           return;
        }

      /* You might think that with TCP we only need one try. However, even
       * when using TCP, servers can time-out our connection just as we're
       * sending a request, or close our connection because they die, or never
       * send us a reply because they get wedged or tickle a bug that drops
       * our request.
       */
    }

  /* If we are here, all attempts to perform query failed. */
  end_query(channel, query, query->error_status, NULL, 0);
}

void ares__send_query(ares_channel channel, struct query *query,
                      struct timeval *now)
{
  struct send_request *sendreq;
  struct server_state *server;
  int timeplus;

  server = &channel->servers[query->server];
  if (query->using_tcp)
    {
      /* Make sure the TCP socket for this server is set up and queue
       * a send request.
       */
      if (server->tcp_socket == ARES_SOCKET_BAD)
        {
          if (open_tcp_socket(channel, server) == -1)
            {
              skip_server(channel, query, query->server);
              next_server(channel, query, now);
              return;
            }
        }
      sendreq = calloc(1, sizeof(struct send_request));
      if (!sendreq)
        {
        end_query(channel, query, ARES_ENOMEM, NULL, 0);
          return;
        }
      /* To make the common case fast, we avoid copies by using the query's
       * tcpbuf for as long as the query is alive. In the rare case where the
       * query ends while it's queued for transmission, then we give the
       * sendreq its own copy of the request packet and put it in
       * sendreq->data_storage.
       */
      sendreq->data_storage = NULL;
      sendreq->data = query->tcpbuf;
      sendreq->len = query->tcplen;
      sendreq->owner_query = query;
      sendreq->next = NULL;
      if (server->qtail)
        server->qtail->next = sendreq;
      else
        {
          SOCK_STATE_CALLBACK(channel, server->tcp_socket, 1, 1);
          server->qhead = sendreq;
        }
      server->qtail = sendreq;
      query->server_info[query->server].tcp_connection_generation =
        server->tcp_connection_generation;
    }
  else
    {
      if (server->udp_socket == ARES_SOCKET_BAD)
        {
          if (open_udp_socket(channel, server) == -1)
            {
              skip_server(channel, query, query->server);
              next_server(channel, query, now);
              return;
            }
        }
      if (swrite(server->udp_socket, query->qbuf, query->qlen) == -1)
        {
          /* FIXME: Handle EAGAIN here since it likely can happen. */
          skip_server(channel, query, query->server);
          next_server(channel, query, now);
          return;
        }
    }
    timeplus = channel->timeout << (query->try_count / channel->nservers);
    timeplus = (timeplus * (9 + (rand () & 7))) / 16;
    query->timeout = *now;
    ares__timeadd(&query->timeout,
                  timeplus);
    /* Keep track of queries bucketed by timeout, so we can process
     * timeout events quickly.
     */
    ares__remove_from_list(&(query->queries_by_timeout));
    ares__insert_in_list(
        &(query->queries_by_timeout),
        &(channel->queries_by_timeout[query->timeout.tv_sec %
                                      ARES_TIMEOUT_TABLE_SIZE]));

    /* Keep track of queries bucketed by server, so we can process server
     * errors quickly.
     */
    ares__remove_from_list(&(query->queries_to_server));
    ares__insert_in_list(&(query->queries_to_server),
                         &(server->queries_to_server));
}

/*
 * setsocknonblock sets the given socket to either blocking or non-blocking
 * mode based on the 'nonblock' boolean argument. This function is highly
 * portable.
 */
static int setsocknonblock(ares_socket_t sockfd,    /* operate on this */
                    int nonblock   /* TRUE or FALSE */)
{
#if defined(USE_BLOCKING_SOCKETS)

  return 0; /* returns success */

#elif defined(HAVE_FCNTL_O_NONBLOCK)

  /* most recent unix versions */
  int flags;
  flags = fcntl(sockfd, F_GETFL, 0);
  if (FALSE != nonblock)
    return fcntl(sockfd, F_SETFL, flags | O_NONBLOCK);
  else
    return fcntl(sockfd, F_SETFL, flags & (~O_NONBLOCK));

#elif defined(HAVE_IOCTL_FIONBIO)

  /* older unix versions */
  int flags = nonblock ? 1 : 0;
  return ioctl(sockfd, FIONBIO, &flags);

#elif defined(HAVE_IOCTLSOCKET_FIONBIO)

#ifdef WATT32
  char flags = nonblock ? 1 : 0;
#else
  /* Windows */
  unsigned long flags = nonblock ? 1UL : 0UL;
#endif
  return ioctlsocket(sockfd, FIONBIO, &flags);

#elif defined(HAVE_IOCTLSOCKET_CAMEL_FIONBIO)

  /* Amiga */
  long flags = nonblock ? 1L : 0L;
  return IoctlSocket(sockfd, FIONBIO, flags);

#elif defined(HAVE_SETSOCKOPT_SO_NONBLOCK)

  /* BeOS */
  long b = nonblock ? 1L : 0L;
  return setsockopt(sockfd, SOL_SOCKET, SO_NONBLOCK, &b, sizeof(b));

#else
#  error "no non-blocking method was found/used/set"
#endif
}

static int configure_socket(ares_socket_t s, int family, ares_channel channel)
{
  union {
    struct sockaddr     sa;
    struct sockaddr_in  sa4;
    struct sockaddr_in6 sa6;
  } local;

  setsocknonblock(s, TRUE);

#if defined(FD_CLOEXEC) && !defined(MSDOS)
  /* Configure the socket fd as close-on-exec. */
  if (fcntl(s, F_SETFD, FD_CLOEXEC) == -1)
    return -1;
#endif

  /* Set the socket's send and receive buffer sizes. */
  if ((channel->socket_send_buffer_size > 0) &&
      setsockopt(s, SOL_SOCKET, SO_SNDBUF,
                 (void *)&channel->socket_send_buffer_size,
                 sizeof(channel->socket_send_buffer_size)) == -1)
    return -1;

  if ((channel->socket_receive_buffer_size > 0) &&
      setsockopt(s, SOL_SOCKET, SO_RCVBUF,
                 (void *)&channel->socket_receive_buffer_size,
                 sizeof(channel->socket_receive_buffer_size)) == -1)
    return -1;

#ifdef SO_BINDTODEVICE
  if (channel->local_dev_name[0]) {
    if (setsockopt(s, SOL_SOCKET, SO_BINDTODEVICE,
                   channel->local_dev_name, sizeof(channel->local_dev_name))) {
      /* Only root can do this, and usually not fatal if it doesn't work, so */
      /* just continue on. */
    }
  }
#endif

  if (family == AF_INET) {
    if (channel->local_ip4) {
      memset(&local.sa4, 0, sizeof(local.sa4));
      local.sa4.sin_family = AF_INET;
      local.sa4.sin_addr.s_addr = htonl(channel->local_ip4);
      if (bind(s, &local.sa, sizeof(local.sa4)) < 0)
        return -1;
    }
  }
  else if (family == AF_INET6) {
    if (memcmp(channel->local_ip6, &ares_in6addr_any,
               sizeof(channel->local_ip6)) != 0) {
      memset(&local.sa6, 0, sizeof(local.sa6));
      local.sa6.sin6_family = AF_INET6;
      memcpy(&local.sa6.sin6_addr, channel->local_ip6,
             sizeof(channel->local_ip6));
      if (bind(s, &local.sa, sizeof(local.sa6)) < 0)
        return -1;
    }
  }

  return 0;
}

static int open_tcp_socket(ares_channel channel, struct server_state *server)
{
  ares_socket_t s;
  int opt;
  ares_socklen_t salen;
  union {
    struct sockaddr_in  sa4;
    struct sockaddr_in6 sa6;
  } saddr;
  struct sockaddr *sa;

  switch (server->addr.family)
    {
      case AF_INET:
        sa = (void *)&saddr.sa4;
        salen = sizeof(saddr.sa4);
        memset(sa, 0, salen);
        saddr.sa4.sin_family = AF_INET;
        saddr.sa4.sin_port = aresx_sitous(channel->tcp_port);
        memcpy(&saddr.sa4.sin_addr, &server->addr.addrV4,
               sizeof(server->addr.addrV4));
        break;
      case AF_INET6:
        sa = (void *)&saddr.sa6;
        salen = sizeof(saddr.sa6);
        memset(sa, 0, salen);
        saddr.sa6.sin6_family = AF_INET6;
        saddr.sa6.sin6_port = aresx_sitous(channel->tcp_port);
        memcpy(&saddr.sa6.sin6_addr, &server->addr.addrV6,
               sizeof(server->addr.addrV6));
        break;
      default:
        return -1;
    }

  /* Acquire a socket. */
  s = socket(server->addr.family, SOCK_STREAM, 0);
  if (s == ARES_SOCKET_BAD)
    return -1;

  /* Configure it. */
  if (configure_socket(s, server->addr.family, channel) < 0)
    {
       sclose(s);
       return -1;
    }

#ifdef TCP_NODELAY
  /*
   * Disable the Nagle algorithm (only relevant for TCP sockets, and thus not
   * in configure_socket). In general, in DNS lookups we're pretty much
   * interested in firing off a single request and then waiting for a reply,
   * so batching isn't very interesting.
   */
  opt = 1;
  if (setsockopt(s, IPPROTO_TCP, TCP_NODELAY,
                 (void *)&opt, sizeof(opt)) == -1)
    {
       sclose(s);
       return -1;
    }
#endif

  /* Connect to the server. */
  if (connect(s, sa, salen) == -1)
    {
      int err = SOCKERRNO;

      if (err != EINPROGRESS && err != EWOULDBLOCK)
        {
          sclose(s);
          return -1;
        }
    }

  if (channel->sock_create_cb)
    {
      int err = channel->sock_create_cb(s, SOCK_STREAM,
                                        channel->sock_create_cb_data);
      if (err < 0)
        {
          sclose(s);
          return err;
        }
    }

  SOCK_STATE_CALLBACK(channel, s, 1, 0);
  server->tcp_buffer_pos = 0;
  server->tcp_socket = s;
  server->tcp_connection_generation = ++channel->tcp_connection_generation;
  return 0;
}

static int open_udp_socket(ares_channel channel, struct server_state *server)
{
  ares_socket_t s;
  ares_socklen_t salen;
  union {
    struct sockaddr_in  sa4;
    struct sockaddr_in6 sa6;
  } saddr;
  struct sockaddr *sa;

  switch (server->addr.family)
    {
      case AF_INET:
        sa = (void *)&saddr.sa4;
        salen = sizeof(saddr.sa4);
        memset(sa, 0, salen);
        saddr.sa4.sin_family = AF_INET;
        saddr.sa4.sin_port = aresx_sitous(channel->udp_port);
        memcpy(&saddr.sa4.sin_addr, &server->addr.addrV4,
               sizeof(server->addr.addrV4));
        break;
      case AF_INET6:
        sa = (void *)&saddr.sa6;
        salen = sizeof(saddr.sa6);
        memset(sa, 0, salen);
        saddr.sa6.sin6_family = AF_INET6;
        saddr.sa6.sin6_port = aresx_sitous(channel->udp_port);
        memcpy(&saddr.sa6.sin6_addr, &server->addr.addrV6,
               sizeof(server->addr.addrV6));
        break;
      default:
        return -1;
    }

  /* Acquire a socket. */
  s = socket(server->addr.family, SOCK_DGRAM, 0);
  if (s == ARES_SOCKET_BAD)
    return -1;

  /* Set the socket non-blocking. */
  if (configure_socket(s, server->addr.family, channel) < 0)
    {
       sclose(s);
       return -1;
    }

  /* Connect to the server. */
  if (connect(s, sa, salen) == -1)
    {
      int err = SOCKERRNO;

      if (err != EINPROGRESS && err != EWOULDBLOCK)
        {
          sclose(s);
          return -1;
        }
    }

  if (channel->sock_create_cb)
    {
      int err = channel->sock_create_cb(s, SOCK_DGRAM,
                                        channel->sock_create_cb_data);
      if (err < 0)
        {
          sclose(s);
          return err;
        }
    }

  SOCK_STATE_CALLBACK(channel, s, 1, 0);

  server->udp_socket = s;
  return 0;
}

static int same_questions(const unsigned char *qbuf, int qlen,
                          const unsigned char *abuf, int alen)
{
  struct {
    const unsigned char *p;
    int qdcount;
    char *name;
    long namelen;
    int type;
    int dnsclass;
  } q, a;
  int i, j;

  if (qlen < HFIXEDSZ || alen < HFIXEDSZ)
    return 0;

  /* Extract qdcount from the request and reply buffers and compare them. */
  q.qdcount = DNS_HEADER_QDCOUNT(qbuf);
  a.qdcount = DNS_HEADER_QDCOUNT(abuf);
  if (q.qdcount != a.qdcount)
    return 0;

  /* For each question in qbuf, find it in abuf. */
  q.p = qbuf + HFIXEDSZ;
  for (i = 0; i < q.qdcount; i++)
    {
      /* Decode the question in the query. */
      if (ares_expand_name(q.p, qbuf, qlen, &q.name, &q.namelen)
          != ARES_SUCCESS)
        return 0;
      q.p += q.namelen;
      if (q.p + QFIXEDSZ > qbuf + qlen)
        {
          free(q.name);
          return 0;
        }
      q.type = DNS_QUESTION_TYPE(q.p);
      q.dnsclass = DNS_QUESTION_CLASS(q.p);
      q.p += QFIXEDSZ;

      /* Search for this question in the answer. */
      a.p = abuf + HFIXEDSZ;
      for (j = 0; j < a.qdcount; j++)
        {
          /* Decode the question in the answer. */
          if (ares_expand_name(a.p, abuf, alen, &a.name, &a.namelen)
              != ARES_SUCCESS)
            {
              free(q.name);
              return 0;
            }
          a.p += a.namelen;
          if (a.p + QFIXEDSZ > abuf + alen)
            {
              free(q.name);
              free(a.name);
              return 0;
            }
          a.type = DNS_QUESTION_TYPE(a.p);
          a.dnsclass = DNS_QUESTION_CLASS(a.p);
          a.p += QFIXEDSZ;

          /* Compare the decoded questions. */
          if (strcasecmp(q.name, a.name) == 0 && q.type == a.type
              && q.dnsclass == a.dnsclass)
            {
              free(a.name);
              break;
            }
          free(a.name);
        }

      free(q.name);
      if (j == a.qdcount)
        return 0;
    }
  return 1;
}

static int same_address(struct sockaddr *sa, struct ares_addr *aa)
{
  void *addr1;
  void *addr2;

  if (sa->sa_family == aa->family)
    {
      switch (aa->family)
        {
          case AF_INET:
            addr1 = &aa->addrV4;
            addr2 = &((struct sockaddr_in *)sa)->sin_addr;
            if (memcmp(addr1, addr2, sizeof(aa->addrV4)) == 0)
              return 1; /* match */
            break;
          case AF_INET6:
            addr1 = &aa->addrV6;
            addr2 = &((struct sockaddr_in6 *)sa)->sin6_addr;
            if (memcmp(addr1, addr2, sizeof(aa->addrV6)) == 0)
              return 1; /* match */
            break;
          default:
            break;
        }
    }
  return 0; /* different */
}

static void end_query (ares_channel channel, struct query *query, int status,
                       unsigned char *abuf, int alen)
{
  int i;

  /* First we check to see if this query ended while one of our send
   * queues still has pointers to it.
   */
  for (i = 0; i < channel->nservers; i++)
    {
      struct server_state *server = &channel->servers[i];
      struct send_request *sendreq;
      for (sendreq = server->qhead; sendreq; sendreq = sendreq->next)
        if (sendreq->owner_query == query)
          {
            sendreq->owner_query = NULL;
            assert(sendreq->data_storage == NULL);
            if (status == ARES_SUCCESS)
              {
                /* We got a reply for this query, but this queued sendreq
                 * points into this soon-to-be-gone query's tcpbuf. Probably
                 * this means we timed out and queued the query for
                 * retransmission, then received a response before actually
                 * retransmitting. This is perfectly fine, so we want to keep
                 * the connection running smoothly if we can. But in the worst
                 * case we may have sent only some prefix of the query, with
                 * some suffix of the query left to send. Also, the buffer may
                 * be queued on multiple queues. To prevent dangling pointers
                 * to the query's tcpbuf and handle these cases, we just give
                 * such sendreqs their own copy of the query packet.
                 */
               sendreq->data_storage = malloc(sendreq->len);
               if (sendreq->data_storage != NULL)
                 {
                   memcpy(sendreq->data_storage, sendreq->data, sendreq->len);
                   sendreq->data = sendreq->data_storage;
                 }
              }
            if ((status != ARES_SUCCESS) || (sendreq->data_storage == NULL))
              {
                /* We encountered an error (probably a timeout, suggesting the
                 * DNS server we're talking to is probably unreachable,
                 * wedged, or severely overloaded) or we couldn't copy the
                 * request, so mark the connection as broken. When we get to
                 * process_broken_connections() we'll close the connection and
                 * try to re-send requests to another server.
                 */
               server->is_broken = 1;
               /* Just to be paranoid, zero out this sendreq... */
               sendreq->data = NULL;
               sendreq->len = 0;
             }
          }
    }

  /* Invoke the callback */
  query->callback(query->arg, status, query->timeouts, abuf, alen);
  ares__free_query(query);

  /* Simple cleanup policy: if no queries are remaining, close all network
   * sockets unless STAYOPEN is set.
   */
  if (!(channel->flags & ARES_FLAG_STAYOPEN) &&
      ares__is_list_empty(&(channel->all_queries)))
    {
      for (i = 0; i < channel->nservers; i++)
        ares__close_sockets(channel, &channel->servers[i]);
    }
}

void ares__free_query(struct query *query)
{
  /* Remove the query from all the lists in which it is linked */
  ares__remove_from_list(&(query->queries_by_qid));
  ares__remove_from_list(&(query->queries_by_timeout));
  ares__remove_from_list(&(query->queries_to_server));
  ares__remove_from_list(&(query->all_queries));
  /* Zero out some important stuff, to help catch bugs */
  query->callback = NULL;
  query->arg = NULL;
  /* Deallocate the memory associated with the query */
  free(query->tcpbuf);
  free(query->server_info);
  free(query);
}