]> git.ipfire.org Git - people/ms/dnsmasq.git/blob - src/forward.c
Tweak EDNS timeout code.
[people/ms/dnsmasq.git] / src / forward.c
1 /* dnsmasq is Copyright (c) 2000-2015 Simon Kelley
2
3 This program is free software; you can redistribute it and/or modify
4 it under the terms of the GNU General Public License as published by
5 the Free Software Foundation; version 2 dated June, 1991, or
6 (at your option) version 3 dated 29 June, 2007.
7
8 This program is distributed in the hope that it will be useful,
9 but WITHOUT ANY WARRANTY; without even the implied warranty of
10 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 GNU General Public License for more details.
12
13 You should have received a copy of the GNU General Public License
14 along with this program. If not, see <http://www.gnu.org/licenses/>.
15 */
16
17 #include "dnsmasq.h"
18
19 static struct frec *lookup_frec(unsigned short id, void *hash);
20 static struct frec *lookup_frec_by_sender(unsigned short id,
21 union mysockaddr *addr,
22 void *hash);
23 static unsigned short get_id(void);
24 static void free_frec(struct frec *f);
25
26 #ifdef HAVE_DNSSEC
27 static int tcp_key_recurse(time_t now, int status, struct dns_header *header, size_t n,
28 int class, char *name, char *keyname, struct server *server, int *keycount);
29 static int do_check_sign(struct frec *forward, int status, time_t now, char *name, char *keyname);
30 static int send_check_sign(struct frec *forward, time_t now, struct dns_header *header, size_t plen,
31 char *name, char *keyname);
32 #endif
33
34
35 /* Send a UDP packet with its source address set as "source"
36 unless nowild is true, when we just send it with the kernel default */
37 int send_from(int fd, int nowild, char *packet, size_t len,
38 union mysockaddr *to, struct all_addr *source,
39 unsigned int iface)
40 {
41 struct msghdr msg;
42 struct iovec iov[1];
43 union {
44 struct cmsghdr align; /* this ensures alignment */
45 #if defined(HAVE_LINUX_NETWORK)
46 char control[CMSG_SPACE(sizeof(struct in_pktinfo))];
47 #elif defined(IP_SENDSRCADDR)
48 char control[CMSG_SPACE(sizeof(struct in_addr))];
49 #endif
50 #ifdef HAVE_IPV6
51 char control6[CMSG_SPACE(sizeof(struct in6_pktinfo))];
52 #endif
53 } control_u;
54
55 iov[0].iov_base = packet;
56 iov[0].iov_len = len;
57
58 msg.msg_control = NULL;
59 msg.msg_controllen = 0;
60 msg.msg_flags = 0;
61 msg.msg_name = to;
62 msg.msg_namelen = sa_len(to);
63 msg.msg_iov = iov;
64 msg.msg_iovlen = 1;
65
66 if (!nowild)
67 {
68 struct cmsghdr *cmptr;
69 msg.msg_control = &control_u;
70 msg.msg_controllen = sizeof(control_u);
71 cmptr = CMSG_FIRSTHDR(&msg);
72
73 if (to->sa.sa_family == AF_INET)
74 {
75 #if defined(HAVE_LINUX_NETWORK)
76 struct in_pktinfo p;
77 p.ipi_ifindex = 0;
78 p.ipi_spec_dst = source->addr.addr4;
79 memcpy(CMSG_DATA(cmptr), &p, sizeof(p));
80 msg.msg_controllen = cmptr->cmsg_len = CMSG_LEN(sizeof(struct in_pktinfo));
81 cmptr->cmsg_level = IPPROTO_IP;
82 cmptr->cmsg_type = IP_PKTINFO;
83 #elif defined(IP_SENDSRCADDR)
84 memcpy(CMSG_DATA(cmptr), &(source->addr.addr4), sizeof(source->addr.addr4));
85 msg.msg_controllen = cmptr->cmsg_len = CMSG_LEN(sizeof(struct in_addr));
86 cmptr->cmsg_level = IPPROTO_IP;
87 cmptr->cmsg_type = IP_SENDSRCADDR;
88 #endif
89 }
90 else
91 #ifdef HAVE_IPV6
92 {
93 struct in6_pktinfo p;
94 p.ipi6_ifindex = iface; /* Need iface for IPv6 to handle link-local addrs */
95 p.ipi6_addr = source->addr.addr6;
96 memcpy(CMSG_DATA(cmptr), &p, sizeof(p));
97 msg.msg_controllen = cmptr->cmsg_len = CMSG_LEN(sizeof(struct in6_pktinfo));
98 cmptr->cmsg_type = daemon->v6pktinfo;
99 cmptr->cmsg_level = IPPROTO_IPV6;
100 }
101 #else
102 (void)iface; /* eliminate warning */
103 #endif
104 }
105
106 while (retry_send(sendmsg(fd, &msg, 0)));
107
108 /* If interface is still in DAD, EINVAL results - ignore that. */
109 if (errno != 0 && errno != EINVAL)
110 {
111 my_syslog(LOG_ERR, _("failed to send packet: %s"), strerror(errno));
112 return 0;
113 }
114
115 return 1;
116 }
117
118 static unsigned int search_servers(time_t now, struct all_addr **addrpp,
119 unsigned int qtype, char *qdomain, int *type, char **domain, int *norebind)
120
121 {
122 /* If the query ends in the domain in one of our servers, set
123 domain to point to that name. We find the largest match to allow both
124 domain.org and sub.domain.org to exist. */
125
126 unsigned int namelen = strlen(qdomain);
127 unsigned int matchlen = 0;
128 struct server *serv;
129 unsigned int flags = 0;
130
131 for (serv = daemon->servers; serv; serv=serv->next)
132 /* domain matches take priority over NODOTS matches */
133 if ((serv->flags & SERV_FOR_NODOTS) && *type != SERV_HAS_DOMAIN && !strchr(qdomain, '.') && namelen != 0)
134 {
135 unsigned int sflag = serv->addr.sa.sa_family == AF_INET ? F_IPV4 : F_IPV6;
136 *type = SERV_FOR_NODOTS;
137 if (serv->flags & SERV_NO_ADDR)
138 flags = F_NXDOMAIN;
139 else if (serv->flags & SERV_LITERAL_ADDRESS)
140 {
141 if (sflag & qtype)
142 {
143 flags = sflag;
144 if (serv->addr.sa.sa_family == AF_INET)
145 *addrpp = (struct all_addr *)&serv->addr.in.sin_addr;
146 #ifdef HAVE_IPV6
147 else
148 *addrpp = (struct all_addr *)&serv->addr.in6.sin6_addr;
149 #endif
150 }
151 else if (!flags || (flags & F_NXDOMAIN))
152 flags = F_NOERR;
153 }
154 }
155 else if (serv->flags & SERV_HAS_DOMAIN)
156 {
157 unsigned int domainlen = strlen(serv->domain);
158 char *matchstart = qdomain + namelen - domainlen;
159 if (namelen >= domainlen &&
160 hostname_isequal(matchstart, serv->domain) &&
161 (domainlen == 0 || namelen == domainlen || *(matchstart-1) == '.' ))
162 {
163 if (serv->flags & SERV_NO_REBIND)
164 *norebind = 1;
165 else
166 {
167 unsigned int sflag = serv->addr.sa.sa_family == AF_INET ? F_IPV4 : F_IPV6;
168 /* implement priority rules for --address and --server for same domain.
169 --address wins if the address is for the correct AF
170 --server wins otherwise. */
171 if (domainlen != 0 && domainlen == matchlen)
172 {
173 if ((serv->flags & SERV_LITERAL_ADDRESS))
174 {
175 if (!(sflag & qtype) && flags == 0)
176 continue;
177 }
178 else
179 {
180 if (flags & (F_IPV4 | F_IPV6))
181 continue;
182 }
183 }
184
185 if (domainlen >= matchlen)
186 {
187 *type = serv->flags & (SERV_HAS_DOMAIN | SERV_USE_RESOLV | SERV_NO_REBIND);
188 *domain = serv->domain;
189 matchlen = domainlen;
190 if (serv->flags & SERV_NO_ADDR)
191 flags = F_NXDOMAIN;
192 else if (serv->flags & SERV_LITERAL_ADDRESS)
193 {
194 if (sflag & qtype)
195 {
196 flags = sflag;
197 if (serv->addr.sa.sa_family == AF_INET)
198 *addrpp = (struct all_addr *)&serv->addr.in.sin_addr;
199 #ifdef HAVE_IPV6
200 else
201 *addrpp = (struct all_addr *)&serv->addr.in6.sin6_addr;
202 #endif
203 }
204 else if (!flags || (flags & F_NXDOMAIN))
205 flags = F_NOERR;
206 }
207 else
208 flags = 0;
209 }
210 }
211 }
212 }
213
214 if (flags == 0 && !(qtype & F_QUERY) &&
215 option_bool(OPT_NODOTS_LOCAL) && !strchr(qdomain, '.') && namelen != 0)
216 /* don't forward A or AAAA queries for simple names, except the empty name */
217 flags = F_NOERR;
218
219 if (flags == F_NXDOMAIN && check_for_local_domain(qdomain, now))
220 flags = F_NOERR;
221
222 if (flags)
223 {
224 int logflags = 0;
225
226 if (flags == F_NXDOMAIN || flags == F_NOERR)
227 logflags = F_NEG | qtype;
228
229 log_query(logflags | flags | F_CONFIG | F_FORWARD, qdomain, *addrpp, NULL);
230 }
231 else if ((*type) & SERV_USE_RESOLV)
232 {
233 *type = 0; /* use normal servers for this domain */
234 *domain = NULL;
235 }
236 return flags;
237 }
238
239 static int forward_query(int udpfd, union mysockaddr *udpaddr,
240 struct all_addr *dst_addr, unsigned int dst_iface,
241 struct dns_header *header, size_t plen, time_t now,
242 struct frec *forward, int ad_reqd, int do_bit)
243 {
244 char *domain = NULL;
245 int type = 0, norebind = 0;
246 struct all_addr *addrp = NULL;
247 unsigned int flags = 0;
248 struct server *start = NULL;
249 #ifdef HAVE_DNSSEC
250 void *hash = hash_questions(header, plen, daemon->namebuff);
251 #else
252 unsigned int crc = questions_crc(header, plen, daemon->namebuff);
253 void *hash = &crc;
254 #endif
255 unsigned int gotname = extract_request(header, plen, daemon->namebuff, NULL);
256 unsigned char *pheader;
257
258 (void)do_bit;
259
260 /* may be no servers available. */
261 if (!daemon->servers)
262 forward = NULL;
263 else if (forward || (hash && (forward = lookup_frec_by_sender(ntohs(header->id), udpaddr, hash))))
264 {
265 /* If we didn't get an answer advertising a maximal packet in EDNS,
266 fall back to 1280, which should work everywhere on IPv6.
267 If that generates an answer, it will become the new default
268 for this server */
269 forward->flags |= FREC_TEST_PKTSZ;
270
271 #ifdef HAVE_DNSSEC
272 /* If we've already got an answer to this query, but we're awaiting keys for validation,
273 there's no point retrying the query, retry the key query instead...... */
274 if (forward->blocking_query)
275 {
276 int fd;
277
278 forward->flags &= ~FREC_TEST_PKTSZ;
279
280 while (forward->blocking_query)
281 forward = forward->blocking_query;
282
283 forward->flags |= FREC_TEST_PKTSZ;
284
285 blockdata_retrieve(forward->stash, forward->stash_len, (void *)header);
286 plen = forward->stash_len;
287
288 if (find_pseudoheader(header, plen, NULL, &pheader, NULL))
289 PUTSHORT((forward->flags & FREC_TEST_PKTSZ) ? SAFE_PKTSZ : forward->sentto->edns_pktsz, pheader);
290
291 if (forward->sentto->addr.sa.sa_family == AF_INET)
292 log_query(F_NOEXTRA | F_DNSSEC | F_IPV4, "retry", (struct all_addr *)&forward->sentto->addr.in.sin_addr, "dnssec");
293 #ifdef HAVE_IPV6
294 else
295 log_query(F_NOEXTRA | F_DNSSEC | F_IPV6, "retry", (struct all_addr *)&forward->sentto->addr.in6.sin6_addr, "dnssec");
296 #endif
297
298 if (forward->sentto->sfd)
299 fd = forward->sentto->sfd->fd;
300 else
301 {
302 #ifdef HAVE_IPV6
303 if (forward->sentto->addr.sa.sa_family == AF_INET6)
304 fd = forward->rfd6->fd;
305 else
306 #endif
307 fd = forward->rfd4->fd;
308 }
309
310 while (retry_send( sendto(fd, (char *)header, plen, 0,
311 &forward->sentto->addr.sa,
312 sa_len(&forward->sentto->addr))));
313
314 return 1;
315 }
316 #endif
317
318 /* retry on existing query, send to all available servers */
319 domain = forward->sentto->domain;
320 forward->sentto->failed_queries++;
321 if (!option_bool(OPT_ORDER))
322 {
323 forward->forwardall = 1;
324 daemon->last_server = NULL;
325 }
326 type = forward->sentto->flags & SERV_TYPE;
327 if (!(start = forward->sentto->next))
328 start = daemon->servers; /* at end of list, recycle */
329 header->id = htons(forward->new_id);
330 }
331 else
332 {
333 if (gotname)
334 flags = search_servers(now, &addrp, gotname, daemon->namebuff, &type, &domain, &norebind);
335
336 if (!flags && !(forward = get_new_frec(now, NULL, 0)))
337 /* table full - server failure. */
338 flags = F_NEG;
339
340 if (forward)
341 {
342 forward->source = *udpaddr;
343 forward->dest = *dst_addr;
344 forward->iface = dst_iface;
345 forward->orig_id = ntohs(header->id);
346 forward->new_id = get_id();
347 forward->fd = udpfd;
348 memcpy(forward->hash, hash, HASH_SIZE);
349 forward->forwardall = 0;
350 forward->flags = 0;
351 if (norebind)
352 forward->flags |= FREC_NOREBIND;
353 if (header->hb4 & HB4_CD)
354 forward->flags |= FREC_CHECKING_DISABLED;
355 if (ad_reqd)
356 forward->flags |= FREC_AD_QUESTION;
357 #ifdef HAVE_DNSSEC
358 forward->work_counter = DNSSEC_WORK;
359 if (do_bit)
360 forward->flags |= FREC_DO_QUESTION;
361 #endif
362
363 header->id = htons(forward->new_id);
364
365 /* In strict_order mode, always try servers in the order
366 specified in resolv.conf, if a domain is given
367 always try all the available servers,
368 otherwise, use the one last known to work. */
369
370 if (type == 0)
371 {
372 if (option_bool(OPT_ORDER))
373 start = daemon->servers;
374 else if (!(start = daemon->last_server) ||
375 daemon->forwardcount++ > FORWARD_TEST ||
376 difftime(now, daemon->forwardtime) > FORWARD_TIME)
377 {
378 start = daemon->servers;
379 forward->forwardall = 1;
380 daemon->forwardcount = 0;
381 daemon->forwardtime = now;
382 }
383 }
384 else
385 {
386 start = daemon->servers;
387 if (!option_bool(OPT_ORDER))
388 forward->forwardall = 1;
389 }
390 }
391 }
392
393 /* check for send errors here (no route to host)
394 if we fail to send to all nameservers, send back an error
395 packet straight away (helps modem users when offline) */
396
397 if (!flags && forward)
398 {
399 struct server *firstsentto = start;
400 int forwarded = 0;
401
402 /* If a query is retried, use the log_id for the retry when logging the answer. */
403 forward->log_id = daemon->log_id;
404
405 if (option_bool(OPT_ADD_MAC))
406 plen = add_mac(header, plen, ((char *) header) + daemon->packet_buff_sz, &forward->source);
407
408 if (option_bool(OPT_CLIENT_SUBNET))
409 {
410 size_t new = add_source_addr(header, plen, ((char *) header) + daemon->packet_buff_sz, &forward->source);
411 if (new != plen)
412 {
413 plen = new;
414 forward->flags |= FREC_HAS_SUBNET;
415 }
416 }
417
418 #ifdef HAVE_DNSSEC
419 if (option_bool(OPT_DNSSEC_VALID))
420 {
421 size_t new_plen = add_do_bit(header, plen, ((char *) header) + daemon->packet_buff_sz);
422
423 /* For debugging, set Checking Disabled, otherwise, have the upstream check too,
424 this allows it to select auth servers when one is returning bad data. */
425 if (option_bool(OPT_DNSSEC_DEBUG))
426 header->hb4 |= HB4_CD;
427
428 if (new_plen != plen)
429 forward->flags |= FREC_ADDED_PHEADER;
430
431 plen = new_plen;
432 }
433 #endif
434
435 while (1)
436 {
437 /* only send to servers dealing with our domain.
438 domain may be NULL, in which case server->domain
439 must be NULL also. */
440
441 if (type == (start->flags & SERV_TYPE) &&
442 (type != SERV_HAS_DOMAIN || hostname_isequal(domain, start->domain)) &&
443 !(start->flags & (SERV_LITERAL_ADDRESS | SERV_LOOP)))
444 {
445 int fd;
446
447 /* find server socket to use, may need to get random one. */
448 if (start->sfd)
449 fd = start->sfd->fd;
450 else
451 {
452 #ifdef HAVE_IPV6
453 if (start->addr.sa.sa_family == AF_INET6)
454 {
455 if (!forward->rfd6 &&
456 !(forward->rfd6 = allocate_rfd(AF_INET6)))
457 break;
458 daemon->rfd_save = forward->rfd6;
459 fd = forward->rfd6->fd;
460 }
461 else
462 #endif
463 {
464 if (!forward->rfd4 &&
465 !(forward->rfd4 = allocate_rfd(AF_INET)))
466 break;
467 daemon->rfd_save = forward->rfd4;
468 fd = forward->rfd4->fd;
469 }
470
471 #ifdef HAVE_CONNTRACK
472 /* Copy connection mark of incoming query to outgoing connection. */
473 if (option_bool(OPT_CONNTRACK))
474 {
475 unsigned int mark;
476 if (get_incoming_mark(&forward->source, &forward->dest, 0, &mark))
477 setsockopt(fd, SOL_SOCKET, SO_MARK, &mark, sizeof(unsigned int));
478 }
479 #endif
480 }
481
482 if (find_pseudoheader(header, plen, NULL, &pheader, NULL))
483 PUTSHORT((forward->flags & FREC_TEST_PKTSZ) ? SAFE_PKTSZ : start->edns_pktsz, pheader);
484
485 if (retry_send(sendto(fd, (char *)header, plen, 0,
486 &start->addr.sa,
487 sa_len(&start->addr))))
488 continue;
489
490 if (errno == 0)
491 {
492 /* Keep info in case we want to re-send this packet */
493 daemon->srv_save = start;
494 daemon->packet_len = plen;
495
496 if (!gotname)
497 strcpy(daemon->namebuff, "query");
498 if (start->addr.sa.sa_family == AF_INET)
499 log_query(F_SERVER | F_IPV4 | F_FORWARD, daemon->namebuff,
500 (struct all_addr *)&start->addr.in.sin_addr, NULL);
501 #ifdef HAVE_IPV6
502 else
503 log_query(F_SERVER | F_IPV6 | F_FORWARD, daemon->namebuff,
504 (struct all_addr *)&start->addr.in6.sin6_addr, NULL);
505 #endif
506 start->queries++;
507 forwarded = 1;
508 forward->sentto = start;
509 if (!forward->forwardall)
510 break;
511 forward->forwardall++;
512 }
513 }
514
515 if (!(start = start->next))
516 start = daemon->servers;
517
518 if (start == firstsentto)
519 break;
520 }
521
522 if (forwarded)
523 return 1;
524
525 /* could not send on, prepare to return */
526 header->id = htons(forward->orig_id);
527 free_frec(forward); /* cancel */
528 }
529
530 /* could not send on, return empty answer or address if known for whole domain */
531 if (udpfd != -1)
532 {
533 plen = setup_reply(header, plen, addrp, flags, daemon->local_ttl);
534 send_from(udpfd, option_bool(OPT_NOWILD) || option_bool(OPT_CLEVERBIND), (char *)header, plen, udpaddr, dst_addr, dst_iface);
535 }
536
537 return 0;
538 }
539
540 static size_t process_reply(struct dns_header *header, time_t now, struct server *server, size_t n, int check_rebind,
541 int no_cache, int cache_secure, int bogusanswer, int ad_reqd, int do_bit, int added_pheader,
542 int check_subnet, union mysockaddr *query_source)
543 {
544 unsigned char *pheader, *sizep;
545 char **sets = 0;
546 int munged = 0, is_sign;
547 size_t plen;
548
549 (void)ad_reqd;
550 (void)do_bit;
551 (void)bogusanswer;
552
553 #ifdef HAVE_IPSET
554 if (daemon->ipsets && extract_request(header, n, daemon->namebuff, NULL))
555 {
556 /* Similar algorithm to search_servers. */
557 struct ipsets *ipset_pos;
558 unsigned int namelen = strlen(daemon->namebuff);
559 unsigned int matchlen = 0;
560 for (ipset_pos = daemon->ipsets; ipset_pos; ipset_pos = ipset_pos->next)
561 {
562 unsigned int domainlen = strlen(ipset_pos->domain);
563 char *matchstart = daemon->namebuff + namelen - domainlen;
564 if (namelen >= domainlen && hostname_isequal(matchstart, ipset_pos->domain) &&
565 (domainlen == 0 || namelen == domainlen || *(matchstart - 1) == '.' ) &&
566 domainlen >= matchlen)
567 {
568 matchlen = domainlen;
569 sets = ipset_pos->sets;
570 }
571 }
572 }
573 #endif
574
575 /* If upstream is advertising a larger UDP packet size
576 than we allow, trim it so that we don't get overlarge
577 requests for the client. We can't do this for signed packets. */
578
579 if ((pheader = find_pseudoheader(header, n, &plen, &sizep, &is_sign)))
580 {
581 unsigned short udpsz;
582 unsigned char *psave = sizep;
583
584 GETSHORT(udpsz, sizep);
585
586 if (!is_sign && udpsz > daemon->edns_pktsz)
587 PUTSHORT(daemon->edns_pktsz, psave);
588
589 if (check_subnet && !check_source(header, plen, pheader, query_source))
590 {
591 my_syslog(LOG_WARNING, _("discarding DNS reply: subnet option mismatch"));
592 return 0;
593 }
594
595 if (added_pheader)
596 {
597 pheader = 0;
598 header->arcount = htons(0);
599 }
600 }
601
602 /* RFC 4035 sect 4.6 para 3 */
603 if (!is_sign && !option_bool(OPT_DNSSEC_PROXY))
604 header->hb4 &= ~HB4_AD;
605
606 if (OPCODE(header) != QUERY || (RCODE(header) != NOERROR && RCODE(header) != NXDOMAIN))
607 return resize_packet(header, n, pheader, plen);
608
609 /* Complain loudly if the upstream server is non-recursive. */
610 if (!(header->hb4 & HB4_RA) && RCODE(header) == NOERROR && ntohs(header->ancount) == 0 &&
611 server && !(server->flags & SERV_WARNED_RECURSIVE))
612 {
613 prettyprint_addr(&server->addr, daemon->namebuff);
614 my_syslog(LOG_WARNING, _("nameserver %s refused to do a recursive query"), daemon->namebuff);
615 if (!option_bool(OPT_LOG))
616 server->flags |= SERV_WARNED_RECURSIVE;
617 }
618
619 if (daemon->bogus_addr && RCODE(header) != NXDOMAIN &&
620 check_for_bogus_wildcard(header, n, daemon->namebuff, daemon->bogus_addr, now))
621 {
622 munged = 1;
623 SET_RCODE(header, NXDOMAIN);
624 header->hb3 &= ~HB3_AA;
625 cache_secure = 0;
626 }
627 else
628 {
629 int doctored = 0;
630
631 if (RCODE(header) == NXDOMAIN &&
632 extract_request(header, n, daemon->namebuff, NULL) &&
633 check_for_local_domain(daemon->namebuff, now))
634 {
635 /* if we forwarded a query for a locally known name (because it was for
636 an unknown type) and the answer is NXDOMAIN, convert that to NODATA,
637 since we know that the domain exists, even if upstream doesn't */
638 munged = 1;
639 header->hb3 |= HB3_AA;
640 SET_RCODE(header, NOERROR);
641 cache_secure = 0;
642 }
643
644 if (extract_addresses(header, n, daemon->namebuff, now, sets, is_sign, check_rebind, no_cache, cache_secure, &doctored))
645 {
646 my_syslog(LOG_WARNING, _("possible DNS-rebind attack detected: %s"), daemon->namebuff);
647 munged = 1;
648 cache_secure = 0;
649 }
650
651 if (doctored)
652 cache_secure = 0;
653 }
654
655 #ifdef HAVE_DNSSEC
656 if (bogusanswer && !(header->hb4 & HB4_CD))
657 {
658 if (!option_bool(OPT_DNSSEC_DEBUG))
659 {
660 /* Bogus reply, turn into SERVFAIL */
661 SET_RCODE(header, SERVFAIL);
662 munged = 1;
663 }
664 }
665
666 if (option_bool(OPT_DNSSEC_VALID))
667 header->hb4 &= ~HB4_AD;
668
669 if (!(header->hb4 & HB4_CD) && ad_reqd && cache_secure)
670 header->hb4 |= HB4_AD;
671
672 /* If the requestor didn't set the DO bit, don't return DNSSEC info. */
673 if (!do_bit)
674 n = filter_rrsigs(header, n);
675 #endif
676
677 /* do this after extract_addresses. Ensure NODATA reply and remove
678 nameserver info. */
679
680 if (munged)
681 {
682 header->ancount = htons(0);
683 header->nscount = htons(0);
684 header->arcount = htons(0);
685 header->hb3 &= ~HB3_TC;
686 }
687
688 /* the bogus-nxdomain stuff, doctor and NXDOMAIN->NODATA munging can all elide
689 sections of the packet. Find the new length here and put back pseudoheader
690 if it was removed. */
691 return resize_packet(header, n, pheader, plen);
692 }
693
694 /* sets new last_server */
695 void reply_query(int fd, int family, time_t now)
696 {
697 /* packet from peer server, extract data for cache, and send to
698 original requester */
699 struct dns_header *header;
700 union mysockaddr serveraddr;
701 struct frec *forward;
702 socklen_t addrlen = sizeof(serveraddr);
703 ssize_t n = recvfrom(fd, daemon->packet, daemon->packet_buff_sz, 0, &serveraddr.sa, &addrlen);
704 size_t nn;
705 struct server *server;
706 void *hash;
707 #ifndef HAVE_DNSSEC
708 unsigned int crc;
709 #endif
710
711 /* packet buffer overwritten */
712 daemon->srv_save = NULL;
713
714 /* Determine the address of the server replying so that we can mark that as good */
715 serveraddr.sa.sa_family = family;
716 #ifdef HAVE_IPV6
717 if (serveraddr.sa.sa_family == AF_INET6)
718 serveraddr.in6.sin6_flowinfo = 0;
719 #endif
720
721 header = (struct dns_header *)daemon->packet;
722
723 if (n < (int)sizeof(struct dns_header) || !(header->hb3 & HB3_QR))
724 return;
725
726 /* spoof check: answer must come from known server, */
727 for (server = daemon->servers; server; server = server->next)
728 if (!(server->flags & (SERV_LITERAL_ADDRESS | SERV_NO_ADDR)) &&
729 sockaddr_isequal(&server->addr, &serveraddr))
730 break;
731
732 if (!server)
733 return;
734
735 #ifdef HAVE_DNSSEC
736 hash = hash_questions(header, n, daemon->namebuff);
737 #else
738 hash = &crc;
739 crc = questions_crc(header, n, daemon->namebuff);
740 #endif
741
742 if (!(forward = lookup_frec(ntohs(header->id), hash)))
743 return;
744
745 /* log_query gets called indirectly all over the place, so
746 pass these in global variables - sorry. */
747 daemon->log_display_id = forward->log_id;
748 daemon->log_source_addr = &forward->source;
749
750 if (daemon->ignore_addr && RCODE(header) == NOERROR &&
751 check_for_ignored_address(header, n, daemon->ignore_addr))
752 return;
753
754 if (RCODE(header) == REFUSED &&
755 !option_bool(OPT_ORDER) &&
756 forward->forwardall == 0)
757 /* for broken servers, attempt to send to another one. */
758 {
759 unsigned char *pheader;
760 size_t plen;
761 int is_sign;
762
763 /* recreate query from reply */
764 pheader = find_pseudoheader(header, (size_t)n, &plen, NULL, &is_sign);
765 if (!is_sign)
766 {
767 header->ancount = htons(0);
768 header->nscount = htons(0);
769 header->arcount = htons(0);
770 if ((nn = resize_packet(header, (size_t)n, pheader, plen)))
771 {
772 header->hb3 &= ~(HB3_QR | HB3_TC);
773 forward_query(-1, NULL, NULL, 0, header, nn, now, forward, 0, 0);
774 return;
775 }
776 }
777 }
778
779 server = forward->sentto;
780 if ((forward->sentto->flags & SERV_TYPE) == 0)
781 {
782 if (RCODE(header) == REFUSED)
783 server = NULL;
784 else
785 {
786 struct server *last_server;
787
788 /* find good server by address if possible, otherwise assume the last one we sent to */
789 for (last_server = daemon->servers; last_server; last_server = last_server->next)
790 if (!(last_server->flags & (SERV_LITERAL_ADDRESS | SERV_HAS_DOMAIN | SERV_FOR_NODOTS | SERV_NO_ADDR)) &&
791 sockaddr_isequal(&last_server->addr, &serveraddr))
792 {
793 server = last_server;
794 break;
795 }
796 }
797 if (!option_bool(OPT_ALL_SERVERS))
798 daemon->last_server = server;
799 }
800
801 /* We tried resending to this server with a smaller maximum size and got an answer.
802 Make that permanent. To avoid reduxing the packet size for an single dropped packet,
803 only do this when we get a truncated answer, or one larger than the safe size. */
804 if (server && (forward->flags & FREC_TEST_PKTSZ) &&
805 ((header->hb3 & HB3_TC) || n >= SAFE_PKTSZ))
806 server->edns_pktsz = SAFE_PKTSZ;
807
808 /* If the answer is an error, keep the forward record in place in case
809 we get a good reply from another server. Kill it when we've
810 had replies from all to avoid filling the forwarding table when
811 everything is broken */
812 if (forward->forwardall == 0 || --forward->forwardall == 1 || RCODE(header) != SERVFAIL)
813 {
814 int check_rebind = 0, no_cache_dnssec = 0, cache_secure = 0, bogusanswer = 0;
815
816 if (option_bool(OPT_NO_REBIND))
817 check_rebind = !(forward->flags & FREC_NOREBIND);
818
819 /* Don't cache replies where DNSSEC validation was turned off, either
820 the upstream server told us so, or the original query specified it. */
821 if ((header->hb4 & HB4_CD) || (forward->flags & FREC_CHECKING_DISABLED))
822 no_cache_dnssec = 1;
823
824 #ifdef HAVE_DNSSEC
825 if (server && option_bool(OPT_DNSSEC_VALID) && !(forward->flags & FREC_CHECKING_DISABLED))
826 {
827 int status;
828
829 /* We've had a reply already, which we're validating. Ignore this duplicate */
830 if (forward->blocking_query)
831 return;
832
833 if (header->hb3 & HB3_TC)
834 {
835 /* Truncated answer can't be validated.
836 If this is an answer to a DNSSEC-generated query, we still
837 need to get the client to retry over TCP, so return
838 an answer with the TC bit set, even if the actual answer fits.
839 */
840 status = STAT_TRUNCATED;
841 }
842 else if (forward->flags & FREC_DNSKEY_QUERY)
843 status = dnssec_validate_by_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
844 else if (forward->flags & FREC_DS_QUERY)
845 {
846 status = dnssec_validate_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
847 /* Provably no DS, everything below is insecure, even if signatures are offered */
848 if (status == STAT_NO_DS)
849 /* We only cache sigs when we've validated a reply.
850 Avoid caching a reply with sigs if there's a vaildated break in the
851 DS chain, so we don't return replies from cache missing sigs. */
852 status = STAT_INSECURE_DS;
853 else if (status == STAT_NO_NS)
854 status = STAT_BOGUS;
855 }
856 else if (forward->flags & FREC_CHECK_NOSIGN)
857 {
858 status = dnssec_validate_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
859 if (status != STAT_NEED_KEY)
860 status = do_check_sign(forward, status, now, daemon->namebuff, daemon->keyname);
861 }
862 else
863 {
864 status = dnssec_validate_reply(now, header, n, daemon->namebuff, daemon->keyname, &forward->class, NULL, NULL);
865 if (status == STAT_NO_SIG)
866 {
867 if (option_bool(OPT_DNSSEC_NO_SIGN))
868 status = send_check_sign(forward, now, header, n, daemon->namebuff, daemon->keyname);
869 else
870 status = STAT_INSECURE;
871 }
872 }
873 /* Can't validate, as we're missing key data. Put this
874 answer aside, whilst we get that. */
875 if (status == STAT_NEED_DS || status == STAT_NEED_DS_NEG || status == STAT_NEED_KEY)
876 {
877 struct frec *new, *orig;
878
879 /* Free any saved query */
880 if (forward->stash)
881 blockdata_free(forward->stash);
882
883 /* Now save reply pending receipt of key data */
884 if (!(forward->stash = blockdata_alloc((char *)header, n)))
885 return;
886 forward->stash_len = n;
887
888 anotherkey:
889 /* Find the original query that started it all.... */
890 for (orig = forward; orig->dependent; orig = orig->dependent);
891
892 if (--orig->work_counter == 0 || !(new = get_new_frec(now, NULL, 1)))
893 status = STAT_INSECURE;
894 else
895 {
896 int fd;
897 struct frec *next = new->next;
898 *new = *forward; /* copy everything, then overwrite */
899 new->next = next;
900 new->blocking_query = NULL;
901 new->sentto = server;
902 new->rfd4 = NULL;
903 new->orig_domain = NULL;
904 #ifdef HAVE_IPV6
905 new->rfd6 = NULL;
906 #endif
907 new->flags &= ~(FREC_DNSKEY_QUERY | FREC_DS_QUERY | FREC_CHECK_NOSIGN);
908
909 new->dependent = forward; /* to find query awaiting new one. */
910 forward->blocking_query = new; /* for garbage cleaning */
911 /* validate routines leave name of required record in daemon->keyname */
912 if (status == STAT_NEED_KEY)
913 {
914 new->flags |= FREC_DNSKEY_QUERY;
915 nn = dnssec_generate_query(header, ((char *) header) + daemon->packet_buff_sz,
916 daemon->keyname, forward->class, T_DNSKEY, &server->addr, server->edns_pktsz);
917 }
918 else
919 {
920 if (status == STAT_NEED_DS_NEG)
921 new->flags |= FREC_CHECK_NOSIGN;
922 else
923 new->flags |= FREC_DS_QUERY;
924 nn = dnssec_generate_query(header,((char *) header) + daemon->packet_buff_sz,
925 daemon->keyname, forward->class, T_DS, &server->addr, server->edns_pktsz);
926 }
927 if ((hash = hash_questions(header, nn, daemon->namebuff)))
928 memcpy(new->hash, hash, HASH_SIZE);
929 new->new_id = get_id();
930 header->id = htons(new->new_id);
931 /* Save query for retransmission */
932 if (!(new->stash = blockdata_alloc((char *)header, nn)))
933 return;
934
935 new->stash_len = nn;
936
937 /* Don't resend this. */
938 daemon->srv_save = NULL;
939
940 if (server->sfd)
941 fd = server->sfd->fd;
942 else
943 {
944 fd = -1;
945 #ifdef HAVE_IPV6
946 if (server->addr.sa.sa_family == AF_INET6)
947 {
948 if (new->rfd6 || (new->rfd6 = allocate_rfd(AF_INET6)))
949 fd = new->rfd6->fd;
950 }
951 else
952 #endif
953 {
954 if (new->rfd4 || (new->rfd4 = allocate_rfd(AF_INET)))
955 fd = new->rfd4->fd;
956 }
957 }
958
959 if (fd != -1)
960 {
961 while (retry_send(sendto(fd, (char *)header, nn, 0,
962 &server->addr.sa,
963 sa_len(&server->addr))));
964 server->queries++;
965 }
966
967 return;
968 }
969 }
970
971 /* Ok, we reached far enough up the chain-of-trust that we can validate something.
972 Now wind back down, pulling back answers which wouldn't previously validate
973 and validate them with the new data. Note that if an answer needs multiple
974 keys to validate, we may find another key is needed, in which case we set off
975 down another branch of the tree. Once we get to the original answer
976 (FREC_DNSSEC_QUERY not set) and it validates, return it to the original requestor. */
977 while (forward->dependent)
978 {
979 struct frec *prev = forward->dependent;
980 free_frec(forward);
981 forward = prev;
982 forward->blocking_query = NULL; /* already gone */
983 blockdata_retrieve(forward->stash, forward->stash_len, (void *)header);
984 n = forward->stash_len;
985
986 if (status == STAT_SECURE)
987 {
988 if (forward->flags & FREC_DNSKEY_QUERY)
989 status = dnssec_validate_by_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
990 else if (forward->flags & FREC_DS_QUERY)
991 {
992 status = dnssec_validate_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
993 /* Provably no DS, everything below is insecure, even if signatures are offered */
994 if (status == STAT_NO_DS)
995 /* We only cache sigs when we've validated a reply.
996 Avoid caching a reply with sigs if there's a vaildated break in the
997 DS chain, so we don't return replies from cache missing sigs. */
998 status = STAT_INSECURE_DS;
999 else if (status == STAT_NO_NS)
1000 status = STAT_BOGUS;
1001 }
1002 else if (forward->flags & FREC_CHECK_NOSIGN)
1003 {
1004 status = dnssec_validate_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
1005 if (status != STAT_NEED_KEY)
1006 status = do_check_sign(forward, status, now, daemon->namebuff, daemon->keyname);
1007 }
1008 else
1009 {
1010 status = dnssec_validate_reply(now, header, n, daemon->namebuff, daemon->keyname, &forward->class, NULL, NULL);
1011 if (status == STAT_NO_SIG)
1012 {
1013 if (option_bool(OPT_DNSSEC_NO_SIGN))
1014 status = send_check_sign(forward, now, header, n, daemon->namebuff, daemon->keyname);
1015 else
1016 status = STAT_INSECURE;
1017 }
1018 }
1019
1020 if (status == STAT_NEED_DS || status == STAT_NEED_DS_NEG || status == STAT_NEED_KEY)
1021 goto anotherkey;
1022 }
1023 }
1024
1025 no_cache_dnssec = 0;
1026
1027 if (status == STAT_INSECURE_DS)
1028 {
1029 /* We only cache sigs when we've validated a reply.
1030 Avoid caching a reply with sigs if there's a vaildated break in the
1031 DS chain, so we don't return replies from cache missing sigs. */
1032 status = STAT_INSECURE;
1033 no_cache_dnssec = 1;
1034 }
1035
1036 if (status == STAT_TRUNCATED)
1037 header->hb3 |= HB3_TC;
1038 else
1039 {
1040 char *result, *domain = "result";
1041
1042 if (forward->work_counter == 0)
1043 {
1044 result = "ABANDONED";
1045 status = STAT_BOGUS;
1046 }
1047 else
1048 result = (status == STAT_SECURE ? "SECURE" : (status == STAT_INSECURE ? "INSECURE" : "BOGUS"));
1049
1050 if (status == STAT_BOGUS && extract_request(header, n, daemon->namebuff, NULL))
1051 domain = daemon->namebuff;
1052
1053 log_query(F_KEYTAG | F_SECSTAT, domain, NULL, result);
1054 }
1055
1056 if (status == STAT_SECURE)
1057 cache_secure = 1;
1058 else if (status == STAT_BOGUS)
1059 {
1060 no_cache_dnssec = 1;
1061 bogusanswer = 1;
1062 }
1063 }
1064 #endif
1065
1066 /* restore CD bit to the value in the query */
1067 if (forward->flags & FREC_CHECKING_DISABLED)
1068 header->hb4 |= HB4_CD;
1069 else
1070 header->hb4 &= ~HB4_CD;
1071
1072 if ((nn = process_reply(header, now, server, (size_t)n, check_rebind, no_cache_dnssec, cache_secure, bogusanswer,
1073 forward->flags & FREC_AD_QUESTION, forward->flags & FREC_DO_QUESTION,
1074 forward->flags & FREC_ADDED_PHEADER, forward->flags & FREC_HAS_SUBNET, &forward->source)))
1075 {
1076 header->id = htons(forward->orig_id);
1077 header->hb4 |= HB4_RA; /* recursion if available */
1078 send_from(forward->fd, option_bool(OPT_NOWILD) || option_bool (OPT_CLEVERBIND), daemon->packet, nn,
1079 &forward->source, &forward->dest, forward->iface);
1080 }
1081 free_frec(forward); /* cancel */
1082 }
1083 }
1084
1085
1086 void receive_query(struct listener *listen, time_t now)
1087 {
1088 struct dns_header *header = (struct dns_header *)daemon->packet;
1089 union mysockaddr source_addr;
1090 unsigned short type;
1091 struct all_addr dst_addr;
1092 struct in_addr netmask, dst_addr_4;
1093 size_t m;
1094 ssize_t n;
1095 int if_index = 0, auth_dns = 0;
1096 #ifdef HAVE_AUTH
1097 int local_auth = 0;
1098 #endif
1099 struct iovec iov[1];
1100 struct msghdr msg;
1101 struct cmsghdr *cmptr;
1102 union {
1103 struct cmsghdr align; /* this ensures alignment */
1104 #ifdef HAVE_IPV6
1105 char control6[CMSG_SPACE(sizeof(struct in6_pktinfo))];
1106 #endif
1107 #if defined(HAVE_LINUX_NETWORK)
1108 char control[CMSG_SPACE(sizeof(struct in_pktinfo))];
1109 #elif defined(IP_RECVDSTADDR) && defined(HAVE_SOLARIS_NETWORK)
1110 char control[CMSG_SPACE(sizeof(struct in_addr)) +
1111 CMSG_SPACE(sizeof(unsigned int))];
1112 #elif defined(IP_RECVDSTADDR)
1113 char control[CMSG_SPACE(sizeof(struct in_addr)) +
1114 CMSG_SPACE(sizeof(struct sockaddr_dl))];
1115 #endif
1116 } control_u;
1117 #ifdef HAVE_IPV6
1118 /* Can always get recvd interface for IPv6 */
1119 int check_dst = !option_bool(OPT_NOWILD) || listen->family == AF_INET6;
1120 #else
1121 int check_dst = !option_bool(OPT_NOWILD);
1122 #endif
1123
1124 /* packet buffer overwritten */
1125 daemon->srv_save = NULL;
1126
1127 dst_addr_4.s_addr = dst_addr.addr.addr4.s_addr = 0;
1128 netmask.s_addr = 0;
1129
1130 if (option_bool(OPT_NOWILD) && listen->iface)
1131 {
1132 auth_dns = listen->iface->dns_auth;
1133
1134 if (listen->family == AF_INET)
1135 {
1136 dst_addr_4 = dst_addr.addr.addr4 = listen->iface->addr.in.sin_addr;
1137 netmask = listen->iface->netmask;
1138 }
1139 }
1140
1141 iov[0].iov_base = daemon->packet;
1142 iov[0].iov_len = daemon->edns_pktsz;
1143
1144 msg.msg_control = control_u.control;
1145 msg.msg_controllen = sizeof(control_u);
1146 msg.msg_flags = 0;
1147 msg.msg_name = &source_addr;
1148 msg.msg_namelen = sizeof(source_addr);
1149 msg.msg_iov = iov;
1150 msg.msg_iovlen = 1;
1151
1152 if ((n = recvmsg(listen->fd, &msg, 0)) == -1)
1153 return;
1154
1155 if (n < (int)sizeof(struct dns_header) ||
1156 (msg.msg_flags & MSG_TRUNC) ||
1157 (header->hb3 & HB3_QR))
1158 return;
1159
1160 source_addr.sa.sa_family = listen->family;
1161
1162 if (listen->family == AF_INET)
1163 {
1164 /* Source-port == 0 is an error, we can't send back to that.
1165 http://www.ietf.org/mail-archive/web/dnsop/current/msg11441.html */
1166 if (source_addr.in.sin_port == 0)
1167 return;
1168 }
1169 #ifdef HAVE_IPV6
1170 else
1171 {
1172 /* Source-port == 0 is an error, we can't send back to that. */
1173 if (source_addr.in6.sin6_port == 0)
1174 return;
1175 source_addr.in6.sin6_flowinfo = 0;
1176 }
1177 #endif
1178
1179 /* We can be configured to only accept queries from at-most-one-hop-away addresses. */
1180 if (option_bool(OPT_LOCAL_SERVICE))
1181 {
1182 struct addrlist *addr;
1183 #ifdef HAVE_IPV6
1184 if (listen->family == AF_INET6)
1185 {
1186 for (addr = daemon->interface_addrs; addr; addr = addr->next)
1187 if ((addr->flags & ADDRLIST_IPV6) &&
1188 is_same_net6(&addr->addr.addr.addr6, &source_addr.in6.sin6_addr, addr->prefixlen))
1189 break;
1190 }
1191 else
1192 #endif
1193 {
1194 struct in_addr netmask;
1195 for (addr = daemon->interface_addrs; addr; addr = addr->next)
1196 {
1197 netmask.s_addr = htonl(~(in_addr_t)0 << (32 - addr->prefixlen));
1198 if (!(addr->flags & ADDRLIST_IPV6) &&
1199 is_same_net(addr->addr.addr.addr4, source_addr.in.sin_addr, netmask))
1200 break;
1201 }
1202 }
1203 if (!addr)
1204 {
1205 static int warned = 0;
1206 if (!warned)
1207 {
1208 my_syslog(LOG_WARNING, _("Ignoring query from non-local network"));
1209 warned = 1;
1210 }
1211 return;
1212 }
1213 }
1214
1215 if (check_dst)
1216 {
1217 struct ifreq ifr;
1218
1219 if (msg.msg_controllen < sizeof(struct cmsghdr))
1220 return;
1221
1222 #if defined(HAVE_LINUX_NETWORK)
1223 if (listen->family == AF_INET)
1224 for (cmptr = CMSG_FIRSTHDR(&msg); cmptr; cmptr = CMSG_NXTHDR(&msg, cmptr))
1225 if (cmptr->cmsg_level == IPPROTO_IP && cmptr->cmsg_type == IP_PKTINFO)
1226 {
1227 union {
1228 unsigned char *c;
1229 struct in_pktinfo *p;
1230 } p;
1231 p.c = CMSG_DATA(cmptr);
1232 dst_addr_4 = dst_addr.addr.addr4 = p.p->ipi_spec_dst;
1233 if_index = p.p->ipi_ifindex;
1234 }
1235 #elif defined(IP_RECVDSTADDR) && defined(IP_RECVIF)
1236 if (listen->family == AF_INET)
1237 {
1238 for (cmptr = CMSG_FIRSTHDR(&msg); cmptr; cmptr = CMSG_NXTHDR(&msg, cmptr))
1239 {
1240 union {
1241 unsigned char *c;
1242 unsigned int *i;
1243 struct in_addr *a;
1244 #ifndef HAVE_SOLARIS_NETWORK
1245 struct sockaddr_dl *s;
1246 #endif
1247 } p;
1248 p.c = CMSG_DATA(cmptr);
1249 if (cmptr->cmsg_level == IPPROTO_IP && cmptr->cmsg_type == IP_RECVDSTADDR)
1250 dst_addr_4 = dst_addr.addr.addr4 = *(p.a);
1251 else if (cmptr->cmsg_level == IPPROTO_IP && cmptr->cmsg_type == IP_RECVIF)
1252 #ifdef HAVE_SOLARIS_NETWORK
1253 if_index = *(p.i);
1254 #else
1255 if_index = p.s->sdl_index;
1256 #endif
1257 }
1258 }
1259 #endif
1260
1261 #ifdef HAVE_IPV6
1262 if (listen->family == AF_INET6)
1263 {
1264 for (cmptr = CMSG_FIRSTHDR(&msg); cmptr; cmptr = CMSG_NXTHDR(&msg, cmptr))
1265 if (cmptr->cmsg_level == IPPROTO_IPV6 && cmptr->cmsg_type == daemon->v6pktinfo)
1266 {
1267 union {
1268 unsigned char *c;
1269 struct in6_pktinfo *p;
1270 } p;
1271 p.c = CMSG_DATA(cmptr);
1272
1273 dst_addr.addr.addr6 = p.p->ipi6_addr;
1274 if_index = p.p->ipi6_ifindex;
1275 }
1276 }
1277 #endif
1278
1279 /* enforce available interface configuration */
1280
1281 if (!indextoname(listen->fd, if_index, ifr.ifr_name))
1282 return;
1283
1284 if (!iface_check(listen->family, &dst_addr, ifr.ifr_name, &auth_dns))
1285 {
1286 if (!option_bool(OPT_CLEVERBIND))
1287 enumerate_interfaces(0);
1288 if (!loopback_exception(listen->fd, listen->family, &dst_addr, ifr.ifr_name) &&
1289 !label_exception(if_index, listen->family, &dst_addr))
1290 return;
1291 }
1292
1293 if (listen->family == AF_INET && option_bool(OPT_LOCALISE))
1294 {
1295 struct irec *iface;
1296
1297 /* get the netmask of the interface whch has the address we were sent to.
1298 This is no neccessarily the interface we arrived on. */
1299
1300 for (iface = daemon->interfaces; iface; iface = iface->next)
1301 if (iface->addr.sa.sa_family == AF_INET &&
1302 iface->addr.in.sin_addr.s_addr == dst_addr_4.s_addr)
1303 break;
1304
1305 /* interface may be new */
1306 if (!iface && !option_bool(OPT_CLEVERBIND))
1307 enumerate_interfaces(0);
1308
1309 for (iface = daemon->interfaces; iface; iface = iface->next)
1310 if (iface->addr.sa.sa_family == AF_INET &&
1311 iface->addr.in.sin_addr.s_addr == dst_addr_4.s_addr)
1312 break;
1313
1314 /* If we failed, abandon localisation */
1315 if (iface)
1316 netmask = iface->netmask;
1317 else
1318 dst_addr_4.s_addr = 0;
1319 }
1320 }
1321
1322 /* log_query gets called indirectly all over the place, so
1323 pass these in global variables - sorry. */
1324 daemon->log_display_id = ++daemon->log_id;
1325 daemon->log_source_addr = &source_addr;
1326
1327 if (extract_request(header, (size_t)n, daemon->namebuff, &type))
1328 {
1329 #ifdef HAVE_AUTH
1330 struct auth_zone *zone;
1331 #endif
1332 char *types = querystr(auth_dns ? "auth" : "query", type);
1333
1334 if (listen->family == AF_INET)
1335 log_query(F_QUERY | F_IPV4 | F_FORWARD, daemon->namebuff,
1336 (struct all_addr *)&source_addr.in.sin_addr, types);
1337 #ifdef HAVE_IPV6
1338 else
1339 log_query(F_QUERY | F_IPV6 | F_FORWARD, daemon->namebuff,
1340 (struct all_addr *)&source_addr.in6.sin6_addr, types);
1341 #endif
1342
1343 #ifdef HAVE_AUTH
1344 /* find queries for zones we're authoritative for, and answer them directly */
1345 if (!auth_dns)
1346 for (zone = daemon->auth_zones; zone; zone = zone->next)
1347 if (in_zone(zone, daemon->namebuff, NULL))
1348 {
1349 auth_dns = 1;
1350 local_auth = 1;
1351 break;
1352 }
1353 #endif
1354
1355 #ifdef HAVE_LOOP
1356 /* Check for forwarding loop */
1357 if (detect_loop(daemon->namebuff, type))
1358 return;
1359 #endif
1360 }
1361
1362 #ifdef HAVE_AUTH
1363 if (auth_dns)
1364 {
1365 m = answer_auth(header, ((char *) header) + daemon->packet_buff_sz, (size_t)n, now, &source_addr, local_auth);
1366 if (m >= 1)
1367 {
1368 send_from(listen->fd, option_bool(OPT_NOWILD) || option_bool(OPT_CLEVERBIND),
1369 (char *)header, m, &source_addr, &dst_addr, if_index);
1370 daemon->auth_answer++;
1371 }
1372 }
1373 else
1374 #endif
1375 {
1376 int ad_reqd, do_bit;
1377 m = answer_request(header, ((char *) header) + daemon->packet_buff_sz, (size_t)n,
1378 dst_addr_4, netmask, now, &ad_reqd, &do_bit);
1379
1380 if (m >= 1)
1381 {
1382 send_from(listen->fd, option_bool(OPT_NOWILD) || option_bool(OPT_CLEVERBIND),
1383 (char *)header, m, &source_addr, &dst_addr, if_index);
1384 daemon->local_answer++;
1385 }
1386 else if (forward_query(listen->fd, &source_addr, &dst_addr, if_index,
1387 header, (size_t)n, now, NULL, ad_reqd, do_bit))
1388 daemon->queries_forwarded++;
1389 else
1390 daemon->local_answer++;
1391 }
1392 }
1393
1394 #ifdef HAVE_DNSSEC
1395
1396 /* UDP: we've got an unsigned answer, return STAT_INSECURE if we can prove there's no DS
1397 and therefore the answer shouldn't be signed, or STAT_BOGUS if it should be, or
1398 STAT_NEED_DS_NEG and keyname if we need to do the query. */
1399 static int send_check_sign(struct frec *forward, time_t now, struct dns_header *header, size_t plen,
1400 char *name, char *keyname)
1401 {
1402 int status = dnssec_chase_cname(now, header, plen, name, keyname);
1403
1404 if (status != STAT_INSECURE)
1405 return status;
1406
1407 /* Store the domain we're trying to check. */
1408 forward->name_start = strlen(name);
1409 forward->name_len = forward->name_start + 1;
1410 if (!(forward->orig_domain = blockdata_alloc(name, forward->name_len)))
1411 return STAT_BOGUS;
1412
1413 return do_check_sign(forward, 0, now, name, keyname);
1414 }
1415
1416 /* We either have a a reply (header non-NULL, or we need to start by looking in the cache */
1417 static int do_check_sign(struct frec *forward, int status, time_t now, char *name, char *keyname)
1418 {
1419 /* get domain we're checking back from blockdata store, it's stored on the original query. */
1420 while (forward->dependent)
1421 forward = forward->dependent;
1422
1423 blockdata_retrieve(forward->orig_domain, forward->name_len, name);
1424
1425 while (1)
1426 {
1427 char *p;
1428
1429 if (status == 0)
1430 {
1431 struct crec *crecp;
1432
1433 /* Haven't received answer, see if in cache */
1434 if (!(crecp = cache_find_by_name(NULL, &name[forward->name_start], now, F_DS)))
1435 {
1436 /* put name of DS record we're missing into keyname */
1437 strcpy(keyname, &name[forward->name_start]);
1438 /* and wait for reply to arrive */
1439 return STAT_NEED_DS_NEG;
1440 }
1441
1442 /* F_DNSSECOK misused in DS cache records to non-existance of NS record */
1443 if (!(crecp->flags & F_NEG))
1444 status = STAT_SECURE;
1445 else if (crecp->flags & F_DNSSECOK)
1446 status = STAT_NO_DS;
1447 else
1448 status = STAT_NO_NS;
1449 }
1450
1451 /* Have entered non-signed part of DNS tree. */
1452 if (status == STAT_NO_DS)
1453 return STAT_INSECURE;
1454
1455 if (status == STAT_BOGUS)
1456 return STAT_BOGUS;
1457
1458 /* There's a proven DS record, or we're within a zone, where there doesn't need
1459 to be a DS record. Add a name and try again.
1460 If we've already tried the whole name, then fail */
1461
1462 if (forward->name_start == 0)
1463 return STAT_BOGUS;
1464
1465 for (p = &name[forward->name_start-2]; (*p != '.') && (p != name); p--);
1466
1467 if (p != name)
1468 p++;
1469
1470 forward->name_start = p - name;
1471 status = 0; /* force to cache when we iterate. */
1472 }
1473 }
1474
1475 /* Move down from the root, until we find a signed non-existance of a DS, in which case
1476 an unsigned answer is OK, or we find a signed DS, in which case there should be
1477 a signature, and the answer is BOGUS */
1478 static int tcp_check_for_unsigned_zone(time_t now, struct dns_header *header, size_t plen, int class, char *name,
1479 char *keyname, struct server *server, int *keycount)
1480 {
1481 size_t m;
1482 unsigned char *packet, *payload;
1483 u16 *length;
1484 int status, name_len;
1485 struct blockdata *block;
1486
1487 char *name_start;
1488
1489 /* Get first insecure entry in CNAME chain */
1490 status = tcp_key_recurse(now, STAT_CHASE_CNAME, header, plen, class, name, keyname, server, keycount);
1491 if (status == STAT_BOGUS)
1492 return STAT_BOGUS;
1493
1494 if (!(packet = whine_malloc(65536 + MAXDNAME + RRFIXEDSZ + sizeof(u16))))
1495 return STAT_BOGUS;
1496
1497 payload = &packet[2];
1498 header = (struct dns_header *)payload;
1499 length = (u16 *)packet;
1500
1501 /* Stash the name away, since the buffer will be trashed when we recurse */
1502 name_len = strlen(name) + 1;
1503 name_start = name + name_len - 1;
1504
1505 if (!(block = blockdata_alloc(name, name_len)))
1506 {
1507 free(packet);
1508 return STAT_BOGUS;
1509 }
1510
1511 while (1)
1512 {
1513 unsigned char c1, c2;
1514 struct crec *crecp;
1515
1516 if (--(*keycount) == 0)
1517 {
1518 free(packet);
1519 blockdata_free(block);
1520 return STAT_BOGUS;
1521 }
1522
1523 while ((crecp = cache_find_by_name(NULL, name_start, now, F_DS)))
1524 {
1525 if ((crecp->flags & F_NEG) && (crecp->flags & F_DNSSECOK))
1526 {
1527 /* Found a secure denial of DS - delegation is indeed insecure */
1528 free(packet);
1529 blockdata_free(block);
1530 return STAT_INSECURE;
1531 }
1532
1533 /* Here, either there's a secure DS, or no NS and no DS, and therefore no delegation.
1534 Add another label and continue. */
1535
1536 if (name_start == name)
1537 {
1538 free(packet);
1539 blockdata_free(block);
1540 return STAT_BOGUS; /* run out of labels */
1541 }
1542
1543 name_start -= 2;
1544 while (*name_start != '.' && name_start != name)
1545 name_start--;
1546 if (name_start != name)
1547 name_start++;
1548 }
1549
1550 /* Can't find it in the cache, have to send a query */
1551
1552 m = dnssec_generate_query(header, ((char *) header) + 65536, name_start, class, T_DS, &server->addr, server->edns_pktsz);
1553
1554 *length = htons(m);
1555
1556 if (read_write(server->tcpfd, packet, m + sizeof(u16), 0) &&
1557 read_write(server->tcpfd, &c1, 1, 1) &&
1558 read_write(server->tcpfd, &c2, 1, 1) &&
1559 read_write(server->tcpfd, payload, (c1 << 8) | c2, 1))
1560 {
1561 m = (c1 << 8) | c2;
1562
1563 /* Note this trashes all three name workspaces */
1564 status = tcp_key_recurse(now, STAT_NEED_DS_NEG, header, m, class, name, keyname, server, keycount);
1565
1566 if (status == STAT_NO_DS)
1567 {
1568 /* Found a secure denial of DS - delegation is indeed insecure */
1569 free(packet);
1570 blockdata_free(block);
1571 return STAT_INSECURE;
1572 }
1573
1574 if (status == STAT_BOGUS)
1575 {
1576 free(packet);
1577 blockdata_free(block);
1578 return STAT_BOGUS;
1579 }
1580
1581 /* Here, either there's a secure DS, or no NS and no DS, and therefore no delegation.
1582 Add another label and continue. */
1583
1584 /* Get name we're checking back. */
1585 blockdata_retrieve(block, name_len, name);
1586
1587 if (name_start == name)
1588 {
1589 free(packet);
1590 blockdata_free(block);
1591 return STAT_BOGUS; /* run out of labels */
1592 }
1593
1594 name_start -= 2;
1595 while (*name_start != '.' && name_start != name)
1596 name_start--;
1597 if (name_start != name)
1598 name_start++;
1599 }
1600 else
1601 {
1602 /* IO failure */
1603 free(packet);
1604 blockdata_free(block);
1605 return STAT_BOGUS; /* run out of labels */
1606 }
1607 }
1608 }
1609
1610 static int tcp_key_recurse(time_t now, int status, struct dns_header *header, size_t n,
1611 int class, char *name, char *keyname, struct server *server, int *keycount)
1612 {
1613 /* Recurse up the key heirarchy */
1614 int new_status;
1615
1616 /* limit the amount of work we do, to avoid cycling forever on loops in the DNS */
1617 if (--(*keycount) == 0)
1618 return STAT_INSECURE;
1619
1620 if (status == STAT_NEED_KEY)
1621 new_status = dnssec_validate_by_ds(now, header, n, name, keyname, class);
1622 else if (status == STAT_NEED_DS || status == STAT_NEED_DS_NEG)
1623 {
1624 new_status = dnssec_validate_ds(now, header, n, name, keyname, class);
1625 if (status == STAT_NEED_DS)
1626 {
1627 if (new_status == STAT_NO_DS)
1628 new_status = STAT_INSECURE_DS;
1629 else if (new_status == STAT_NO_NS)
1630 new_status = STAT_BOGUS;
1631 }
1632 }
1633 else if (status == STAT_CHASE_CNAME)
1634 new_status = dnssec_chase_cname(now, header, n, name, keyname);
1635 else
1636 {
1637 new_status = dnssec_validate_reply(now, header, n, name, keyname, &class, NULL, NULL);
1638
1639 if (new_status == STAT_NO_SIG)
1640 {
1641 if (option_bool(OPT_DNSSEC_NO_SIGN))
1642 new_status = tcp_check_for_unsigned_zone(now, header, n, class, name, keyname, server, keycount);
1643 else
1644 new_status = STAT_INSECURE;
1645 }
1646 }
1647
1648 /* Can't validate because we need a key/DS whose name now in keyname.
1649 Make query for same, and recurse to validate */
1650 if (new_status == STAT_NEED_DS || new_status == STAT_NEED_KEY)
1651 {
1652 size_t m;
1653 unsigned char *packet = whine_malloc(65536 + MAXDNAME + RRFIXEDSZ + sizeof(u16));
1654 unsigned char *payload = &packet[2];
1655 struct dns_header *new_header = (struct dns_header *)payload;
1656 u16 *length = (u16 *)packet;
1657 unsigned char c1, c2;
1658
1659 if (!packet)
1660 return STAT_INSECURE;
1661
1662 another_tcp_key:
1663 m = dnssec_generate_query(new_header, ((char *) new_header) + 65536, keyname, class,
1664 new_status == STAT_NEED_KEY ? T_DNSKEY : T_DS, &server->addr, server->edns_pktsz);
1665
1666 *length = htons(m);
1667
1668 if (!read_write(server->tcpfd, packet, m + sizeof(u16), 0) ||
1669 !read_write(server->tcpfd, &c1, 1, 1) ||
1670 !read_write(server->tcpfd, &c2, 1, 1) ||
1671 !read_write(server->tcpfd, payload, (c1 << 8) | c2, 1))
1672 new_status = STAT_INSECURE;
1673 else
1674 {
1675 m = (c1 << 8) | c2;
1676
1677 new_status = tcp_key_recurse(now, new_status, new_header, m, class, name, keyname, server, keycount);
1678
1679 if (new_status == STAT_SECURE)
1680 {
1681 /* Reached a validated record, now try again at this level.
1682 Note that we may get ANOTHER NEED_* if an answer needs more than one key.
1683 If so, go round again. */
1684
1685 if (status == STAT_NEED_KEY)
1686 new_status = dnssec_validate_by_ds(now, header, n, name, keyname, class);
1687 else if (status == STAT_NEED_DS || status == STAT_NEED_DS_NEG)
1688 {
1689 new_status = dnssec_validate_ds(now, header, n, name, keyname, class);
1690 if (status == STAT_NEED_DS)
1691 {
1692 if (new_status == STAT_NO_DS)
1693 new_status = STAT_INSECURE_DS;
1694 else if (new_status == STAT_NO_NS)
1695 new_status = STAT_BOGUS; /* Validated no DS */
1696 }
1697 }
1698 else if (status == STAT_CHASE_CNAME)
1699 new_status = dnssec_chase_cname(now, header, n, name, keyname);
1700 else
1701 {
1702 new_status = dnssec_validate_reply(now, header, n, name, keyname, &class, NULL, NULL);
1703
1704 if (new_status == STAT_NO_SIG)
1705 {
1706 if (option_bool(OPT_DNSSEC_NO_SIGN))
1707 new_status = tcp_check_for_unsigned_zone(now, header, n, class, name, keyname, server, keycount);
1708 else
1709 new_status = STAT_INSECURE;
1710 }
1711 }
1712
1713 if (new_status == STAT_NEED_DS || new_status == STAT_NEED_KEY)
1714 goto another_tcp_key;
1715 }
1716 }
1717
1718 free(packet);
1719 }
1720 return new_status;
1721 }
1722 #endif
1723
1724
1725 /* The daemon forks before calling this: it should deal with one connection,
1726 blocking as neccessary, and then return. Note, need to be a bit careful
1727 about resources for debug mode, when the fork is suppressed: that's
1728 done by the caller. */
1729 unsigned char *tcp_request(int confd, time_t now,
1730 union mysockaddr *local_addr, struct in_addr netmask, int auth_dns)
1731 {
1732 size_t size = 0;
1733 int norebind = 0;
1734 #ifdef HAVE_AUTH
1735 int local_auth = 0;
1736 #endif
1737 int checking_disabled, ad_question, do_bit, added_pheader = 0;
1738 int check_subnet, no_cache_dnssec = 0, cache_secure = 0, bogusanswer = 0;
1739 size_t m;
1740 unsigned short qtype;
1741 unsigned int gotname;
1742 unsigned char c1, c2;
1743 /* Max TCP packet + slop + size */
1744 unsigned char *packet = whine_malloc(65536 + MAXDNAME + RRFIXEDSZ + sizeof(u16));
1745 unsigned char *payload = &packet[2];
1746 /* largest field in header is 16-bits, so this is still sufficiently aligned */
1747 struct dns_header *header = (struct dns_header *)payload;
1748 u16 *length = (u16 *)packet;
1749 struct server *last_server;
1750 struct in_addr dst_addr_4;
1751 union mysockaddr peer_addr;
1752 socklen_t peer_len = sizeof(union mysockaddr);
1753 int query_count = 0;
1754
1755 if (getpeername(confd, (struct sockaddr *)&peer_addr, &peer_len) == -1)
1756 return packet;
1757
1758 /* We can be configured to only accept queries from at-most-one-hop-away addresses. */
1759 if (option_bool(OPT_LOCAL_SERVICE))
1760 {
1761 struct addrlist *addr;
1762 #ifdef HAVE_IPV6
1763 if (peer_addr.sa.sa_family == AF_INET6)
1764 {
1765 for (addr = daemon->interface_addrs; addr; addr = addr->next)
1766 if ((addr->flags & ADDRLIST_IPV6) &&
1767 is_same_net6(&addr->addr.addr.addr6, &peer_addr.in6.sin6_addr, addr->prefixlen))
1768 break;
1769 }
1770 else
1771 #endif
1772 {
1773 struct in_addr netmask;
1774 for (addr = daemon->interface_addrs; addr; addr = addr->next)
1775 {
1776 netmask.s_addr = htonl(~(in_addr_t)0 << (32 - addr->prefixlen));
1777 if (!(addr->flags & ADDRLIST_IPV6) &&
1778 is_same_net(addr->addr.addr.addr4, peer_addr.in.sin_addr, netmask))
1779 break;
1780 }
1781 }
1782 if (!addr)
1783 {
1784 my_syslog(LOG_WARNING, _("Ignoring query from non-local network"));
1785 return packet;
1786 }
1787 }
1788
1789 while (1)
1790 {
1791 if (query_count == TCP_MAX_QUERIES ||
1792 !packet ||
1793 !read_write(confd, &c1, 1, 1) || !read_write(confd, &c2, 1, 1) ||
1794 !(size = c1 << 8 | c2) ||
1795 !read_write(confd, payload, size, 1))
1796 return packet;
1797
1798 if (size < (int)sizeof(struct dns_header))
1799 continue;
1800
1801 query_count++;
1802
1803 /* log_query gets called indirectly all over the place, so
1804 pass these in global variables - sorry. */
1805 daemon->log_display_id = ++daemon->log_id;
1806 daemon->log_source_addr = &peer_addr;
1807
1808 check_subnet = 0;
1809
1810 /* save state of "cd" flag in query */
1811 if ((checking_disabled = header->hb4 & HB4_CD))
1812 no_cache_dnssec = 1;
1813
1814 if ((gotname = extract_request(header, (unsigned int)size, daemon->namebuff, &qtype)))
1815 {
1816 #ifdef HAVE_AUTH
1817 struct auth_zone *zone;
1818 #endif
1819 char *types = querystr(auth_dns ? "auth" : "query", qtype);
1820
1821 if (peer_addr.sa.sa_family == AF_INET)
1822 log_query(F_QUERY | F_IPV4 | F_FORWARD, daemon->namebuff,
1823 (struct all_addr *)&peer_addr.in.sin_addr, types);
1824 #ifdef HAVE_IPV6
1825 else
1826 log_query(F_QUERY | F_IPV6 | F_FORWARD, daemon->namebuff,
1827 (struct all_addr *)&peer_addr.in6.sin6_addr, types);
1828 #endif
1829
1830 #ifdef HAVE_AUTH
1831 /* find queries for zones we're authoritative for, and answer them directly */
1832 if (!auth_dns)
1833 for (zone = daemon->auth_zones; zone; zone = zone->next)
1834 if (in_zone(zone, daemon->namebuff, NULL))
1835 {
1836 auth_dns = 1;
1837 local_auth = 1;
1838 break;
1839 }
1840 #endif
1841 }
1842
1843 if (local_addr->sa.sa_family == AF_INET)
1844 dst_addr_4 = local_addr->in.sin_addr;
1845 else
1846 dst_addr_4.s_addr = 0;
1847
1848 #ifdef HAVE_AUTH
1849 if (auth_dns)
1850 m = answer_auth(header, ((char *) header) + 65536, (size_t)size, now, &peer_addr, local_auth);
1851 else
1852 #endif
1853 {
1854 /* m > 0 if answered from cache */
1855 m = answer_request(header, ((char *) header) + 65536, (size_t)size,
1856 dst_addr_4, netmask, now, &ad_question, &do_bit);
1857
1858 /* Do this by steam now we're not in the select() loop */
1859 check_log_writer(NULL);
1860
1861 if (m == 0)
1862 {
1863 unsigned int flags = 0;
1864 struct all_addr *addrp = NULL;
1865 int type = 0;
1866 char *domain = NULL;
1867
1868 if (option_bool(OPT_ADD_MAC))
1869 size = add_mac(header, size, ((char *) header) + 65536, &peer_addr);
1870
1871 if (option_bool(OPT_CLIENT_SUBNET))
1872 {
1873 size_t new = add_source_addr(header, size, ((char *) header) + 65536, &peer_addr);
1874 if (size != new)
1875 {
1876 size = new;
1877 check_subnet = 1;
1878 }
1879 }
1880
1881 if (gotname)
1882 flags = search_servers(now, &addrp, gotname, daemon->namebuff, &type, &domain, &norebind);
1883
1884 if (type != 0 || option_bool(OPT_ORDER) || !daemon->last_server)
1885 last_server = daemon->servers;
1886 else
1887 last_server = daemon->last_server;
1888
1889 if (!flags && last_server)
1890 {
1891 struct server *firstsendto = NULL;
1892 #ifdef HAVE_DNSSEC
1893 unsigned char *newhash, hash[HASH_SIZE];
1894 if ((newhash = hash_questions(header, (unsigned int)size, daemon->namebuff)))
1895 memcpy(hash, newhash, HASH_SIZE);
1896 else
1897 memset(hash, 0, HASH_SIZE);
1898 #else
1899 unsigned int crc = questions_crc(header, (unsigned int)size, daemon->namebuff);
1900 #endif
1901 /* Loop round available servers until we succeed in connecting to one.
1902 Note that this code subtley ensures that consecutive queries on this connection
1903 which can go to the same server, do so. */
1904 while (1)
1905 {
1906 if (!firstsendto)
1907 firstsendto = last_server;
1908 else
1909 {
1910 if (!(last_server = last_server->next))
1911 last_server = daemon->servers;
1912
1913 if (last_server == firstsendto)
1914 break;
1915 }
1916
1917 /* server for wrong domain */
1918 if (type != (last_server->flags & SERV_TYPE) ||
1919 (type == SERV_HAS_DOMAIN && !hostname_isequal(domain, last_server->domain)) ||
1920 (last_server->flags & (SERV_LITERAL_ADDRESS | SERV_LOOP)))
1921 continue;
1922
1923 if (last_server->tcpfd == -1)
1924 {
1925 if ((last_server->tcpfd = socket(last_server->addr.sa.sa_family, SOCK_STREAM, 0)) == -1)
1926 continue;
1927
1928 #ifdef HAVE_CONNTRACK
1929 /* Copy connection mark of incoming query to outgoing connection. */
1930 if (option_bool(OPT_CONNTRACK))
1931 {
1932 unsigned int mark;
1933 struct all_addr local;
1934 #ifdef HAVE_IPV6
1935 if (local_addr->sa.sa_family == AF_INET6)
1936 local.addr.addr6 = local_addr->in6.sin6_addr;
1937 else
1938 #endif
1939 local.addr.addr4 = local_addr->in.sin_addr;
1940
1941 if (get_incoming_mark(&peer_addr, &local, 1, &mark))
1942 setsockopt(last_server->tcpfd, SOL_SOCKET, SO_MARK, &mark, sizeof(unsigned int));
1943 }
1944 #endif
1945
1946 if ((!local_bind(last_server->tcpfd, &last_server->source_addr, last_server->interface, 1) ||
1947 connect(last_server->tcpfd, &last_server->addr.sa, sa_len(&last_server->addr)) == -1))
1948 {
1949 close(last_server->tcpfd);
1950 last_server->tcpfd = -1;
1951 continue;
1952 }
1953
1954 #ifdef HAVE_DNSSEC
1955 if (option_bool(OPT_DNSSEC_VALID))
1956 {
1957 size_t new_size = add_do_bit(header, size, ((char *) header) + 65536);
1958
1959 /* For debugging, set Checking Disabled, otherwise, have the upstream check too,
1960 this allows it to select auth servers when one is returning bad data. */
1961 if (option_bool(OPT_DNSSEC_DEBUG))
1962 header->hb4 |= HB4_CD;
1963
1964 if (size != new_size)
1965 added_pheader = 1;
1966
1967 size = new_size;
1968 }
1969 #endif
1970 }
1971
1972 *length = htons(size);
1973
1974 /* get query name again for logging - may have been overwritten */
1975 if (!(gotname = extract_request(header, (unsigned int)size, daemon->namebuff, &qtype)))
1976 strcpy(daemon->namebuff, "query");
1977
1978 if (!read_write(last_server->tcpfd, packet, size + sizeof(u16), 0) ||
1979 !read_write(last_server->tcpfd, &c1, 1, 1) ||
1980 !read_write(last_server->tcpfd, &c2, 1, 1) ||
1981 !read_write(last_server->tcpfd, payload, (c1 << 8) | c2, 1))
1982 {
1983 close(last_server->tcpfd);
1984 last_server->tcpfd = -1;
1985 continue;
1986 }
1987
1988 m = (c1 << 8) | c2;
1989
1990 if (last_server->addr.sa.sa_family == AF_INET)
1991 log_query(F_SERVER | F_IPV4 | F_FORWARD, daemon->namebuff,
1992 (struct all_addr *)&last_server->addr.in.sin_addr, NULL);
1993 #ifdef HAVE_IPV6
1994 else
1995 log_query(F_SERVER | F_IPV6 | F_FORWARD, daemon->namebuff,
1996 (struct all_addr *)&last_server->addr.in6.sin6_addr, NULL);
1997 #endif
1998
1999 #ifdef HAVE_DNSSEC
2000 if (option_bool(OPT_DNSSEC_VALID) && !checking_disabled)
2001 {
2002 int keycount = DNSSEC_WORK; /* Limit to number of DNSSEC questions, to catch loops and avoid filling cache. */
2003 int status = tcp_key_recurse(now, STAT_TRUNCATED, header, m, 0, daemon->namebuff, daemon->keyname, last_server, &keycount);
2004 char *result, *domain = "result";
2005
2006 if (status == STAT_INSECURE_DS)
2007 {
2008 /* We only cache sigs when we've validated a reply.
2009 Avoid caching a reply with sigs if there's a vaildated break in the
2010 DS chain, so we don't return replies from cache missing sigs. */
2011 status = STAT_INSECURE;
2012 no_cache_dnssec = 1;
2013 }
2014
2015 if (keycount == 0)
2016 {
2017 result = "ABANDONED";
2018 status = STAT_BOGUS;
2019 }
2020 else
2021 result = (status == STAT_SECURE ? "SECURE" : (status == STAT_INSECURE ? "INSECURE" : "BOGUS"));
2022
2023 if (status == STAT_BOGUS && extract_request(header, m, daemon->namebuff, NULL))
2024 domain = daemon->namebuff;
2025
2026 log_query(F_KEYTAG | F_SECSTAT, domain, NULL, result);
2027
2028 if (status == STAT_BOGUS)
2029 {
2030 no_cache_dnssec = 1;
2031 bogusanswer = 1;
2032 }
2033
2034 if (status == STAT_SECURE)
2035 cache_secure = 1;
2036 }
2037 #endif
2038
2039 /* restore CD bit to the value in the query */
2040 if (checking_disabled)
2041 header->hb4 |= HB4_CD;
2042 else
2043 header->hb4 &= ~HB4_CD;
2044
2045 /* There's no point in updating the cache, since this process will exit and
2046 lose the information after a few queries. We make this call for the alias and
2047 bogus-nxdomain side-effects. */
2048 /* If the crc of the question section doesn't match the crc we sent, then
2049 someone might be attempting to insert bogus values into the cache by
2050 sending replies containing questions and bogus answers. */
2051 #ifdef HAVE_DNSSEC
2052 newhash = hash_questions(header, (unsigned int)m, daemon->namebuff);
2053 if (!newhash || memcmp(hash, newhash, HASH_SIZE) != 0)
2054 {
2055 m = 0;
2056 break;
2057 }
2058 #else
2059 if (crc != questions_crc(header, (unsigned int)m, daemon->namebuff))
2060 {
2061 m = 0;
2062 break;
2063 }
2064 #endif
2065
2066 m = process_reply(header, now, last_server, (unsigned int)m,
2067 option_bool(OPT_NO_REBIND) && !norebind, no_cache_dnssec, cache_secure, bogusanswer,
2068 ad_question, do_bit, added_pheader, check_subnet, &peer_addr);
2069
2070 break;
2071 }
2072 }
2073
2074 /* In case of local answer or no connections made. */
2075 if (m == 0)
2076 m = setup_reply(header, (unsigned int)size, addrp, flags, daemon->local_ttl);
2077 }
2078 }
2079
2080 check_log_writer(NULL);
2081
2082 *length = htons(m);
2083
2084 if (m == 0 || !read_write(confd, packet, m + sizeof(u16), 0))
2085 return packet;
2086 }
2087 }
2088
2089 static struct frec *allocate_frec(time_t now)
2090 {
2091 struct frec *f;
2092
2093 if ((f = (struct frec *)whine_malloc(sizeof(struct frec))))
2094 {
2095 f->next = daemon->frec_list;
2096 f->time = now;
2097 f->sentto = NULL;
2098 f->rfd4 = NULL;
2099 f->flags = 0;
2100 #ifdef HAVE_IPV6
2101 f->rfd6 = NULL;
2102 #endif
2103 #ifdef HAVE_DNSSEC
2104 f->dependent = NULL;
2105 f->blocking_query = NULL;
2106 f->stash = NULL;
2107 f->orig_domain = NULL;
2108 #endif
2109 daemon->frec_list = f;
2110 }
2111
2112 return f;
2113 }
2114
2115 struct randfd *allocate_rfd(int family)
2116 {
2117 static int finger = 0;
2118 int i;
2119
2120 /* limit the number of sockets we have open to avoid starvation of
2121 (eg) TFTP. Once we have a reasonable number, randomness should be OK */
2122
2123 for (i = 0; i < RANDOM_SOCKS; i++)
2124 if (daemon->randomsocks[i].refcount == 0)
2125 {
2126 if ((daemon->randomsocks[i].fd = random_sock(family)) == -1)
2127 break;
2128
2129 daemon->randomsocks[i].refcount = 1;
2130 daemon->randomsocks[i].family = family;
2131 return &daemon->randomsocks[i];
2132 }
2133
2134 /* No free ones or cannot get new socket, grab an existing one */
2135 for (i = 0; i < RANDOM_SOCKS; i++)
2136 {
2137 int j = (i+finger) % RANDOM_SOCKS;
2138 if (daemon->randomsocks[j].refcount != 0 &&
2139 daemon->randomsocks[j].family == family &&
2140 daemon->randomsocks[j].refcount != 0xffff)
2141 {
2142 finger = j;
2143 daemon->randomsocks[j].refcount++;
2144 return &daemon->randomsocks[j];
2145 }
2146 }
2147
2148 return NULL; /* doom */
2149 }
2150
2151 void free_rfd(struct randfd *rfd)
2152 {
2153 if (rfd && --(rfd->refcount) == 0)
2154 close(rfd->fd);
2155 }
2156
2157 static void free_frec(struct frec *f)
2158 {
2159 free_rfd(f->rfd4);
2160 f->rfd4 = NULL;
2161 f->sentto = NULL;
2162 f->flags = 0;
2163
2164 #ifdef HAVE_IPV6
2165 free_rfd(f->rfd6);
2166 f->rfd6 = NULL;
2167 #endif
2168
2169 #ifdef HAVE_DNSSEC
2170 if (f->stash)
2171 {
2172 blockdata_free(f->stash);
2173 f->stash = NULL;
2174 }
2175
2176 if (f->orig_domain)
2177 {
2178 blockdata_free(f->orig_domain);
2179 f->orig_domain = NULL;
2180 }
2181
2182 /* Anything we're waiting on is pointless now, too */
2183 if (f->blocking_query)
2184 free_frec(f->blocking_query);
2185 f->blocking_query = NULL;
2186 f->dependent = NULL;
2187 #endif
2188 }
2189
2190 /* if wait==NULL return a free or older than TIMEOUT record.
2191 else return *wait zero if one available, or *wait is delay to
2192 when the oldest in-use record will expire. Impose an absolute
2193 limit of 4*TIMEOUT before we wipe things (for random sockets).
2194 If force is set, always return a result, even if we have
2195 to allocate above the limit. */
2196 struct frec *get_new_frec(time_t now, int *wait, int force)
2197 {
2198 struct frec *f, *oldest, *target;
2199 int count;
2200
2201 if (wait)
2202 *wait = 0;
2203
2204 for (f = daemon->frec_list, oldest = NULL, target = NULL, count = 0; f; f = f->next, count++)
2205 if (!f->sentto)
2206 target = f;
2207 else
2208 {
2209 if (difftime(now, f->time) >= 4*TIMEOUT)
2210 {
2211 free_frec(f);
2212 target = f;
2213 }
2214
2215 if (!oldest || difftime(f->time, oldest->time) <= 0)
2216 oldest = f;
2217 }
2218
2219 if (target)
2220 {
2221 target->time = now;
2222 return target;
2223 }
2224
2225 /* can't find empty one, use oldest if there is one
2226 and it's older than timeout */
2227 if (oldest && ((int)difftime(now, oldest->time)) >= TIMEOUT)
2228 {
2229 /* keep stuff for twice timeout if we can by allocating a new
2230 record instead */
2231 if (difftime(now, oldest->time) < 2*TIMEOUT &&
2232 count <= daemon->ftabsize &&
2233 (f = allocate_frec(now)))
2234 return f;
2235
2236 if (!wait)
2237 {
2238 free_frec(oldest);
2239 oldest->time = now;
2240 }
2241 return oldest;
2242 }
2243
2244 /* none available, calculate time 'till oldest record expires */
2245 if (!force && count > daemon->ftabsize)
2246 {
2247 static time_t last_log = 0;
2248
2249 if (oldest && wait)
2250 *wait = oldest->time + (time_t)TIMEOUT - now;
2251
2252 if ((int)difftime(now, last_log) > 5)
2253 {
2254 last_log = now;
2255 my_syslog(LOG_WARNING, _("Maximum number of concurrent DNS queries reached (max: %d)"), daemon->ftabsize);
2256 }
2257
2258 return NULL;
2259 }
2260
2261 if (!(f = allocate_frec(now)) && wait)
2262 /* wait one second on malloc failure */
2263 *wait = 1;
2264
2265 return f; /* OK if malloc fails and this is NULL */
2266 }
2267
2268 /* crc is all-ones if not known. */
2269 static struct frec *lookup_frec(unsigned short id, void *hash)
2270 {
2271 struct frec *f;
2272
2273 for(f = daemon->frec_list; f; f = f->next)
2274 if (f->sentto && f->new_id == id &&
2275 (!hash || memcmp(hash, f->hash, HASH_SIZE) == 0))
2276 return f;
2277
2278 return NULL;
2279 }
2280
2281 static struct frec *lookup_frec_by_sender(unsigned short id,
2282 union mysockaddr *addr,
2283 void *hash)
2284 {
2285 struct frec *f;
2286
2287 for(f = daemon->frec_list; f; f = f->next)
2288 if (f->sentto &&
2289 f->orig_id == id &&
2290 memcmp(hash, f->hash, HASH_SIZE) == 0 &&
2291 sockaddr_isequal(&f->source, addr))
2292 return f;
2293
2294 return NULL;
2295 }
2296
2297 /* Send query packet again, if we can. */
2298 void resend_query()
2299 {
2300 if (daemon->srv_save)
2301 {
2302 int fd;
2303
2304 if (daemon->srv_save->sfd)
2305 fd = daemon->srv_save->sfd->fd;
2306 else if (daemon->rfd_save && daemon->rfd_save->refcount != 0)
2307 fd = daemon->rfd_save->fd;
2308 else
2309 return;
2310
2311 while(retry_send(sendto(fd, daemon->packet, daemon->packet_len, 0,
2312 &daemon->srv_save->addr.sa,
2313 sa_len(&daemon->srv_save->addr))));
2314 }
2315 }
2316
2317 /* A server record is going away, remove references to it */
2318 void server_gone(struct server *server)
2319 {
2320 struct frec *f;
2321
2322 for (f = daemon->frec_list; f; f = f->next)
2323 if (f->sentto && f->sentto == server)
2324 free_frec(f);
2325
2326 if (daemon->last_server == server)
2327 daemon->last_server = NULL;
2328
2329 if (daemon->srv_save == server)
2330 daemon->srv_save = NULL;
2331 }
2332
2333 /* return unique random ids. */
2334 static unsigned short get_id(void)
2335 {
2336 unsigned short ret = 0;
2337
2338 do
2339 ret = rand16();
2340 while (lookup_frec(ret, NULL));
2341
2342 return ret;
2343 }
2344
2345
2346
2347
2348