]> git.ipfire.org Git - people/ms/dnsmasq.git/blob - src/forward.c
Correctly sanitise DNS header bits in answer when recreating query for retry.
[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_AA | HB3_TC);
773 header->hb4 &= ~(HB4_RA | HB4_RCODE);
774 forward_query(-1, NULL, NULL, 0, header, nn, now, forward, 0, 0);
775 return;
776 }
777 }
778 }
779
780 server = forward->sentto;
781 if ((forward->sentto->flags & SERV_TYPE) == 0)
782 {
783 if (RCODE(header) == REFUSED)
784 server = NULL;
785 else
786 {
787 struct server *last_server;
788
789 /* find good server by address if possible, otherwise assume the last one we sent to */
790 for (last_server = daemon->servers; last_server; last_server = last_server->next)
791 if (!(last_server->flags & (SERV_LITERAL_ADDRESS | SERV_HAS_DOMAIN | SERV_FOR_NODOTS | SERV_NO_ADDR)) &&
792 sockaddr_isequal(&last_server->addr, &serveraddr))
793 {
794 server = last_server;
795 break;
796 }
797 }
798 if (!option_bool(OPT_ALL_SERVERS))
799 daemon->last_server = server;
800 }
801
802 /* We tried resending to this server with a smaller maximum size and got an answer.
803 Make that permanent. To avoid reduxing the packet size for an single dropped packet,
804 only do this when we get a truncated answer, or one larger than the safe size. */
805 if (server && (forward->flags & FREC_TEST_PKTSZ) &&
806 ((header->hb3 & HB3_TC) || n >= SAFE_PKTSZ))
807 server->edns_pktsz = SAFE_PKTSZ;
808
809 /* If the answer is an error, keep the forward record in place in case
810 we get a good reply from another server. Kill it when we've
811 had replies from all to avoid filling the forwarding table when
812 everything is broken */
813 if (forward->forwardall == 0 || --forward->forwardall == 1 || RCODE(header) != SERVFAIL)
814 {
815 int check_rebind = 0, no_cache_dnssec = 0, cache_secure = 0, bogusanswer = 0;
816
817 if (option_bool(OPT_NO_REBIND))
818 check_rebind = !(forward->flags & FREC_NOREBIND);
819
820 /* Don't cache replies where DNSSEC validation was turned off, either
821 the upstream server told us so, or the original query specified it. */
822 if ((header->hb4 & HB4_CD) || (forward->flags & FREC_CHECKING_DISABLED))
823 no_cache_dnssec = 1;
824
825 #ifdef HAVE_DNSSEC
826 if (server && option_bool(OPT_DNSSEC_VALID) && !(forward->flags & FREC_CHECKING_DISABLED))
827 {
828 int status;
829
830 /* We've had a reply already, which we're validating. Ignore this duplicate */
831 if (forward->blocking_query)
832 return;
833
834 if (header->hb3 & HB3_TC)
835 {
836 /* Truncated answer can't be validated.
837 If this is an answer to a DNSSEC-generated query, we still
838 need to get the client to retry over TCP, so return
839 an answer with the TC bit set, even if the actual answer fits.
840 */
841 status = STAT_TRUNCATED;
842 }
843 else if (forward->flags & FREC_DNSKEY_QUERY)
844 status = dnssec_validate_by_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
845 else if (forward->flags & FREC_DS_QUERY)
846 {
847 status = dnssec_validate_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
848 /* Provably no DS, everything below is insecure, even if signatures are offered */
849 if (status == STAT_NO_DS)
850 /* We only cache sigs when we've validated a reply.
851 Avoid caching a reply with sigs if there's a vaildated break in the
852 DS chain, so we don't return replies from cache missing sigs. */
853 status = STAT_INSECURE_DS;
854 else if (status == STAT_NO_NS)
855 status = STAT_BOGUS;
856 }
857 else if (forward->flags & FREC_CHECK_NOSIGN)
858 {
859 status = dnssec_validate_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
860 if (status != STAT_NEED_KEY)
861 status = do_check_sign(forward, status, now, daemon->namebuff, daemon->keyname);
862 }
863 else
864 {
865 status = dnssec_validate_reply(now, header, n, daemon->namebuff, daemon->keyname, &forward->class, NULL, NULL);
866 if (status == STAT_NO_SIG)
867 {
868 if (option_bool(OPT_DNSSEC_NO_SIGN))
869 status = send_check_sign(forward, now, header, n, daemon->namebuff, daemon->keyname);
870 else
871 status = STAT_INSECURE;
872 }
873 }
874 /* Can't validate, as we're missing key data. Put this
875 answer aside, whilst we get that. */
876 if (status == STAT_NEED_DS || status == STAT_NEED_DS_NEG || status == STAT_NEED_KEY)
877 {
878 struct frec *new, *orig;
879
880 /* Free any saved query */
881 if (forward->stash)
882 blockdata_free(forward->stash);
883
884 /* Now save reply pending receipt of key data */
885 if (!(forward->stash = blockdata_alloc((char *)header, n)))
886 return;
887 forward->stash_len = n;
888
889 anotherkey:
890 /* Find the original query that started it all.... */
891 for (orig = forward; orig->dependent; orig = orig->dependent);
892
893 if (--orig->work_counter == 0 || !(new = get_new_frec(now, NULL, 1)))
894 status = STAT_INSECURE;
895 else
896 {
897 int fd;
898 struct frec *next = new->next;
899 *new = *forward; /* copy everything, then overwrite */
900 new->next = next;
901 new->blocking_query = NULL;
902 new->sentto = server;
903 new->rfd4 = NULL;
904 new->orig_domain = NULL;
905 #ifdef HAVE_IPV6
906 new->rfd6 = NULL;
907 #endif
908 new->flags &= ~(FREC_DNSKEY_QUERY | FREC_DS_QUERY | FREC_CHECK_NOSIGN);
909
910 new->dependent = forward; /* to find query awaiting new one. */
911 forward->blocking_query = new; /* for garbage cleaning */
912 /* validate routines leave name of required record in daemon->keyname */
913 if (status == STAT_NEED_KEY)
914 {
915 new->flags |= FREC_DNSKEY_QUERY;
916 nn = dnssec_generate_query(header, ((char *) header) + daemon->packet_buff_sz,
917 daemon->keyname, forward->class, T_DNSKEY, &server->addr, server->edns_pktsz);
918 }
919 else
920 {
921 if (status == STAT_NEED_DS_NEG)
922 new->flags |= FREC_CHECK_NOSIGN;
923 else
924 new->flags |= FREC_DS_QUERY;
925 nn = dnssec_generate_query(header,((char *) header) + daemon->packet_buff_sz,
926 daemon->keyname, forward->class, T_DS, &server->addr, server->edns_pktsz);
927 }
928 if ((hash = hash_questions(header, nn, daemon->namebuff)))
929 memcpy(new->hash, hash, HASH_SIZE);
930 new->new_id = get_id();
931 header->id = htons(new->new_id);
932 /* Save query for retransmission */
933 if (!(new->stash = blockdata_alloc((char *)header, nn)))
934 return;
935
936 new->stash_len = nn;
937
938 /* Don't resend this. */
939 daemon->srv_save = NULL;
940
941 if (server->sfd)
942 fd = server->sfd->fd;
943 else
944 {
945 fd = -1;
946 #ifdef HAVE_IPV6
947 if (server->addr.sa.sa_family == AF_INET6)
948 {
949 if (new->rfd6 || (new->rfd6 = allocate_rfd(AF_INET6)))
950 fd = new->rfd6->fd;
951 }
952 else
953 #endif
954 {
955 if (new->rfd4 || (new->rfd4 = allocate_rfd(AF_INET)))
956 fd = new->rfd4->fd;
957 }
958 }
959
960 if (fd != -1)
961 {
962 while (retry_send(sendto(fd, (char *)header, nn, 0,
963 &server->addr.sa,
964 sa_len(&server->addr))));
965 server->queries++;
966 }
967
968 return;
969 }
970 }
971
972 /* Ok, we reached far enough up the chain-of-trust that we can validate something.
973 Now wind back down, pulling back answers which wouldn't previously validate
974 and validate them with the new data. Note that if an answer needs multiple
975 keys to validate, we may find another key is needed, in which case we set off
976 down another branch of the tree. Once we get to the original answer
977 (FREC_DNSSEC_QUERY not set) and it validates, return it to the original requestor. */
978 while (forward->dependent)
979 {
980 struct frec *prev = forward->dependent;
981 free_frec(forward);
982 forward = prev;
983 forward->blocking_query = NULL; /* already gone */
984 blockdata_retrieve(forward->stash, forward->stash_len, (void *)header);
985 n = forward->stash_len;
986
987 if (status == STAT_SECURE)
988 {
989 if (forward->flags & FREC_DNSKEY_QUERY)
990 status = dnssec_validate_by_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
991 else if (forward->flags & FREC_DS_QUERY)
992 {
993 status = dnssec_validate_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
994 /* Provably no DS, everything below is insecure, even if signatures are offered */
995 if (status == STAT_NO_DS)
996 /* We only cache sigs when we've validated a reply.
997 Avoid caching a reply with sigs if there's a vaildated break in the
998 DS chain, so we don't return replies from cache missing sigs. */
999 status = STAT_INSECURE_DS;
1000 else if (status == STAT_NO_NS)
1001 status = STAT_BOGUS;
1002 }
1003 else if (forward->flags & FREC_CHECK_NOSIGN)
1004 {
1005 status = dnssec_validate_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
1006 if (status != STAT_NEED_KEY)
1007 status = do_check_sign(forward, status, now, daemon->namebuff, daemon->keyname);
1008 }
1009 else
1010 {
1011 status = dnssec_validate_reply(now, header, n, daemon->namebuff, daemon->keyname, &forward->class, NULL, NULL);
1012 if (status == STAT_NO_SIG)
1013 {
1014 if (option_bool(OPT_DNSSEC_NO_SIGN))
1015 status = send_check_sign(forward, now, header, n, daemon->namebuff, daemon->keyname);
1016 else
1017 status = STAT_INSECURE;
1018 }
1019 }
1020
1021 if (status == STAT_NEED_DS || status == STAT_NEED_DS_NEG || status == STAT_NEED_KEY)
1022 goto anotherkey;
1023 }
1024 }
1025
1026 no_cache_dnssec = 0;
1027
1028 if (status == STAT_INSECURE_DS)
1029 {
1030 /* We only cache sigs when we've validated a reply.
1031 Avoid caching a reply with sigs if there's a vaildated break in the
1032 DS chain, so we don't return replies from cache missing sigs. */
1033 status = STAT_INSECURE;
1034 no_cache_dnssec = 1;
1035 }
1036
1037 if (status == STAT_TRUNCATED)
1038 header->hb3 |= HB3_TC;
1039 else
1040 {
1041 char *result, *domain = "result";
1042
1043 if (forward->work_counter == 0)
1044 {
1045 result = "ABANDONED";
1046 status = STAT_BOGUS;
1047 }
1048 else
1049 result = (status == STAT_SECURE ? "SECURE" : (status == STAT_INSECURE ? "INSECURE" : "BOGUS"));
1050
1051 if (status == STAT_BOGUS && extract_request(header, n, daemon->namebuff, NULL))
1052 domain = daemon->namebuff;
1053
1054 log_query(F_KEYTAG | F_SECSTAT, domain, NULL, result);
1055 }
1056
1057 if (status == STAT_SECURE)
1058 cache_secure = 1;
1059 else if (status == STAT_BOGUS)
1060 {
1061 no_cache_dnssec = 1;
1062 bogusanswer = 1;
1063 }
1064 }
1065 #endif
1066
1067 /* restore CD bit to the value in the query */
1068 if (forward->flags & FREC_CHECKING_DISABLED)
1069 header->hb4 |= HB4_CD;
1070 else
1071 header->hb4 &= ~HB4_CD;
1072
1073 if ((nn = process_reply(header, now, server, (size_t)n, check_rebind, no_cache_dnssec, cache_secure, bogusanswer,
1074 forward->flags & FREC_AD_QUESTION, forward->flags & FREC_DO_QUESTION,
1075 forward->flags & FREC_ADDED_PHEADER, forward->flags & FREC_HAS_SUBNET, &forward->source)))
1076 {
1077 header->id = htons(forward->orig_id);
1078 header->hb4 |= HB4_RA; /* recursion if available */
1079 send_from(forward->fd, option_bool(OPT_NOWILD) || option_bool (OPT_CLEVERBIND), daemon->packet, nn,
1080 &forward->source, &forward->dest, forward->iface);
1081 }
1082 free_frec(forward); /* cancel */
1083 }
1084 }
1085
1086
1087 void receive_query(struct listener *listen, time_t now)
1088 {
1089 struct dns_header *header = (struct dns_header *)daemon->packet;
1090 union mysockaddr source_addr;
1091 unsigned short type;
1092 struct all_addr dst_addr;
1093 struct in_addr netmask, dst_addr_4;
1094 size_t m;
1095 ssize_t n;
1096 int if_index = 0, auth_dns = 0;
1097 #ifdef HAVE_AUTH
1098 int local_auth = 0;
1099 #endif
1100 struct iovec iov[1];
1101 struct msghdr msg;
1102 struct cmsghdr *cmptr;
1103 union {
1104 struct cmsghdr align; /* this ensures alignment */
1105 #ifdef HAVE_IPV6
1106 char control6[CMSG_SPACE(sizeof(struct in6_pktinfo))];
1107 #endif
1108 #if defined(HAVE_LINUX_NETWORK)
1109 char control[CMSG_SPACE(sizeof(struct in_pktinfo))];
1110 #elif defined(IP_RECVDSTADDR) && defined(HAVE_SOLARIS_NETWORK)
1111 char control[CMSG_SPACE(sizeof(struct in_addr)) +
1112 CMSG_SPACE(sizeof(unsigned int))];
1113 #elif defined(IP_RECVDSTADDR)
1114 char control[CMSG_SPACE(sizeof(struct in_addr)) +
1115 CMSG_SPACE(sizeof(struct sockaddr_dl))];
1116 #endif
1117 } control_u;
1118 #ifdef HAVE_IPV6
1119 /* Can always get recvd interface for IPv6 */
1120 int check_dst = !option_bool(OPT_NOWILD) || listen->family == AF_INET6;
1121 #else
1122 int check_dst = !option_bool(OPT_NOWILD);
1123 #endif
1124
1125 /* packet buffer overwritten */
1126 daemon->srv_save = NULL;
1127
1128 dst_addr_4.s_addr = dst_addr.addr.addr4.s_addr = 0;
1129 netmask.s_addr = 0;
1130
1131 if (option_bool(OPT_NOWILD) && listen->iface)
1132 {
1133 auth_dns = listen->iface->dns_auth;
1134
1135 if (listen->family == AF_INET)
1136 {
1137 dst_addr_4 = dst_addr.addr.addr4 = listen->iface->addr.in.sin_addr;
1138 netmask = listen->iface->netmask;
1139 }
1140 }
1141
1142 iov[0].iov_base = daemon->packet;
1143 iov[0].iov_len = daemon->edns_pktsz;
1144
1145 msg.msg_control = control_u.control;
1146 msg.msg_controllen = sizeof(control_u);
1147 msg.msg_flags = 0;
1148 msg.msg_name = &source_addr;
1149 msg.msg_namelen = sizeof(source_addr);
1150 msg.msg_iov = iov;
1151 msg.msg_iovlen = 1;
1152
1153 if ((n = recvmsg(listen->fd, &msg, 0)) == -1)
1154 return;
1155
1156 if (n < (int)sizeof(struct dns_header) ||
1157 (msg.msg_flags & MSG_TRUNC) ||
1158 (header->hb3 & HB3_QR))
1159 return;
1160
1161 source_addr.sa.sa_family = listen->family;
1162
1163 if (listen->family == AF_INET)
1164 {
1165 /* Source-port == 0 is an error, we can't send back to that.
1166 http://www.ietf.org/mail-archive/web/dnsop/current/msg11441.html */
1167 if (source_addr.in.sin_port == 0)
1168 return;
1169 }
1170 #ifdef HAVE_IPV6
1171 else
1172 {
1173 /* Source-port == 0 is an error, we can't send back to that. */
1174 if (source_addr.in6.sin6_port == 0)
1175 return;
1176 source_addr.in6.sin6_flowinfo = 0;
1177 }
1178 #endif
1179
1180 /* We can be configured to only accept queries from at-most-one-hop-away addresses. */
1181 if (option_bool(OPT_LOCAL_SERVICE))
1182 {
1183 struct addrlist *addr;
1184 #ifdef HAVE_IPV6
1185 if (listen->family == AF_INET6)
1186 {
1187 for (addr = daemon->interface_addrs; addr; addr = addr->next)
1188 if ((addr->flags & ADDRLIST_IPV6) &&
1189 is_same_net6(&addr->addr.addr.addr6, &source_addr.in6.sin6_addr, addr->prefixlen))
1190 break;
1191 }
1192 else
1193 #endif
1194 {
1195 struct in_addr netmask;
1196 for (addr = daemon->interface_addrs; addr; addr = addr->next)
1197 {
1198 netmask.s_addr = htonl(~(in_addr_t)0 << (32 - addr->prefixlen));
1199 if (!(addr->flags & ADDRLIST_IPV6) &&
1200 is_same_net(addr->addr.addr.addr4, source_addr.in.sin_addr, netmask))
1201 break;
1202 }
1203 }
1204 if (!addr)
1205 {
1206 static int warned = 0;
1207 if (!warned)
1208 {
1209 my_syslog(LOG_WARNING, _("Ignoring query from non-local network"));
1210 warned = 1;
1211 }
1212 return;
1213 }
1214 }
1215
1216 if (check_dst)
1217 {
1218 struct ifreq ifr;
1219
1220 if (msg.msg_controllen < sizeof(struct cmsghdr))
1221 return;
1222
1223 #if defined(HAVE_LINUX_NETWORK)
1224 if (listen->family == AF_INET)
1225 for (cmptr = CMSG_FIRSTHDR(&msg); cmptr; cmptr = CMSG_NXTHDR(&msg, cmptr))
1226 if (cmptr->cmsg_level == IPPROTO_IP && cmptr->cmsg_type == IP_PKTINFO)
1227 {
1228 union {
1229 unsigned char *c;
1230 struct in_pktinfo *p;
1231 } p;
1232 p.c = CMSG_DATA(cmptr);
1233 dst_addr_4 = dst_addr.addr.addr4 = p.p->ipi_spec_dst;
1234 if_index = p.p->ipi_ifindex;
1235 }
1236 #elif defined(IP_RECVDSTADDR) && defined(IP_RECVIF)
1237 if (listen->family == AF_INET)
1238 {
1239 for (cmptr = CMSG_FIRSTHDR(&msg); cmptr; cmptr = CMSG_NXTHDR(&msg, cmptr))
1240 {
1241 union {
1242 unsigned char *c;
1243 unsigned int *i;
1244 struct in_addr *a;
1245 #ifndef HAVE_SOLARIS_NETWORK
1246 struct sockaddr_dl *s;
1247 #endif
1248 } p;
1249 p.c = CMSG_DATA(cmptr);
1250 if (cmptr->cmsg_level == IPPROTO_IP && cmptr->cmsg_type == IP_RECVDSTADDR)
1251 dst_addr_4 = dst_addr.addr.addr4 = *(p.a);
1252 else if (cmptr->cmsg_level == IPPROTO_IP && cmptr->cmsg_type == IP_RECVIF)
1253 #ifdef HAVE_SOLARIS_NETWORK
1254 if_index = *(p.i);
1255 #else
1256 if_index = p.s->sdl_index;
1257 #endif
1258 }
1259 }
1260 #endif
1261
1262 #ifdef HAVE_IPV6
1263 if (listen->family == AF_INET6)
1264 {
1265 for (cmptr = CMSG_FIRSTHDR(&msg); cmptr; cmptr = CMSG_NXTHDR(&msg, cmptr))
1266 if (cmptr->cmsg_level == IPPROTO_IPV6 && cmptr->cmsg_type == daemon->v6pktinfo)
1267 {
1268 union {
1269 unsigned char *c;
1270 struct in6_pktinfo *p;
1271 } p;
1272 p.c = CMSG_DATA(cmptr);
1273
1274 dst_addr.addr.addr6 = p.p->ipi6_addr;
1275 if_index = p.p->ipi6_ifindex;
1276 }
1277 }
1278 #endif
1279
1280 /* enforce available interface configuration */
1281
1282 if (!indextoname(listen->fd, if_index, ifr.ifr_name))
1283 return;
1284
1285 if (!iface_check(listen->family, &dst_addr, ifr.ifr_name, &auth_dns))
1286 {
1287 if (!option_bool(OPT_CLEVERBIND))
1288 enumerate_interfaces(0);
1289 if (!loopback_exception(listen->fd, listen->family, &dst_addr, ifr.ifr_name) &&
1290 !label_exception(if_index, listen->family, &dst_addr))
1291 return;
1292 }
1293
1294 if (listen->family == AF_INET && option_bool(OPT_LOCALISE))
1295 {
1296 struct irec *iface;
1297
1298 /* get the netmask of the interface whch has the address we were sent to.
1299 This is no neccessarily the interface we arrived on. */
1300
1301 for (iface = daemon->interfaces; iface; iface = iface->next)
1302 if (iface->addr.sa.sa_family == AF_INET &&
1303 iface->addr.in.sin_addr.s_addr == dst_addr_4.s_addr)
1304 break;
1305
1306 /* interface may be new */
1307 if (!iface && !option_bool(OPT_CLEVERBIND))
1308 enumerate_interfaces(0);
1309
1310 for (iface = daemon->interfaces; iface; iface = iface->next)
1311 if (iface->addr.sa.sa_family == AF_INET &&
1312 iface->addr.in.sin_addr.s_addr == dst_addr_4.s_addr)
1313 break;
1314
1315 /* If we failed, abandon localisation */
1316 if (iface)
1317 netmask = iface->netmask;
1318 else
1319 dst_addr_4.s_addr = 0;
1320 }
1321 }
1322
1323 /* log_query gets called indirectly all over the place, so
1324 pass these in global variables - sorry. */
1325 daemon->log_display_id = ++daemon->log_id;
1326 daemon->log_source_addr = &source_addr;
1327
1328 if (extract_request(header, (size_t)n, daemon->namebuff, &type))
1329 {
1330 #ifdef HAVE_AUTH
1331 struct auth_zone *zone;
1332 #endif
1333 char *types = querystr(auth_dns ? "auth" : "query", type);
1334
1335 if (listen->family == AF_INET)
1336 log_query(F_QUERY | F_IPV4 | F_FORWARD, daemon->namebuff,
1337 (struct all_addr *)&source_addr.in.sin_addr, types);
1338 #ifdef HAVE_IPV6
1339 else
1340 log_query(F_QUERY | F_IPV6 | F_FORWARD, daemon->namebuff,
1341 (struct all_addr *)&source_addr.in6.sin6_addr, types);
1342 #endif
1343
1344 #ifdef HAVE_AUTH
1345 /* find queries for zones we're authoritative for, and answer them directly */
1346 if (!auth_dns)
1347 for (zone = daemon->auth_zones; zone; zone = zone->next)
1348 if (in_zone(zone, daemon->namebuff, NULL))
1349 {
1350 auth_dns = 1;
1351 local_auth = 1;
1352 break;
1353 }
1354 #endif
1355
1356 #ifdef HAVE_LOOP
1357 /* Check for forwarding loop */
1358 if (detect_loop(daemon->namebuff, type))
1359 return;
1360 #endif
1361 }
1362
1363 #ifdef HAVE_AUTH
1364 if (auth_dns)
1365 {
1366 m = answer_auth(header, ((char *) header) + daemon->packet_buff_sz, (size_t)n, now, &source_addr, local_auth);
1367 if (m >= 1)
1368 {
1369 send_from(listen->fd, option_bool(OPT_NOWILD) || option_bool(OPT_CLEVERBIND),
1370 (char *)header, m, &source_addr, &dst_addr, if_index);
1371 daemon->auth_answer++;
1372 }
1373 }
1374 else
1375 #endif
1376 {
1377 int ad_reqd, do_bit;
1378 m = answer_request(header, ((char *) header) + daemon->packet_buff_sz, (size_t)n,
1379 dst_addr_4, netmask, now, &ad_reqd, &do_bit);
1380
1381 if (m >= 1)
1382 {
1383 send_from(listen->fd, option_bool(OPT_NOWILD) || option_bool(OPT_CLEVERBIND),
1384 (char *)header, m, &source_addr, &dst_addr, if_index);
1385 daemon->local_answer++;
1386 }
1387 else if (forward_query(listen->fd, &source_addr, &dst_addr, if_index,
1388 header, (size_t)n, now, NULL, ad_reqd, do_bit))
1389 daemon->queries_forwarded++;
1390 else
1391 daemon->local_answer++;
1392 }
1393 }
1394
1395 #ifdef HAVE_DNSSEC
1396
1397 /* UDP: we've got an unsigned answer, return STAT_INSECURE if we can prove there's no DS
1398 and therefore the answer shouldn't be signed, or STAT_BOGUS if it should be, or
1399 STAT_NEED_DS_NEG and keyname if we need to do the query. */
1400 static int send_check_sign(struct frec *forward, time_t now, struct dns_header *header, size_t plen,
1401 char *name, char *keyname)
1402 {
1403 int status = dnssec_chase_cname(now, header, plen, name, keyname);
1404
1405 if (status != STAT_INSECURE)
1406 return status;
1407
1408 /* Store the domain we're trying to check. */
1409 forward->name_start = strlen(name);
1410 forward->name_len = forward->name_start + 1;
1411 if (!(forward->orig_domain = blockdata_alloc(name, forward->name_len)))
1412 return STAT_BOGUS;
1413
1414 return do_check_sign(forward, 0, now, name, keyname);
1415 }
1416
1417 /* We either have a a reply (header non-NULL, or we need to start by looking in the cache */
1418 static int do_check_sign(struct frec *forward, int status, time_t now, char *name, char *keyname)
1419 {
1420 /* get domain we're checking back from blockdata store, it's stored on the original query. */
1421 while (forward->dependent)
1422 forward = forward->dependent;
1423
1424 blockdata_retrieve(forward->orig_domain, forward->name_len, name);
1425
1426 while (1)
1427 {
1428 char *p;
1429
1430 if (status == 0)
1431 {
1432 struct crec *crecp;
1433
1434 /* Haven't received answer, see if in cache */
1435 if (!(crecp = cache_find_by_name(NULL, &name[forward->name_start], now, F_DS)))
1436 {
1437 /* put name of DS record we're missing into keyname */
1438 strcpy(keyname, &name[forward->name_start]);
1439 /* and wait for reply to arrive */
1440 return STAT_NEED_DS_NEG;
1441 }
1442
1443 /* F_DNSSECOK misused in DS cache records to non-existance of NS record */
1444 if (!(crecp->flags & F_NEG))
1445 status = STAT_SECURE;
1446 else if (crecp->flags & F_DNSSECOK)
1447 status = STAT_NO_DS;
1448 else
1449 status = STAT_NO_NS;
1450 }
1451
1452 /* Have entered non-signed part of DNS tree. */
1453 if (status == STAT_NO_DS)
1454 return STAT_INSECURE;
1455
1456 if (status == STAT_BOGUS)
1457 return STAT_BOGUS;
1458
1459 /* There's a proven DS record, or we're within a zone, where there doesn't need
1460 to be a DS record. Add a name and try again.
1461 If we've already tried the whole name, then fail */
1462
1463 if (forward->name_start == 0)
1464 return STAT_BOGUS;
1465
1466 for (p = &name[forward->name_start-2]; (*p != '.') && (p != name); p--);
1467
1468 if (p != name)
1469 p++;
1470
1471 forward->name_start = p - name;
1472 status = 0; /* force to cache when we iterate. */
1473 }
1474 }
1475
1476 /* Move down from the root, until we find a signed non-existance of a DS, in which case
1477 an unsigned answer is OK, or we find a signed DS, in which case there should be
1478 a signature, and the answer is BOGUS */
1479 static int tcp_check_for_unsigned_zone(time_t now, struct dns_header *header, size_t plen, int class, char *name,
1480 char *keyname, struct server *server, int *keycount)
1481 {
1482 size_t m;
1483 unsigned char *packet, *payload;
1484 u16 *length;
1485 int status, name_len;
1486 struct blockdata *block;
1487
1488 char *name_start;
1489
1490 /* Get first insecure entry in CNAME chain */
1491 status = tcp_key_recurse(now, STAT_CHASE_CNAME, header, plen, class, name, keyname, server, keycount);
1492 if (status == STAT_BOGUS)
1493 return STAT_BOGUS;
1494
1495 if (!(packet = whine_malloc(65536 + MAXDNAME + RRFIXEDSZ + sizeof(u16))))
1496 return STAT_BOGUS;
1497
1498 payload = &packet[2];
1499 header = (struct dns_header *)payload;
1500 length = (u16 *)packet;
1501
1502 /* Stash the name away, since the buffer will be trashed when we recurse */
1503 name_len = strlen(name) + 1;
1504 name_start = name + name_len - 1;
1505
1506 if (!(block = blockdata_alloc(name, name_len)))
1507 {
1508 free(packet);
1509 return STAT_BOGUS;
1510 }
1511
1512 while (1)
1513 {
1514 unsigned char c1, c2;
1515 struct crec *crecp;
1516
1517 if (--(*keycount) == 0)
1518 {
1519 free(packet);
1520 blockdata_free(block);
1521 return STAT_BOGUS;
1522 }
1523
1524 while ((crecp = cache_find_by_name(NULL, name_start, now, F_DS)))
1525 {
1526 if ((crecp->flags & F_NEG) && (crecp->flags & F_DNSSECOK))
1527 {
1528 /* Found a secure denial of DS - delegation is indeed insecure */
1529 free(packet);
1530 blockdata_free(block);
1531 return STAT_INSECURE;
1532 }
1533
1534 /* Here, either there's a secure DS, or no NS and no DS, and therefore no delegation.
1535 Add another label and continue. */
1536
1537 if (name_start == name)
1538 {
1539 free(packet);
1540 blockdata_free(block);
1541 return STAT_BOGUS; /* run out of labels */
1542 }
1543
1544 name_start -= 2;
1545 while (*name_start != '.' && name_start != name)
1546 name_start--;
1547 if (name_start != name)
1548 name_start++;
1549 }
1550
1551 /* Can't find it in the cache, have to send a query */
1552
1553 m = dnssec_generate_query(header, ((char *) header) + 65536, name_start, class, T_DS, &server->addr, server->edns_pktsz);
1554
1555 *length = htons(m);
1556
1557 if (read_write(server->tcpfd, packet, m + sizeof(u16), 0) &&
1558 read_write(server->tcpfd, &c1, 1, 1) &&
1559 read_write(server->tcpfd, &c2, 1, 1) &&
1560 read_write(server->tcpfd, payload, (c1 << 8) | c2, 1))
1561 {
1562 m = (c1 << 8) | c2;
1563
1564 /* Note this trashes all three name workspaces */
1565 status = tcp_key_recurse(now, STAT_NEED_DS_NEG, header, m, class, name, keyname, server, keycount);
1566
1567 if (status == STAT_NO_DS)
1568 {
1569 /* Found a secure denial of DS - delegation is indeed insecure */
1570 free(packet);
1571 blockdata_free(block);
1572 return STAT_INSECURE;
1573 }
1574
1575 if (status == STAT_BOGUS)
1576 {
1577 free(packet);
1578 blockdata_free(block);
1579 return STAT_BOGUS;
1580 }
1581
1582 /* Here, either there's a secure DS, or no NS and no DS, and therefore no delegation.
1583 Add another label and continue. */
1584
1585 /* Get name we're checking back. */
1586 blockdata_retrieve(block, name_len, name);
1587
1588 if (name_start == name)
1589 {
1590 free(packet);
1591 blockdata_free(block);
1592 return STAT_BOGUS; /* run out of labels */
1593 }
1594
1595 name_start -= 2;
1596 while (*name_start != '.' && name_start != name)
1597 name_start--;
1598 if (name_start != name)
1599 name_start++;
1600 }
1601 else
1602 {
1603 /* IO failure */
1604 free(packet);
1605 blockdata_free(block);
1606 return STAT_BOGUS; /* run out of labels */
1607 }
1608 }
1609 }
1610
1611 static int tcp_key_recurse(time_t now, int status, struct dns_header *header, size_t n,
1612 int class, char *name, char *keyname, struct server *server, int *keycount)
1613 {
1614 /* Recurse up the key heirarchy */
1615 int new_status;
1616
1617 /* limit the amount of work we do, to avoid cycling forever on loops in the DNS */
1618 if (--(*keycount) == 0)
1619 return STAT_INSECURE;
1620
1621 if (status == STAT_NEED_KEY)
1622 new_status = dnssec_validate_by_ds(now, header, n, name, keyname, class);
1623 else if (status == STAT_NEED_DS || status == STAT_NEED_DS_NEG)
1624 {
1625 new_status = dnssec_validate_ds(now, header, n, name, keyname, class);
1626 if (status == STAT_NEED_DS)
1627 {
1628 if (new_status == STAT_NO_DS)
1629 new_status = STAT_INSECURE_DS;
1630 else if (new_status == STAT_NO_NS)
1631 new_status = STAT_BOGUS;
1632 }
1633 }
1634 else if (status == STAT_CHASE_CNAME)
1635 new_status = dnssec_chase_cname(now, header, n, name, keyname);
1636 else
1637 {
1638 new_status = dnssec_validate_reply(now, header, n, name, keyname, &class, NULL, NULL);
1639
1640 if (new_status == STAT_NO_SIG)
1641 {
1642 if (option_bool(OPT_DNSSEC_NO_SIGN))
1643 new_status = tcp_check_for_unsigned_zone(now, header, n, class, name, keyname, server, keycount);
1644 else
1645 new_status = STAT_INSECURE;
1646 }
1647 }
1648
1649 /* Can't validate because we need a key/DS whose name now in keyname.
1650 Make query for same, and recurse to validate */
1651 if (new_status == STAT_NEED_DS || new_status == STAT_NEED_KEY)
1652 {
1653 size_t m;
1654 unsigned char *packet = whine_malloc(65536 + MAXDNAME + RRFIXEDSZ + sizeof(u16));
1655 unsigned char *payload = &packet[2];
1656 struct dns_header *new_header = (struct dns_header *)payload;
1657 u16 *length = (u16 *)packet;
1658 unsigned char c1, c2;
1659
1660 if (!packet)
1661 return STAT_INSECURE;
1662
1663 another_tcp_key:
1664 m = dnssec_generate_query(new_header, ((char *) new_header) + 65536, keyname, class,
1665 new_status == STAT_NEED_KEY ? T_DNSKEY : T_DS, &server->addr, server->edns_pktsz);
1666
1667 *length = htons(m);
1668
1669 if (!read_write(server->tcpfd, packet, m + sizeof(u16), 0) ||
1670 !read_write(server->tcpfd, &c1, 1, 1) ||
1671 !read_write(server->tcpfd, &c2, 1, 1) ||
1672 !read_write(server->tcpfd, payload, (c1 << 8) | c2, 1))
1673 new_status = STAT_INSECURE;
1674 else
1675 {
1676 m = (c1 << 8) | c2;
1677
1678 new_status = tcp_key_recurse(now, new_status, new_header, m, class, name, keyname, server, keycount);
1679
1680 if (new_status == STAT_SECURE)
1681 {
1682 /* Reached a validated record, now try again at this level.
1683 Note that we may get ANOTHER NEED_* if an answer needs more than one key.
1684 If so, go round again. */
1685
1686 if (status == STAT_NEED_KEY)
1687 new_status = dnssec_validate_by_ds(now, header, n, name, keyname, class);
1688 else if (status == STAT_NEED_DS || status == STAT_NEED_DS_NEG)
1689 {
1690 new_status = dnssec_validate_ds(now, header, n, name, keyname, class);
1691 if (status == STAT_NEED_DS)
1692 {
1693 if (new_status == STAT_NO_DS)
1694 new_status = STAT_INSECURE_DS;
1695 else if (new_status == STAT_NO_NS)
1696 new_status = STAT_BOGUS; /* Validated no DS */
1697 }
1698 }
1699 else if (status == STAT_CHASE_CNAME)
1700 new_status = dnssec_chase_cname(now, header, n, name, keyname);
1701 else
1702 {
1703 new_status = dnssec_validate_reply(now, header, n, name, keyname, &class, NULL, NULL);
1704
1705 if (new_status == STAT_NO_SIG)
1706 {
1707 if (option_bool(OPT_DNSSEC_NO_SIGN))
1708 new_status = tcp_check_for_unsigned_zone(now, header, n, class, name, keyname, server, keycount);
1709 else
1710 new_status = STAT_INSECURE;
1711 }
1712 }
1713
1714 if (new_status == STAT_NEED_DS || new_status == STAT_NEED_KEY)
1715 goto another_tcp_key;
1716 }
1717 }
1718
1719 free(packet);
1720 }
1721 return new_status;
1722 }
1723 #endif
1724
1725
1726 /* The daemon forks before calling this: it should deal with one connection,
1727 blocking as neccessary, and then return. Note, need to be a bit careful
1728 about resources for debug mode, when the fork is suppressed: that's
1729 done by the caller. */
1730 unsigned char *tcp_request(int confd, time_t now,
1731 union mysockaddr *local_addr, struct in_addr netmask, int auth_dns)
1732 {
1733 size_t size = 0;
1734 int norebind = 0;
1735 #ifdef HAVE_AUTH
1736 int local_auth = 0;
1737 #endif
1738 int checking_disabled, ad_question, do_bit, added_pheader = 0;
1739 int check_subnet, no_cache_dnssec = 0, cache_secure = 0, bogusanswer = 0;
1740 size_t m;
1741 unsigned short qtype;
1742 unsigned int gotname;
1743 unsigned char c1, c2;
1744 /* Max TCP packet + slop + size */
1745 unsigned char *packet = whine_malloc(65536 + MAXDNAME + RRFIXEDSZ + sizeof(u16));
1746 unsigned char *payload = &packet[2];
1747 /* largest field in header is 16-bits, so this is still sufficiently aligned */
1748 struct dns_header *header = (struct dns_header *)payload;
1749 u16 *length = (u16 *)packet;
1750 struct server *last_server;
1751 struct in_addr dst_addr_4;
1752 union mysockaddr peer_addr;
1753 socklen_t peer_len = sizeof(union mysockaddr);
1754 int query_count = 0;
1755
1756 if (getpeername(confd, (struct sockaddr *)&peer_addr, &peer_len) == -1)
1757 return packet;
1758
1759 /* We can be configured to only accept queries from at-most-one-hop-away addresses. */
1760 if (option_bool(OPT_LOCAL_SERVICE))
1761 {
1762 struct addrlist *addr;
1763 #ifdef HAVE_IPV6
1764 if (peer_addr.sa.sa_family == AF_INET6)
1765 {
1766 for (addr = daemon->interface_addrs; addr; addr = addr->next)
1767 if ((addr->flags & ADDRLIST_IPV6) &&
1768 is_same_net6(&addr->addr.addr.addr6, &peer_addr.in6.sin6_addr, addr->prefixlen))
1769 break;
1770 }
1771 else
1772 #endif
1773 {
1774 struct in_addr netmask;
1775 for (addr = daemon->interface_addrs; addr; addr = addr->next)
1776 {
1777 netmask.s_addr = htonl(~(in_addr_t)0 << (32 - addr->prefixlen));
1778 if (!(addr->flags & ADDRLIST_IPV6) &&
1779 is_same_net(addr->addr.addr.addr4, peer_addr.in.sin_addr, netmask))
1780 break;
1781 }
1782 }
1783 if (!addr)
1784 {
1785 my_syslog(LOG_WARNING, _("Ignoring query from non-local network"));
1786 return packet;
1787 }
1788 }
1789
1790 while (1)
1791 {
1792 if (query_count == TCP_MAX_QUERIES ||
1793 !packet ||
1794 !read_write(confd, &c1, 1, 1) || !read_write(confd, &c2, 1, 1) ||
1795 !(size = c1 << 8 | c2) ||
1796 !read_write(confd, payload, size, 1))
1797 return packet;
1798
1799 if (size < (int)sizeof(struct dns_header))
1800 continue;
1801
1802 query_count++;
1803
1804 /* log_query gets called indirectly all over the place, so
1805 pass these in global variables - sorry. */
1806 daemon->log_display_id = ++daemon->log_id;
1807 daemon->log_source_addr = &peer_addr;
1808
1809 check_subnet = 0;
1810
1811 /* save state of "cd" flag in query */
1812 if ((checking_disabled = header->hb4 & HB4_CD))
1813 no_cache_dnssec = 1;
1814
1815 if ((gotname = extract_request(header, (unsigned int)size, daemon->namebuff, &qtype)))
1816 {
1817 #ifdef HAVE_AUTH
1818 struct auth_zone *zone;
1819 #endif
1820 char *types = querystr(auth_dns ? "auth" : "query", qtype);
1821
1822 if (peer_addr.sa.sa_family == AF_INET)
1823 log_query(F_QUERY | F_IPV4 | F_FORWARD, daemon->namebuff,
1824 (struct all_addr *)&peer_addr.in.sin_addr, types);
1825 #ifdef HAVE_IPV6
1826 else
1827 log_query(F_QUERY | F_IPV6 | F_FORWARD, daemon->namebuff,
1828 (struct all_addr *)&peer_addr.in6.sin6_addr, types);
1829 #endif
1830
1831 #ifdef HAVE_AUTH
1832 /* find queries for zones we're authoritative for, and answer them directly */
1833 if (!auth_dns)
1834 for (zone = daemon->auth_zones; zone; zone = zone->next)
1835 if (in_zone(zone, daemon->namebuff, NULL))
1836 {
1837 auth_dns = 1;
1838 local_auth = 1;
1839 break;
1840 }
1841 #endif
1842 }
1843
1844 if (local_addr->sa.sa_family == AF_INET)
1845 dst_addr_4 = local_addr->in.sin_addr;
1846 else
1847 dst_addr_4.s_addr = 0;
1848
1849 #ifdef HAVE_AUTH
1850 if (auth_dns)
1851 m = answer_auth(header, ((char *) header) + 65536, (size_t)size, now, &peer_addr, local_auth);
1852 else
1853 #endif
1854 {
1855 /* m > 0 if answered from cache */
1856 m = answer_request(header, ((char *) header) + 65536, (size_t)size,
1857 dst_addr_4, netmask, now, &ad_question, &do_bit);
1858
1859 /* Do this by steam now we're not in the select() loop */
1860 check_log_writer(NULL);
1861
1862 if (m == 0)
1863 {
1864 unsigned int flags = 0;
1865 struct all_addr *addrp = NULL;
1866 int type = 0;
1867 char *domain = NULL;
1868
1869 if (option_bool(OPT_ADD_MAC))
1870 size = add_mac(header, size, ((char *) header) + 65536, &peer_addr);
1871
1872 if (option_bool(OPT_CLIENT_SUBNET))
1873 {
1874 size_t new = add_source_addr(header, size, ((char *) header) + 65536, &peer_addr);
1875 if (size != new)
1876 {
1877 size = new;
1878 check_subnet = 1;
1879 }
1880 }
1881
1882 if (gotname)
1883 flags = search_servers(now, &addrp, gotname, daemon->namebuff, &type, &domain, &norebind);
1884
1885 if (type != 0 || option_bool(OPT_ORDER) || !daemon->last_server)
1886 last_server = daemon->servers;
1887 else
1888 last_server = daemon->last_server;
1889
1890 if (!flags && last_server)
1891 {
1892 struct server *firstsendto = NULL;
1893 #ifdef HAVE_DNSSEC
1894 unsigned char *newhash, hash[HASH_SIZE];
1895 if ((newhash = hash_questions(header, (unsigned int)size, daemon->namebuff)))
1896 memcpy(hash, newhash, HASH_SIZE);
1897 else
1898 memset(hash, 0, HASH_SIZE);
1899 #else
1900 unsigned int crc = questions_crc(header, (unsigned int)size, daemon->namebuff);
1901 #endif
1902 /* Loop round available servers until we succeed in connecting to one.
1903 Note that this code subtley ensures that consecutive queries on this connection
1904 which can go to the same server, do so. */
1905 while (1)
1906 {
1907 if (!firstsendto)
1908 firstsendto = last_server;
1909 else
1910 {
1911 if (!(last_server = last_server->next))
1912 last_server = daemon->servers;
1913
1914 if (last_server == firstsendto)
1915 break;
1916 }
1917
1918 /* server for wrong domain */
1919 if (type != (last_server->flags & SERV_TYPE) ||
1920 (type == SERV_HAS_DOMAIN && !hostname_isequal(domain, last_server->domain)) ||
1921 (last_server->flags & (SERV_LITERAL_ADDRESS | SERV_LOOP)))
1922 continue;
1923
1924 if (last_server->tcpfd == -1)
1925 {
1926 if ((last_server->tcpfd = socket(last_server->addr.sa.sa_family, SOCK_STREAM, 0)) == -1)
1927 continue;
1928
1929 #ifdef HAVE_CONNTRACK
1930 /* Copy connection mark of incoming query to outgoing connection. */
1931 if (option_bool(OPT_CONNTRACK))
1932 {
1933 unsigned int mark;
1934 struct all_addr local;
1935 #ifdef HAVE_IPV6
1936 if (local_addr->sa.sa_family == AF_INET6)
1937 local.addr.addr6 = local_addr->in6.sin6_addr;
1938 else
1939 #endif
1940 local.addr.addr4 = local_addr->in.sin_addr;
1941
1942 if (get_incoming_mark(&peer_addr, &local, 1, &mark))
1943 setsockopt(last_server->tcpfd, SOL_SOCKET, SO_MARK, &mark, sizeof(unsigned int));
1944 }
1945 #endif
1946
1947 if ((!local_bind(last_server->tcpfd, &last_server->source_addr, last_server->interface, 1) ||
1948 connect(last_server->tcpfd, &last_server->addr.sa, sa_len(&last_server->addr)) == -1))
1949 {
1950 close(last_server->tcpfd);
1951 last_server->tcpfd = -1;
1952 continue;
1953 }
1954
1955 #ifdef HAVE_DNSSEC
1956 if (option_bool(OPT_DNSSEC_VALID))
1957 {
1958 size_t new_size = add_do_bit(header, size, ((char *) header) + 65536);
1959
1960 /* For debugging, set Checking Disabled, otherwise, have the upstream check too,
1961 this allows it to select auth servers when one is returning bad data. */
1962 if (option_bool(OPT_DNSSEC_DEBUG))
1963 header->hb4 |= HB4_CD;
1964
1965 if (size != new_size)
1966 added_pheader = 1;
1967
1968 size = new_size;
1969 }
1970 #endif
1971 }
1972
1973 *length = htons(size);
1974
1975 /* get query name again for logging - may have been overwritten */
1976 if (!(gotname = extract_request(header, (unsigned int)size, daemon->namebuff, &qtype)))
1977 strcpy(daemon->namebuff, "query");
1978
1979 if (!read_write(last_server->tcpfd, packet, size + sizeof(u16), 0) ||
1980 !read_write(last_server->tcpfd, &c1, 1, 1) ||
1981 !read_write(last_server->tcpfd, &c2, 1, 1) ||
1982 !read_write(last_server->tcpfd, payload, (c1 << 8) | c2, 1))
1983 {
1984 close(last_server->tcpfd);
1985 last_server->tcpfd = -1;
1986 continue;
1987 }
1988
1989 m = (c1 << 8) | c2;
1990
1991 if (last_server->addr.sa.sa_family == AF_INET)
1992 log_query(F_SERVER | F_IPV4 | F_FORWARD, daemon->namebuff,
1993 (struct all_addr *)&last_server->addr.in.sin_addr, NULL);
1994 #ifdef HAVE_IPV6
1995 else
1996 log_query(F_SERVER | F_IPV6 | F_FORWARD, daemon->namebuff,
1997 (struct all_addr *)&last_server->addr.in6.sin6_addr, NULL);
1998 #endif
1999
2000 #ifdef HAVE_DNSSEC
2001 if (option_bool(OPT_DNSSEC_VALID) && !checking_disabled)
2002 {
2003 int keycount = DNSSEC_WORK; /* Limit to number of DNSSEC questions, to catch loops and avoid filling cache. */
2004 int status = tcp_key_recurse(now, STAT_TRUNCATED, header, m, 0, daemon->namebuff, daemon->keyname, last_server, &keycount);
2005 char *result, *domain = "result";
2006
2007 if (status == STAT_INSECURE_DS)
2008 {
2009 /* We only cache sigs when we've validated a reply.
2010 Avoid caching a reply with sigs if there's a vaildated break in the
2011 DS chain, so we don't return replies from cache missing sigs. */
2012 status = STAT_INSECURE;
2013 no_cache_dnssec = 1;
2014 }
2015
2016 if (keycount == 0)
2017 {
2018 result = "ABANDONED";
2019 status = STAT_BOGUS;
2020 }
2021 else
2022 result = (status == STAT_SECURE ? "SECURE" : (status == STAT_INSECURE ? "INSECURE" : "BOGUS"));
2023
2024 if (status == STAT_BOGUS && extract_request(header, m, daemon->namebuff, NULL))
2025 domain = daemon->namebuff;
2026
2027 log_query(F_KEYTAG | F_SECSTAT, domain, NULL, result);
2028
2029 if (status == STAT_BOGUS)
2030 {
2031 no_cache_dnssec = 1;
2032 bogusanswer = 1;
2033 }
2034
2035 if (status == STAT_SECURE)
2036 cache_secure = 1;
2037 }
2038 #endif
2039
2040 /* restore CD bit to the value in the query */
2041 if (checking_disabled)
2042 header->hb4 |= HB4_CD;
2043 else
2044 header->hb4 &= ~HB4_CD;
2045
2046 /* There's no point in updating the cache, since this process will exit and
2047 lose the information after a few queries. We make this call for the alias and
2048 bogus-nxdomain side-effects. */
2049 /* If the crc of the question section doesn't match the crc we sent, then
2050 someone might be attempting to insert bogus values into the cache by
2051 sending replies containing questions and bogus answers. */
2052 #ifdef HAVE_DNSSEC
2053 newhash = hash_questions(header, (unsigned int)m, daemon->namebuff);
2054 if (!newhash || memcmp(hash, newhash, HASH_SIZE) != 0)
2055 {
2056 m = 0;
2057 break;
2058 }
2059 #else
2060 if (crc != questions_crc(header, (unsigned int)m, daemon->namebuff))
2061 {
2062 m = 0;
2063 break;
2064 }
2065 #endif
2066
2067 m = process_reply(header, now, last_server, (unsigned int)m,
2068 option_bool(OPT_NO_REBIND) && !norebind, no_cache_dnssec, cache_secure, bogusanswer,
2069 ad_question, do_bit, added_pheader, check_subnet, &peer_addr);
2070
2071 break;
2072 }
2073 }
2074
2075 /* In case of local answer or no connections made. */
2076 if (m == 0)
2077 m = setup_reply(header, (unsigned int)size, addrp, flags, daemon->local_ttl);
2078 }
2079 }
2080
2081 check_log_writer(NULL);
2082
2083 *length = htons(m);
2084
2085 if (m == 0 || !read_write(confd, packet, m + sizeof(u16), 0))
2086 return packet;
2087 }
2088 }
2089
2090 static struct frec *allocate_frec(time_t now)
2091 {
2092 struct frec *f;
2093
2094 if ((f = (struct frec *)whine_malloc(sizeof(struct frec))))
2095 {
2096 f->next = daemon->frec_list;
2097 f->time = now;
2098 f->sentto = NULL;
2099 f->rfd4 = NULL;
2100 f->flags = 0;
2101 #ifdef HAVE_IPV6
2102 f->rfd6 = NULL;
2103 #endif
2104 #ifdef HAVE_DNSSEC
2105 f->dependent = NULL;
2106 f->blocking_query = NULL;
2107 f->stash = NULL;
2108 f->orig_domain = NULL;
2109 #endif
2110 daemon->frec_list = f;
2111 }
2112
2113 return f;
2114 }
2115
2116 struct randfd *allocate_rfd(int family)
2117 {
2118 static int finger = 0;
2119 int i;
2120
2121 /* limit the number of sockets we have open to avoid starvation of
2122 (eg) TFTP. Once we have a reasonable number, randomness should be OK */
2123
2124 for (i = 0; i < RANDOM_SOCKS; i++)
2125 if (daemon->randomsocks[i].refcount == 0)
2126 {
2127 if ((daemon->randomsocks[i].fd = random_sock(family)) == -1)
2128 break;
2129
2130 daemon->randomsocks[i].refcount = 1;
2131 daemon->randomsocks[i].family = family;
2132 return &daemon->randomsocks[i];
2133 }
2134
2135 /* No free ones or cannot get new socket, grab an existing one */
2136 for (i = 0; i < RANDOM_SOCKS; i++)
2137 {
2138 int j = (i+finger) % RANDOM_SOCKS;
2139 if (daemon->randomsocks[j].refcount != 0 &&
2140 daemon->randomsocks[j].family == family &&
2141 daemon->randomsocks[j].refcount != 0xffff)
2142 {
2143 finger = j;
2144 daemon->randomsocks[j].refcount++;
2145 return &daemon->randomsocks[j];
2146 }
2147 }
2148
2149 return NULL; /* doom */
2150 }
2151
2152 void free_rfd(struct randfd *rfd)
2153 {
2154 if (rfd && --(rfd->refcount) == 0)
2155 close(rfd->fd);
2156 }
2157
2158 static void free_frec(struct frec *f)
2159 {
2160 free_rfd(f->rfd4);
2161 f->rfd4 = NULL;
2162 f->sentto = NULL;
2163 f->flags = 0;
2164
2165 #ifdef HAVE_IPV6
2166 free_rfd(f->rfd6);
2167 f->rfd6 = NULL;
2168 #endif
2169
2170 #ifdef HAVE_DNSSEC
2171 if (f->stash)
2172 {
2173 blockdata_free(f->stash);
2174 f->stash = NULL;
2175 }
2176
2177 if (f->orig_domain)
2178 {
2179 blockdata_free(f->orig_domain);
2180 f->orig_domain = NULL;
2181 }
2182
2183 /* Anything we're waiting on is pointless now, too */
2184 if (f->blocking_query)
2185 free_frec(f->blocking_query);
2186 f->blocking_query = NULL;
2187 f->dependent = NULL;
2188 #endif
2189 }
2190
2191 /* if wait==NULL return a free or older than TIMEOUT record.
2192 else return *wait zero if one available, or *wait is delay to
2193 when the oldest in-use record will expire. Impose an absolute
2194 limit of 4*TIMEOUT before we wipe things (for random sockets).
2195 If force is set, always return a result, even if we have
2196 to allocate above the limit. */
2197 struct frec *get_new_frec(time_t now, int *wait, int force)
2198 {
2199 struct frec *f, *oldest, *target;
2200 int count;
2201
2202 if (wait)
2203 *wait = 0;
2204
2205 for (f = daemon->frec_list, oldest = NULL, target = NULL, count = 0; f; f = f->next, count++)
2206 if (!f->sentto)
2207 target = f;
2208 else
2209 {
2210 if (difftime(now, f->time) >= 4*TIMEOUT)
2211 {
2212 free_frec(f);
2213 target = f;
2214 }
2215
2216 if (!oldest || difftime(f->time, oldest->time) <= 0)
2217 oldest = f;
2218 }
2219
2220 if (target)
2221 {
2222 target->time = now;
2223 return target;
2224 }
2225
2226 /* can't find empty one, use oldest if there is one
2227 and it's older than timeout */
2228 if (oldest && ((int)difftime(now, oldest->time)) >= TIMEOUT)
2229 {
2230 /* keep stuff for twice timeout if we can by allocating a new
2231 record instead */
2232 if (difftime(now, oldest->time) < 2*TIMEOUT &&
2233 count <= daemon->ftabsize &&
2234 (f = allocate_frec(now)))
2235 return f;
2236
2237 if (!wait)
2238 {
2239 free_frec(oldest);
2240 oldest->time = now;
2241 }
2242 return oldest;
2243 }
2244
2245 /* none available, calculate time 'till oldest record expires */
2246 if (!force && count > daemon->ftabsize)
2247 {
2248 static time_t last_log = 0;
2249
2250 if (oldest && wait)
2251 *wait = oldest->time + (time_t)TIMEOUT - now;
2252
2253 if ((int)difftime(now, last_log) > 5)
2254 {
2255 last_log = now;
2256 my_syslog(LOG_WARNING, _("Maximum number of concurrent DNS queries reached (max: %d)"), daemon->ftabsize);
2257 }
2258
2259 return NULL;
2260 }
2261
2262 if (!(f = allocate_frec(now)) && wait)
2263 /* wait one second on malloc failure */
2264 *wait = 1;
2265
2266 return f; /* OK if malloc fails and this is NULL */
2267 }
2268
2269 /* crc is all-ones if not known. */
2270 static struct frec *lookup_frec(unsigned short id, void *hash)
2271 {
2272 struct frec *f;
2273
2274 for(f = daemon->frec_list; f; f = f->next)
2275 if (f->sentto && f->new_id == id &&
2276 (!hash || memcmp(hash, f->hash, HASH_SIZE) == 0))
2277 return f;
2278
2279 return NULL;
2280 }
2281
2282 static struct frec *lookup_frec_by_sender(unsigned short id,
2283 union mysockaddr *addr,
2284 void *hash)
2285 {
2286 struct frec *f;
2287
2288 for(f = daemon->frec_list; f; f = f->next)
2289 if (f->sentto &&
2290 f->orig_id == id &&
2291 memcmp(hash, f->hash, HASH_SIZE) == 0 &&
2292 sockaddr_isequal(&f->source, addr))
2293 return f;
2294
2295 return NULL;
2296 }
2297
2298 /* Send query packet again, if we can. */
2299 void resend_query()
2300 {
2301 if (daemon->srv_save)
2302 {
2303 int fd;
2304
2305 if (daemon->srv_save->sfd)
2306 fd = daemon->srv_save->sfd->fd;
2307 else if (daemon->rfd_save && daemon->rfd_save->refcount != 0)
2308 fd = daemon->rfd_save->fd;
2309 else
2310 return;
2311
2312 while(retry_send(sendto(fd, daemon->packet, daemon->packet_len, 0,
2313 &daemon->srv_save->addr.sa,
2314 sa_len(&daemon->srv_save->addr))));
2315 }
2316 }
2317
2318 /* A server record is going away, remove references to it */
2319 void server_gone(struct server *server)
2320 {
2321 struct frec *f;
2322
2323 for (f = daemon->frec_list; f; f = f->next)
2324 if (f->sentto && f->sentto == server)
2325 free_frec(f);
2326
2327 if (daemon->last_server == server)
2328 daemon->last_server = NULL;
2329
2330 if (daemon->srv_save == server)
2331 daemon->srv_save = NULL;
2332 }
2333
2334 /* return unique random ids. */
2335 static unsigned short get_id(void)
2336 {
2337 unsigned short ret = 0;
2338
2339 do
2340 ret = rand16();
2341 while (lookup_frec(ret, NULL));
2342
2343 return ret;
2344 }
2345
2346
2347
2348
2349