]> git.ipfire.org Git - people/ms/dnsmasq.git/blob - src/forward.c
Fix new poll() code for helper pipe. Removed CPU-spin.
[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_SIG)
855 {
856 if (option_bool(OPT_DNSSEC_NO_SIGN))
857 {
858 status = send_check_sign(forward, now, header, n, daemon->namebuff, daemon->keyname);
859 if (status == STAT_INSECURE)
860 status = STAT_INSECURE_DS;
861 }
862 else
863 status = STAT_INSECURE_DS;
864 }
865 else if (status == STAT_NO_NS)
866 status = STAT_BOGUS;
867 }
868 else if (forward->flags & FREC_CHECK_NOSIGN)
869 {
870 status = dnssec_validate_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
871 if (status != STAT_NEED_KEY)
872 status = do_check_sign(forward, status, now, daemon->namebuff, daemon->keyname);
873 }
874 else
875 {
876 status = dnssec_validate_reply(now, header, n, daemon->namebuff, daemon->keyname, &forward->class, NULL, NULL);
877 if (status == STAT_NO_SIG)
878 {
879 if (option_bool(OPT_DNSSEC_NO_SIGN))
880 status = send_check_sign(forward, now, header, n, daemon->namebuff, daemon->keyname);
881 else
882 status = STAT_INSECURE;
883 }
884 }
885 /* Can't validate, as we're missing key data. Put this
886 answer aside, whilst we get that. */
887 if (status == STAT_NEED_DS || status == STAT_NEED_DS_NEG || status == STAT_NEED_KEY)
888 {
889 struct frec *new, *orig;
890
891 /* Free any saved query */
892 if (forward->stash)
893 blockdata_free(forward->stash);
894
895 /* Now save reply pending receipt of key data */
896 if (!(forward->stash = blockdata_alloc((char *)header, n)))
897 return;
898 forward->stash_len = n;
899
900 anotherkey:
901 /* Find the original query that started it all.... */
902 for (orig = forward; orig->dependent; orig = orig->dependent);
903
904 if (--orig->work_counter == 0 || !(new = get_new_frec(now, NULL, 1)))
905 status = STAT_INSECURE;
906 else
907 {
908 int fd;
909 struct frec *next = new->next;
910 *new = *forward; /* copy everything, then overwrite */
911 new->next = next;
912 new->blocking_query = NULL;
913 new->sentto = server;
914 new->rfd4 = NULL;
915 new->orig_domain = NULL;
916 #ifdef HAVE_IPV6
917 new->rfd6 = NULL;
918 #endif
919 new->flags &= ~(FREC_DNSKEY_QUERY | FREC_DS_QUERY | FREC_CHECK_NOSIGN);
920
921 new->dependent = forward; /* to find query awaiting new one. */
922 forward->blocking_query = new; /* for garbage cleaning */
923 /* validate routines leave name of required record in daemon->keyname */
924 if (status == STAT_NEED_KEY)
925 {
926 new->flags |= FREC_DNSKEY_QUERY;
927 nn = dnssec_generate_query(header, ((char *) header) + daemon->packet_buff_sz,
928 daemon->keyname, forward->class, T_DNSKEY, &server->addr, server->edns_pktsz);
929 }
930 else
931 {
932 if (status == STAT_NEED_DS_NEG)
933 new->flags |= FREC_CHECK_NOSIGN;
934 else
935 new->flags |= FREC_DS_QUERY;
936 nn = dnssec_generate_query(header,((char *) header) + daemon->packet_buff_sz,
937 daemon->keyname, forward->class, T_DS, &server->addr, server->edns_pktsz);
938 }
939 if ((hash = hash_questions(header, nn, daemon->namebuff)))
940 memcpy(new->hash, hash, HASH_SIZE);
941 new->new_id = get_id();
942 header->id = htons(new->new_id);
943 /* Save query for retransmission */
944 if (!(new->stash = blockdata_alloc((char *)header, nn)))
945 return;
946
947 new->stash_len = nn;
948
949 /* Don't resend this. */
950 daemon->srv_save = NULL;
951
952 if (server->sfd)
953 fd = server->sfd->fd;
954 else
955 {
956 fd = -1;
957 #ifdef HAVE_IPV6
958 if (server->addr.sa.sa_family == AF_INET6)
959 {
960 if (new->rfd6 || (new->rfd6 = allocate_rfd(AF_INET6)))
961 fd = new->rfd6->fd;
962 }
963 else
964 #endif
965 {
966 if (new->rfd4 || (new->rfd4 = allocate_rfd(AF_INET)))
967 fd = new->rfd4->fd;
968 }
969 }
970
971 if (fd != -1)
972 {
973 while (retry_send(sendto(fd, (char *)header, nn, 0,
974 &server->addr.sa,
975 sa_len(&server->addr))));
976 server->queries++;
977 }
978
979 return;
980 }
981 }
982
983 /* Ok, we reached far enough up the chain-of-trust that we can validate something.
984 Now wind back down, pulling back answers which wouldn't previously validate
985 and validate them with the new data. Note that if an answer needs multiple
986 keys to validate, we may find another key is needed, in which case we set off
987 down another branch of the tree. Once we get to the original answer
988 (FREC_DNSSEC_QUERY not set) and it validates, return it to the original requestor. */
989 while (forward->dependent)
990 {
991 struct frec *prev = forward->dependent;
992 free_frec(forward);
993 forward = prev;
994 forward->blocking_query = NULL; /* already gone */
995 blockdata_retrieve(forward->stash, forward->stash_len, (void *)header);
996 n = forward->stash_len;
997
998 if (status == STAT_SECURE)
999 {
1000 if (forward->flags & FREC_DNSKEY_QUERY)
1001 status = dnssec_validate_by_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
1002 else if (forward->flags & FREC_DS_QUERY)
1003 {
1004 status = dnssec_validate_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
1005 /* Provably no DS, everything below is insecure, even if signatures are offered */
1006 if (status == STAT_NO_DS)
1007 /* We only cache sigs when we've validated a reply.
1008 Avoid caching a reply with sigs if there's a vaildated break in the
1009 DS chain, so we don't return replies from cache missing sigs. */
1010 status = STAT_INSECURE_DS;
1011 else if (status == STAT_NO_SIG)
1012 {
1013 if (option_bool(OPT_DNSSEC_NO_SIGN))
1014 {
1015 status = send_check_sign(forward, now, header, n, daemon->namebuff, daemon->keyname);
1016 if (status == STAT_INSECURE)
1017 status = STAT_INSECURE_DS;
1018 }
1019 else
1020 status = STAT_INSECURE_DS;
1021 }
1022 else if (status == STAT_NO_NS)
1023 status = STAT_BOGUS;
1024 }
1025 else if (forward->flags & FREC_CHECK_NOSIGN)
1026 {
1027 status = dnssec_validate_ds(now, header, n, daemon->namebuff, daemon->keyname, forward->class);
1028 if (status != STAT_NEED_KEY)
1029 status = do_check_sign(forward, status, now, daemon->namebuff, daemon->keyname);
1030 }
1031 else
1032 {
1033 status = dnssec_validate_reply(now, header, n, daemon->namebuff, daemon->keyname, &forward->class, NULL, NULL);
1034 if (status == STAT_NO_SIG)
1035 {
1036 if (option_bool(OPT_DNSSEC_NO_SIGN))
1037 status = send_check_sign(forward, now, header, n, daemon->namebuff, daemon->keyname);
1038 else
1039 status = STAT_INSECURE;
1040 }
1041 }
1042
1043 if (status == STAT_NEED_DS || status == STAT_NEED_DS_NEG || status == STAT_NEED_KEY)
1044 goto anotherkey;
1045 }
1046 }
1047
1048 no_cache_dnssec = 0;
1049
1050 if (status == STAT_INSECURE_DS)
1051 {
1052 /* We only cache sigs when we've validated a reply.
1053 Avoid caching a reply with sigs if there's a vaildated break in the
1054 DS chain, so we don't return replies from cache missing sigs. */
1055 status = STAT_INSECURE;
1056 no_cache_dnssec = 1;
1057 }
1058
1059 if (status == STAT_TRUNCATED)
1060 header->hb3 |= HB3_TC;
1061 else
1062 {
1063 char *result, *domain = "result";
1064
1065 if (forward->work_counter == 0)
1066 {
1067 result = "ABANDONED";
1068 status = STAT_BOGUS;
1069 }
1070 else
1071 result = (status == STAT_SECURE ? "SECURE" : (status == STAT_INSECURE ? "INSECURE" : "BOGUS"));
1072
1073 if (status == STAT_BOGUS && extract_request(header, n, daemon->namebuff, NULL))
1074 domain = daemon->namebuff;
1075
1076 log_query(F_KEYTAG | F_SECSTAT, domain, NULL, result);
1077 }
1078
1079 if (status == STAT_SECURE)
1080 cache_secure = 1;
1081 else if (status == STAT_BOGUS)
1082 {
1083 no_cache_dnssec = 1;
1084 bogusanswer = 1;
1085 }
1086 }
1087 #endif
1088
1089 /* restore CD bit to the value in the query */
1090 if (forward->flags & FREC_CHECKING_DISABLED)
1091 header->hb4 |= HB4_CD;
1092 else
1093 header->hb4 &= ~HB4_CD;
1094
1095 if ((nn = process_reply(header, now, server, (size_t)n, check_rebind, no_cache_dnssec, cache_secure, bogusanswer,
1096 forward->flags & FREC_AD_QUESTION, forward->flags & FREC_DO_QUESTION,
1097 forward->flags & FREC_ADDED_PHEADER, forward->flags & FREC_HAS_SUBNET, &forward->source)))
1098 {
1099 header->id = htons(forward->orig_id);
1100 header->hb4 |= HB4_RA; /* recursion if available */
1101 send_from(forward->fd, option_bool(OPT_NOWILD) || option_bool (OPT_CLEVERBIND), daemon->packet, nn,
1102 &forward->source, &forward->dest, forward->iface);
1103 }
1104 free_frec(forward); /* cancel */
1105 }
1106 }
1107
1108
1109 void receive_query(struct listener *listen, time_t now)
1110 {
1111 struct dns_header *header = (struct dns_header *)daemon->packet;
1112 union mysockaddr source_addr;
1113 unsigned short type;
1114 struct all_addr dst_addr;
1115 struct in_addr netmask, dst_addr_4;
1116 size_t m;
1117 ssize_t n;
1118 int if_index = 0, auth_dns = 0;
1119 #ifdef HAVE_AUTH
1120 int local_auth = 0;
1121 #endif
1122 struct iovec iov[1];
1123 struct msghdr msg;
1124 struct cmsghdr *cmptr;
1125 union {
1126 struct cmsghdr align; /* this ensures alignment */
1127 #ifdef HAVE_IPV6
1128 char control6[CMSG_SPACE(sizeof(struct in6_pktinfo))];
1129 #endif
1130 #if defined(HAVE_LINUX_NETWORK)
1131 char control[CMSG_SPACE(sizeof(struct in_pktinfo))];
1132 #elif defined(IP_RECVDSTADDR) && defined(HAVE_SOLARIS_NETWORK)
1133 char control[CMSG_SPACE(sizeof(struct in_addr)) +
1134 CMSG_SPACE(sizeof(unsigned int))];
1135 #elif defined(IP_RECVDSTADDR)
1136 char control[CMSG_SPACE(sizeof(struct in_addr)) +
1137 CMSG_SPACE(sizeof(struct sockaddr_dl))];
1138 #endif
1139 } control_u;
1140 #ifdef HAVE_IPV6
1141 /* Can always get recvd interface for IPv6 */
1142 int check_dst = !option_bool(OPT_NOWILD) || listen->family == AF_INET6;
1143 #else
1144 int check_dst = !option_bool(OPT_NOWILD);
1145 #endif
1146
1147 /* packet buffer overwritten */
1148 daemon->srv_save = NULL;
1149
1150 dst_addr_4.s_addr = dst_addr.addr.addr4.s_addr = 0;
1151 netmask.s_addr = 0;
1152
1153 if (option_bool(OPT_NOWILD) && listen->iface)
1154 {
1155 auth_dns = listen->iface->dns_auth;
1156
1157 if (listen->family == AF_INET)
1158 {
1159 dst_addr_4 = dst_addr.addr.addr4 = listen->iface->addr.in.sin_addr;
1160 netmask = listen->iface->netmask;
1161 }
1162 }
1163
1164 iov[0].iov_base = daemon->packet;
1165 iov[0].iov_len = daemon->edns_pktsz;
1166
1167 msg.msg_control = control_u.control;
1168 msg.msg_controllen = sizeof(control_u);
1169 msg.msg_flags = 0;
1170 msg.msg_name = &source_addr;
1171 msg.msg_namelen = sizeof(source_addr);
1172 msg.msg_iov = iov;
1173 msg.msg_iovlen = 1;
1174
1175 if ((n = recvmsg(listen->fd, &msg, 0)) == -1)
1176 return;
1177
1178 if (n < (int)sizeof(struct dns_header) ||
1179 (msg.msg_flags & MSG_TRUNC) ||
1180 (header->hb3 & HB3_QR))
1181 return;
1182
1183 source_addr.sa.sa_family = listen->family;
1184
1185 if (listen->family == AF_INET)
1186 {
1187 /* Source-port == 0 is an error, we can't send back to that.
1188 http://www.ietf.org/mail-archive/web/dnsop/current/msg11441.html */
1189 if (source_addr.in.sin_port == 0)
1190 return;
1191 }
1192 #ifdef HAVE_IPV6
1193 else
1194 {
1195 /* Source-port == 0 is an error, we can't send back to that. */
1196 if (source_addr.in6.sin6_port == 0)
1197 return;
1198 source_addr.in6.sin6_flowinfo = 0;
1199 }
1200 #endif
1201
1202 /* We can be configured to only accept queries from at-most-one-hop-away addresses. */
1203 if (option_bool(OPT_LOCAL_SERVICE))
1204 {
1205 struct addrlist *addr;
1206 #ifdef HAVE_IPV6
1207 if (listen->family == AF_INET6)
1208 {
1209 for (addr = daemon->interface_addrs; addr; addr = addr->next)
1210 if ((addr->flags & ADDRLIST_IPV6) &&
1211 is_same_net6(&addr->addr.addr.addr6, &source_addr.in6.sin6_addr, addr->prefixlen))
1212 break;
1213 }
1214 else
1215 #endif
1216 {
1217 struct in_addr netmask;
1218 for (addr = daemon->interface_addrs; addr; addr = addr->next)
1219 {
1220 netmask.s_addr = htonl(~(in_addr_t)0 << (32 - addr->prefixlen));
1221 if (!(addr->flags & ADDRLIST_IPV6) &&
1222 is_same_net(addr->addr.addr.addr4, source_addr.in.sin_addr, netmask))
1223 break;
1224 }
1225 }
1226 if (!addr)
1227 {
1228 static int warned = 0;
1229 if (!warned)
1230 {
1231 my_syslog(LOG_WARNING, _("Ignoring query from non-local network"));
1232 warned = 1;
1233 }
1234 return;
1235 }
1236 }
1237
1238 if (check_dst)
1239 {
1240 struct ifreq ifr;
1241
1242 if (msg.msg_controllen < sizeof(struct cmsghdr))
1243 return;
1244
1245 #if defined(HAVE_LINUX_NETWORK)
1246 if (listen->family == AF_INET)
1247 for (cmptr = CMSG_FIRSTHDR(&msg); cmptr; cmptr = CMSG_NXTHDR(&msg, cmptr))
1248 if (cmptr->cmsg_level == IPPROTO_IP && cmptr->cmsg_type == IP_PKTINFO)
1249 {
1250 union {
1251 unsigned char *c;
1252 struct in_pktinfo *p;
1253 } p;
1254 p.c = CMSG_DATA(cmptr);
1255 dst_addr_4 = dst_addr.addr.addr4 = p.p->ipi_spec_dst;
1256 if_index = p.p->ipi_ifindex;
1257 }
1258 #elif defined(IP_RECVDSTADDR) && defined(IP_RECVIF)
1259 if (listen->family == AF_INET)
1260 {
1261 for (cmptr = CMSG_FIRSTHDR(&msg); cmptr; cmptr = CMSG_NXTHDR(&msg, cmptr))
1262 {
1263 union {
1264 unsigned char *c;
1265 unsigned int *i;
1266 struct in_addr *a;
1267 #ifndef HAVE_SOLARIS_NETWORK
1268 struct sockaddr_dl *s;
1269 #endif
1270 } p;
1271 p.c = CMSG_DATA(cmptr);
1272 if (cmptr->cmsg_level == IPPROTO_IP && cmptr->cmsg_type == IP_RECVDSTADDR)
1273 dst_addr_4 = dst_addr.addr.addr4 = *(p.a);
1274 else if (cmptr->cmsg_level == IPPROTO_IP && cmptr->cmsg_type == IP_RECVIF)
1275 #ifdef HAVE_SOLARIS_NETWORK
1276 if_index = *(p.i);
1277 #else
1278 if_index = p.s->sdl_index;
1279 #endif
1280 }
1281 }
1282 #endif
1283
1284 #ifdef HAVE_IPV6
1285 if (listen->family == AF_INET6)
1286 {
1287 for (cmptr = CMSG_FIRSTHDR(&msg); cmptr; cmptr = CMSG_NXTHDR(&msg, cmptr))
1288 if (cmptr->cmsg_level == IPPROTO_IPV6 && cmptr->cmsg_type == daemon->v6pktinfo)
1289 {
1290 union {
1291 unsigned char *c;
1292 struct in6_pktinfo *p;
1293 } p;
1294 p.c = CMSG_DATA(cmptr);
1295
1296 dst_addr.addr.addr6 = p.p->ipi6_addr;
1297 if_index = p.p->ipi6_ifindex;
1298 }
1299 }
1300 #endif
1301
1302 /* enforce available interface configuration */
1303
1304 if (!indextoname(listen->fd, if_index, ifr.ifr_name))
1305 return;
1306
1307 if (!iface_check(listen->family, &dst_addr, ifr.ifr_name, &auth_dns))
1308 {
1309 if (!option_bool(OPT_CLEVERBIND))
1310 enumerate_interfaces(0);
1311 if (!loopback_exception(listen->fd, listen->family, &dst_addr, ifr.ifr_name) &&
1312 !label_exception(if_index, listen->family, &dst_addr))
1313 return;
1314 }
1315
1316 if (listen->family == AF_INET && option_bool(OPT_LOCALISE))
1317 {
1318 struct irec *iface;
1319
1320 /* get the netmask of the interface whch has the address we were sent to.
1321 This is no neccessarily the interface we arrived on. */
1322
1323 for (iface = daemon->interfaces; iface; iface = iface->next)
1324 if (iface->addr.sa.sa_family == AF_INET &&
1325 iface->addr.in.sin_addr.s_addr == dst_addr_4.s_addr)
1326 break;
1327
1328 /* interface may be new */
1329 if (!iface && !option_bool(OPT_CLEVERBIND))
1330 enumerate_interfaces(0);
1331
1332 for (iface = daemon->interfaces; iface; iface = iface->next)
1333 if (iface->addr.sa.sa_family == AF_INET &&
1334 iface->addr.in.sin_addr.s_addr == dst_addr_4.s_addr)
1335 break;
1336
1337 /* If we failed, abandon localisation */
1338 if (iface)
1339 netmask = iface->netmask;
1340 else
1341 dst_addr_4.s_addr = 0;
1342 }
1343 }
1344
1345 /* log_query gets called indirectly all over the place, so
1346 pass these in global variables - sorry. */
1347 daemon->log_display_id = ++daemon->log_id;
1348 daemon->log_source_addr = &source_addr;
1349
1350 if (extract_request(header, (size_t)n, daemon->namebuff, &type))
1351 {
1352 #ifdef HAVE_AUTH
1353 struct auth_zone *zone;
1354 #endif
1355 char *types = querystr(auth_dns ? "auth" : "query", type);
1356
1357 if (listen->family == AF_INET)
1358 log_query(F_QUERY | F_IPV4 | F_FORWARD, daemon->namebuff,
1359 (struct all_addr *)&source_addr.in.sin_addr, types);
1360 #ifdef HAVE_IPV6
1361 else
1362 log_query(F_QUERY | F_IPV6 | F_FORWARD, daemon->namebuff,
1363 (struct all_addr *)&source_addr.in6.sin6_addr, types);
1364 #endif
1365
1366 #ifdef HAVE_AUTH
1367 /* find queries for zones we're authoritative for, and answer them directly */
1368 if (!auth_dns)
1369 for (zone = daemon->auth_zones; zone; zone = zone->next)
1370 if (in_zone(zone, daemon->namebuff, NULL))
1371 {
1372 auth_dns = 1;
1373 local_auth = 1;
1374 break;
1375 }
1376 #endif
1377
1378 #ifdef HAVE_LOOP
1379 /* Check for forwarding loop */
1380 if (detect_loop(daemon->namebuff, type))
1381 return;
1382 #endif
1383 }
1384
1385 #ifdef HAVE_AUTH
1386 if (auth_dns)
1387 {
1388 m = answer_auth(header, ((char *) header) + daemon->packet_buff_sz, (size_t)n, now, &source_addr, local_auth);
1389 if (m >= 1)
1390 {
1391 send_from(listen->fd, option_bool(OPT_NOWILD) || option_bool(OPT_CLEVERBIND),
1392 (char *)header, m, &source_addr, &dst_addr, if_index);
1393 daemon->auth_answer++;
1394 }
1395 }
1396 else
1397 #endif
1398 {
1399 int ad_reqd, do_bit;
1400 m = answer_request(header, ((char *) header) + daemon->packet_buff_sz, (size_t)n,
1401 dst_addr_4, netmask, now, &ad_reqd, &do_bit);
1402
1403 if (m >= 1)
1404 {
1405 send_from(listen->fd, option_bool(OPT_NOWILD) || option_bool(OPT_CLEVERBIND),
1406 (char *)header, m, &source_addr, &dst_addr, if_index);
1407 daemon->local_answer++;
1408 }
1409 else if (forward_query(listen->fd, &source_addr, &dst_addr, if_index,
1410 header, (size_t)n, now, NULL, ad_reqd, do_bit))
1411 daemon->queries_forwarded++;
1412 else
1413 daemon->local_answer++;
1414 }
1415 }
1416
1417 #ifdef HAVE_DNSSEC
1418
1419 /* UDP: we've got an unsigned answer, return STAT_INSECURE if we can prove there's no DS
1420 and therefore the answer shouldn't be signed, or STAT_BOGUS if it should be, or
1421 STAT_NEED_DS_NEG and keyname if we need to do the query. */
1422 static int send_check_sign(struct frec *forward, time_t now, struct dns_header *header, size_t plen,
1423 char *name, char *keyname)
1424 {
1425 int status = dnssec_chase_cname(now, header, plen, name, keyname);
1426
1427 if (status != STAT_INSECURE)
1428 return status;
1429
1430 /* Store the domain we're trying to check. */
1431 forward->name_start = strlen(name);
1432 forward->name_len = forward->name_start + 1;
1433 if (!(forward->orig_domain = blockdata_alloc(name, forward->name_len)))
1434 return STAT_BOGUS;
1435
1436 return do_check_sign(forward, 0, now, name, keyname);
1437 }
1438
1439 /* We either have a a reply (header non-NULL, or we need to start by looking in the cache */
1440 static int do_check_sign(struct frec *forward, int status, time_t now, char *name, char *keyname)
1441 {
1442 /* get domain we're checking back from blockdata store, it's stored on the original query. */
1443 while (forward->dependent && !forward->orig_domain)
1444 forward = forward->dependent;
1445
1446 blockdata_retrieve(forward->orig_domain, forward->name_len, name);
1447
1448 while (1)
1449 {
1450 char *p;
1451
1452 if (status == 0)
1453 {
1454 struct crec *crecp;
1455
1456 /* Haven't received answer, see if in cache */
1457 if (!(crecp = cache_find_by_name(NULL, &name[forward->name_start], now, F_DS)))
1458 {
1459 /* put name of DS record we're missing into keyname */
1460 strcpy(keyname, &name[forward->name_start]);
1461 /* and wait for reply to arrive */
1462 return STAT_NEED_DS_NEG;
1463 }
1464
1465 /* F_DNSSECOK misused in DS cache records to non-existance of NS record */
1466 if (!(crecp->flags & F_NEG))
1467 status = STAT_SECURE;
1468 else if (crecp->flags & F_DNSSECOK)
1469 status = STAT_NO_DS;
1470 else
1471 status = STAT_NO_NS;
1472 }
1473
1474 /* Have entered non-signed part of DNS tree. */
1475 if (status == STAT_NO_DS)
1476 return forward->dependent ? STAT_INSECURE_DS : STAT_INSECURE;
1477
1478 if (status == STAT_BOGUS)
1479 return STAT_BOGUS;
1480
1481 if (status == STAT_NO_SIG && *keyname != 0)
1482 {
1483 /* There is a validated CNAME chain that doesn't end in a DS record. Start
1484 the search again in that domain. */
1485 blockdata_free(forward->orig_domain);
1486 forward->name_start = strlen(keyname);
1487 forward->name_len = forward->name_start + 1;
1488 if (!(forward->orig_domain = blockdata_alloc(keyname, forward->name_len)))
1489 return STAT_BOGUS;
1490
1491 strcpy(name, keyname);
1492 status = 0; /* force to cache when we iterate. */
1493 continue;
1494 }
1495
1496 /* There's a proven DS record, or we're within a zone, where there doesn't need
1497 to be a DS record. Add a name and try again.
1498 If we've already tried the whole name, then fail */
1499
1500 if (forward->name_start == 0)
1501 return STAT_BOGUS;
1502
1503 for (p = &name[forward->name_start-2]; (*p != '.') && (p != name); p--);
1504
1505 if (p != name)
1506 p++;
1507
1508 forward->name_start = p - name;
1509 status = 0; /* force to cache when we iterate. */
1510 }
1511 }
1512
1513 /* Move down from the root, until we find a signed non-existance of a DS, in which case
1514 an unsigned answer is OK, or we find a signed DS, in which case there should be
1515 a signature, and the answer is BOGUS */
1516 static int tcp_check_for_unsigned_zone(time_t now, struct dns_header *header, size_t plen, int class, char *name,
1517 char *keyname, struct server *server, int *keycount)
1518 {
1519 size_t m;
1520 unsigned char *packet, *payload;
1521 u16 *length;
1522 int status, name_len;
1523 struct blockdata *block;
1524
1525 char *name_start;
1526
1527 /* Get first insecure entry in CNAME chain */
1528 status = tcp_key_recurse(now, STAT_CHASE_CNAME, header, plen, class, name, keyname, server, keycount);
1529 if (status == STAT_BOGUS)
1530 return STAT_BOGUS;
1531
1532 if (!(packet = whine_malloc(65536 + MAXDNAME + RRFIXEDSZ + sizeof(u16))))
1533 return STAT_BOGUS;
1534
1535 payload = &packet[2];
1536 header = (struct dns_header *)payload;
1537 length = (u16 *)packet;
1538
1539 /* Stash the name away, since the buffer will be trashed when we recurse */
1540 name_len = strlen(name) + 1;
1541 name_start = name + name_len - 1;
1542
1543 if (!(block = blockdata_alloc(name, name_len)))
1544 {
1545 free(packet);
1546 return STAT_BOGUS;
1547 }
1548
1549 while (1)
1550 {
1551 unsigned char c1, c2;
1552 struct crec *crecp;
1553
1554 if (--(*keycount) == 0)
1555 {
1556 free(packet);
1557 blockdata_free(block);
1558 return STAT_BOGUS;
1559 }
1560
1561 while ((crecp = cache_find_by_name(NULL, name_start, now, F_DS)))
1562 {
1563 if ((crecp->flags & F_NEG) && (crecp->flags & F_DNSSECOK))
1564 {
1565 /* Found a secure denial of DS - delegation is indeed insecure */
1566 free(packet);
1567 blockdata_free(block);
1568 return STAT_INSECURE;
1569 }
1570
1571 /* Here, either there's a secure DS, or no NS and no DS, and therefore no delegation.
1572 Add another label and continue. */
1573
1574 if (name_start == name)
1575 {
1576 free(packet);
1577 blockdata_free(block);
1578 return STAT_BOGUS; /* run out of labels */
1579 }
1580
1581 name_start -= 2;
1582 while (*name_start != '.' && name_start != name)
1583 name_start--;
1584 if (name_start != name)
1585 name_start++;
1586 }
1587
1588 /* Can't find it in the cache, have to send a query */
1589
1590 m = dnssec_generate_query(header, ((char *) header) + 65536, name_start, class, T_DS, &server->addr, server->edns_pktsz);
1591
1592 *length = htons(m);
1593
1594 if (read_write(server->tcpfd, packet, m + sizeof(u16), 0) &&
1595 read_write(server->tcpfd, &c1, 1, 1) &&
1596 read_write(server->tcpfd, &c2, 1, 1) &&
1597 read_write(server->tcpfd, payload, (c1 << 8) | c2, 1))
1598 {
1599 m = (c1 << 8) | c2;
1600
1601 /* Note this trashes all three name workspaces */
1602 status = tcp_key_recurse(now, STAT_NEED_DS_NEG, header, m, class, name, keyname, server, keycount);
1603
1604 if (status == STAT_NO_DS)
1605 {
1606 /* Found a secure denial of DS - delegation is indeed insecure */
1607 free(packet);
1608 blockdata_free(block);
1609 return STAT_INSECURE;
1610 }
1611
1612 if (status == STAT_NO_SIG && *keyname != 0)
1613 {
1614 /* There is a validated CNAME chain that doesn't end in a DS record. Start
1615 the search again in that domain. */
1616 blockdata_free(block);
1617 name_len = strlen(keyname) + 1;
1618 name_start = name + name_len - 1;
1619
1620 if (!(block = blockdata_alloc(keyname, name_len)))
1621 return STAT_BOGUS;
1622
1623 strcpy(name, keyname);
1624 continue;
1625 }
1626
1627 if (status == STAT_BOGUS)
1628 {
1629 free(packet);
1630 blockdata_free(block);
1631 return STAT_BOGUS;
1632 }
1633
1634 /* Here, either there's a secure DS, or no NS and no DS, and therefore no delegation.
1635 Add another label and continue. */
1636
1637 /* Get name we're checking back. */
1638 blockdata_retrieve(block, name_len, name);
1639
1640 if (name_start == name)
1641 {
1642 free(packet);
1643 blockdata_free(block);
1644 return STAT_BOGUS; /* run out of labels */
1645 }
1646
1647 name_start -= 2;
1648 while (*name_start != '.' && name_start != name)
1649 name_start--;
1650 if (name_start != name)
1651 name_start++;
1652 }
1653 else
1654 {
1655 /* IO failure */
1656 free(packet);
1657 blockdata_free(block);
1658 return STAT_BOGUS; /* run out of labels */
1659 }
1660 }
1661 }
1662
1663 static int tcp_key_recurse(time_t now, int status, struct dns_header *header, size_t n,
1664 int class, char *name, char *keyname, struct server *server, int *keycount)
1665 {
1666 /* Recurse up the key heirarchy */
1667 int new_status;
1668
1669 /* limit the amount of work we do, to avoid cycling forever on loops in the DNS */
1670 if (--(*keycount) == 0)
1671 return STAT_INSECURE;
1672
1673 if (status == STAT_NEED_KEY)
1674 new_status = dnssec_validate_by_ds(now, header, n, name, keyname, class);
1675 else if (status == STAT_NEED_DS || status == STAT_NEED_DS_NEG)
1676 {
1677 new_status = dnssec_validate_ds(now, header, n, name, keyname, class);
1678 if (status == STAT_NEED_DS)
1679 {
1680 if (new_status == STAT_NO_DS)
1681 new_status = STAT_INSECURE_DS;
1682 if (new_status == STAT_NO_SIG)
1683 {
1684 if (option_bool(OPT_DNSSEC_NO_SIGN))
1685 {
1686 new_status = tcp_check_for_unsigned_zone(now, header, n, class, name, keyname, server, keycount);
1687 if (new_status == STAT_INSECURE)
1688 new_status = STAT_INSECURE_DS;
1689 }
1690 else
1691 new_status = STAT_INSECURE_DS;
1692 }
1693 else if (new_status == STAT_NO_NS)
1694 new_status = STAT_BOGUS;
1695 }
1696 }
1697 else if (status == STAT_CHASE_CNAME)
1698 new_status = dnssec_chase_cname(now, header, n, name, keyname);
1699 else
1700 {
1701 new_status = dnssec_validate_reply(now, header, n, name, keyname, &class, NULL, NULL);
1702
1703 if (new_status == STAT_NO_SIG)
1704 {
1705 if (option_bool(OPT_DNSSEC_NO_SIGN))
1706 new_status = tcp_check_for_unsigned_zone(now, header, n, class, name, keyname, server, keycount);
1707 else
1708 new_status = STAT_INSECURE;
1709 }
1710 }
1711
1712 /* Can't validate because we need a key/DS whose name now in keyname.
1713 Make query for same, and recurse to validate */
1714 if (new_status == STAT_NEED_DS || new_status == STAT_NEED_KEY)
1715 {
1716 size_t m;
1717 unsigned char *packet = whine_malloc(65536 + MAXDNAME + RRFIXEDSZ + sizeof(u16));
1718 unsigned char *payload = &packet[2];
1719 struct dns_header *new_header = (struct dns_header *)payload;
1720 u16 *length = (u16 *)packet;
1721 unsigned char c1, c2;
1722
1723 if (!packet)
1724 return STAT_INSECURE;
1725
1726 another_tcp_key:
1727 m = dnssec_generate_query(new_header, ((char *) new_header) + 65536, keyname, class,
1728 new_status == STAT_NEED_KEY ? T_DNSKEY : T_DS, &server->addr, server->edns_pktsz);
1729
1730 *length = htons(m);
1731
1732 if (!read_write(server->tcpfd, packet, m + sizeof(u16), 0) ||
1733 !read_write(server->tcpfd, &c1, 1, 1) ||
1734 !read_write(server->tcpfd, &c2, 1, 1) ||
1735 !read_write(server->tcpfd, payload, (c1 << 8) | c2, 1))
1736 new_status = STAT_INSECURE;
1737 else
1738 {
1739 m = (c1 << 8) | c2;
1740
1741 new_status = tcp_key_recurse(now, new_status, new_header, m, class, name, keyname, server, keycount);
1742
1743 if (new_status == STAT_SECURE)
1744 {
1745 /* Reached a validated record, now try again at this level.
1746 Note that we may get ANOTHER NEED_* if an answer needs more than one key.
1747 If so, go round again. */
1748
1749 if (status == STAT_NEED_KEY)
1750 new_status = dnssec_validate_by_ds(now, header, n, name, keyname, class);
1751 else if (status == STAT_NEED_DS || status == STAT_NEED_DS_NEG)
1752 {
1753 new_status = dnssec_validate_ds(now, header, n, name, keyname, class);
1754 if (status == STAT_NEED_DS)
1755 {
1756 if (new_status == STAT_NO_DS)
1757 new_status = STAT_INSECURE_DS;
1758 else if (new_status == STAT_NO_SIG)
1759 {
1760 if (option_bool(OPT_DNSSEC_NO_SIGN))
1761 {
1762 new_status = tcp_check_for_unsigned_zone(now, header, n, class, name, keyname, server, keycount);
1763 if (new_status == STAT_INSECURE)
1764 new_status = STAT_INSECURE_DS;
1765 }
1766 else
1767 new_status = STAT_INSECURE_DS;
1768 }
1769 else if (new_status == STAT_NO_NS)
1770 new_status = STAT_BOGUS;
1771 }
1772 }
1773 else if (status == STAT_CHASE_CNAME)
1774 new_status = dnssec_chase_cname(now, header, n, name, keyname);
1775 else
1776 {
1777 new_status = dnssec_validate_reply(now, header, n, name, keyname, &class, NULL, NULL);
1778
1779 if (new_status == STAT_NO_SIG)
1780 {
1781 if (option_bool(OPT_DNSSEC_NO_SIGN))
1782 new_status = tcp_check_for_unsigned_zone(now, header, n, class, name, keyname, server, keycount);
1783 else
1784 new_status = STAT_INSECURE;
1785 }
1786 }
1787
1788 if (new_status == STAT_NEED_DS || new_status == STAT_NEED_KEY)
1789 goto another_tcp_key;
1790 }
1791 }
1792
1793 free(packet);
1794 }
1795 return new_status;
1796 }
1797 #endif
1798
1799
1800 /* The daemon forks before calling this: it should deal with one connection,
1801 blocking as neccessary, and then return. Note, need to be a bit careful
1802 about resources for debug mode, when the fork is suppressed: that's
1803 done by the caller. */
1804 unsigned char *tcp_request(int confd, time_t now,
1805 union mysockaddr *local_addr, struct in_addr netmask, int auth_dns)
1806 {
1807 size_t size = 0;
1808 int norebind = 0;
1809 #ifdef HAVE_AUTH
1810 int local_auth = 0;
1811 #endif
1812 int checking_disabled, ad_question, do_bit, added_pheader = 0;
1813 int check_subnet, no_cache_dnssec = 0, cache_secure = 0, bogusanswer = 0;
1814 size_t m;
1815 unsigned short qtype;
1816 unsigned int gotname;
1817 unsigned char c1, c2;
1818 /* Max TCP packet + slop + size */
1819 unsigned char *packet = whine_malloc(65536 + MAXDNAME + RRFIXEDSZ + sizeof(u16));
1820 unsigned char *payload = &packet[2];
1821 /* largest field in header is 16-bits, so this is still sufficiently aligned */
1822 struct dns_header *header = (struct dns_header *)payload;
1823 u16 *length = (u16 *)packet;
1824 struct server *last_server;
1825 struct in_addr dst_addr_4;
1826 union mysockaddr peer_addr;
1827 socklen_t peer_len = sizeof(union mysockaddr);
1828 int query_count = 0;
1829
1830 if (getpeername(confd, (struct sockaddr *)&peer_addr, &peer_len) == -1)
1831 return packet;
1832
1833 /* We can be configured to only accept queries from at-most-one-hop-away addresses. */
1834 if (option_bool(OPT_LOCAL_SERVICE))
1835 {
1836 struct addrlist *addr;
1837 #ifdef HAVE_IPV6
1838 if (peer_addr.sa.sa_family == AF_INET6)
1839 {
1840 for (addr = daemon->interface_addrs; addr; addr = addr->next)
1841 if ((addr->flags & ADDRLIST_IPV6) &&
1842 is_same_net6(&addr->addr.addr.addr6, &peer_addr.in6.sin6_addr, addr->prefixlen))
1843 break;
1844 }
1845 else
1846 #endif
1847 {
1848 struct in_addr netmask;
1849 for (addr = daemon->interface_addrs; addr; addr = addr->next)
1850 {
1851 netmask.s_addr = htonl(~(in_addr_t)0 << (32 - addr->prefixlen));
1852 if (!(addr->flags & ADDRLIST_IPV6) &&
1853 is_same_net(addr->addr.addr.addr4, peer_addr.in.sin_addr, netmask))
1854 break;
1855 }
1856 }
1857 if (!addr)
1858 {
1859 my_syslog(LOG_WARNING, _("Ignoring query from non-local network"));
1860 return packet;
1861 }
1862 }
1863
1864 while (1)
1865 {
1866 if (query_count == TCP_MAX_QUERIES ||
1867 !packet ||
1868 !read_write(confd, &c1, 1, 1) || !read_write(confd, &c2, 1, 1) ||
1869 !(size = c1 << 8 | c2) ||
1870 !read_write(confd, payload, size, 1))
1871 return packet;
1872
1873 if (size < (int)sizeof(struct dns_header))
1874 continue;
1875
1876 query_count++;
1877
1878 /* log_query gets called indirectly all over the place, so
1879 pass these in global variables - sorry. */
1880 daemon->log_display_id = ++daemon->log_id;
1881 daemon->log_source_addr = &peer_addr;
1882
1883 check_subnet = 0;
1884
1885 /* save state of "cd" flag in query */
1886 if ((checking_disabled = header->hb4 & HB4_CD))
1887 no_cache_dnssec = 1;
1888
1889 if ((gotname = extract_request(header, (unsigned int)size, daemon->namebuff, &qtype)))
1890 {
1891 #ifdef HAVE_AUTH
1892 struct auth_zone *zone;
1893 #endif
1894 char *types = querystr(auth_dns ? "auth" : "query", qtype);
1895
1896 if (peer_addr.sa.sa_family == AF_INET)
1897 log_query(F_QUERY | F_IPV4 | F_FORWARD, daemon->namebuff,
1898 (struct all_addr *)&peer_addr.in.sin_addr, types);
1899 #ifdef HAVE_IPV6
1900 else
1901 log_query(F_QUERY | F_IPV6 | F_FORWARD, daemon->namebuff,
1902 (struct all_addr *)&peer_addr.in6.sin6_addr, types);
1903 #endif
1904
1905 #ifdef HAVE_AUTH
1906 /* find queries for zones we're authoritative for, and answer them directly */
1907 if (!auth_dns)
1908 for (zone = daemon->auth_zones; zone; zone = zone->next)
1909 if (in_zone(zone, daemon->namebuff, NULL))
1910 {
1911 auth_dns = 1;
1912 local_auth = 1;
1913 break;
1914 }
1915 #endif
1916 }
1917
1918 if (local_addr->sa.sa_family == AF_INET)
1919 dst_addr_4 = local_addr->in.sin_addr;
1920 else
1921 dst_addr_4.s_addr = 0;
1922
1923 #ifdef HAVE_AUTH
1924 if (auth_dns)
1925 m = answer_auth(header, ((char *) header) + 65536, (size_t)size, now, &peer_addr, local_auth);
1926 else
1927 #endif
1928 {
1929 /* m > 0 if answered from cache */
1930 m = answer_request(header, ((char *) header) + 65536, (size_t)size,
1931 dst_addr_4, netmask, now, &ad_question, &do_bit);
1932
1933 /* Do this by steam now we're not in the select() loop */
1934 check_log_writer(1);
1935
1936 if (m == 0)
1937 {
1938 unsigned int flags = 0;
1939 struct all_addr *addrp = NULL;
1940 int type = 0;
1941 char *domain = NULL;
1942
1943 if (option_bool(OPT_ADD_MAC))
1944 size = add_mac(header, size, ((char *) header) + 65536, &peer_addr);
1945
1946 if (option_bool(OPT_CLIENT_SUBNET))
1947 {
1948 size_t new = add_source_addr(header, size, ((char *) header) + 65536, &peer_addr);
1949 if (size != new)
1950 {
1951 size = new;
1952 check_subnet = 1;
1953 }
1954 }
1955
1956 if (gotname)
1957 flags = search_servers(now, &addrp, gotname, daemon->namebuff, &type, &domain, &norebind);
1958
1959 if (type != 0 || option_bool(OPT_ORDER) || !daemon->last_server)
1960 last_server = daemon->servers;
1961 else
1962 last_server = daemon->last_server;
1963
1964 if (!flags && last_server)
1965 {
1966 struct server *firstsendto = NULL;
1967 #ifdef HAVE_DNSSEC
1968 unsigned char *newhash, hash[HASH_SIZE];
1969 if ((newhash = hash_questions(header, (unsigned int)size, daemon->namebuff)))
1970 memcpy(hash, newhash, HASH_SIZE);
1971 else
1972 memset(hash, 0, HASH_SIZE);
1973 #else
1974 unsigned int crc = questions_crc(header, (unsigned int)size, daemon->namebuff);
1975 #endif
1976 /* Loop round available servers until we succeed in connecting to one.
1977 Note that this code subtley ensures that consecutive queries on this connection
1978 which can go to the same server, do so. */
1979 while (1)
1980 {
1981 if (!firstsendto)
1982 firstsendto = last_server;
1983 else
1984 {
1985 if (!(last_server = last_server->next))
1986 last_server = daemon->servers;
1987
1988 if (last_server == firstsendto)
1989 break;
1990 }
1991
1992 /* server for wrong domain */
1993 if (type != (last_server->flags & SERV_TYPE) ||
1994 (type == SERV_HAS_DOMAIN && !hostname_isequal(domain, last_server->domain)) ||
1995 (last_server->flags & (SERV_LITERAL_ADDRESS | SERV_LOOP)))
1996 continue;
1997
1998 if (last_server->tcpfd == -1)
1999 {
2000 if ((last_server->tcpfd = socket(last_server->addr.sa.sa_family, SOCK_STREAM, 0)) == -1)
2001 continue;
2002
2003 #ifdef HAVE_CONNTRACK
2004 /* Copy connection mark of incoming query to outgoing connection. */
2005 if (option_bool(OPT_CONNTRACK))
2006 {
2007 unsigned int mark;
2008 struct all_addr local;
2009 #ifdef HAVE_IPV6
2010 if (local_addr->sa.sa_family == AF_INET6)
2011 local.addr.addr6 = local_addr->in6.sin6_addr;
2012 else
2013 #endif
2014 local.addr.addr4 = local_addr->in.sin_addr;
2015
2016 if (get_incoming_mark(&peer_addr, &local, 1, &mark))
2017 setsockopt(last_server->tcpfd, SOL_SOCKET, SO_MARK, &mark, sizeof(unsigned int));
2018 }
2019 #endif
2020
2021 if ((!local_bind(last_server->tcpfd, &last_server->source_addr, last_server->interface, 1) ||
2022 connect(last_server->tcpfd, &last_server->addr.sa, sa_len(&last_server->addr)) == -1))
2023 {
2024 close(last_server->tcpfd);
2025 last_server->tcpfd = -1;
2026 continue;
2027 }
2028
2029 #ifdef HAVE_DNSSEC
2030 if (option_bool(OPT_DNSSEC_VALID))
2031 {
2032 size_t new_size = add_do_bit(header, size, ((char *) header) + 65536);
2033
2034 /* For debugging, set Checking Disabled, otherwise, have the upstream check too,
2035 this allows it to select auth servers when one is returning bad data. */
2036 if (option_bool(OPT_DNSSEC_DEBUG))
2037 header->hb4 |= HB4_CD;
2038
2039 if (size != new_size)
2040 added_pheader = 1;
2041
2042 size = new_size;
2043 }
2044 #endif
2045 }
2046
2047 *length = htons(size);
2048
2049 /* get query name again for logging - may have been overwritten */
2050 if (!(gotname = extract_request(header, (unsigned int)size, daemon->namebuff, &qtype)))
2051 strcpy(daemon->namebuff, "query");
2052
2053 if (!read_write(last_server->tcpfd, packet, size + sizeof(u16), 0) ||
2054 !read_write(last_server->tcpfd, &c1, 1, 1) ||
2055 !read_write(last_server->tcpfd, &c2, 1, 1) ||
2056 !read_write(last_server->tcpfd, payload, (c1 << 8) | c2, 1))
2057 {
2058 close(last_server->tcpfd);
2059 last_server->tcpfd = -1;
2060 continue;
2061 }
2062
2063 m = (c1 << 8) | c2;
2064
2065 if (last_server->addr.sa.sa_family == AF_INET)
2066 log_query(F_SERVER | F_IPV4 | F_FORWARD, daemon->namebuff,
2067 (struct all_addr *)&last_server->addr.in.sin_addr, NULL);
2068 #ifdef HAVE_IPV6
2069 else
2070 log_query(F_SERVER | F_IPV6 | F_FORWARD, daemon->namebuff,
2071 (struct all_addr *)&last_server->addr.in6.sin6_addr, NULL);
2072 #endif
2073
2074 #ifdef HAVE_DNSSEC
2075 if (option_bool(OPT_DNSSEC_VALID) && !checking_disabled)
2076 {
2077 int keycount = DNSSEC_WORK; /* Limit to number of DNSSEC questions, to catch loops and avoid filling cache. */
2078 int status = tcp_key_recurse(now, STAT_TRUNCATED, header, m, 0, daemon->namebuff, daemon->keyname, last_server, &keycount);
2079 char *result, *domain = "result";
2080
2081 if (status == STAT_INSECURE_DS)
2082 {
2083 /* We only cache sigs when we've validated a reply.
2084 Avoid caching a reply with sigs if there's a vaildated break in the
2085 DS chain, so we don't return replies from cache missing sigs. */
2086 status = STAT_INSECURE;
2087 no_cache_dnssec = 1;
2088 }
2089
2090 if (keycount == 0)
2091 {
2092 result = "ABANDONED";
2093 status = STAT_BOGUS;
2094 }
2095 else
2096 result = (status == STAT_SECURE ? "SECURE" : (status == STAT_INSECURE ? "INSECURE" : "BOGUS"));
2097
2098 if (status == STAT_BOGUS && extract_request(header, m, daemon->namebuff, NULL))
2099 domain = daemon->namebuff;
2100
2101 log_query(F_KEYTAG | F_SECSTAT, domain, NULL, result);
2102
2103 if (status == STAT_BOGUS)
2104 {
2105 no_cache_dnssec = 1;
2106 bogusanswer = 1;
2107 }
2108
2109 if (status == STAT_SECURE)
2110 cache_secure = 1;
2111 }
2112 #endif
2113
2114 /* restore CD bit to the value in the query */
2115 if (checking_disabled)
2116 header->hb4 |= HB4_CD;
2117 else
2118 header->hb4 &= ~HB4_CD;
2119
2120 /* There's no point in updating the cache, since this process will exit and
2121 lose the information after a few queries. We make this call for the alias and
2122 bogus-nxdomain side-effects. */
2123 /* If the crc of the question section doesn't match the crc we sent, then
2124 someone might be attempting to insert bogus values into the cache by
2125 sending replies containing questions and bogus answers. */
2126 #ifdef HAVE_DNSSEC
2127 newhash = hash_questions(header, (unsigned int)m, daemon->namebuff);
2128 if (!newhash || memcmp(hash, newhash, HASH_SIZE) != 0)
2129 {
2130 m = 0;
2131 break;
2132 }
2133 #else
2134 if (crc != questions_crc(header, (unsigned int)m, daemon->namebuff))
2135 {
2136 m = 0;
2137 break;
2138 }
2139 #endif
2140
2141 m = process_reply(header, now, last_server, (unsigned int)m,
2142 option_bool(OPT_NO_REBIND) && !norebind, no_cache_dnssec, cache_secure, bogusanswer,
2143 ad_question, do_bit, added_pheader, check_subnet, &peer_addr);
2144
2145 break;
2146 }
2147 }
2148
2149 /* In case of local answer or no connections made. */
2150 if (m == 0)
2151 m = setup_reply(header, (unsigned int)size, addrp, flags, daemon->local_ttl);
2152 }
2153 }
2154
2155 check_log_writer(1);
2156
2157 *length = htons(m);
2158
2159 if (m == 0 || !read_write(confd, packet, m + sizeof(u16), 0))
2160 return packet;
2161 }
2162 }
2163
2164 static struct frec *allocate_frec(time_t now)
2165 {
2166 struct frec *f;
2167
2168 if ((f = (struct frec *)whine_malloc(sizeof(struct frec))))
2169 {
2170 f->next = daemon->frec_list;
2171 f->time = now;
2172 f->sentto = NULL;
2173 f->rfd4 = NULL;
2174 f->flags = 0;
2175 #ifdef HAVE_IPV6
2176 f->rfd6 = NULL;
2177 #endif
2178 #ifdef HAVE_DNSSEC
2179 f->dependent = NULL;
2180 f->blocking_query = NULL;
2181 f->stash = NULL;
2182 f->orig_domain = NULL;
2183 #endif
2184 daemon->frec_list = f;
2185 }
2186
2187 return f;
2188 }
2189
2190 struct randfd *allocate_rfd(int family)
2191 {
2192 static int finger = 0;
2193 int i;
2194
2195 /* limit the number of sockets we have open to avoid starvation of
2196 (eg) TFTP. Once we have a reasonable number, randomness should be OK */
2197
2198 for (i = 0; i < RANDOM_SOCKS; i++)
2199 if (daemon->randomsocks[i].refcount == 0)
2200 {
2201 if ((daemon->randomsocks[i].fd = random_sock(family)) == -1)
2202 break;
2203
2204 daemon->randomsocks[i].refcount = 1;
2205 daemon->randomsocks[i].family = family;
2206 return &daemon->randomsocks[i];
2207 }
2208
2209 /* No free ones or cannot get new socket, grab an existing one */
2210 for (i = 0; i < RANDOM_SOCKS; i++)
2211 {
2212 int j = (i+finger) % RANDOM_SOCKS;
2213 if (daemon->randomsocks[j].refcount != 0 &&
2214 daemon->randomsocks[j].family == family &&
2215 daemon->randomsocks[j].refcount != 0xffff)
2216 {
2217 finger = j;
2218 daemon->randomsocks[j].refcount++;
2219 return &daemon->randomsocks[j];
2220 }
2221 }
2222
2223 return NULL; /* doom */
2224 }
2225
2226 void free_rfd(struct randfd *rfd)
2227 {
2228 if (rfd && --(rfd->refcount) == 0)
2229 close(rfd->fd);
2230 }
2231
2232 static void free_frec(struct frec *f)
2233 {
2234 free_rfd(f->rfd4);
2235 f->rfd4 = NULL;
2236 f->sentto = NULL;
2237 f->flags = 0;
2238
2239 #ifdef HAVE_IPV6
2240 free_rfd(f->rfd6);
2241 f->rfd6 = NULL;
2242 #endif
2243
2244 #ifdef HAVE_DNSSEC
2245 if (f->stash)
2246 {
2247 blockdata_free(f->stash);
2248 f->stash = NULL;
2249 }
2250
2251 if (f->orig_domain)
2252 {
2253 blockdata_free(f->orig_domain);
2254 f->orig_domain = NULL;
2255 }
2256
2257 /* Anything we're waiting on is pointless now, too */
2258 if (f->blocking_query)
2259 free_frec(f->blocking_query);
2260 f->blocking_query = NULL;
2261 f->dependent = NULL;
2262 #endif
2263 }
2264
2265 /* if wait==NULL return a free or older than TIMEOUT record.
2266 else return *wait zero if one available, or *wait is delay to
2267 when the oldest in-use record will expire. Impose an absolute
2268 limit of 4*TIMEOUT before we wipe things (for random sockets).
2269 If force is set, always return a result, even if we have
2270 to allocate above the limit. */
2271 struct frec *get_new_frec(time_t now, int *wait, int force)
2272 {
2273 struct frec *f, *oldest, *target;
2274 int count;
2275
2276 if (wait)
2277 *wait = 0;
2278
2279 for (f = daemon->frec_list, oldest = NULL, target = NULL, count = 0; f; f = f->next, count++)
2280 if (!f->sentto)
2281 target = f;
2282 else
2283 {
2284 if (difftime(now, f->time) >= 4*TIMEOUT)
2285 {
2286 free_frec(f);
2287 target = f;
2288 }
2289
2290 if (!oldest || difftime(f->time, oldest->time) <= 0)
2291 oldest = f;
2292 }
2293
2294 if (target)
2295 {
2296 target->time = now;
2297 return target;
2298 }
2299
2300 /* can't find empty one, use oldest if there is one
2301 and it's older than timeout */
2302 if (oldest && ((int)difftime(now, oldest->time)) >= TIMEOUT)
2303 {
2304 /* keep stuff for twice timeout if we can by allocating a new
2305 record instead */
2306 if (difftime(now, oldest->time) < 2*TIMEOUT &&
2307 count <= daemon->ftabsize &&
2308 (f = allocate_frec(now)))
2309 return f;
2310
2311 if (!wait)
2312 {
2313 free_frec(oldest);
2314 oldest->time = now;
2315 }
2316 return oldest;
2317 }
2318
2319 /* none available, calculate time 'till oldest record expires */
2320 if (!force && count > daemon->ftabsize)
2321 {
2322 static time_t last_log = 0;
2323
2324 if (oldest && wait)
2325 *wait = oldest->time + (time_t)TIMEOUT - now;
2326
2327 if ((int)difftime(now, last_log) > 5)
2328 {
2329 last_log = now;
2330 my_syslog(LOG_WARNING, _("Maximum number of concurrent DNS queries reached (max: %d)"), daemon->ftabsize);
2331 }
2332
2333 return NULL;
2334 }
2335
2336 if (!(f = allocate_frec(now)) && wait)
2337 /* wait one second on malloc failure */
2338 *wait = 1;
2339
2340 return f; /* OK if malloc fails and this is NULL */
2341 }
2342
2343 /* crc is all-ones if not known. */
2344 static struct frec *lookup_frec(unsigned short id, void *hash)
2345 {
2346 struct frec *f;
2347
2348 for(f = daemon->frec_list; f; f = f->next)
2349 if (f->sentto && f->new_id == id &&
2350 (!hash || memcmp(hash, f->hash, HASH_SIZE) == 0))
2351 return f;
2352
2353 return NULL;
2354 }
2355
2356 static struct frec *lookup_frec_by_sender(unsigned short id,
2357 union mysockaddr *addr,
2358 void *hash)
2359 {
2360 struct frec *f;
2361
2362 for(f = daemon->frec_list; f; f = f->next)
2363 if (f->sentto &&
2364 f->orig_id == id &&
2365 memcmp(hash, f->hash, HASH_SIZE) == 0 &&
2366 sockaddr_isequal(&f->source, addr))
2367 return f;
2368
2369 return NULL;
2370 }
2371
2372 /* Send query packet again, if we can. */
2373 void resend_query()
2374 {
2375 if (daemon->srv_save)
2376 {
2377 int fd;
2378
2379 if (daemon->srv_save->sfd)
2380 fd = daemon->srv_save->sfd->fd;
2381 else if (daemon->rfd_save && daemon->rfd_save->refcount != 0)
2382 fd = daemon->rfd_save->fd;
2383 else
2384 return;
2385
2386 while(retry_send(sendto(fd, daemon->packet, daemon->packet_len, 0,
2387 &daemon->srv_save->addr.sa,
2388 sa_len(&daemon->srv_save->addr))));
2389 }
2390 }
2391
2392 /* A server record is going away, remove references to it */
2393 void server_gone(struct server *server)
2394 {
2395 struct frec *f;
2396
2397 for (f = daemon->frec_list; f; f = f->next)
2398 if (f->sentto && f->sentto == server)
2399 free_frec(f);
2400
2401 if (daemon->last_server == server)
2402 daemon->last_server = NULL;
2403
2404 if (daemon->srv_save == server)
2405 daemon->srv_save = NULL;
2406 }
2407
2408 /* return unique random ids. */
2409 static unsigned short get_id(void)
2410 {
2411 unsigned short ret = 0;
2412
2413 do
2414 ret = rand16();
2415 while (lookup_frec(ret, NULL));
2416
2417 return ret;
2418 }
2419
2420
2421
2422
2423