2 * iterator/iter_utils.c - iterative resolver module utility functions.
4 * Copyright (c) 2007, NLnet Labs. All rights reserved.
6 * This software is open source.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * Redistributions of source code must retain the above copyright notice,
13 * this list of conditions and the following disclaimer.
15 * Redistributions in binary form must reproduce the above copyright notice,
16 * this list of conditions and the following disclaimer in the documentation
17 * and/or other materials provided with the distribution.
19 * Neither the name of the NLNET LABS nor the names of its contributors may
20 * be used to endorse or promote products derived from this software without
21 * specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
24 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
25 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
26 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
27 * HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
28 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
29 * TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
30 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
39 * This file contains functions to assist the iterator module.
40 * Configuration options. Forward zones.
43 #include "iterator/iter_utils.h"
44 #include "iterator/iterator.h"
45 #include "iterator/iter_hints.h"
46 #include "iterator/iter_fwd.h"
47 #include "iterator/iter_donotq.h"
48 #include "iterator/iter_delegpt.h"
49 #include "iterator/iter_priv.h"
50 #include "services/cache/infra.h"
51 #include "services/cache/dns.h"
52 #include "services/cache/rrset.h"
53 #include "util/net_help.h"
54 #include "util/module.h"
56 #include "util/config_file.h"
57 #include "util/regional.h"
58 #include "util/data/msgparse.h"
59 #include "util/data/dname.h"
60 #include "util/random.h"
61 #include "util/fptr_wlist.h"
62 #include "validator/val_anchor.h"
63 #include "validator/val_kcache.h"
64 #include "validator/val_kentry.h"
65 #include "validator/val_utils.h"
66 #include "validator/val_sigcrypt.h"
67 #include "sldns/sbuffer.h"
68 #include "sldns/str2wire.h"
70 /** time when nameserver glue is said to be 'recent' */
71 #define SUSPICION_RECENT_EXPIRY 86400
72 /** penalty to validation failed blacklisted IPs */
73 #define BLACKLIST_PENALTY (USEFUL_SERVER_TOP_TIMEOUT*4)
75 /** fillup fetch policy array */
77 fetch_fill(struct iter_env* ie, const char* str)
79 char* s = (char*)str, *e;
81 for(i=0; i<ie->max_dependency_depth+1; i++) {
82 ie->target_fetch_policy[i] = strtol(s, &e, 10);
84 fatal_exit("cannot parse fetch policy number %s", s);
89 /** Read config string that represents the target fetch policy */
91 read_fetch_policy(struct iter_env* ie, const char* str)
93 int count = cfg_count_numbers(str);
95 log_err("Cannot parse target fetch policy: \"%s\"", str);
98 ie->max_dependency_depth = count - 1;
99 ie->target_fetch_policy = (int*)calloc(
100 (size_t)ie->max_dependency_depth+1, sizeof(int));
101 if(!ie->target_fetch_policy) {
102 log_err("alloc fetch policy: out of memory");
109 /** apply config caps whitelist items to name tree */
111 caps_white_apply_cfg(rbtree_type* ntree, struct config_file* cfg)
113 struct config_strlist* p;
114 for(p=cfg->caps_whitelist; p; p=p->next) {
115 struct name_tree_node* n;
117 uint8_t* nm = sldns_str2wire_dname(p->str, &len);
119 log_err("could not parse %s", p->str);
122 n = (struct name_tree_node*)calloc(1, sizeof(*n));
124 log_err("out of memory");
131 n->labs = dname_count_labels(nm);
132 n->dclass = LDNS_RR_CLASS_IN;
133 if(!name_tree_insert(ntree, n, nm, len, n->labs, n->dclass)) {
134 /* duplicate element ignored, idempotent */
139 name_tree_init_parents(ntree);
144 iter_apply_cfg(struct iter_env* iter_env, struct config_file* cfg)
147 /* target fetch policy */
148 if(!read_fetch_policy(iter_env, cfg->target_fetch_policy))
150 for(i=0; i<iter_env->max_dependency_depth+1; i++)
151 verbose(VERB_QUERY, "target fetch policy for level %d is %d",
152 i, iter_env->target_fetch_policy[i]);
154 if(!iter_env->donotq)
155 iter_env->donotq = donotq_create();
156 if(!iter_env->donotq || !donotq_apply_cfg(iter_env->donotq, cfg)) {
157 log_err("Could not set donotqueryaddresses");
161 iter_env->priv = priv_create();
162 if(!iter_env->priv || !priv_apply_cfg(iter_env->priv, cfg)) {
163 log_err("Could not set private addresses");
166 if(cfg->caps_whitelist) {
167 if(!iter_env->caps_white)
168 iter_env->caps_white = rbtree_create(name_tree_compare);
169 if(!iter_env->caps_white || !caps_white_apply_cfg(
170 iter_env->caps_white, cfg)) {
171 log_err("Could not set capsforid whitelist");
176 iter_env->supports_ipv6 = cfg->do_ip6;
177 iter_env->supports_ipv4 = cfg->do_ip4;
181 /** filter out unsuitable targets
182 * @param iter_env: iterator environment with ipv6-support flag.
183 * @param env: module environment with infra cache.
184 * @param name: zone name
185 * @param namelen: length of name
186 * @param qtype: query type (host order).
187 * @param now: current time
188 * @param a: address in delegation point we are examining.
189 * @return an integer that signals the target suitability.
191 * -1: The address should be omitted from the list.
193 * o The address is bogus (DNSSEC validation failure).
194 * o Listed as donotquery
195 * o is ipv6 but no ipv6 support (in operating system).
196 * o is ipv4 but no ipv4 support (in operating system).
198 * Otherwise, an rtt in milliseconds.
199 * 0 .. USEFUL_SERVER_TOP_TIMEOUT-1
200 * The roundtrip time timeout estimate. less than 2 minutes.
201 * Note that util/rtt.c has a MIN_TIMEOUT of 50 msec, thus
202 * values 0 .. 49 are not used, unless that is changed.
203 * USEFUL_SERVER_TOP_TIMEOUT
204 * This value exactly is given for unresponsive blacklisted.
205 * USEFUL_SERVER_TOP_TIMEOUT+1
206 * For non-blacklisted servers: huge timeout, but has traffic.
207 * USEFUL_SERVER_TOP_TIMEOUT*1 ..
208 * parent-side lame servers get this penalty. A dispreferential
209 * server. (lame in delegpt).
210 * USEFUL_SERVER_TOP_TIMEOUT*2 ..
211 * dnsseclame servers get penalty
212 * USEFUL_SERVER_TOP_TIMEOUT*3 ..
213 * recursion lame servers get penalty
214 * UNKNOWN_SERVER_NICENESS
215 * If no information is known about the server, this is
216 * returned. 376 msec or so.
217 * +BLACKLIST_PENALTY (of USEFUL_TOP_TIMEOUT*4) for dnssec failed IPs.
219 * When a final value is chosen that is dnsseclame ; dnsseclameness checking
220 * is turned off (so we do not discard the reply).
221 * When a final value is chosen that is recursionlame; RD bit is set on query.
222 * Because of the numbers this means recursionlame also have dnssec lameness
223 * checking turned off.
226 iter_filter_unsuitable(struct iter_env* iter_env, struct module_env* env,
227 uint8_t* name, size_t namelen, uint16_t qtype, time_t now,
228 struct delegpt_addr* a)
230 int rtt, lame, reclame, dnsseclame;
232 return -1; /* address of server is bogus */
233 if(donotq_lookup(iter_env->donotq, &a->addr, a->addrlen)) {
234 log_addr(VERB_ALGO, "skip addr on the donotquery list",
235 &a->addr, a->addrlen);
236 return -1; /* server is on the donotquery list */
238 if(!iter_env->supports_ipv6 && addr_is_ip6(&a->addr, a->addrlen)) {
239 return -1; /* there is no ip6 available */
241 if(!iter_env->supports_ipv4 && !addr_is_ip6(&a->addr, a->addrlen)) {
242 return -1; /* there is no ip4 available */
244 /* check lameness - need zone , class info */
245 if(infra_get_lame_rtt(env->infra_cache, &a->addr, a->addrlen,
246 name, namelen, qtype, &lame, &dnsseclame, &reclame,
248 log_addr(VERB_ALGO, "servselect", &a->addr, a->addrlen);
249 verbose(VERB_ALGO, " rtt=%d%s%s%s%s", rtt,
251 dnsseclame?" DNSSEC_LAME":"",
252 reclame?" REC_LAME":"",
253 a->lame?" ADDR_LAME":"");
255 return -1; /* server is lame */
256 else if(rtt >= USEFUL_SERVER_TOP_TIMEOUT)
257 /* server is unresponsive,
258 * we used to return TOP_TIMEOUT, but fairly useless,
259 * because if == TOP_TIMEOUT is dropped because
260 * blacklisted later, instead, remove it here, so
261 * other choices (that are not blacklisted) can be
264 /* select remainder from worst to best */
266 return rtt+USEFUL_SERVER_TOP_TIMEOUT*3; /* nonpref */
267 else if(dnsseclame || a->dnsseclame)
268 return rtt+USEFUL_SERVER_TOP_TIMEOUT*2; /* nonpref */
270 return rtt+USEFUL_SERVER_TOP_TIMEOUT+1; /* nonpref */
273 /* no server information present */
275 return UNKNOWN_SERVER_NICENESS+USEFUL_SERVER_TOP_TIMEOUT*2; /* nonpref */
277 return USEFUL_SERVER_TOP_TIMEOUT+1+UNKNOWN_SERVER_NICENESS; /* nonpref */
278 return UNKNOWN_SERVER_NICENESS;
281 /** lookup RTT information, and also store fastest rtt (if any) */
283 iter_fill_rtt(struct iter_env* iter_env, struct module_env* env,
284 uint8_t* name, size_t namelen, uint16_t qtype, time_t now,
285 struct delegpt* dp, int* best_rtt, struct sock_list* blacklist)
288 struct delegpt_addr* a;
290 return 0; /* NS bogus, all bogus, nothing found */
291 for(a=dp->result_list; a; a = a->next_result) {
292 a->sel_rtt = iter_filter_unsuitable(iter_env, env,
293 name, namelen, qtype, now, a);
294 if(a->sel_rtt != -1) {
295 if(sock_list_find(blacklist, &a->addr, a->addrlen))
296 a->sel_rtt += BLACKLIST_PENALTY;
299 *best_rtt = a->sel_rtt;
301 } else if(a->sel_rtt < *best_rtt) {
302 *best_rtt = a->sel_rtt;
309 /** filter the address list, putting best targets at front,
310 * returns number of best targets (or 0, no suitable targets) */
312 iter_filter_order(struct iter_env* iter_env, struct module_env* env,
313 uint8_t* name, size_t namelen, uint16_t qtype, time_t now,
314 struct delegpt* dp, int* selected_rtt, int open_target,
315 struct sock_list* blacklist, time_t prefetch)
317 int got_num = 0, low_rtt = 0, swap_to_front, rtt_band = RTT_BAND;
318 struct delegpt_addr* a, *n, *prev=NULL;
320 /* fillup sel_rtt and find best rtt in the bunch */
321 got_num = iter_fill_rtt(iter_env, env, name, namelen, qtype, now, dp,
322 &low_rtt, blacklist);
325 if(low_rtt >= USEFUL_SERVER_TOP_TIMEOUT &&
326 (delegpt_count_missing_targets(dp) > 0 || open_target > 0)) {
327 verbose(VERB_ALGO, "Bad choices, trying to get more choice");
328 return 0; /* we want more choice. The best choice is a bad one.
329 return 0 to force the caller to fetch more */
332 if(env->cfg->low_rtt_permil != 0 && prefetch == 0 &&
333 low_rtt < env->cfg->low_rtt &&
334 ub_random_max(env->rnd, 1000) < env->cfg->low_rtt_permil) {
335 /* the query is not prefetch, but for a downstream client,
336 * there is a low_rtt (fast) server. We choose that x% of the
338 /* pick rtt numbers from 0..LOWBAND_RTT */
339 rtt_band = env->cfg->low_rtt - low_rtt;
345 /* skip unsuitable targets */
346 if(a->sel_rtt == -1) {
351 /* classify the server address and determine what to do */
353 if(a->sel_rtt >= low_rtt && a->sel_rtt - low_rtt <= rtt_band) {
356 } else if(a->sel_rtt<low_rtt && low_rtt-a->sel_rtt<=rtt_band) {
360 /* swap to front if necessary, or move to next result */
361 if(swap_to_front && prev) {
363 prev->next_result = n;
364 a->next_result = dp->result_list;
372 *selected_rtt = low_rtt;
374 if (env->cfg->prefer_ip6) {
378 int attempt = -1; /* filter to make sure addresses have
379 less attempts on them than the first, to force round
380 robin when all the IPv6 addresses fail */
381 int num4ok = 0; /* number ip4 at low attempt count */
385 for(i = 0; i < got_num; i++) {
387 if(a->addr.ss_family != AF_INET6 && attempt == -1) {
388 /* if we only have ip4 at low attempt count,
389 * then ip6 is failing, and we need to
390 * select one of the remaining IPv4 addrs */
391 attempt = a->attempts;
393 num4_lowrtt = a->sel_rtt;
394 } else if(a->addr.ss_family != AF_INET6 && attempt == a->attempts) {
396 if(num4_lowrtt == 0 || a->sel_rtt < num4_lowrtt) {
397 num4_lowrtt = a->sel_rtt;
400 if(a->addr.ss_family == AF_INET6) {
402 attempt = a->attempts;
403 } else if(a->attempts > attempt) {
408 if(low_rtt6 == 0 || a->sel_rtt < low_rtt6) {
409 low_rtt6 = a->sel_rtt;
412 /* swap to front if IPv6, or move to next result */
413 if(swap_to_front && prev) {
415 prev->next_result = n;
416 a->next_result = dp->result_list;
426 *selected_rtt = low_rtt6;
427 } else if(num4ok > 0) {
429 *selected_rtt = num4_lowrtt;
436 iter_server_selection(struct iter_env* iter_env,
437 struct module_env* env, struct delegpt* dp,
438 uint8_t* name, size_t namelen, uint16_t qtype, int* dnssec_lame,
439 int* chase_to_rd, int open_target, struct sock_list* blacklist,
444 struct delegpt_addr* a, *prev;
445 int num = iter_filter_order(iter_env, env, name, namelen, qtype,
446 *env->now, dp, &selrtt, open_target, blacklist, prefetch);
450 verbose(VERB_ALGO, "selrtt %d", selrtt);
451 if(selrtt > BLACKLIST_PENALTY) {
452 if(selrtt-BLACKLIST_PENALTY > USEFUL_SERVER_TOP_TIMEOUT*3) {
453 verbose(VERB_ALGO, "chase to "
454 "blacklisted recursion lame server");
457 if(selrtt-BLACKLIST_PENALTY > USEFUL_SERVER_TOP_TIMEOUT*2) {
458 verbose(VERB_ALGO, "chase to "
459 "blacklisted dnssec lame server");
463 if(selrtt > USEFUL_SERVER_TOP_TIMEOUT*3) {
464 verbose(VERB_ALGO, "chase to recursion lame server");
467 if(selrtt > USEFUL_SERVER_TOP_TIMEOUT*2) {
468 verbose(VERB_ALGO, "chase to dnssec lame server");
471 if(selrtt == USEFUL_SERVER_TOP_TIMEOUT) {
472 verbose(VERB_ALGO, "chase to blacklisted lame server");
479 if(++a->attempts < OUTBOUND_MSG_RETRY)
481 dp->result_list = a->next_result;
485 /* randomly select a target from the list */
487 /* grab secure random number, to pick unexpected server.
488 * also we need it to be threadsafe. */
489 sel = ub_random_max(env->rnd, num);
492 while(sel > 0 && a) {
497 if(!a) /* robustness */
499 if(++a->attempts < OUTBOUND_MSG_RETRY)
501 /* remove it from the delegation point result list */
503 prev->next_result = a->next_result;
504 else dp->result_list = a->next_result;
509 dns_alloc_msg(sldns_buffer* pkt, struct msg_parse* msg,
510 struct regional* region)
512 struct dns_msg* m = (struct dns_msg*)regional_alloc(region,
513 sizeof(struct dns_msg));
516 memset(m, 0, sizeof(*m));
517 if(!parse_create_msg(pkt, msg, NULL, &m->qinfo, &m->rep, region)) {
518 log_err("malloc failure: allocating incoming dns_msg");
525 dns_copy_msg(struct dns_msg* from, struct regional* region)
527 struct dns_msg* m = (struct dns_msg*)regional_alloc(region,
528 sizeof(struct dns_msg));
531 m->qinfo = from->qinfo;
532 if(!(m->qinfo.qname = regional_alloc_init(region, from->qinfo.qname,
533 from->qinfo.qname_len)))
535 if(!(m->rep = reply_info_copy(from->rep, NULL, region)))
541 iter_dns_store(struct module_env* env, struct query_info* msgqinf,
542 struct reply_info* msgrep, int is_referral, time_t leeway, int pside,
543 struct regional* region, uint16_t flags)
545 if(!dns_cache_store(env, msgqinf, msgrep, is_referral, leeway,
546 pside, region, flags))
547 log_err("out of memory: cannot store data in cache");
551 iter_ns_probability(struct ub_randstate* rnd, int n, int m)
554 if(n == m) /* 100% chance */
556 /* we do not need secure random numbers here, but
557 * we do need it to be threadsafe, so we use this */
558 sel = ub_random_max(rnd, m);
562 /** detect dependency cycle for query and target */
564 causes_cycle(struct module_qstate* qstate, uint8_t* name, size_t namelen,
565 uint16_t t, uint16_t c)
567 struct query_info qinf;
569 qinf.qname_len = namelen;
572 qinf.local_alias = NULL;
573 fptr_ok(fptr_whitelist_modenv_detect_cycle(
574 qstate->env->detect_cycle));
575 return (*qstate->env->detect_cycle)(qstate, &qinf,
576 (uint16_t)(BIT_RD|BIT_CD), qstate->is_priming,
581 iter_mark_cycle_targets(struct module_qstate* qstate, struct delegpt* dp)
583 struct delegpt_ns* ns;
584 for(ns = dp->nslist; ns; ns = ns->next) {
587 /* see if this ns as target causes dependency cycle */
588 if(causes_cycle(qstate, ns->name, ns->namelen,
589 LDNS_RR_TYPE_AAAA, qstate->qinfo.qclass) ||
590 causes_cycle(qstate, ns->name, ns->namelen,
591 LDNS_RR_TYPE_A, qstate->qinfo.qclass)) {
592 log_nametypeclass(VERB_QUERY, "skipping target due "
593 "to dependency cycle (harden-glue: no may "
594 "fix some of the cycles)",
595 ns->name, LDNS_RR_TYPE_A,
596 qstate->qinfo.qclass);
603 iter_mark_pside_cycle_targets(struct module_qstate* qstate, struct delegpt* dp)
605 struct delegpt_ns* ns;
606 for(ns = dp->nslist; ns; ns = ns->next) {
607 if(ns->done_pside4 && ns->done_pside6)
609 /* see if this ns as target causes dependency cycle */
610 if(causes_cycle(qstate, ns->name, ns->namelen,
611 LDNS_RR_TYPE_A, qstate->qinfo.qclass)) {
612 log_nametypeclass(VERB_QUERY, "skipping target due "
613 "to dependency cycle", ns->name,
614 LDNS_RR_TYPE_A, qstate->qinfo.qclass);
617 if(causes_cycle(qstate, ns->name, ns->namelen,
618 LDNS_RR_TYPE_AAAA, qstate->qinfo.qclass)) {
619 log_nametypeclass(VERB_QUERY, "skipping target due "
620 "to dependency cycle", ns->name,
621 LDNS_RR_TYPE_AAAA, qstate->qinfo.qclass);
628 iter_dp_is_useless(struct query_info* qinfo, uint16_t qflags,
631 struct delegpt_ns* ns;
634 * o no addresses are provided.
635 * o all NS items are required glue.
638 * o no addresses are provided.
639 * o the query is for one of the nameservers in dp,
640 * and that nameserver is a glue-name for this dp.
644 /* either available or unused targets */
645 if(dp->usable_list || dp->result_list)
648 /* see if query is for one of the nameservers, which is glue */
649 if( (qinfo->qtype == LDNS_RR_TYPE_A ||
650 qinfo->qtype == LDNS_RR_TYPE_AAAA) &&
651 dname_subdomain_c(qinfo->qname, dp->name) &&
652 delegpt_find_ns(dp, qinfo->qname, qinfo->qname_len))
655 for(ns = dp->nslist; ns; ns = ns->next) {
656 if(ns->resolved) /* skip failed targets */
658 if(!dname_subdomain_c(ns->name, dp->name))
659 return 0; /* one address is not required glue */
665 iter_qname_indicates_dnssec(struct module_env* env, struct query_info *qinfo)
667 struct trust_anchor* a;
668 if(!env || !env->anchors || !qinfo || !qinfo->qname)
670 /* a trust anchor exists above the name? */
671 if((a=anchors_lookup(env->anchors, qinfo->qname, qinfo->qname_len,
673 if(a->numDS == 0 && a->numDNSKEY == 0) {
674 /* insecure trust point */
675 lock_basic_unlock(&a->lock);
678 lock_basic_unlock(&a->lock);
681 /* no trust anchor above it. */
686 iter_indicates_dnssec(struct module_env* env, struct delegpt* dp,
687 struct dns_msg* msg, uint16_t dclass)
689 struct trust_anchor* a;
690 /* information not available, !env->anchors can be common */
691 if(!env || !env->anchors || !dp || !dp->name)
693 /* a trust anchor exists with this name, RRSIGs expected */
694 if((a=anchor_find(env->anchors, dp->name, dp->namelabs, dp->namelen,
696 if(a->numDS == 0 && a->numDNSKEY == 0) {
697 /* insecure trust point */
698 lock_basic_unlock(&a->lock);
701 lock_basic_unlock(&a->lock);
704 /* see if DS rrset was given, in AUTH section */
705 if(msg && msg->rep &&
706 reply_find_rrset_section_ns(msg->rep, dp->name, dp->namelen,
707 LDNS_RR_TYPE_DS, dclass))
709 /* look in key cache */
711 struct key_entry_key* kk = key_cache_obtain(env->key_cache,
712 dp->name, dp->namelen, dclass, env->scratch, *env->now);
714 if(query_dname_compare(kk->name, dp->name) == 0) {
715 if(key_entry_isgood(kk) || key_entry_isbad(kk)) {
716 regional_free_all(env->scratch);
718 } else if(key_entry_isnull(kk)) {
719 regional_free_all(env->scratch);
723 regional_free_all(env->scratch);
730 iter_msg_has_dnssec(struct dns_msg* msg)
733 if(!msg || !msg->rep)
735 for(i=0; i<msg->rep->an_numrrsets + msg->rep->ns_numrrsets; i++) {
736 if(((struct packed_rrset_data*)msg->rep->rrsets[i]->
737 entry.data)->rrsig_count > 0)
740 /* empty message has no DNSSEC info, with DNSSEC the reply is
741 * not empty (NSEC) */
745 int iter_msg_from_zone(struct dns_msg* msg, struct delegpt* dp,
746 enum response_type type, uint16_t dclass)
748 if(!msg || !dp || !msg->rep || !dp->name)
750 /* SOA RRset - always from reply zone */
751 if(reply_find_rrset_section_an(msg->rep, dp->name, dp->namelen,
752 LDNS_RR_TYPE_SOA, dclass) ||
753 reply_find_rrset_section_ns(msg->rep, dp->name, dp->namelen,
754 LDNS_RR_TYPE_SOA, dclass))
756 if(type == RESPONSE_TYPE_REFERRAL) {
758 /* if it adds a single label, i.e. we expect .com,
759 * and referral to example.com. NS ... , then origin zone
760 * is .com. For a referral to sub.example.com. NS ... then
761 * we do not know, since example.com. may be in between. */
762 for(i=0; i<msg->rep->an_numrrsets+msg->rep->ns_numrrsets;
764 struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
765 if(ntohs(s->rk.type) == LDNS_RR_TYPE_NS &&
766 ntohs(s->rk.rrset_class) == dclass) {
767 int l = dname_count_labels(s->rk.dname);
768 if(l == dp->namelabs + 1 &&
769 dname_strict_subdomain(s->rk.dname,
770 l, dp->name, dp->namelabs))
776 log_assert(type==RESPONSE_TYPE_ANSWER || type==RESPONSE_TYPE_CNAME);
777 /* not a referral, and not lame delegation (upwards), so,
778 * any NS rrset must be from the zone itself */
779 if(reply_find_rrset_section_an(msg->rep, dp->name, dp->namelen,
780 LDNS_RR_TYPE_NS, dclass) ||
781 reply_find_rrset_section_ns(msg->rep, dp->name, dp->namelen,
782 LDNS_RR_TYPE_NS, dclass))
784 /* a DNSKEY set is expected at the zone apex as well */
785 /* this is for 'minimal responses' for DNSKEYs */
786 if(reply_find_rrset_section_an(msg->rep, dp->name, dp->namelen,
787 LDNS_RR_TYPE_DNSKEY, dclass))
793 * check equality of two rrsets
796 * @return true if equal
799 rrset_equal(struct ub_packed_rrset_key* k1, struct ub_packed_rrset_key* k2)
801 struct packed_rrset_data* d1 = (struct packed_rrset_data*)
803 struct packed_rrset_data* d2 = (struct packed_rrset_data*)
806 if(k1->rk.dname_len != k2->rk.dname_len ||
807 k1->rk.flags != k2->rk.flags ||
808 k1->rk.type != k2->rk.type ||
809 k1->rk.rrset_class != k2->rk.rrset_class ||
810 query_dname_compare(k1->rk.dname, k2->rk.dname) != 0)
812 if( /* do not check ttl: d1->ttl != d2->ttl || */
813 d1->count != d2->count ||
814 d1->rrsig_count != d2->rrsig_count ||
815 d1->trust != d2->trust ||
816 d1->security != d2->security)
818 t = d1->count + d1->rrsig_count;
820 if(d1->rr_len[i] != d2->rr_len[i] ||
821 /* no ttl check: d1->rr_ttl[i] != d2->rr_ttl[i] ||*/
822 memcmp(d1->rr_data[i], d2->rr_data[i],
830 reply_equal(struct reply_info* p, struct reply_info* q, struct regional* region)
833 if(p->flags != q->flags ||
834 p->qdcount != q->qdcount ||
835 /* do not check TTL, this may differ */
838 p->prefetch_ttl != q->prefetch_ttl ||
840 p->security != q->security ||
841 p->an_numrrsets != q->an_numrrsets ||
842 p->ns_numrrsets != q->ns_numrrsets ||
843 p->ar_numrrsets != q->ar_numrrsets ||
844 p->rrset_count != q->rrset_count)
846 for(i=0; i<p->rrset_count; i++) {
847 if(!rrset_equal(p->rrsets[i], q->rrsets[i])) {
848 if(!rrset_canonical_equal(region, p->rrsets[i],
850 regional_free_all(region);
853 regional_free_all(region);
860 caps_strip_reply(struct reply_info* rep)
864 /* see if message is a referral, in which case the additional and
865 * NS record cannot be removed */
866 /* referrals have the AA flag unset (strict check, not elsewhere in
867 * unbound, but for 0x20 this is very convenient). */
868 if(!(rep->flags&BIT_AA))
870 /* remove the additional section from the reply */
871 if(rep->ar_numrrsets != 0) {
872 verbose(VERB_ALGO, "caps fallback: removing additional section");
873 rep->rrset_count -= rep->ar_numrrsets;
874 rep->ar_numrrsets = 0;
876 /* is there an NS set in the authority section to remove? */
877 /* the failure case (Cisco firewalls) only has one rrset in authsec */
878 for(i=rep->an_numrrsets; i<rep->an_numrrsets+rep->ns_numrrsets; i++) {
879 struct ub_packed_rrset_key* s = rep->rrsets[i];
880 if(ntohs(s->rk.type) == LDNS_RR_TYPE_NS) {
881 /* remove NS rrset and break from loop (loop limits
883 /* move last rrset into this position (there is no
884 * additional section any more) */
885 verbose(VERB_ALGO, "caps fallback: removing NS rrset");
886 if(i < rep->rrset_count-1)
887 rep->rrsets[i]=rep->rrsets[rep->rrset_count-1];
889 rep->ns_numrrsets --;
895 int caps_failed_rcode(struct reply_info* rep)
897 return !(FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NOERROR ||
898 FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NXDOMAIN);
902 iter_store_parentside_rrset(struct module_env* env,
903 struct ub_packed_rrset_key* rrset)
905 struct rrset_ref ref;
906 rrset = packed_rrset_copy_alloc(rrset, env->alloc, *env->now);
908 log_err("malloc failure in store_parentside_rrset");
911 rrset->rk.flags |= PACKED_RRSET_PARENT_SIDE;
912 rrset->entry.hash = rrset_key_hash(&rrset->rk);
915 /* ignore ret: if it was in the cache, ref updated */
916 (void)rrset_cache_update(env->rrset_cache, &ref, env->alloc, *env->now);
919 /** fetch NS record from reply, if any */
920 static struct ub_packed_rrset_key*
921 reply_get_NS_rrset(struct reply_info* rep)
924 for(i=0; i<rep->rrset_count; i++) {
925 if(rep->rrsets[i]->rk.type == htons(LDNS_RR_TYPE_NS)) {
926 return rep->rrsets[i];
933 iter_store_parentside_NS(struct module_env* env, struct reply_info* rep)
935 struct ub_packed_rrset_key* rrset = reply_get_NS_rrset(rep);
937 log_rrset_key(VERB_ALGO, "store parent-side NS", rrset);
938 iter_store_parentside_rrset(env, rrset);
942 void iter_store_parentside_neg(struct module_env* env,
943 struct query_info* qinfo, struct reply_info* rep)
945 /* TTL: NS from referral in iq->deleg_msg,
946 * or first RR from iq->response,
947 * or servfail5secs if !iq->response */
948 time_t ttl = NORR_TTL;
949 struct ub_packed_rrset_key* neg;
950 struct packed_rrset_data* newd;
952 struct ub_packed_rrset_key* rrset = reply_get_NS_rrset(rep);
953 if(!rrset && rep->rrset_count != 0) rrset = rep->rrsets[0];
954 if(rrset) ttl = ub_packed_rrset_ttl(rrset);
956 /* create empty rrset to store */
957 neg = (struct ub_packed_rrset_key*)regional_alloc(env->scratch,
958 sizeof(struct ub_packed_rrset_key));
960 log_err("out of memory in store_parentside_neg");
963 memset(&neg->entry, 0, sizeof(neg->entry));
964 neg->entry.key = neg;
965 neg->rk.type = htons(qinfo->qtype);
966 neg->rk.rrset_class = htons(qinfo->qclass);
968 neg->rk.dname = regional_alloc_init(env->scratch, qinfo->qname,
971 log_err("out of memory in store_parentside_neg");
974 neg->rk.dname_len = qinfo->qname_len;
975 neg->entry.hash = rrset_key_hash(&neg->rk);
976 newd = (struct packed_rrset_data*)regional_alloc_zero(env->scratch,
977 sizeof(struct packed_rrset_data) + sizeof(size_t) +
978 sizeof(uint8_t*) + sizeof(time_t) + sizeof(uint16_t));
980 log_err("out of memory in store_parentside_neg");
983 neg->entry.data = newd;
985 /* entry must have one RR, otherwise not valid in cache.
986 * put in one RR with empty rdata: those are ignored as nameserver */
988 newd->rrsig_count = 0;
989 newd->trust = rrset_trust_ans_noAA;
990 newd->rr_len = (size_t*)((uint8_t*)newd +
991 sizeof(struct packed_rrset_data));
992 newd->rr_len[0] = 0 /* zero len rdata */ + sizeof(uint16_t);
993 packed_rrset_ptr_fixup(newd);
994 newd->rr_ttl[0] = newd->ttl;
995 sldns_write_uint16(newd->rr_data[0], 0 /* zero len rdata */);
997 log_rrset_key(VERB_ALGO, "store parent-side negative", neg);
998 iter_store_parentside_rrset(env, neg);
1002 iter_lookup_parent_NS_from_cache(struct module_env* env, struct delegpt* dp,
1003 struct regional* region, struct query_info* qinfo)
1005 struct ub_packed_rrset_key* akey;
1006 akey = rrset_cache_lookup(env->rrset_cache, dp->name,
1007 dp->namelen, LDNS_RR_TYPE_NS, qinfo->qclass,
1008 PACKED_RRSET_PARENT_SIDE, *env->now, 0);
1010 log_rrset_key(VERB_ALGO, "found parent-side NS in cache", akey);
1011 dp->has_parent_side_NS = 1;
1012 /* and mark the new names as lame */
1013 if(!delegpt_rrset_add_ns(dp, region, akey, 1)) {
1014 lock_rw_unlock(&akey->entry.lock);
1017 lock_rw_unlock(&akey->entry.lock);
1022 int iter_lookup_parent_glue_from_cache(struct module_env* env,
1023 struct delegpt* dp, struct regional* region, struct query_info* qinfo)
1025 struct ub_packed_rrset_key* akey;
1026 struct delegpt_ns* ns;
1027 size_t num = delegpt_count_targets(dp);
1028 for(ns = dp->nslist; ns; ns = ns->next) {
1029 /* get cached parentside A */
1030 akey = rrset_cache_lookup(env->rrset_cache, ns->name,
1031 ns->namelen, LDNS_RR_TYPE_A, qinfo->qclass,
1032 PACKED_RRSET_PARENT_SIDE, *env->now, 0);
1034 log_rrset_key(VERB_ALGO, "found parent-side", akey);
1035 ns->done_pside4 = 1;
1036 /* a negative-cache-element has no addresses it adds */
1037 if(!delegpt_add_rrset_A(dp, region, akey, 1))
1038 log_err("malloc failure in lookup_parent_glue");
1039 lock_rw_unlock(&akey->entry.lock);
1041 /* get cached parentside AAAA */
1042 akey = rrset_cache_lookup(env->rrset_cache, ns->name,
1043 ns->namelen, LDNS_RR_TYPE_AAAA, qinfo->qclass,
1044 PACKED_RRSET_PARENT_SIDE, *env->now, 0);
1046 log_rrset_key(VERB_ALGO, "found parent-side", akey);
1047 ns->done_pside6 = 1;
1048 /* a negative-cache-element has no addresses it adds */
1049 if(!delegpt_add_rrset_AAAA(dp, region, akey, 1))
1050 log_err("malloc failure in lookup_parent_glue");
1051 lock_rw_unlock(&akey->entry.lock);
1054 /* see if new (but lame) addresses have become available */
1055 return delegpt_count_targets(dp) != num;
1059 iter_get_next_root(struct iter_hints* hints, struct iter_forwards* fwd,
1062 uint16_t c1 = *c, c2 = *c;
1063 int r1 = hints_next_root(hints, &c1);
1064 int r2 = forwards_next_root(fwd, &c2);
1065 if(!r1 && !r2) /* got none, end of list */
1067 else if(!r1) /* got one, return that */
1071 else if(c1 < c2) /* got both take smallest */
1078 iter_scrub_ds(struct dns_msg* msg, struct ub_packed_rrset_key* ns, uint8_t* z)
1080 /* Only the DS record for the delegation itself is expected.
1081 * We allow DS for everything between the bailiwick and the
1082 * zonecut, thus DS records must be at or above the zonecut.
1083 * And the DS records must be below the server authority zone.
1084 * The answer section is already scrubbed. */
1085 size_t i = msg->rep->an_numrrsets;
1086 while(i < (msg->rep->an_numrrsets + msg->rep->ns_numrrsets)) {
1087 struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
1088 if(ntohs(s->rk.type) == LDNS_RR_TYPE_DS &&
1089 (!ns || !dname_subdomain_c(ns->rk.dname, s->rk.dname)
1090 || query_dname_compare(z, s->rk.dname) == 0)) {
1091 log_nametypeclass(VERB_ALGO, "removing irrelevant DS",
1092 s->rk.dname, ntohs(s->rk.type),
1093 ntohs(s->rk.rrset_class));
1094 memmove(msg->rep->rrsets+i, msg->rep->rrsets+i+1,
1095 sizeof(struct ub_packed_rrset_key*) *
1096 (msg->rep->rrset_count-i-1));
1097 msg->rep->ns_numrrsets--;
1098 msg->rep->rrset_count--;
1099 /* stay at same i, but new record */
1106 void iter_dec_attempts(struct delegpt* dp, int d)
1108 struct delegpt_addr* a;
1109 for(a=dp->target_list; a; a = a->next_target) {
1110 if(a->attempts >= OUTBOUND_MSG_RETRY) {
1111 /* add back to result list */
1112 a->next_result = dp->result_list;
1113 dp->result_list = a;
1117 else a->attempts = 0;
1121 void iter_merge_retry_counts(struct delegpt* dp, struct delegpt* old)
1123 struct delegpt_addr* a, *o, *prev;
1124 for(a=dp->target_list; a; a = a->next_target) {
1125 o = delegpt_find_addr(old, &a->addr, a->addrlen);
1127 log_addr(VERB_ALGO, "copy attempt count previous dp",
1128 &a->addr, a->addrlen);
1129 a->attempts = o->attempts;
1133 a = dp->usable_list;
1135 if(a->attempts >= OUTBOUND_MSG_RETRY) {
1136 log_addr(VERB_ALGO, "remove from usable list dp",
1137 &a->addr, a->addrlen);
1138 /* remove from result list */
1140 prev->next_usable = a->next_usable;
1141 else dp->usable_list = a->next_usable;
1142 /* prev stays the same */
1152 iter_ds_toolow(struct dns_msg* msg, struct delegpt* dp)
1154 /* if for query example.com, there is example.com SOA or a subdomain
1155 * of example.com, then we are too low and need to fetch NS. */
1157 /* if we have a DNAME or CNAME we are probably wrong */
1158 /* if we have a qtype DS in the answer section, its fine */
1159 for(i=0; i < msg->rep->an_numrrsets; i++) {
1160 struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
1161 if(ntohs(s->rk.type) == LDNS_RR_TYPE_DNAME ||
1162 ntohs(s->rk.type) == LDNS_RR_TYPE_CNAME) {
1163 /* not the right answer, maybe too low, check the
1164 * RRSIG signer name (if there is any) for a hint
1165 * that it is from the dp zone anyway */
1168 val_find_rrset_signer(s, &sname, &slen);
1169 if(sname && query_dname_compare(dp->name, sname)==0)
1170 return 0; /* it is fine, from the right dp */
1173 if(ntohs(s->rk.type) == LDNS_RR_TYPE_DS)
1174 return 0; /* fine, we have a DS record */
1176 for(i=msg->rep->an_numrrsets;
1177 i < msg->rep->an_numrrsets + msg->rep->ns_numrrsets; i++) {
1178 struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
1179 if(ntohs(s->rk.type) == LDNS_RR_TYPE_SOA) {
1180 if(dname_subdomain_c(s->rk.dname, msg->qinfo.qname))
1181 return 1; /* point is too low */
1182 if(query_dname_compare(s->rk.dname, dp->name)==0)
1183 return 0; /* right dp */
1185 if(ntohs(s->rk.type) == LDNS_RR_TYPE_NSEC ||
1186 ntohs(s->rk.type) == LDNS_RR_TYPE_NSEC3) {
1189 val_find_rrset_signer(s, &sname, &slen);
1190 if(sname && query_dname_compare(dp->name, sname)==0)
1191 return 0; /* it is fine, from the right dp */
1195 /* we do not know */
1199 int iter_dp_cangodown(struct query_info* qinfo, struct delegpt* dp)
1201 /* no delegation point, do not see how we can go down,
1202 * robust check, it should really exist */
1205 /* see if dp equals the qname, then we cannot go down further */
1206 if(query_dname_compare(qinfo->qname, dp->name) == 0)
1208 /* if dp is one label above the name we also cannot go down further */
1209 if(dname_count_labels(qinfo->qname) == dp->namelabs+1)