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) {
380 for(i = 0; i < got_num; i++) {
382 if(a->addr.ss_family == AF_INET6) {
385 if(low_rtt6 == 0 || a->sel_rtt < low_rtt6) {
386 low_rtt6 = a->sel_rtt;
389 /* swap to front if IPv6, or move to next result */
390 if(swap_to_front && prev) {
392 prev->next_result = n;
393 a->next_result = dp->result_list;
403 *selected_rtt = low_rtt6;
410 iter_server_selection(struct iter_env* iter_env,
411 struct module_env* env, struct delegpt* dp,
412 uint8_t* name, size_t namelen, uint16_t qtype, int* dnssec_lame,
413 int* chase_to_rd, int open_target, struct sock_list* blacklist,
418 struct delegpt_addr* a, *prev;
419 int num = iter_filter_order(iter_env, env, name, namelen, qtype,
420 *env->now, dp, &selrtt, open_target, blacklist, prefetch);
424 verbose(VERB_ALGO, "selrtt %d", selrtt);
425 if(selrtt > BLACKLIST_PENALTY) {
426 if(selrtt-BLACKLIST_PENALTY > USEFUL_SERVER_TOP_TIMEOUT*3) {
427 verbose(VERB_ALGO, "chase to "
428 "blacklisted recursion lame server");
431 if(selrtt-BLACKLIST_PENALTY > USEFUL_SERVER_TOP_TIMEOUT*2) {
432 verbose(VERB_ALGO, "chase to "
433 "blacklisted dnssec lame server");
437 if(selrtt > USEFUL_SERVER_TOP_TIMEOUT*3) {
438 verbose(VERB_ALGO, "chase to recursion lame server");
441 if(selrtt > USEFUL_SERVER_TOP_TIMEOUT*2) {
442 verbose(VERB_ALGO, "chase to dnssec lame server");
445 if(selrtt == USEFUL_SERVER_TOP_TIMEOUT) {
446 verbose(VERB_ALGO, "chase to blacklisted lame server");
453 if(++a->attempts < OUTBOUND_MSG_RETRY)
455 dp->result_list = a->next_result;
459 /* randomly select a target from the list */
461 /* grab secure random number, to pick unexpected server.
462 * also we need it to be threadsafe. */
463 sel = ub_random_max(env->rnd, num);
466 while(sel > 0 && a) {
471 if(!a) /* robustness */
473 if(++a->attempts < OUTBOUND_MSG_RETRY)
475 /* remove it from the delegation point result list */
477 prev->next_result = a->next_result;
478 else dp->result_list = a->next_result;
483 dns_alloc_msg(sldns_buffer* pkt, struct msg_parse* msg,
484 struct regional* region)
486 struct dns_msg* m = (struct dns_msg*)regional_alloc(region,
487 sizeof(struct dns_msg));
490 memset(m, 0, sizeof(*m));
491 if(!parse_create_msg(pkt, msg, NULL, &m->qinfo, &m->rep, region)) {
492 log_err("malloc failure: allocating incoming dns_msg");
499 dns_copy_msg(struct dns_msg* from, struct regional* region)
501 struct dns_msg* m = (struct dns_msg*)regional_alloc(region,
502 sizeof(struct dns_msg));
505 m->qinfo = from->qinfo;
506 if(!(m->qinfo.qname = regional_alloc_init(region, from->qinfo.qname,
507 from->qinfo.qname_len)))
509 if(!(m->rep = reply_info_copy(from->rep, NULL, region)))
515 iter_dns_store(struct module_env* env, struct query_info* msgqinf,
516 struct reply_info* msgrep, int is_referral, time_t leeway, int pside,
517 struct regional* region, uint16_t flags)
519 if(!dns_cache_store(env, msgqinf, msgrep, is_referral, leeway,
520 pside, region, flags))
521 log_err("out of memory: cannot store data in cache");
525 iter_ns_probability(struct ub_randstate* rnd, int n, int m)
528 if(n == m) /* 100% chance */
530 /* we do not need secure random numbers here, but
531 * we do need it to be threadsafe, so we use this */
532 sel = ub_random_max(rnd, m);
536 /** detect dependency cycle for query and target */
538 causes_cycle(struct module_qstate* qstate, uint8_t* name, size_t namelen,
539 uint16_t t, uint16_t c)
541 struct query_info qinf;
543 qinf.qname_len = namelen;
546 qinf.local_alias = NULL;
547 fptr_ok(fptr_whitelist_modenv_detect_cycle(
548 qstate->env->detect_cycle));
549 return (*qstate->env->detect_cycle)(qstate, &qinf,
550 (uint16_t)(BIT_RD|BIT_CD), qstate->is_priming,
555 iter_mark_cycle_targets(struct module_qstate* qstate, struct delegpt* dp)
557 struct delegpt_ns* ns;
558 for(ns = dp->nslist; ns; ns = ns->next) {
561 /* see if this ns as target causes dependency cycle */
562 if(causes_cycle(qstate, ns->name, ns->namelen,
563 LDNS_RR_TYPE_AAAA, qstate->qinfo.qclass) ||
564 causes_cycle(qstate, ns->name, ns->namelen,
565 LDNS_RR_TYPE_A, qstate->qinfo.qclass)) {
566 log_nametypeclass(VERB_QUERY, "skipping target due "
567 "to dependency cycle (harden-glue: no may "
568 "fix some of the cycles)",
569 ns->name, LDNS_RR_TYPE_A,
570 qstate->qinfo.qclass);
577 iter_mark_pside_cycle_targets(struct module_qstate* qstate, struct delegpt* dp)
579 struct delegpt_ns* ns;
580 for(ns = dp->nslist; ns; ns = ns->next) {
581 if(ns->done_pside4 && ns->done_pside6)
583 /* see if this ns as target causes dependency cycle */
584 if(causes_cycle(qstate, ns->name, ns->namelen,
585 LDNS_RR_TYPE_A, qstate->qinfo.qclass)) {
586 log_nametypeclass(VERB_QUERY, "skipping target due "
587 "to dependency cycle", ns->name,
588 LDNS_RR_TYPE_A, qstate->qinfo.qclass);
591 if(causes_cycle(qstate, ns->name, ns->namelen,
592 LDNS_RR_TYPE_AAAA, qstate->qinfo.qclass)) {
593 log_nametypeclass(VERB_QUERY, "skipping target due "
594 "to dependency cycle", ns->name,
595 LDNS_RR_TYPE_AAAA, qstate->qinfo.qclass);
602 iter_dp_is_useless(struct query_info* qinfo, uint16_t qflags,
605 struct delegpt_ns* ns;
608 * o no addresses are provided.
609 * o all NS items are required glue.
612 * o no addresses are provided.
613 * o the query is for one of the nameservers in dp,
614 * and that nameserver is a glue-name for this dp.
618 /* either available or unused targets */
619 if(dp->usable_list || dp->result_list)
622 /* see if query is for one of the nameservers, which is glue */
623 if( (qinfo->qtype == LDNS_RR_TYPE_A ||
624 qinfo->qtype == LDNS_RR_TYPE_AAAA) &&
625 dname_subdomain_c(qinfo->qname, dp->name) &&
626 delegpt_find_ns(dp, qinfo->qname, qinfo->qname_len))
629 for(ns = dp->nslist; ns; ns = ns->next) {
630 if(ns->resolved) /* skip failed targets */
632 if(!dname_subdomain_c(ns->name, dp->name))
633 return 0; /* one address is not required glue */
639 iter_qname_indicates_dnssec(struct module_env* env, struct query_info *qinfo)
641 struct trust_anchor* a;
642 if(!env || !env->anchors || !qinfo || !qinfo->qname)
644 /* a trust anchor exists above the name? */
645 if((a=anchors_lookup(env->anchors, qinfo->qname, qinfo->qname_len,
647 if(a->numDS == 0 && a->numDNSKEY == 0) {
648 /* insecure trust point */
649 lock_basic_unlock(&a->lock);
652 lock_basic_unlock(&a->lock);
655 /* no trust anchor above it. */
660 iter_indicates_dnssec(struct module_env* env, struct delegpt* dp,
661 struct dns_msg* msg, uint16_t dclass)
663 struct trust_anchor* a;
664 /* information not available, !env->anchors can be common */
665 if(!env || !env->anchors || !dp || !dp->name)
667 /* a trust anchor exists with this name, RRSIGs expected */
668 if((a=anchor_find(env->anchors, dp->name, dp->namelabs, dp->namelen,
670 if(a->numDS == 0 && a->numDNSKEY == 0) {
671 /* insecure trust point */
672 lock_basic_unlock(&a->lock);
675 lock_basic_unlock(&a->lock);
678 /* see if DS rrset was given, in AUTH section */
679 if(msg && msg->rep &&
680 reply_find_rrset_section_ns(msg->rep, dp->name, dp->namelen,
681 LDNS_RR_TYPE_DS, dclass))
683 /* look in key cache */
685 struct key_entry_key* kk = key_cache_obtain(env->key_cache,
686 dp->name, dp->namelen, dclass, env->scratch, *env->now);
688 if(query_dname_compare(kk->name, dp->name) == 0) {
689 if(key_entry_isgood(kk) || key_entry_isbad(kk)) {
690 regional_free_all(env->scratch);
692 } else if(key_entry_isnull(kk)) {
693 regional_free_all(env->scratch);
697 regional_free_all(env->scratch);
704 iter_msg_has_dnssec(struct dns_msg* msg)
707 if(!msg || !msg->rep)
709 for(i=0; i<msg->rep->an_numrrsets + msg->rep->ns_numrrsets; i++) {
710 if(((struct packed_rrset_data*)msg->rep->rrsets[i]->
711 entry.data)->rrsig_count > 0)
714 /* empty message has no DNSSEC info, with DNSSEC the reply is
715 * not empty (NSEC) */
719 int iter_msg_from_zone(struct dns_msg* msg, struct delegpt* dp,
720 enum response_type type, uint16_t dclass)
722 if(!msg || !dp || !msg->rep || !dp->name)
724 /* SOA RRset - always from reply zone */
725 if(reply_find_rrset_section_an(msg->rep, dp->name, dp->namelen,
726 LDNS_RR_TYPE_SOA, dclass) ||
727 reply_find_rrset_section_ns(msg->rep, dp->name, dp->namelen,
728 LDNS_RR_TYPE_SOA, dclass))
730 if(type == RESPONSE_TYPE_REFERRAL) {
732 /* if it adds a single label, i.e. we expect .com,
733 * and referral to example.com. NS ... , then origin zone
734 * is .com. For a referral to sub.example.com. NS ... then
735 * we do not know, since example.com. may be in between. */
736 for(i=0; i<msg->rep->an_numrrsets+msg->rep->ns_numrrsets;
738 struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
739 if(ntohs(s->rk.type) == LDNS_RR_TYPE_NS &&
740 ntohs(s->rk.rrset_class) == dclass) {
741 int l = dname_count_labels(s->rk.dname);
742 if(l == dp->namelabs + 1 &&
743 dname_strict_subdomain(s->rk.dname,
744 l, dp->name, dp->namelabs))
750 log_assert(type==RESPONSE_TYPE_ANSWER || type==RESPONSE_TYPE_CNAME);
751 /* not a referral, and not lame delegation (upwards), so,
752 * any NS rrset must be from the zone itself */
753 if(reply_find_rrset_section_an(msg->rep, dp->name, dp->namelen,
754 LDNS_RR_TYPE_NS, dclass) ||
755 reply_find_rrset_section_ns(msg->rep, dp->name, dp->namelen,
756 LDNS_RR_TYPE_NS, dclass))
758 /* a DNSKEY set is expected at the zone apex as well */
759 /* this is for 'minimal responses' for DNSKEYs */
760 if(reply_find_rrset_section_an(msg->rep, dp->name, dp->namelen,
761 LDNS_RR_TYPE_DNSKEY, dclass))
767 * check equality of two rrsets
770 * @return true if equal
773 rrset_equal(struct ub_packed_rrset_key* k1, struct ub_packed_rrset_key* k2)
775 struct packed_rrset_data* d1 = (struct packed_rrset_data*)
777 struct packed_rrset_data* d2 = (struct packed_rrset_data*)
780 if(k1->rk.dname_len != k2->rk.dname_len ||
781 k1->rk.flags != k2->rk.flags ||
782 k1->rk.type != k2->rk.type ||
783 k1->rk.rrset_class != k2->rk.rrset_class ||
784 query_dname_compare(k1->rk.dname, k2->rk.dname) != 0)
786 if( /* do not check ttl: d1->ttl != d2->ttl || */
787 d1->count != d2->count ||
788 d1->rrsig_count != d2->rrsig_count ||
789 d1->trust != d2->trust ||
790 d1->security != d2->security)
792 t = d1->count + d1->rrsig_count;
794 if(d1->rr_len[i] != d2->rr_len[i] ||
795 /* no ttl check: d1->rr_ttl[i] != d2->rr_ttl[i] ||*/
796 memcmp(d1->rr_data[i], d2->rr_data[i],
804 reply_equal(struct reply_info* p, struct reply_info* q, struct regional* region)
807 if(p->flags != q->flags ||
808 p->qdcount != q->qdcount ||
809 /* do not check TTL, this may differ */
812 p->prefetch_ttl != q->prefetch_ttl ||
814 p->security != q->security ||
815 p->an_numrrsets != q->an_numrrsets ||
816 p->ns_numrrsets != q->ns_numrrsets ||
817 p->ar_numrrsets != q->ar_numrrsets ||
818 p->rrset_count != q->rrset_count)
820 for(i=0; i<p->rrset_count; i++) {
821 if(!rrset_equal(p->rrsets[i], q->rrsets[i])) {
822 if(!rrset_canonical_equal(region, p->rrsets[i],
824 regional_free_all(region);
827 regional_free_all(region);
834 caps_strip_reply(struct reply_info* rep)
838 /* see if message is a referral, in which case the additional and
839 * NS record cannot be removed */
840 /* referrals have the AA flag unset (strict check, not elsewhere in
841 * unbound, but for 0x20 this is very convenient). */
842 if(!(rep->flags&BIT_AA))
844 /* remove the additional section from the reply */
845 if(rep->ar_numrrsets != 0) {
846 verbose(VERB_ALGO, "caps fallback: removing additional section");
847 rep->rrset_count -= rep->ar_numrrsets;
848 rep->ar_numrrsets = 0;
850 /* is there an NS set in the authority section to remove? */
851 /* the failure case (Cisco firewalls) only has one rrset in authsec */
852 for(i=rep->an_numrrsets; i<rep->an_numrrsets+rep->ns_numrrsets; i++) {
853 struct ub_packed_rrset_key* s = rep->rrsets[i];
854 if(ntohs(s->rk.type) == LDNS_RR_TYPE_NS) {
855 /* remove NS rrset and break from loop (loop limits
857 /* move last rrset into this position (there is no
858 * additional section any more) */
859 verbose(VERB_ALGO, "caps fallback: removing NS rrset");
860 if(i < rep->rrset_count-1)
861 rep->rrsets[i]=rep->rrsets[rep->rrset_count-1];
863 rep->ns_numrrsets --;
869 int caps_failed_rcode(struct reply_info* rep)
871 return !(FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NOERROR ||
872 FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NXDOMAIN);
876 iter_store_parentside_rrset(struct module_env* env,
877 struct ub_packed_rrset_key* rrset)
879 struct rrset_ref ref;
880 rrset = packed_rrset_copy_alloc(rrset, env->alloc, *env->now);
882 log_err("malloc failure in store_parentside_rrset");
885 rrset->rk.flags |= PACKED_RRSET_PARENT_SIDE;
886 rrset->entry.hash = rrset_key_hash(&rrset->rk);
889 /* ignore ret: if it was in the cache, ref updated */
890 (void)rrset_cache_update(env->rrset_cache, &ref, env->alloc, *env->now);
893 /** fetch NS record from reply, if any */
894 static struct ub_packed_rrset_key*
895 reply_get_NS_rrset(struct reply_info* rep)
898 for(i=0; i<rep->rrset_count; i++) {
899 if(rep->rrsets[i]->rk.type == htons(LDNS_RR_TYPE_NS)) {
900 return rep->rrsets[i];
907 iter_store_parentside_NS(struct module_env* env, struct reply_info* rep)
909 struct ub_packed_rrset_key* rrset = reply_get_NS_rrset(rep);
911 log_rrset_key(VERB_ALGO, "store parent-side NS", rrset);
912 iter_store_parentside_rrset(env, rrset);
916 void iter_store_parentside_neg(struct module_env* env,
917 struct query_info* qinfo, struct reply_info* rep)
919 /* TTL: NS from referral in iq->deleg_msg,
920 * or first RR from iq->response,
921 * or servfail5secs if !iq->response */
922 time_t ttl = NORR_TTL;
923 struct ub_packed_rrset_key* neg;
924 struct packed_rrset_data* newd;
926 struct ub_packed_rrset_key* rrset = reply_get_NS_rrset(rep);
927 if(!rrset && rep->rrset_count != 0) rrset = rep->rrsets[0];
928 if(rrset) ttl = ub_packed_rrset_ttl(rrset);
930 /* create empty rrset to store */
931 neg = (struct ub_packed_rrset_key*)regional_alloc(env->scratch,
932 sizeof(struct ub_packed_rrset_key));
934 log_err("out of memory in store_parentside_neg");
937 memset(&neg->entry, 0, sizeof(neg->entry));
938 neg->entry.key = neg;
939 neg->rk.type = htons(qinfo->qtype);
940 neg->rk.rrset_class = htons(qinfo->qclass);
942 neg->rk.dname = regional_alloc_init(env->scratch, qinfo->qname,
945 log_err("out of memory in store_parentside_neg");
948 neg->rk.dname_len = qinfo->qname_len;
949 neg->entry.hash = rrset_key_hash(&neg->rk);
950 newd = (struct packed_rrset_data*)regional_alloc_zero(env->scratch,
951 sizeof(struct packed_rrset_data) + sizeof(size_t) +
952 sizeof(uint8_t*) + sizeof(time_t) + sizeof(uint16_t));
954 log_err("out of memory in store_parentside_neg");
957 neg->entry.data = newd;
959 /* entry must have one RR, otherwise not valid in cache.
960 * put in one RR with empty rdata: those are ignored as nameserver */
962 newd->rrsig_count = 0;
963 newd->trust = rrset_trust_ans_noAA;
964 newd->rr_len = (size_t*)((uint8_t*)newd +
965 sizeof(struct packed_rrset_data));
966 newd->rr_len[0] = 0 /* zero len rdata */ + sizeof(uint16_t);
967 packed_rrset_ptr_fixup(newd);
968 newd->rr_ttl[0] = newd->ttl;
969 sldns_write_uint16(newd->rr_data[0], 0 /* zero len rdata */);
971 log_rrset_key(VERB_ALGO, "store parent-side negative", neg);
972 iter_store_parentside_rrset(env, neg);
976 iter_lookup_parent_NS_from_cache(struct module_env* env, struct delegpt* dp,
977 struct regional* region, struct query_info* qinfo)
979 struct ub_packed_rrset_key* akey;
980 akey = rrset_cache_lookup(env->rrset_cache, dp->name,
981 dp->namelen, LDNS_RR_TYPE_NS, qinfo->qclass,
982 PACKED_RRSET_PARENT_SIDE, *env->now, 0);
984 log_rrset_key(VERB_ALGO, "found parent-side NS in cache", akey);
985 dp->has_parent_side_NS = 1;
986 /* and mark the new names as lame */
987 if(!delegpt_rrset_add_ns(dp, region, akey, 1)) {
988 lock_rw_unlock(&akey->entry.lock);
991 lock_rw_unlock(&akey->entry.lock);
996 int iter_lookup_parent_glue_from_cache(struct module_env* env,
997 struct delegpt* dp, struct regional* region, struct query_info* qinfo)
999 struct ub_packed_rrset_key* akey;
1000 struct delegpt_ns* ns;
1001 size_t num = delegpt_count_targets(dp);
1002 for(ns = dp->nslist; ns; ns = ns->next) {
1003 /* get cached parentside A */
1004 akey = rrset_cache_lookup(env->rrset_cache, ns->name,
1005 ns->namelen, LDNS_RR_TYPE_A, qinfo->qclass,
1006 PACKED_RRSET_PARENT_SIDE, *env->now, 0);
1008 log_rrset_key(VERB_ALGO, "found parent-side", akey);
1009 ns->done_pside4 = 1;
1010 /* a negative-cache-element has no addresses it adds */
1011 if(!delegpt_add_rrset_A(dp, region, akey, 1))
1012 log_err("malloc failure in lookup_parent_glue");
1013 lock_rw_unlock(&akey->entry.lock);
1015 /* get cached parentside AAAA */
1016 akey = rrset_cache_lookup(env->rrset_cache, ns->name,
1017 ns->namelen, LDNS_RR_TYPE_AAAA, qinfo->qclass,
1018 PACKED_RRSET_PARENT_SIDE, *env->now, 0);
1020 log_rrset_key(VERB_ALGO, "found parent-side", akey);
1021 ns->done_pside6 = 1;
1022 /* a negative-cache-element has no addresses it adds */
1023 if(!delegpt_add_rrset_AAAA(dp, region, akey, 1))
1024 log_err("malloc failure in lookup_parent_glue");
1025 lock_rw_unlock(&akey->entry.lock);
1028 /* see if new (but lame) addresses have become available */
1029 return delegpt_count_targets(dp) != num;
1033 iter_get_next_root(struct iter_hints* hints, struct iter_forwards* fwd,
1036 uint16_t c1 = *c, c2 = *c;
1037 int r1 = hints_next_root(hints, &c1);
1038 int r2 = forwards_next_root(fwd, &c2);
1039 if(!r1 && !r2) /* got none, end of list */
1041 else if(!r1) /* got one, return that */
1045 else if(c1 < c2) /* got both take smallest */
1052 iter_scrub_ds(struct dns_msg* msg, struct ub_packed_rrset_key* ns, uint8_t* z)
1054 /* Only the DS record for the delegation itself is expected.
1055 * We allow DS for everything between the bailiwick and the
1056 * zonecut, thus DS records must be at or above the zonecut.
1057 * And the DS records must be below the server authority zone.
1058 * The answer section is already scrubbed. */
1059 size_t i = msg->rep->an_numrrsets;
1060 while(i < (msg->rep->an_numrrsets + msg->rep->ns_numrrsets)) {
1061 struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
1062 if(ntohs(s->rk.type) == LDNS_RR_TYPE_DS &&
1063 (!ns || !dname_subdomain_c(ns->rk.dname, s->rk.dname)
1064 || query_dname_compare(z, s->rk.dname) == 0)) {
1065 log_nametypeclass(VERB_ALGO, "removing irrelevant DS",
1066 s->rk.dname, ntohs(s->rk.type),
1067 ntohs(s->rk.rrset_class));
1068 memmove(msg->rep->rrsets+i, msg->rep->rrsets+i+1,
1069 sizeof(struct ub_packed_rrset_key*) *
1070 (msg->rep->rrset_count-i-1));
1071 msg->rep->ns_numrrsets--;
1072 msg->rep->rrset_count--;
1073 /* stay at same i, but new record */
1080 void iter_dec_attempts(struct delegpt* dp, int d)
1082 struct delegpt_addr* a;
1083 for(a=dp->target_list; a; a = a->next_target) {
1084 if(a->attempts >= OUTBOUND_MSG_RETRY) {
1085 /* add back to result list */
1086 a->next_result = dp->result_list;
1087 dp->result_list = a;
1091 else a->attempts = 0;
1095 void iter_merge_retry_counts(struct delegpt* dp, struct delegpt* old)
1097 struct delegpt_addr* a, *o, *prev;
1098 for(a=dp->target_list; a; a = a->next_target) {
1099 o = delegpt_find_addr(old, &a->addr, a->addrlen);
1101 log_addr(VERB_ALGO, "copy attempt count previous dp",
1102 &a->addr, a->addrlen);
1103 a->attempts = o->attempts;
1107 a = dp->usable_list;
1109 if(a->attempts >= OUTBOUND_MSG_RETRY) {
1110 log_addr(VERB_ALGO, "remove from usable list dp",
1111 &a->addr, a->addrlen);
1112 /* remove from result list */
1114 prev->next_usable = a->next_usable;
1115 else dp->usable_list = a->next_usable;
1116 /* prev stays the same */
1126 iter_ds_toolow(struct dns_msg* msg, struct delegpt* dp)
1128 /* if for query example.com, there is example.com SOA or a subdomain
1129 * of example.com, then we are too low and need to fetch NS. */
1131 /* if we have a DNAME or CNAME we are probably wrong */
1132 /* if we have a qtype DS in the answer section, its fine */
1133 for(i=0; i < msg->rep->an_numrrsets; i++) {
1134 struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
1135 if(ntohs(s->rk.type) == LDNS_RR_TYPE_DNAME ||
1136 ntohs(s->rk.type) == LDNS_RR_TYPE_CNAME) {
1137 /* not the right answer, maybe too low, check the
1138 * RRSIG signer name (if there is any) for a hint
1139 * that it is from the dp zone anyway */
1142 val_find_rrset_signer(s, &sname, &slen);
1143 if(sname && query_dname_compare(dp->name, sname)==0)
1144 return 0; /* it is fine, from the right dp */
1147 if(ntohs(s->rk.type) == LDNS_RR_TYPE_DS)
1148 return 0; /* fine, we have a DS record */
1150 for(i=msg->rep->an_numrrsets;
1151 i < msg->rep->an_numrrsets + msg->rep->ns_numrrsets; i++) {
1152 struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
1153 if(ntohs(s->rk.type) == LDNS_RR_TYPE_SOA) {
1154 if(dname_subdomain_c(s->rk.dname, msg->qinfo.qname))
1155 return 1; /* point is too low */
1156 if(query_dname_compare(s->rk.dname, dp->name)==0)
1157 return 0; /* right dp */
1159 if(ntohs(s->rk.type) == LDNS_RR_TYPE_NSEC ||
1160 ntohs(s->rk.type) == LDNS_RR_TYPE_NSEC3) {
1163 val_find_rrset_signer(s, &sname, &slen);
1164 if(sname && query_dname_compare(dp->name, sname)==0)
1165 return 0; /* it is fine, from the right dp */
1169 /* we do not know */
1173 int iter_dp_cangodown(struct query_info* qinfo, struct delegpt* dp)
1175 /* no delegation point, do not see how we can go down,
1176 * robust check, it should really exist */
1179 /* see if dp equals the qname, then we cannot go down further */
1180 if(query_dname_compare(qinfo->qname, dp->name) == 0)
1182 /* if dp is one label above the name we also cannot go down further */
1183 if(dname_count_labels(qinfo->qname) == dp->namelabs+1)