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,
286 size_t* num_suitable_results)
289 struct delegpt_addr* a;
290 *num_suitable_results = 0;
293 return 0; /* NS bogus, all bogus, nothing found */
294 for(a=dp->result_list; a; a = a->next_result) {
295 a->sel_rtt = iter_filter_unsuitable(iter_env, env,
296 name, namelen, qtype, now, a);
297 if(a->sel_rtt != -1) {
298 if(sock_list_find(blacklist, &a->addr, a->addrlen))
299 a->sel_rtt += BLACKLIST_PENALTY;
302 *best_rtt = a->sel_rtt;
304 } else if(a->sel_rtt < *best_rtt) {
305 *best_rtt = a->sel_rtt;
307 (*num_suitable_results)++;
313 /** compare two rtts, return -1, 0 or 1 */
315 rtt_compare(const void* x, const void* y)
317 if(*(int*)x == *(int*)y)
319 if(*(int*)x > *(int*)y)
324 /** get RTT for the Nth fastest server */
326 nth_rtt(struct delegpt_addr* result_list, size_t num_results, size_t n)
330 int* rtt_list, *rtt_index;
332 if(num_results < 1 || n >= num_results) {
336 rtt_list = calloc(num_results, sizeof(int));
338 log_err("malloc failure: allocating rtt_list");
341 rtt_index = rtt_list;
343 for(i=0; i<num_results && result_list; i++) {
344 if(result_list->sel_rtt != -1) {
345 *rtt_index = result_list->sel_rtt;
348 result_list=result_list->next_result;
350 qsort(rtt_list, num_results, sizeof(*rtt_list), rtt_compare);
353 rtt_band = rtt_list[n-1];
359 /** filter the address list, putting best targets at front,
360 * returns number of best targets (or 0, no suitable targets) */
362 iter_filter_order(struct iter_env* iter_env, struct module_env* env,
363 uint8_t* name, size_t namelen, uint16_t qtype, time_t now,
364 struct delegpt* dp, int* selected_rtt, int open_target,
365 struct sock_list* blacklist, time_t prefetch)
367 int got_num = 0, low_rtt = 0, swap_to_front, rtt_band = RTT_BAND, nth;
369 struct delegpt_addr* a, *n, *prev=NULL;
371 /* fillup sel_rtt and find best rtt in the bunch */
372 got_num = iter_fill_rtt(iter_env, env, name, namelen, qtype, now, dp,
373 &low_rtt, blacklist, &num_results);
376 if(low_rtt >= USEFUL_SERVER_TOP_TIMEOUT &&
377 (delegpt_count_missing_targets(dp) > 0 || open_target > 0)) {
378 verbose(VERB_ALGO, "Bad choices, trying to get more choice");
379 return 0; /* we want more choice. The best choice is a bad one.
380 return 0 to force the caller to fetch more */
383 if(env->cfg->fast_server_permil != 0 && prefetch == 0 &&
384 num_results > env->cfg->fast_server_num &&
385 ub_random_max(env->rnd, 1000) < env->cfg->fast_server_permil) {
386 /* the query is not prefetch, but for a downstream client,
387 * there are more servers available then the fastest N we want
388 * to choose from. Limit our choice to the fastest servers. */
389 nth = nth_rtt(dp->result_list, num_results,
390 env->cfg->fast_server_num);
392 rtt_band = nth - low_rtt;
393 if(rtt_band > RTT_BAND)
401 /* skip unsuitable targets */
402 if(a->sel_rtt == -1) {
407 /* classify the server address and determine what to do */
409 if(a->sel_rtt >= low_rtt && a->sel_rtt - low_rtt <= rtt_band) {
412 } else if(a->sel_rtt<low_rtt && low_rtt-a->sel_rtt<=rtt_band) {
416 /* swap to front if necessary, or move to next result */
417 if(swap_to_front && prev) {
419 prev->next_result = n;
420 a->next_result = dp->result_list;
428 *selected_rtt = low_rtt;
430 if (env->cfg->prefer_ip6) {
434 int attempt = -1; /* filter to make sure addresses have
435 less attempts on them than the first, to force round
436 robin when all the IPv6 addresses fail */
437 int num4ok = 0; /* number ip4 at low attempt count */
441 for(i = 0; i < got_num; i++) {
443 if(a->addr.ss_family != AF_INET6 && attempt == -1) {
444 /* if we only have ip4 at low attempt count,
445 * then ip6 is failing, and we need to
446 * select one of the remaining IPv4 addrs */
447 attempt = a->attempts;
449 num4_lowrtt = a->sel_rtt;
450 } else if(a->addr.ss_family != AF_INET6 && attempt == a->attempts) {
452 if(num4_lowrtt == 0 || a->sel_rtt < num4_lowrtt) {
453 num4_lowrtt = a->sel_rtt;
456 if(a->addr.ss_family == AF_INET6) {
458 attempt = a->attempts;
459 } else if(a->attempts > attempt) {
464 if(low_rtt6 == 0 || a->sel_rtt < low_rtt6) {
465 low_rtt6 = a->sel_rtt;
468 /* swap to front if IPv6, or move to next result */
469 if(swap_to_front && prev) {
471 prev->next_result = n;
472 a->next_result = dp->result_list;
482 *selected_rtt = low_rtt6;
483 } else if(num4ok > 0) {
485 *selected_rtt = num4_lowrtt;
487 } else if (env->cfg->prefer_ip4) {
491 int attempt = -1; /* filter to make sure addresses have
492 less attempts on them than the first, to force round
493 robin when all the IPv4 addresses fail */
494 int num6ok = 0; /* number ip6 at low attempt count */
498 for(i = 0; i < got_num; i++) {
500 if(a->addr.ss_family != AF_INET && attempt == -1) {
501 /* if we only have ip6 at low attempt count,
502 * then ip4 is failing, and we need to
503 * select one of the remaining IPv6 addrs */
504 attempt = a->attempts;
506 num6_lowrtt = a->sel_rtt;
507 } else if(a->addr.ss_family != AF_INET && attempt == a->attempts) {
509 if(num6_lowrtt == 0 || a->sel_rtt < num6_lowrtt) {
510 num6_lowrtt = a->sel_rtt;
513 if(a->addr.ss_family == AF_INET) {
515 attempt = a->attempts;
516 } else if(a->attempts > attempt) {
521 if(low_rtt4 == 0 || a->sel_rtt < low_rtt4) {
522 low_rtt4 = a->sel_rtt;
525 /* swap to front if IPv4, or move to next result */
526 if(swap_to_front && prev) {
528 prev->next_result = n;
529 a->next_result = dp->result_list;
539 *selected_rtt = low_rtt4;
540 } else if(num6ok > 0) {
542 *selected_rtt = num6_lowrtt;
549 iter_server_selection(struct iter_env* iter_env,
550 struct module_env* env, struct delegpt* dp,
551 uint8_t* name, size_t namelen, uint16_t qtype, int* dnssec_lame,
552 int* chase_to_rd, int open_target, struct sock_list* blacklist,
557 struct delegpt_addr* a, *prev;
558 int num = iter_filter_order(iter_env, env, name, namelen, qtype,
559 *env->now, dp, &selrtt, open_target, blacklist, prefetch);
563 verbose(VERB_ALGO, "selrtt %d", selrtt);
564 if(selrtt > BLACKLIST_PENALTY) {
565 if(selrtt-BLACKLIST_PENALTY > USEFUL_SERVER_TOP_TIMEOUT*3) {
566 verbose(VERB_ALGO, "chase to "
567 "blacklisted recursion lame server");
570 if(selrtt-BLACKLIST_PENALTY > USEFUL_SERVER_TOP_TIMEOUT*2) {
571 verbose(VERB_ALGO, "chase to "
572 "blacklisted dnssec lame server");
576 if(selrtt > USEFUL_SERVER_TOP_TIMEOUT*3) {
577 verbose(VERB_ALGO, "chase to recursion lame server");
580 if(selrtt > USEFUL_SERVER_TOP_TIMEOUT*2) {
581 verbose(VERB_ALGO, "chase to dnssec lame server");
584 if(selrtt == USEFUL_SERVER_TOP_TIMEOUT) {
585 verbose(VERB_ALGO, "chase to blacklisted lame server");
592 if(++a->attempts < OUTBOUND_MSG_RETRY)
594 dp->result_list = a->next_result;
598 /* randomly select a target from the list */
600 /* grab secure random number, to pick unexpected server.
601 * also we need it to be threadsafe. */
602 sel = ub_random_max(env->rnd, num);
605 while(sel > 0 && a) {
610 if(!a) /* robustness */
612 if(++a->attempts < OUTBOUND_MSG_RETRY)
614 /* remove it from the delegation point result list */
616 prev->next_result = a->next_result;
617 else dp->result_list = a->next_result;
622 dns_alloc_msg(sldns_buffer* pkt, struct msg_parse* msg,
623 struct regional* region)
625 struct dns_msg* m = (struct dns_msg*)regional_alloc(region,
626 sizeof(struct dns_msg));
629 memset(m, 0, sizeof(*m));
630 if(!parse_create_msg(pkt, msg, NULL, &m->qinfo, &m->rep, region)) {
631 log_err("malloc failure: allocating incoming dns_msg");
638 dns_copy_msg(struct dns_msg* from, struct regional* region)
640 struct dns_msg* m = (struct dns_msg*)regional_alloc(region,
641 sizeof(struct dns_msg));
644 m->qinfo = from->qinfo;
645 if(!(m->qinfo.qname = regional_alloc_init(region, from->qinfo.qname,
646 from->qinfo.qname_len)))
648 if(!(m->rep = reply_info_copy(from->rep, NULL, region)))
654 iter_dns_store(struct module_env* env, struct query_info* msgqinf,
655 struct reply_info* msgrep, int is_referral, time_t leeway, int pside,
656 struct regional* region, uint16_t flags)
658 if(!dns_cache_store(env, msgqinf, msgrep, is_referral, leeway,
659 pside, region, flags))
660 log_err("out of memory: cannot store data in cache");
664 iter_ns_probability(struct ub_randstate* rnd, int n, int m)
667 if(n == m) /* 100% chance */
669 /* we do not need secure random numbers here, but
670 * we do need it to be threadsafe, so we use this */
671 sel = ub_random_max(rnd, m);
675 /** detect dependency cycle for query and target */
677 causes_cycle(struct module_qstate* qstate, uint8_t* name, size_t namelen,
678 uint16_t t, uint16_t c)
680 struct query_info qinf;
682 qinf.qname_len = namelen;
685 qinf.local_alias = NULL;
686 fptr_ok(fptr_whitelist_modenv_detect_cycle(
687 qstate->env->detect_cycle));
688 return (*qstate->env->detect_cycle)(qstate, &qinf,
689 (uint16_t)(BIT_RD|BIT_CD), qstate->is_priming,
694 iter_mark_cycle_targets(struct module_qstate* qstate, struct delegpt* dp)
696 struct delegpt_ns* ns;
697 for(ns = dp->nslist; ns; ns = ns->next) {
700 /* see if this ns as target causes dependency cycle */
701 if(causes_cycle(qstate, ns->name, ns->namelen,
702 LDNS_RR_TYPE_AAAA, qstate->qinfo.qclass) ||
703 causes_cycle(qstate, ns->name, ns->namelen,
704 LDNS_RR_TYPE_A, qstate->qinfo.qclass)) {
705 log_nametypeclass(VERB_QUERY, "skipping target due "
706 "to dependency cycle (harden-glue: no may "
707 "fix some of the cycles)",
708 ns->name, LDNS_RR_TYPE_A,
709 qstate->qinfo.qclass);
716 iter_mark_pside_cycle_targets(struct module_qstate* qstate, struct delegpt* dp)
718 struct delegpt_ns* ns;
719 for(ns = dp->nslist; ns; ns = ns->next) {
720 if(ns->done_pside4 && ns->done_pside6)
722 /* see if this ns as target causes dependency cycle */
723 if(causes_cycle(qstate, ns->name, ns->namelen,
724 LDNS_RR_TYPE_A, qstate->qinfo.qclass)) {
725 log_nametypeclass(VERB_QUERY, "skipping target due "
726 "to dependency cycle", ns->name,
727 LDNS_RR_TYPE_A, qstate->qinfo.qclass);
730 if(causes_cycle(qstate, ns->name, ns->namelen,
731 LDNS_RR_TYPE_AAAA, qstate->qinfo.qclass)) {
732 log_nametypeclass(VERB_QUERY, "skipping target due "
733 "to dependency cycle", ns->name,
734 LDNS_RR_TYPE_AAAA, qstate->qinfo.qclass);
741 iter_dp_is_useless(struct query_info* qinfo, uint16_t qflags,
744 struct delegpt_ns* ns;
747 * o no addresses are provided.
748 * o all NS items are required glue.
751 * o no addresses are provided.
752 * o the query is for one of the nameservers in dp,
753 * and that nameserver is a glue-name for this dp.
757 /* either available or unused targets */
758 if(dp->usable_list || dp->result_list)
761 /* see if query is for one of the nameservers, which is glue */
762 if( (qinfo->qtype == LDNS_RR_TYPE_A ||
763 qinfo->qtype == LDNS_RR_TYPE_AAAA) &&
764 dname_subdomain_c(qinfo->qname, dp->name) &&
765 delegpt_find_ns(dp, qinfo->qname, qinfo->qname_len))
768 for(ns = dp->nslist; ns; ns = ns->next) {
769 if(ns->resolved) /* skip failed targets */
771 if(!dname_subdomain_c(ns->name, dp->name))
772 return 0; /* one address is not required glue */
778 iter_qname_indicates_dnssec(struct module_env* env, struct query_info *qinfo)
780 struct trust_anchor* a;
781 if(!env || !env->anchors || !qinfo || !qinfo->qname)
783 /* a trust anchor exists above the name? */
784 if((a=anchors_lookup(env->anchors, qinfo->qname, qinfo->qname_len,
786 if(a->numDS == 0 && a->numDNSKEY == 0) {
787 /* insecure trust point */
788 lock_basic_unlock(&a->lock);
791 lock_basic_unlock(&a->lock);
794 /* no trust anchor above it. */
799 iter_indicates_dnssec(struct module_env* env, struct delegpt* dp,
800 struct dns_msg* msg, uint16_t dclass)
802 struct trust_anchor* a;
803 /* information not available, !env->anchors can be common */
804 if(!env || !env->anchors || !dp || !dp->name)
806 /* a trust anchor exists with this name, RRSIGs expected */
807 if((a=anchor_find(env->anchors, dp->name, dp->namelabs, dp->namelen,
809 if(a->numDS == 0 && a->numDNSKEY == 0) {
810 /* insecure trust point */
811 lock_basic_unlock(&a->lock);
814 lock_basic_unlock(&a->lock);
817 /* see if DS rrset was given, in AUTH section */
818 if(msg && msg->rep &&
819 reply_find_rrset_section_ns(msg->rep, dp->name, dp->namelen,
820 LDNS_RR_TYPE_DS, dclass))
822 /* look in key cache */
824 struct key_entry_key* kk = key_cache_obtain(env->key_cache,
825 dp->name, dp->namelen, dclass, env->scratch, *env->now);
827 if(query_dname_compare(kk->name, dp->name) == 0) {
828 if(key_entry_isgood(kk) || key_entry_isbad(kk)) {
829 regional_free_all(env->scratch);
831 } else if(key_entry_isnull(kk)) {
832 regional_free_all(env->scratch);
836 regional_free_all(env->scratch);
843 iter_msg_has_dnssec(struct dns_msg* msg)
846 if(!msg || !msg->rep)
848 for(i=0; i<msg->rep->an_numrrsets + msg->rep->ns_numrrsets; i++) {
849 if(((struct packed_rrset_data*)msg->rep->rrsets[i]->
850 entry.data)->rrsig_count > 0)
853 /* empty message has no DNSSEC info, with DNSSEC the reply is
854 * not empty (NSEC) */
858 int iter_msg_from_zone(struct dns_msg* msg, struct delegpt* dp,
859 enum response_type type, uint16_t dclass)
861 if(!msg || !dp || !msg->rep || !dp->name)
863 /* SOA RRset - always from reply zone */
864 if(reply_find_rrset_section_an(msg->rep, dp->name, dp->namelen,
865 LDNS_RR_TYPE_SOA, dclass) ||
866 reply_find_rrset_section_ns(msg->rep, dp->name, dp->namelen,
867 LDNS_RR_TYPE_SOA, dclass))
869 if(type == RESPONSE_TYPE_REFERRAL) {
871 /* if it adds a single label, i.e. we expect .com,
872 * and referral to example.com. NS ... , then origin zone
873 * is .com. For a referral to sub.example.com. NS ... then
874 * we do not know, since example.com. may be in between. */
875 for(i=0; i<msg->rep->an_numrrsets+msg->rep->ns_numrrsets;
877 struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
878 if(ntohs(s->rk.type) == LDNS_RR_TYPE_NS &&
879 ntohs(s->rk.rrset_class) == dclass) {
880 int l = dname_count_labels(s->rk.dname);
881 if(l == dp->namelabs + 1 &&
882 dname_strict_subdomain(s->rk.dname,
883 l, dp->name, dp->namelabs))
889 log_assert(type==RESPONSE_TYPE_ANSWER || type==RESPONSE_TYPE_CNAME);
890 /* not a referral, and not lame delegation (upwards), so,
891 * any NS rrset must be from the zone itself */
892 if(reply_find_rrset_section_an(msg->rep, dp->name, dp->namelen,
893 LDNS_RR_TYPE_NS, dclass) ||
894 reply_find_rrset_section_ns(msg->rep, dp->name, dp->namelen,
895 LDNS_RR_TYPE_NS, dclass))
897 /* a DNSKEY set is expected at the zone apex as well */
898 /* this is for 'minimal responses' for DNSKEYs */
899 if(reply_find_rrset_section_an(msg->rep, dp->name, dp->namelen,
900 LDNS_RR_TYPE_DNSKEY, dclass))
906 * check equality of two rrsets
909 * @return true if equal
912 rrset_equal(struct ub_packed_rrset_key* k1, struct ub_packed_rrset_key* k2)
914 struct packed_rrset_data* d1 = (struct packed_rrset_data*)
916 struct packed_rrset_data* d2 = (struct packed_rrset_data*)
919 if(k1->rk.dname_len != k2->rk.dname_len ||
920 k1->rk.flags != k2->rk.flags ||
921 k1->rk.type != k2->rk.type ||
922 k1->rk.rrset_class != k2->rk.rrset_class ||
923 query_dname_compare(k1->rk.dname, k2->rk.dname) != 0)
925 if( /* do not check ttl: d1->ttl != d2->ttl || */
926 d1->count != d2->count ||
927 d1->rrsig_count != d2->rrsig_count ||
928 d1->trust != d2->trust ||
929 d1->security != d2->security)
931 t = d1->count + d1->rrsig_count;
933 if(d1->rr_len[i] != d2->rr_len[i] ||
934 /* no ttl check: d1->rr_ttl[i] != d2->rr_ttl[i] ||*/
935 memcmp(d1->rr_data[i], d2->rr_data[i],
942 /** compare rrsets and sort canonically. Compares rrset name, type, class.
943 * return 0 if equal, +1 if x > y, and -1 if x < y.
946 rrset_canonical_sort_cmp(const void* x, const void* y)
948 struct ub_packed_rrset_key* rrx = *(struct ub_packed_rrset_key**)x;
949 struct ub_packed_rrset_key* rry = *(struct ub_packed_rrset_key**)y;
950 int r = dname_canonical_compare(rrx->rk.dname, rry->rk.dname);
953 if(rrx->rk.type != rry->rk.type) {
954 if(ntohs(rrx->rk.type) > ntohs(rry->rk.type))
958 if(rrx->rk.rrset_class != rry->rk.rrset_class) {
959 if(ntohs(rrx->rk.rrset_class) > ntohs(rry->rk.rrset_class))
967 reply_equal(struct reply_info* p, struct reply_info* q, struct regional* region)
970 struct ub_packed_rrset_key** sorted_p, **sorted_q;
971 if(p->flags != q->flags ||
972 p->qdcount != q->qdcount ||
973 /* do not check TTL, this may differ */
976 p->prefetch_ttl != q->prefetch_ttl ||
978 p->security != q->security ||
979 p->an_numrrsets != q->an_numrrsets ||
980 p->ns_numrrsets != q->ns_numrrsets ||
981 p->ar_numrrsets != q->ar_numrrsets ||
982 p->rrset_count != q->rrset_count)
984 /* sort the rrsets in the authority and additional sections before
985 * compare, the query and answer sections are ordered in the sequence
986 * they should have (eg. one after the other for aliases). */
987 sorted_p = (struct ub_packed_rrset_key**)regional_alloc_init(
988 region, p->rrsets, sizeof(*sorted_p)*p->rrset_count);
989 if(!sorted_p) return 0;
990 log_assert(p->an_numrrsets + p->ns_numrrsets + p->ar_numrrsets <=
992 qsort(sorted_p + p->an_numrrsets, p->ns_numrrsets,
993 sizeof(*sorted_p), rrset_canonical_sort_cmp);
994 qsort(sorted_p + p->an_numrrsets + p->ns_numrrsets, p->ar_numrrsets,
995 sizeof(*sorted_p), rrset_canonical_sort_cmp);
997 sorted_q = (struct ub_packed_rrset_key**)regional_alloc_init(
998 region, q->rrsets, sizeof(*sorted_q)*q->rrset_count);
1000 regional_free_all(region);
1003 log_assert(q->an_numrrsets + q->ns_numrrsets + q->ar_numrrsets <=
1005 qsort(sorted_q + q->an_numrrsets, q->ns_numrrsets,
1006 sizeof(*sorted_q), rrset_canonical_sort_cmp);
1007 qsort(sorted_q + q->an_numrrsets + q->ns_numrrsets, q->ar_numrrsets,
1008 sizeof(*sorted_q), rrset_canonical_sort_cmp);
1010 /* compare the rrsets */
1011 for(i=0; i<p->rrset_count; i++) {
1012 if(!rrset_equal(sorted_p[i], sorted_q[i])) {
1013 if(!rrset_canonical_equal(region, sorted_p[i],
1015 regional_free_all(region);
1020 regional_free_all(region);
1025 caps_strip_reply(struct reply_info* rep)
1029 /* see if message is a referral, in which case the additional and
1030 * NS record cannot be removed */
1031 /* referrals have the AA flag unset (strict check, not elsewhere in
1032 * unbound, but for 0x20 this is very convenient). */
1033 if(!(rep->flags&BIT_AA))
1035 /* remove the additional section from the reply */
1036 if(rep->ar_numrrsets != 0) {
1037 verbose(VERB_ALGO, "caps fallback: removing additional section");
1038 rep->rrset_count -= rep->ar_numrrsets;
1039 rep->ar_numrrsets = 0;
1041 /* is there an NS set in the authority section to remove? */
1042 /* the failure case (Cisco firewalls) only has one rrset in authsec */
1043 for(i=rep->an_numrrsets; i<rep->an_numrrsets+rep->ns_numrrsets; i++) {
1044 struct ub_packed_rrset_key* s = rep->rrsets[i];
1045 if(ntohs(s->rk.type) == LDNS_RR_TYPE_NS) {
1046 /* remove NS rrset and break from loop (loop limits
1048 /* move last rrset into this position (there is no
1049 * additional section any more) */
1050 verbose(VERB_ALGO, "caps fallback: removing NS rrset");
1051 if(i < rep->rrset_count-1)
1052 rep->rrsets[i]=rep->rrsets[rep->rrset_count-1];
1053 rep->rrset_count --;
1054 rep->ns_numrrsets --;
1060 int caps_failed_rcode(struct reply_info* rep)
1062 return !(FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NOERROR ||
1063 FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NXDOMAIN);
1067 iter_store_parentside_rrset(struct module_env* env,
1068 struct ub_packed_rrset_key* rrset)
1070 struct rrset_ref ref;
1071 rrset = packed_rrset_copy_alloc(rrset, env->alloc, *env->now);
1073 log_err("malloc failure in store_parentside_rrset");
1076 rrset->rk.flags |= PACKED_RRSET_PARENT_SIDE;
1077 rrset->entry.hash = rrset_key_hash(&rrset->rk);
1080 /* ignore ret: if it was in the cache, ref updated */
1081 (void)rrset_cache_update(env->rrset_cache, &ref, env->alloc, *env->now);
1084 /** fetch NS record from reply, if any */
1085 static struct ub_packed_rrset_key*
1086 reply_get_NS_rrset(struct reply_info* rep)
1089 for(i=0; i<rep->rrset_count; i++) {
1090 if(rep->rrsets[i]->rk.type == htons(LDNS_RR_TYPE_NS)) {
1091 return rep->rrsets[i];
1098 iter_store_parentside_NS(struct module_env* env, struct reply_info* rep)
1100 struct ub_packed_rrset_key* rrset = reply_get_NS_rrset(rep);
1102 log_rrset_key(VERB_ALGO, "store parent-side NS", rrset);
1103 iter_store_parentside_rrset(env, rrset);
1107 void iter_store_parentside_neg(struct module_env* env,
1108 struct query_info* qinfo, struct reply_info* rep)
1110 /* TTL: NS from referral in iq->deleg_msg,
1111 * or first RR from iq->response,
1112 * or servfail5secs if !iq->response */
1113 time_t ttl = NORR_TTL;
1114 struct ub_packed_rrset_key* neg;
1115 struct packed_rrset_data* newd;
1117 struct ub_packed_rrset_key* rrset = reply_get_NS_rrset(rep);
1118 if(!rrset && rep->rrset_count != 0) rrset = rep->rrsets[0];
1119 if(rrset) ttl = ub_packed_rrset_ttl(rrset);
1121 /* create empty rrset to store */
1122 neg = (struct ub_packed_rrset_key*)regional_alloc(env->scratch,
1123 sizeof(struct ub_packed_rrset_key));
1125 log_err("out of memory in store_parentside_neg");
1128 memset(&neg->entry, 0, sizeof(neg->entry));
1129 neg->entry.key = neg;
1130 neg->rk.type = htons(qinfo->qtype);
1131 neg->rk.rrset_class = htons(qinfo->qclass);
1133 neg->rk.dname = regional_alloc_init(env->scratch, qinfo->qname,
1135 if(!neg->rk.dname) {
1136 log_err("out of memory in store_parentside_neg");
1139 neg->rk.dname_len = qinfo->qname_len;
1140 neg->entry.hash = rrset_key_hash(&neg->rk);
1141 newd = (struct packed_rrset_data*)regional_alloc_zero(env->scratch,
1142 sizeof(struct packed_rrset_data) + sizeof(size_t) +
1143 sizeof(uint8_t*) + sizeof(time_t) + sizeof(uint16_t));
1145 log_err("out of memory in store_parentside_neg");
1148 neg->entry.data = newd;
1150 /* entry must have one RR, otherwise not valid in cache.
1151 * put in one RR with empty rdata: those are ignored as nameserver */
1153 newd->rrsig_count = 0;
1154 newd->trust = rrset_trust_ans_noAA;
1155 newd->rr_len = (size_t*)((uint8_t*)newd +
1156 sizeof(struct packed_rrset_data));
1157 newd->rr_len[0] = 0 /* zero len rdata */ + sizeof(uint16_t);
1158 packed_rrset_ptr_fixup(newd);
1159 newd->rr_ttl[0] = newd->ttl;
1160 sldns_write_uint16(newd->rr_data[0], 0 /* zero len rdata */);
1162 log_rrset_key(VERB_ALGO, "store parent-side negative", neg);
1163 iter_store_parentside_rrset(env, neg);
1167 iter_lookup_parent_NS_from_cache(struct module_env* env, struct delegpt* dp,
1168 struct regional* region, struct query_info* qinfo)
1170 struct ub_packed_rrset_key* akey;
1171 akey = rrset_cache_lookup(env->rrset_cache, dp->name,
1172 dp->namelen, LDNS_RR_TYPE_NS, qinfo->qclass,
1173 PACKED_RRSET_PARENT_SIDE, *env->now, 0);
1175 log_rrset_key(VERB_ALGO, "found parent-side NS in cache", akey);
1176 dp->has_parent_side_NS = 1;
1177 /* and mark the new names as lame */
1178 if(!delegpt_rrset_add_ns(dp, region, akey, 1)) {
1179 lock_rw_unlock(&akey->entry.lock);
1182 lock_rw_unlock(&akey->entry.lock);
1187 int iter_lookup_parent_glue_from_cache(struct module_env* env,
1188 struct delegpt* dp, struct regional* region, struct query_info* qinfo)
1190 struct ub_packed_rrset_key* akey;
1191 struct delegpt_ns* ns;
1192 size_t num = delegpt_count_targets(dp);
1193 for(ns = dp->nslist; ns; ns = ns->next) {
1194 /* get cached parentside A */
1195 akey = rrset_cache_lookup(env->rrset_cache, ns->name,
1196 ns->namelen, LDNS_RR_TYPE_A, qinfo->qclass,
1197 PACKED_RRSET_PARENT_SIDE, *env->now, 0);
1199 log_rrset_key(VERB_ALGO, "found parent-side", akey);
1200 ns->done_pside4 = 1;
1201 /* a negative-cache-element has no addresses it adds */
1202 if(!delegpt_add_rrset_A(dp, region, akey, 1, NULL))
1203 log_err("malloc failure in lookup_parent_glue");
1204 lock_rw_unlock(&akey->entry.lock);
1206 /* get cached parentside AAAA */
1207 akey = rrset_cache_lookup(env->rrset_cache, ns->name,
1208 ns->namelen, LDNS_RR_TYPE_AAAA, qinfo->qclass,
1209 PACKED_RRSET_PARENT_SIDE, *env->now, 0);
1211 log_rrset_key(VERB_ALGO, "found parent-side", akey);
1212 ns->done_pside6 = 1;
1213 /* a negative-cache-element has no addresses it adds */
1214 if(!delegpt_add_rrset_AAAA(dp, region, akey, 1, NULL))
1215 log_err("malloc failure in lookup_parent_glue");
1216 lock_rw_unlock(&akey->entry.lock);
1219 /* see if new (but lame) addresses have become available */
1220 return delegpt_count_targets(dp) != num;
1224 iter_get_next_root(struct iter_hints* hints, struct iter_forwards* fwd,
1227 uint16_t c1 = *c, c2 = *c;
1228 int r1 = hints_next_root(hints, &c1);
1229 int r2 = forwards_next_root(fwd, &c2);
1230 if(!r1 && !r2) /* got none, end of list */
1232 else if(!r1) /* got one, return that */
1236 else if(c1 < c2) /* got both take smallest */
1243 iter_scrub_ds(struct dns_msg* msg, struct ub_packed_rrset_key* ns, uint8_t* z)
1245 /* Only the DS record for the delegation itself is expected.
1246 * We allow DS for everything between the bailiwick and the
1247 * zonecut, thus DS records must be at or above the zonecut.
1248 * And the DS records must be below the server authority zone.
1249 * The answer section is already scrubbed. */
1250 size_t i = msg->rep->an_numrrsets;
1251 while(i < (msg->rep->an_numrrsets + msg->rep->ns_numrrsets)) {
1252 struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
1253 if(ntohs(s->rk.type) == LDNS_RR_TYPE_DS &&
1254 (!ns || !dname_subdomain_c(ns->rk.dname, s->rk.dname)
1255 || query_dname_compare(z, s->rk.dname) == 0)) {
1256 log_nametypeclass(VERB_ALGO, "removing irrelevant DS",
1257 s->rk.dname, ntohs(s->rk.type),
1258 ntohs(s->rk.rrset_class));
1259 memmove(msg->rep->rrsets+i, msg->rep->rrsets+i+1,
1260 sizeof(struct ub_packed_rrset_key*) *
1261 (msg->rep->rrset_count-i-1));
1262 msg->rep->ns_numrrsets--;
1263 msg->rep->rrset_count--;
1264 /* stay at same i, but new record */
1272 iter_scrub_nxdomain(struct dns_msg* msg)
1274 if(msg->rep->an_numrrsets == 0)
1277 memmove(msg->rep->rrsets, msg->rep->rrsets+msg->rep->an_numrrsets,
1278 sizeof(struct ub_packed_rrset_key*) *
1279 (msg->rep->rrset_count-msg->rep->an_numrrsets));
1280 msg->rep->rrset_count -= msg->rep->an_numrrsets;
1281 msg->rep->an_numrrsets = 0;
1284 void iter_dec_attempts(struct delegpt* dp, int d)
1286 struct delegpt_addr* a;
1287 for(a=dp->target_list; a; a = a->next_target) {
1288 if(a->attempts >= OUTBOUND_MSG_RETRY) {
1289 /* add back to result list */
1290 a->next_result = dp->result_list;
1291 dp->result_list = a;
1295 else a->attempts = 0;
1299 void iter_merge_retry_counts(struct delegpt* dp, struct delegpt* old)
1301 struct delegpt_addr* a, *o, *prev;
1302 for(a=dp->target_list; a; a = a->next_target) {
1303 o = delegpt_find_addr(old, &a->addr, a->addrlen);
1305 log_addr(VERB_ALGO, "copy attempt count previous dp",
1306 &a->addr, a->addrlen);
1307 a->attempts = o->attempts;
1311 a = dp->usable_list;
1313 if(a->attempts >= OUTBOUND_MSG_RETRY) {
1314 log_addr(VERB_ALGO, "remove from usable list dp",
1315 &a->addr, a->addrlen);
1316 /* remove from result list */
1318 prev->next_usable = a->next_usable;
1319 else dp->usable_list = a->next_usable;
1320 /* prev stays the same */
1330 iter_ds_toolow(struct dns_msg* msg, struct delegpt* dp)
1332 /* if for query example.com, there is example.com SOA or a subdomain
1333 * of example.com, then we are too low and need to fetch NS. */
1335 /* if we have a DNAME or CNAME we are probably wrong */
1336 /* if we have a qtype DS in the answer section, its fine */
1337 for(i=0; i < msg->rep->an_numrrsets; i++) {
1338 struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
1339 if(ntohs(s->rk.type) == LDNS_RR_TYPE_DNAME ||
1340 ntohs(s->rk.type) == LDNS_RR_TYPE_CNAME) {
1341 /* not the right answer, maybe too low, check the
1342 * RRSIG signer name (if there is any) for a hint
1343 * that it is from the dp zone anyway */
1346 val_find_rrset_signer(s, &sname, &slen);
1347 if(sname && query_dname_compare(dp->name, sname)==0)
1348 return 0; /* it is fine, from the right dp */
1351 if(ntohs(s->rk.type) == LDNS_RR_TYPE_DS)
1352 return 0; /* fine, we have a DS record */
1354 for(i=msg->rep->an_numrrsets;
1355 i < msg->rep->an_numrrsets + msg->rep->ns_numrrsets; i++) {
1356 struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
1357 if(ntohs(s->rk.type) == LDNS_RR_TYPE_SOA) {
1358 if(dname_subdomain_c(s->rk.dname, msg->qinfo.qname))
1359 return 1; /* point is too low */
1360 if(query_dname_compare(s->rk.dname, dp->name)==0)
1361 return 0; /* right dp */
1363 if(ntohs(s->rk.type) == LDNS_RR_TYPE_NSEC ||
1364 ntohs(s->rk.type) == LDNS_RR_TYPE_NSEC3) {
1367 val_find_rrset_signer(s, &sname, &slen);
1368 if(sname && query_dname_compare(dp->name, sname)==0)
1369 return 0; /* it is fine, from the right dp */
1373 /* we do not know */
1377 int iter_dp_cangodown(struct query_info* qinfo, struct delegpt* dp)
1379 /* no delegation point, do not see how we can go down,
1380 * robust check, it should really exist */
1383 /* see if dp equals the qname, then we cannot go down further */
1384 if(query_dname_compare(qinfo->qname, dp->name) == 0)
1386 /* if dp is one label above the name we also cannot go down further */
1387 if(dname_count_labels(qinfo->qname) == dp->namelabs+1)
1393 iter_stub_fwd_no_cache(struct module_qstate *qstate, struct query_info *qinf)
1395 struct iter_hints_stub *stub;
1398 /* Check for stub. */
1399 stub = hints_lookup_stub(qstate->env->hints, qinf->qname,
1400 qinf->qclass, NULL);
1401 dp = forwards_lookup(qstate->env->fwds, qinf->qname, qinf->qclass);
1403 /* see if forward or stub is more pertinent */
1404 if(stub && stub->dp && dp) {
1405 if(dname_strict_subdomain(dp->name, dp->namelabs,
1406 stub->dp->name, stub->dp->namelabs)) {
1407 stub = NULL; /* ignore stub, forward is lower */
1409 dp = NULL; /* ignore forward, stub is lower */
1414 if (stub != NULL && stub->dp != NULL) {
1415 if(stub->dp->no_cache) {
1418 dname_str(qinf->qname, qname);
1419 dname_str(stub->dp->name, dpname);
1420 verbose(VERB_ALGO, "stub for %s %s has no_cache", qname, dpname);
1422 return (stub->dp->no_cache);
1425 /* Check for forward. */
1430 dname_str(qinf->qname, qname);
1431 dname_str(dp->name, dpname);
1432 verbose(VERB_ALGO, "forward for %s %s has no_cache", qname, dpname);
1434 return (dp->no_cache);